Vous êtes sur la page 1sur 96

Assessing the impact of late Pleistocene megafaunal extinctions on global vegetation and climate

Marc-Olivier BRAULT

Master of Science

Department of Atmospheric and Oceanic Sciences

McGill University Montreal, Quebec, Canada June 28, 2012

A thesis submitted to the Faculty of Graduate and Postdoctoral Studies in partial fulfillment of the requirements for the degree of Master of Science Marc-Olivier BRAULT, June 2012. All rights reserved

iii

ABSTRACT

The end of the Pleistocene marked a turning point for the Earth system, as climate gradually emerged from millennia of severe glaciation in the Northern Hemisphere. It is widely known that the deglacial climate change then was accompanied by an unprecedented decline in many species of large terrestrial mammals, featuring among others the near-total eradication of the woolly mammoth. Due to a herbivorous diet that involved the grazing of a large number of trees, their extinction is thought to have contributed to the rapid and well-documented expansion of dwarf deciduous trees in Siberia and Beringia, which in turn would have resulted in a significant reduction in surface albedo, leading to an increase in global temperature. In this study, we use the UVic ESCM to simulate various scenarios of the megafaunal extinctions, ranging from the catastrophic to more realistic cases, in order to quantify their potential impact on the climate system, and investigate the associated biogeophysical feedbacks between the growing vegetation and rising temperatures. The more realistic experiments include sensitivity tests based on the timing of extinction, tree clearance ration, and size of habitat, as well as a gradual extinction and a simulation involving free (non-prescribed) atmospheric CO2. Overall, most of the paleoclimate simulations and the sensitivity tests yield results that correspond well with our intuition. For the maximum impact scenario, we obtain a surface albedo increase of 0.006, which translates into a global warming of 0.175 C; these numbers are comparable in magnitude to those in similar studies.

iv

ABRG

La fin de lpoque du Plistocne est une tape importante de lhistoire climatique de la Terre. En effet, cest lors de cette priode mouvemente que notre plante sest pour une ultime fois libre des conditions glaciales qui perduraient depuis des dizaines de millnaires, et souvent marques par la prsence dimposante calottes glaciaires dans lhmisphre nord. Il est bien connu que ce changement climatique fut galement accompagn dun dclin sans prcdent de plusieurs espces de grands mammifres terrestres, y compris une extermination rapide et brutale du mammouth laineux. En raison dune dite compose en partie de vgtaux provenant darbres prolifiques durant cette priode, il y a de fortes raisons de croire que les ceux-ci auraient pu contribuer au maintien dune faible densit forestire au sein de leur habitat. Par consquent, leur extinction aurait contribu une rapide mergence dune varit de petits arbres feuillus tant en Sibrie quen Bringie, provoquant par la mme occasion une rduction considrable de lalbdo de surface, qui son tour aurait entrain une augmentation globale de la temprature. Lobjectif vis par cette tude est de quantifier limpact potentiel quaurai t pu avoir une extinction majeure de la mgafaune sur le climat de la Terre, par le biais dune modification de la carte vgtale menant une hausse de la temprature. Afin dexaminer en dtail la rtroaction de processus biogophysiques ce changement de temprature, nous employons le modle de complexit intermdiaire de lUniversit de Victoria (UVic) avec des scnarios plus ou moins ralistes, dont une catastrophe aux proportions exagres servant dterminer les limites de que peut offrir le modle UVic. Parmi les cas plus terre--terre figurent quelques tests de sensibilit mens sur des paramtres tels que le taux de dboisement des mammouths, la grandeur de leur habitat, ainsi que lanne de leur extinction. Dautres expriences ayant t menes portent sur un talement graduel dun dclin des populations de mgaherbivores, ainsi quune simulation laissant libre cours aux changes de carbone entre latmosphre et les autres constituants du systme climatique, en autres mots une libre variation du niveau de CO2 dans latmosphre. En gnral, nous obtenons des rsultats qui se conforment assez bien avec ceux dtudes similaires. Dans le cas dun scnario catastrophique, nous enregistrons une baisse de lalbdo terrestre quivalent un peu moins de 0.006, donnant lieu une hausse de la temprature se chiffrant 0.175C globalement. Quant aux expriences plus ralistes, les rsultats en trs grande majorit confirment notre intuition.

ACKNOWLEDGEMENTS

First and foremost, this project would not have been possible without the support and guidance of Dr. Lawrence Mysak. Since the day he introduced me to the topic, he has contributed to the project in a variety of ways, from our numerous meetings, the friendly advice, and through his careful editing of this thesis and other paperwork. I am also appreciative of all the opportunities he offered me, and for getting me to meet with all sorts of new people. But above all else, I must commend his positive energy, unshakeable enthusiasm, and the patience which he has shown me over the course of the past year. I owe many thanks to Dr. Damon Matthews for providing assistance with the UVic model and for giving crucial suggestions that helped bring the project forward. Our meetings were few and with often with very short notice, but somehow he always managed to make me put things into perspective, and find answers to many of my questions. I am also indebted towards Dr. Jaime Palter, who agreed to act as supervisor to this project within the Department of Atmospheric and Oceanic Sciences. Her involvement with the project at different levels, especially in the writing of this thesis, is greatly appreciated. I am much obliged towards my good friend and fellow graduate student Christopher Simmons, who so generously offered his own time when I needed it the most. In providing me with an IDL script to simulate the megafaunal extinctions within the UVic model, he effectively put me on the right track to get started with the experimentation. Besides that, our discussions were always interesting and constructive, and they often helped me clarify things about the model and its underlying physics. A special mention should also be given to the AOS network administrator Michael Havas, who frequently aided me in the constant fight against my greatest foe computers! I was especially impressed when I sent a complaint on a Saturday evening, only to find that on the following Sunday the problem had already been resolved! This work has been funded by scholarships awarded to Marc-Olivier Brault by the Natural Sciences and Engineering Research Council (NSERC), the Global Environmental and Climate Change Centre (GEC3), and an NSERC Discovery grant awarded to Lawrence Mysak. I am thankful for this financial support. Finally, much love towards my immediate family, who as usual went beyond the call of duty in their unconditional support and faith in me throughout all these years. The endless hours spent on the phone and the countless Ottawa-Montreal trips testify to their patience and generosity; they are the very reason I have come this far. To all these people, those few words cannot even begin to express my gratitude.

Contents
ABSTRACT ABRG ACKNOWLEDGEMENTS LIST OF TABLES AND FIGURES 1 2 INTRODUCTION CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS 2.1 2.2 iii iv v xi 1 7

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Forests as interactive components of the climate system . . . . . 8


2.2.1 2.2.2 2.2.3 Early work: biosphere-atmosphere interaction in the tropics . . . 8 Biogeophysical mechanism in high latitude forests . . . . . . . . 9 Investigating the climatic impacts of boreal deforestation: the major numerical experiments . . . . . . . . . . . . . . . . . . . . 10 Climatic impact of the global forest cover changes. . . . . . . . 13

2.2.4

2.3

Climate-biosphere interactions in paleoclimate simulations . 15


2.3.1 2.3.2 2.3.3 2.3.4 The Mid-Holocene . . . . . . . . . . . . . . . . . . . . . . . . . . 15 The Last Glacial Maximum . . . . . . . . . . . . . . . . . . . . . 17 Earlier periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Stability of the climate-vegetation system . . . . . . . . . . . . . 21

2.4

Present-day interactions between climate and the biosphere: analyzing vegetation response to climate change . . . . . . . . . 23
vii

CONTENTS 2.4.1 2.4.2 2.4.3

viii Global vegetation feedback to increases in atmospheric CO2 . . 23 Climate response to high-latitude afforestation . . . . . . . . . . 24 Climate response to anthropogenic land cover change . . . . . . 25

2.5 3

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 27

MODEL DESCRIPTION 3.1 3.2

Earth system Models of Intermediate Complexity . . . . . . . . 27 General description of the UVic ESCM . . . . . . . . . . . . . . . 28
3.2.1 3.2.2 3.2.3 3.2.4 Atmosphere . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Sea ice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Ocean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Coupling strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.3

Recent additions and improvements to the model . . . . . . . . 33


3.3.1 3.3.2 Enhanced radiative transfer model . . . . . . . . . . . . . . . . . 33 Land surface scheme . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.4

Description of the vegetation module . . . . . . . . . . . . . . . . 35


3.4.1 3.4.2 3.4.3 3.4.4 3.4.5 3.4.6 3.4.7 Evolution of vegetation modeling. . . . . . . . . . . . . . . . . . 35 An overview of Dynamic Global Vegetation Models (DGVMs) 36 The Plant Functional Type (PFT) approach . . . . . . . . . . . . 38 General description of TRIFFID . . . . . . . . . . . . . . . . . . 38 Vegetation dynamics . . . . . . . . . . . . . . . . . . . . . . . . . 39 Leaf phenology and soil carbon . . . . . . . . . . . . . . . . . . . 40 Biophysical parameters in MOSES-2 . . . . . . . . . . . . . . . 41

CONTENTS 3.4.8

ix Coupling with the UVic ESCM . . . . . . . . . . . . . . . . . . . 42

RESULTS OF THE TRANSIENT SIMULATIONS 4.1 4.2

43

An overview of the original study by Doughty et al. (2010) . . 43 Description of the present experiment . . . . . . . . . . . . . . . 45
4.2.1 4.2.2 Differences with the original study . . . . . . . . . . . . . . . . . 45 Experimental approach . . . . . . . . . . . . . . . . . . . . . . . . 46

4.3

The maximum impact scenario . . . . . . . . . . . . . . . . . . . . 47


4.3.1 4.3.2 4.3.3 4.3.4 4.3.5 Short description and parameter tuning . . . . . . . . . . . . . . 47 Vegetation and surface albedo changes . . . . . . . . . . . . . . 48 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Precipitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 Sea ice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.4

A set of more realistic experiments. . . . . . . . . . . . . . . . . . 59


4.4.1 4.4.2 4.4.3 4.4.4 Description of the experiments . . . . . . . . . . . . . . . . . . . 59 Sensitivity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 Gradual extinction experiment . . . . . . . . . . . . . . . . . . . 64 Free CO2 experiment . . . . . . . . . . . . . . . . . . . . . . . . . 66

CONCLUSIONS 5.1 5.2

69

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 73

REFERENCES

List of tables and figures


Tables

4.1

List of experiments used in the sensitivity study and their parameterizations. Results from entries in bold are shown in Figure 4.13 in the form of a world map of temperature anomalies 500 years after the prescribed extinction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Figures

4.1

Change in vegetation fraction over the mammoth habitat (all land north of 30N) simulated by the UVic ESCM in the context of a maximum impact scenario. This figure and every subsequent one represent the difference between a simulation where mammoths go extinct, and a simulation where their extinction does not occur. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Change in surface albedo over the mammoth habitat (all land north of 30N) simulated by the UVic ESCM in the context of a maximum impact scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 (a) World map showing the spatial distribution of albedo changes after 500 years of climate model simulations; (b) Map depicting the size and location of ice sheets at the end of the simulation. . . . . . . . . . . . . . . . . . 51 Annual cycle of land surface albedo anomaly in the Northern Hemisphere during the last year of climate model simulations. Solid line represent positive contours, while dotted lines represent negative values. On the abscissa, months are displayed from January to December according to their numerical order. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Globally-averaged temperature increase due to biogeophysical effects only, in the context of a maximum impact scenario. . . . . . . . . . . . . . . . . . . . 53 (a) Zonally-averaged temperature difference between the extinction and no-extinction runs; (b) spatial distribution of the temperature anomaly. The dotted lines represent 0.05C isotherms. . . . . . . . . . . . . . . . . . . . . 54 Zonally-averaged, annual cycle of temperature anomalies over the northern Hemisphere. The contour interval of the isotherms is 0.1C. On the abscissa, months are displayed from January to December in their numerical order. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 x

4.2

4.3

4.4

4.5

4.6

4.7

LIST OF TABLES AND FIGURES 4.8

xi

Variations in 14C anomaly as a function of depth. This particular snapshot is taken in the Weddell Sea, in the middle of the cold anomaly in Fig. 4.6(b), and averaged for the entire last year of the simulation. . . . . . . . 56 Change in total precipitation rates, shown for land only and land + ocean. . . . 57 Annual cycle of precipitation anomalies in the Northern Hemisphere during the last year of model simulations. Solid lines represent positive contours, while dotted lines represent negative values. The contour interval is in units of 10-7 kg m2 s-1. On the abscissa, months are displayed from January to December according to their numerical order.. . . . . . . . . . 58 Sea-ice thickness anomaly. Left panel : Global change in sea ice volume, over the course of the simulation. Right panel : Thickness anomaly over the Arctic Ocean. This particular snapshot represents the 5-day average of days 255-260 (out of 365) of simulation year -11500, or 500 years after the extinction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Results of the sensitivity tests, presented here as a timeseries of temperature anomalies. The maximum impact scenario is shown in red for the sake of comparison. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 Spatial distribution of temperature anomalies for various simulations in the set of sensitivity experiments. The number besides each panel refers to the that of the specific experiment in Table 4.1. All of these figures are one-year averaged differences in temperature between the simulation and a related no extinction simulation with similar parameterizations. The year of averaging is 500 yrs after extinction. . . . . . . . . . . . . . . . . . . . . 62 Results of the gradual extinction experiments, presented in the form of temperature-albedo graphs. The four panels represent each of the individual simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 A selection of results from the free CO2 experiment. (a) a comparison of the temperature anomaly between the free and prescribed CO2 experiments; (b) difference in atmospheric CO2 between the two simulations; (c) change in total soil carbon resulting from the vegetation change; (d) carbon flux from the atmosphere to the land (since it is mostly negative, it indicates a land-to-atmosphere flux. . . . . . . . . . . . . . . . . . . 67

4.9 4.10

4.11

4.12

4.13

4.14

4.15

Chapter 1
Introduction
The spatial and temporal fluctuations in climate have attracted the attention of a number of scientific minds over the course of the last century. On the one hand, studies of short-term future variations of global climate and its implications on the human society are becoming increasingly popular. On the other hand, many have turned to the study of paleoclimates, not only in order to explore the rich history of Earths climatic evolution, but also as a means to provide more information to the former group by relating past climatic phenomena to similar occurrences in present-day or near-future climatic conditions. In particular, the past few decades have seen a considerable improvement of research tools for climate studies, leading to the emergence of several research efforts to document and describe some of the most intriguing events or periods in the Earths climatic history. Different means could be employed for this end: whereas some used an exhaustive analysis of various proxies to reconstruct past states of the Earth system, others would draw conclusions based on simulations of the major interactions and feedback mechanisms obtained through climate modeling. While somewhat different in their methodological approach, both strategies aimed towards a better understanding of climate system evolution at all time scales, and its sensitivity to external forcings. Extensive analyses of various climate proxies, notably Greenland ice cores and deep sea sediment records, revealed that the Earths climatic history through the greater part of the Quaternary (beginning approx. 1.8 My ago) was characterized by colder temperatures, lower sea levels and severe glaciations in the Northern Hemisphere (NH). Extended periods of glacial climate were punctuated by brief interglacials marked by a return to temperate conditions in the NH, forming a cycle of NH glaciations that repeated itself over time with striking periodicity. The pioneering work of Hays et al. (1976) and Imbrie et al. (1980) identified the Milankovitch cycles of solar insolation (periodic variations in the Earths orbital cycle that lead to latitudinallydependent seasonal changes in incoming solar radiation) as the main source of geologic-scale 1

CHAPTER 1. INTRODUCTION

climate variability during the Quaternary, establishing the framework for the astronomical theory of climate. However, if orbital forcing is now generally recognized as the underlying cause behind the glacial-interglacial cycles, it alone comes far from explaining many aspects of the temperature and ice-volume profiles obtained from climate proxies, especially at subMilankovitch timescales (<20 ky). Even today, the diagnosis of physical and biogeochemical processes and feedback mechanisms behind deglacial climate change (and its opposite, glacial inception) remains one of the most challenging problems in paleoclimate research. The final millennia of the Pleistocene epoch (which lasted from the start of the Quaternary until about 11.7 ky ago) marked an important transitional period in the climatic timeline as the Earth emerged from the latest recorded bout of widespread NH glaciations and entered the warm and stable conditions of the current Holocene epoch (11.7 kyr ago present). This transition involved major changes in the land surface configuration. Ice sheets which had dominated the continental landscape at the Last Glacial Maximum (LGM, 21 ky BP) began to recede, and eventually disappeared from the mainland. In their wake came tundra vegetation a combination of cold-adapted short grasses lichens, and mosses which in turn would be replaced by boreal forests of evergreen needleleaf trees and dwarf deciduous trees, in locations where climate became favorable to the maintenance of such ecosystems. These changes in the land surface had a profound impact on the climate of the late Pleistocene, often acting as a positive feedback to global warming and reinforcing the positive energy imbalance. Of course, the last Pleistocene deglaciation becomes especially important relative to other similar occurrences in the cycle of NH glaciations in light of the events that followed it, notably the rise of the human civilization and the onset of the anthropogenic era, both of which occurred during the relatively short time span of the Holocene. In any assessment of present-day or near-future climate it is impossible to avoid the effects of human activity, especially in regards to an increase in atmospheric carbon dioxide of unparalleled abruptness. The Holocene

interglacial is also unusual in its length, and some would argue that a delay in the next glacial

CHAPTER 1. INTRODUCTION

inception (and a possible termination of the Quaternary glacial cycles) would be directly attributable to the impacts of increased greenhouse gas levels on the Earths energy imbalance (see, for example, Mysak, 2008). Another uncharacteristic aspect of the late Pleistocene deglaciation is that it coincided with the extinction of at least 34 genera of megafaunal mammals (also called Late Quaternary Extinctions, or LQE), one of the most significant shocks on faunal biodiversity during the past 55 million years (Koch and Barnosky, 2006). The mass extinction was a discontinuous event spread over 50,000 years (and thus not entirely constrained within the time frame of deglacial climate change), and consisting of a series of short-term diachronous pulses; nonetheless, it is generally recognized that most of the extinctions did not continue into the Holocene (Barnosky et al., 2004). Numerous theories and hypotheses have been put forward to provide a tentative explanation for the rapid decline of the Pleistocene megafauna. Most of these hypotheses would fall into one of two categories: those that favored environmental causes (for example, Thomas et al., 2004), and those who insisted on the role of human intervention (for example, Alroy, 2001; Wroe et al., 2004). Included among the former category were topics such as: a direct or indirect impact of climate change (e.g. through a change in vegetation that would reduce access to optimal food, or a loss of habitat due to the rise in sea level), a change in population dynamics leading to overwhelming competition, and a regional catastrophe (bolide impact!). Proposals in the latter category emphasized the role of man through various scenarios: an artificial modification of the habitat, the introduction of new predators and alien diseases, or any form of the overkill hypothesis (Webb and Barnosky, 1989). For many years there was no perceived middle-ground between the two set of hypotheses, and a fierce debate raged between proponents of either faction (Barnosky et al., 2004). Detractors of the overkill hypothesis argued that some of the extinct species included mammal and avian genera that were not vulnerable to hunting (i.e., not attractive to the presumed hunters),

CHAPTER 1. INTRODUCTION

and that in any case the evidence supporting the systematic hunting of megafauna by human tribes was defined based on ambiguous parameters, and was at best inconclusive. On the other hand, many criticized the environmental hypotheses because they could not explain what was so different about the late Pleistocene that would have driven such a large number of species to the brink of extinction, whereas previous deglaciations had witnessed nothing of the sort. Theories giving most of the credit to climate change also failed to explain why extinction patterns were localized; indeed, the presumed date of extinction for each species varied inconsistently with geographical location, and in some cases a descendant species was shown to have survived many thousand years into the Holocene after relocating to a remote location (for example, a smaller version of the woolly mammoth, often called dwarf mammoth, is believed to have survived until, 2000 BC on small islands off the Siberian coast). Recent investigations have reinforced the claim of human intervention in the catastrophe. Using evidence from paleontology, climatology, archaeology and ecology, it was determined that early human tribes likely had a role in the extinction of some species, with a strong level of confidence for human activity in North America, Africa, and Australia (Barnosky et al., 2004; Koch and Barnosky, 2006). The evidence also appeared to be stronger on islands where humans were known to have settled. However, it was noted that humans could not be responsible for extinctions everywhere on the planet, and that it would be oversimplistic to pretend that hunting alone could have caused the eradication of so many species prior to the Holocene. Instead, the authors wrote off the human factor as an additional stress on the endangered species, which when combined with a rapidly evolving environmental context, would have driven them to famine, exhaustion, and eventually extinction. As a result, while the debate is still ongoing, many have adopted this point of view, accepting that in the end the LQE are likely a combination of both natural and anthropogenic factors. Some of the larger extinct genera, known to have a strong interaction with the surrounding vegetation, have long been thought to play a central role in the mass extinction

CHAPTER 1. INTRODUCTION

because their departure would have triggered a positive feedback from the vegetation, further aggravating the situation for other threatened species (Owen-Smith, 1987). Among these

terrestrial megaherbivores, the case for human implication in the extinction of the woolly mammoth is especially strong due to their body size, exceptionally slow gestation period, and abundance of archaeological evidence found at Paleolithic sites in Siberia (Guthrie, 2006). Due to the perceived resemblance with their elephant successors, as well as strong evidence for the inclusion of various Pleistocene tree species into their diet, there is compelling reason to believe that mammoths played a dominant role in the maintenance of grasslands over the expansion of trees in the Eurasian taiga much like elephants are maintaining the African savanna and therefore their extinction would have triggered a significant recovery of forest biomes at the NH northern latitudes. In such a mindset, should the suspicions of human involvement in the mammoth extinction happen to be well-founded, the combination of all of the above would have the surprising consequence of redefining the onset of anthropogenic influences on climate. It is not the first time that scientists challenge the idea that the Anthropocene is entirely constrained within the past two hundred years. In a novel paper, Ruddiman (2003) proposed an alternative explanation to the observed positive trend in greenhouse gas levels during the midHolocene (which should have been negative in casual circumstances), by linking them with the start of agriculture in Eurasia, and thus associating the increases in CO2 and CH4 to related activities such as forest clearance (starting 8000 years ago) and rice irrigation (starting 5000 years ago). In a similar manner, Doughty et al. (2010) proposed that the start of the

anthropogenic era should be pushed back an additional several thousand years, by associating the megafaunal extinctions (and the likely involvement of human hunters) with significant modifications in the distribution of terrestrial vegetation. In particular, they suggest that the extinctions played a pivotal role in the rapid expansion of Betula trees in Siberia and parts of Beringia, and that an increase in surface darkness led to a significant warming over these regions. In their paper, this assertion is backed up by a combination of proxy data analysis and climate model simulations.

CHAPTER 1. INTRODUCTION

Following on from the work of Doughty et al. (2010), the main focus of this thesis is to simulate the deglacial climate of the late Pleistocene in the context of megafaunal extinctions for the purpose of obtaining a quantitative measurement of the latters impact on the climate system. However, our objective varies from that of Doughty et al. (2010): whereas the aim of the original paper was to provide solid argumentation in support of the authors novel claim on the first potential case of human-induced global warming, in this thesis we propose an in-depth assessment of biophysical interactions between the fauna, flora, and climate. Concurrently, we wish to extend the modeling effort presented in Doughty et al. (2010), by executing long-term transient simulations of the Earth system with the University of Victoria Earth System Climate Model (henceforth UVic ESCM), a fully coupled global climate model of intermediate complexity which includes, among others, a dynamical treatment of vegetation feedbacks. The impacts of the megafaunal extinctions are to be prescribed directly into the models vegetation component, first as geographically-dependent perturbation that reduces tree cover (while the mammoth are alive and roaming the land), and then as a release of that perturbation (when they go extinct), with the subsequent recovery of forest biomes acting as the main driving force for climate change. Most of our analysis will focus on a single scenario of the most extreme catastrophe, which we dubbed maximum impact scenario, and whose purpose is to quantif y the largest response that can be obtained from the UVic as a result of the megafaunal extinctions. However, due to the obvious lack of realism of the latter case, we have also included results from simulations that represent a more likely outcome. The thesis is structured as follows. Chapter 2 reviews the available literature on climatevegetation interactions, with a particular emphasis on boreal forest feedbacks and paleoclimate studies. Chapter 3 describes the climate model used in this study, as well as the dynamical vegetation model involved in the simulations. In Chapter 4, we present the experimental context, the methodology, and our analysis of the model output. Finally, the thesis conclusions are given in Chapter 5.

Chapter 2
Climate-vegetation interactions and feedbacks
2.1 Introduction
The deep and complex role of vegetation within the Earth system provides one of the finest examples of climate-biosphere interactions, involving a set of biogeophysical and biogeochemical processes that couple it with various components of the climate system. The relationship is twofold: on the one hand, climate (as defined by the annual average in air temperature and precipitation) has long been known as a prime factor in determining the spatial coverage and distribution of vegetation as well as the structural and phenological properties of plants. In fact, the first systems of climate classification used vegetation almost exclusively in their definitions of climatic zones, because flora was thought of as an exact mirror of temperature and precipitation patterns (Kppen, 1936). On the other hand, it has become increasingly clear in recent decades that vegetation dynamics comprises a major climate forcing, exerting its influence through biogeophysical processes which alter the radiative, hydrological and turbulent properties of the land surface and through biogeochemical effects which modify the atmospheric gas composition (carbon dioxide, methane and nitrogen dioxide, to name a few), ocean chemistry and soil carbon content (Kabat et al., 2004). The purpose of this literature review is to gain a better understanding of climatevegetation interactions. The chapter first discusses in section 2.2 the basic physical concepts involved in climate-biosphere interactions, with a special focus on Arctic-boreal vegetation; it also reviews numerical modeling papers which discuss boreal and global deforestation experiments. Section 2.3 covers some of the literature on the role of high northern vegetation on past climatic changes such as the mid-Holocene climatic optimum and the Last Glacial Maximum. Finally, section 2.4 deals with contemporary issues surrounding forest vegetation as a component of the climate system. It should also be noted that, although biogeophysical and

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

biogeochemical processes are equally important in the broad spectrum of climate-vegetation interactions, the former is central to this study and therefore the main topic of this chapter.

2.2

Forests as interactive components of the climate system

2.2.1 Early work: atmosphere-biosphere interaction in the tropics


The study of climate-vegetation interactions has attracted an increasing amount of interest over the past few decades in the climate modeling community. Charney et al. (1975) were the first to investigate feedback mechanisms between land surface processes and the climate system. In a pioneering study, they used a then state-of-the-art General Circulation Model (GCM) to simulate the climate response to a decrease in vegetation cover in the Sahara region (parameterized as an increase in surface albedo). The model output revealed a significant decrease in rainfall caused by the reduced surface heating, which led them to conclude that land surface processes could be responsible for the self-induction of low-latitude deserts. Another emerging issue at the time was the possible climatic impacts of tropical land cover changes. Some of the earliest modeling studies on climate-vegetation interactions (e.g., Potter et al., 1975; Dickinson and Henderson-Sellers, 1988) were concerned with the short-term impacts of large-scale deforestation in the Amazonian rainforest. In particular, Dickinson and Henderson-Sellers (1988) observed that a replacement of tropical vegetation with impoverished grassland resulted in warmer temperatures and a notoriously drier soil, which would not only be detrimental to the survival of any remaining woodland, thus igniting a potentially irreversible feedback between climate and vegetation loss, but would also compromise the very motivation behind this massive deforestation that is, to create more space for arable land. Further studies (Shukla et al., 1990; Nobre et al., 1991, Henderson-Sellers et al., 1993) confirmed the establishment of warmer, drier conditions, along with significant alterations in evaporation and net surface radiation, and found that a reduction in vegetation cover would lead to the

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

development of a lengthy dry season in the affected regions, creating conditions similar to those that are thought to have prevailed in the tropics during the last major glaciation. The causes behind this climatic response to tropical deforestation are well known. Vegetation in general, and especially broadleaved evergreen trees, contributes to moisture recycling in many different ways. Aside from direct evapotranspiration, plants can also extract additional water from deep soil layers as well as increase surface roughness (and therefore atmospheric turbulence), both of which act to further increase water vapor input to the atmosphere (Meir et al., 2006). Consequently, plant vegetation adds moisture to the surrounding environment and promotes ambient air cooling through evaporative latent heat release, the sum of which indirectly contributes to creating a cool, moist boundary layer that enhances precipitation (Bonan, 2008). It comes as no surprise, then, that the loss of these processes upon deforestation along with a corresponding reduction in carbon sequestration results in warmer, drier conditions locally which can also act as a major perturbation on atmospheric dynamics in the tropics.

2.2.2 Biogeophysical mechanism in high-latitude forests


Climate-vegetation interactions in high-latitude woodlands are dominated by radiative feedbacks, rather than hydrologic processes. This can be partially attributed to the limited amount of moisture in these regions, but it is mainly due to the very large difference in surface reflectivity between the dark forest canopy and bare or grass-covered ground, most of which becomes snow-covered during the winter and early spring. Data analyses from flux tower measurements in the mid-latitudes (Betts and Ball, 1997) reveal that surface albedo over the forests in winter can be as low as 0.3 (a little higher for deciduous trees), which is quite a contrast to that of bare soil, which can exceed 0.8 in the wake of a decent snowfall. Such a massive difference in surface reflectivity, due to a masking of snow-covered ground by tree cover, leads to an important radiative feedback between vegetation and surface air temperature. The expanding vegetation cover (often expressed as leaf area index) reduces surface albedo

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

10

during the cold season, which in turn favors an early spring snow melt, resulting in warmer temperatures that further enhance the spread of vegetation. Therefore, it is widely understood that high-latitude forest vegetation provides a significant warming contribution on both local and global scales, and that any successful model simulation of high-latitude seasonal variability must include these land surface processes in order to accurately reproduce the annual cycle of temperature change (Wilson et al., 1987). Since land cover changes are also an important issue in boreal ecosystems, many studies have been made to better assess and quantify climate-vegetation feedbacks in high latitudes. Among the different modeling options, large-scale deforestation experiments remain the most popular as they allow climate models to isolate the individual contribution of the removed vegetation in the context of realistic scenarios of medium-range future climate change. The radiative feedback in boreal forests was first introduced using simple energy balance climate models (Otterman et al., 1984, Harvey, 1988). The goal of both these studies was to assess the sensitivity of boreal forest species to climate change and the potential impacts of forest removal on climate. Both found a significant hemispheric cooling in the absence of snow masking by forests, as well as increased climate sensitivity to solar forcing and external perturbations.

2.2.3 Investigating the climatic impacts of boreal deforestation: the major numerical experiments
The climatic impact of high-latitude vegetation has been analyzed extensively with a variety of numerical models, and the examples in peer-reviewed literature are plentiful. A better assessment of climatic feedbacks to ecological processes has been made possible through the use of increasingly complex representations of land-atmosphere interactions, including among others a better representation of land surface processes and an improved parameterization of vegetation feedbacks. Among the earlier work, Thomas and Rowntree (1992) used the UKMO (United Kingdom Meteorological Office) GCM to show that an increased wintertime and springtime

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

11

surface albedo associated with boreal deforestation resulted in a significant cool perturbation to the surface energy balance, despite the fact that only half of the change in surface reflectivity is transferred to planetary albedo due to cloud cover. The overall effect of boreal vegetation on the global heat budget was estimated to be equivalent to that of doubling CO2 levels, with a complete removal of the boreal forest yielding a net cooling effect of up to 2.8C. Bonan et al. (1992) employed a more explicit approach by removing all forest vegetation poleward of 45N in the NCAR climate model CCM1, which also initiated a substantial spring cooling in the high latitudes. In addition, the coupled model revealed that sea ice-albedo

feedbacks amplified and extended the effect well beyond the deforested area, while cool anomalies tended to persist throughout the entire year in many locations due to the strong thermal inertia of oceanic basins. Due to these results, they reasoned that climate feedbacks associated with boreal deforestation could create unfavorable environmental conditions, irreversibly turning the tides against eventual forest regeneration. Chalita and le Treut (1994) examined the impact of increased albedo in the LMD (Laboratoire de Mtorologie Dynamique) Regional GCM, and argued that the cold perturbation associated with higher surface albedo could modify soil moisture so as to enhance summer precipitation in Europe. This result is interesting because it contradicts the notion that

precipitation increases monotonously with temperature (at high latitudes), at least locally. Although regional model do account for far more processes than their global counterpart, it is surprising to find that none of the global studies seem to acknowledge an increase in summer precipitation over Europe due to the increased surface albedo. Two subsequent model studies used the same experimental setup as Bonan et al. (1992) in climate models GENESIS (Bonan et al., 1995) and ARPEGE (Douville and Royer, 1997). Results from both experiments clearly indicated that deforestation at high latitudes cools the surface air and decreases latent heat flux and atmospheric moisture at all times of the year. Additionally, results from the latter suggest mid to high latitude deforestation could produce

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

12

significant climatic trends in other parts of the globe, notably a shift in the Asian monsoon and the African ITCZ. Snyder et al. (2004), in experiments with the coupled atmosphere-biosphere model CCM3-IBIS, compared the climatic impact of the removal of six major vegetation biomes and determined that boreal forests produce the largest temperature signal, even surpassing in intensity some important shorter-term climate oscillations such as ENSO. Their results also underlined several changes in the summertime hydrologic cycle including a decrease in evapotranspiration, atmospheric moisture and precipitation and confirmed the potential of high latitude vegetation to influence climate remotely, all largely in agreement with much of the earlier GCM work focused on the radiative effects of high-latitude vegetation changes. Climate-vegetation interactions have also been investigated with the far better resolved regional climate models (RCM), allowing a more diverse representation of sub-continent-scale (e.g., orographic) and therefore a better assessment of the impact of localized land surface perturbations. For example, Heck et al. (2001) studied the climatic impact of regional-scale deforestation in Europe, and found that the climate sensitivity to vegetation changes occurred in two distinct phases: a cool, wet spring, followed by a warm, dry summer. These results would imply that hydrologic processes override radiative feedbacks during summer. As another

example, Notaro and Liu (2008) examined vegetation feedbacks in Asiatic Russia with a combined statistical-dynamical approach, and both methodologies supported a year-round positive feedback of forest cover on both temperature and precipitation. Some of the interesting consequences of the increased surface albedo include an extended snow season and increased atmospheric stability, which in turn act to enhance the Siberian High and reduce convective precipitation. The latter, combined with an expected decrease in plant transpiration and moisture recycling because of the sparser vegetation, point toward a significant decrease in precipitation.

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

13

2.2.4 Climatic impact of the global forest cover changes


The previous three subsections have highlighted a strong competition between two biogeophysical feedback mechanisms namely, surface reflectivity and evapotranspiration both of which are driven by the absorption of energy by the land surface. Since these two processes directly oppose each other, it is clear that the prevalence of one over the other will determine the impact (e.g. a warming or a cooling trend) of vegetation cover on climate in a particular region at a given time of the year. For example, reflectivity usually dominates in areas of high seasonal variability, low precipitation and sparse vegetation cover, whereas evapotranspiration has a dominant effect in the densely vegetated tropical areas, but can also be important during the warm season in other parts of the world. This creates a diversity of ecological responses to climate forcings, which becomes especially important not only for the global evaluation of the climate impacts of forests, but also in the context of climate change mitigation efforts for example, it is useful to know that afforestation would provide the greatest climate benefit when concentrated in tropical regions (see Bonan,, 2008). There have been a number of global-scale climate-biosphere investigations in order to evaluate the full impact of the worlds forests on climate and determine which feedback mechanism dominates the temperature and precipitation signal on a global scale. The idea was first initiated in a numerical study with the atmospheric GCM ECHAM4 (see Fraedrich et al., 1999; Kleidon et al., 2000), which compared the two opposite extremes of the vegetation spectrum: a fully vegetated green world, and a desert world devoid of vegetation. Among the many substantial differences between the two simulations, the green planet featured twice as much precipitation worldwide and tripled land evapotranspiration, resulting in a significant surface cooling from latent heat release many times compensating for the increase in absorbed solar irradiation at the surface. A major criticism of the previous experiment, however, was its prescription of sea surface temperatures and sea ice coverage, which were seen as a constraint on the studys ability

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

14

land cover change effects. Subsequent model investigations with some level of ocean dynamics included would prove these reservations to be well-founded. In particular, Fraedrich et al. (2005) conducted a series of sensitivity tests with the Planet Simulator (which involves a mixed-layer ocean and thermodynamic sea ice model) and found the green world to be warmer than the desert world. While still displaying regional pockets that experienced cooling, the new simulations clearly showed that the global temperature response to increased tree cover was being dominated by radiative effects. Another study (Gibbard et al., 2005) used a coupled AGCM-slab ocean model and reached similar results, evaluating the temperature difference between a forest world and a grass world to be approximately 1.7C, along with a change in surface albedo of 0.07. Interestingly, an assumption of increased carbon sequestration has been agreed to not affect the warming trend in the long term, because a change in surface albedo is perceived as permanent whereas the anomalous carbon levels eventually vanish once the model equilibrates with the ocean, and ultimately the sediment components. Other similar experiments have since confirmed that the warming effect of forests at mid to high latitudes dominates over the cooling effect of latent heat fluxes in the tropical forests (Bala et al., 2007; Brovkin et al., 2009; Davin and de Noblet-Ducoudr, 2010). In particular, the last study explicitly demonstrates that the climate response to changing surface albedo is the most important biogeophysical effect of land cover change. In regards to the unequivocal role of ocean dynamics in the success of recent investigations, it has been suggested that the ocean might be unresponsive to nonradiative forcings (such as perturbations in the hydrologic cycle), which would explain why the inclusion of an interactive ocean module only appears to strengthen warm anomalies brought on by decreased surface albedo.

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

15

2.3

Climate-biosphere interactions in paleoclimate simulations

2.3.1 The Mid-Holocene


The past few decades of climate research have highlighted boreal forest-climate interactions as an amplifier of externally-driven climate change, both in the present and in the past. Nowhere is this truer than for the case of the mid-Holocene climatic optimum (about 6000 years ago). Reconstructions indicate warmer-than-present temperatures globally, most of which can be attributed to the higher total insolation received during this period. However, orbital forcing alone fails to explain why largest temperature departures (with regards to present-day conditions) occur in the spring, at which time the Earth reaches the aphelion and it thus farthest away from the Sun. The first suggestion implicating climate-vegetation feedbacks in this discrepancy can be traced back to the early work of Foley (1994), who first discovered, through the use of a processbased ecosystem model, that the terrestrial biosphere responded to mid-Holocene warming with a significant expansion of the boreal forest in high latitudes and an expansion of grassland in subtropical Africa, both supported by palaeobotanical evidence. Building on this knowledge, Foley et al. (1994) used a set of climate simulations and integrated palaeobotanical data to show that these vegetation feedbacks could help account for the additional warming indicated by the proxies. These findings were soon reinforced by subsequent experiments including an exhaustive study with two coupled AGCM-slab ocean models (TEMPO authors,, 1996), which added that the simulated vegetation feedbacks contributed to as much as 50% of the temperature increase that drove the northward boreal forest expansion (relative to present day distributions). The latter idea, however, was not shared by Texier et al. (1997), who argued instead for a more secondary role of vegetation feedbacks as an amplifier of orbitally induced climate and

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

16

vegetation changes, perhaps, but not the main factor that would reconcile simulated with observed conditions. A study with a fully coupled atmosphere-ocean-vegetation model (Gallimore et al., 2005) further corroborated the above results, finding a climate-vegetation response of comparable intensity to that of Texier et al. (1997). The northern expansion of both the taiga at highlatitudes and grassland over low-latitude deserts was reproduced, however orbital forcing remained the dominant cause of temperature change. The positive vegetation feedback in their simulation was not uniform however, as the southern extent of the taiga also retreated north due to severe water limitations, leaving previously wooded areas in the form of grassland and therefore more susceptible to late spring snowmelt. In another study, Wang et al. (2005) examined the climate system response to changes in both orbital parameters and ice sheet configuration in a number of sensitivity experiments with the McGill Paleoclimate Model (MPM), greened with the dynamic global vegetation module VECODE. They found that orbital forcing together with strong vegetation-albedo feedbacks induced by a retreating Laurentide Ice Sheet were mostly responsible for the warming trend in the mid-Holocene, and as a response the northern limit of the boreal forest moved northward during this period. However, declining summer insolation reversed that trend in the following centuries and a gradually cooling climate forced the boreal forest to retreat further south. As it became increasingly clear that mid-Holocene warming was influenced in some manner by vegetation feedbacks, a new experimental setup emerged which identifies and isolates three individual contributions to the climate signal: vegetation-atmosphere interaction, atmosphere-ocean interaction, and a synergy that arises from the coupling of these two processes. This approach was initiated by Ganopolski et al. (1998), who found a significant contribution from vegetation-atmosphere interactions, but concluded that overall agreement with paleodata is very weak without the synergy between vegetation-atmosphere and atmosphere-ocean interactions.

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

17

Following on these footsteps, Crucifix et al. (2002) and Wolfhart et al. (2004) initiated several climate simulations with different Earth system Models of Intermediate Complexity (EMICs), and found that most of the warming of the Northern Hemisphere during the midHolocene could be attributed to vegetation-atmosphere interactions. Since, in both cases, ocean and vegetation feedbacks often displayed opposite impacts on continental temperature trends, a strong synergy was deemed unlikely. In a more recent set of experimentations with an updated version of the ECHAM AGCM, Otto et al. (2009) could reproduce neither the strong vegetation-atmosphere interaction featured in the previous two studies, nor the strong synergy found by Ganopolski et al. (1998), prompting the authors to suggest that the observed mid-Holocene warming signal was dominated by the contribution from atmosphere-ocean interactions. These findings were further strengthened by a full investigation of forest-albedo feedbacks in the mid-Holocene (Otto et al., 2011), which found that factors to which the intensity of spring warming was most sensitive to (such as the parameterization of snow albedo) had little impact on boreal forest cover. Because of these latest developments, it is now believed that vegetation-atmosphere interactions have been overestimated in early climate simulations of the mid-Holocene.

2.3.2 The Last Glacial Maximum


Among the many different climatic periods of interest, the Last Glacial Maximum (~21000 years BP) has also gathered considerable interest because there are signs of major vegetation changes (as indicated by various palaeobotanical records). This period offers a unique opportunity to test model performance in the context of severe glaciation in the Northern Hemisphere and provide additional insight on atmosphere-biosphere interactions, especially with regards to feedback mechanisms between the massive continental ice sheets and a rapidly evolving land surface cover. Its relatively recent time frame (when considering the availability of proxy data records) also contributes to make it an attractive candidate for paleoclimate modeling.

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

18

The Last Glacial Maximum (LGM) was first investigated within the framework of the Paleoclimate Model Intercomparison Project (PMIP): for example, Lorentz et al. (1997) found major discrepancies with geologic reconstructions, especially in sub-glacial high latitudes. In an effort to better represent land surface patterns, Kubatzki et al. (1998) added a vegetation module to their climate model, and were much more successful in obtaining a better representation of the LGM climate. The model performance was especially improved in the Siberian region, where atmosphere-only simulations tended to overestimate the ability of cold-adapted vegetation to resist the harsh winter cold. These results were soon followed by another study (Levis et al., 1999), which also found that LGM conditions conducive to the southward migration of the tree line, replacing much of the high-latitude forestry with tundra. Their model output also hinted at a reduction of tropical forest cover in favor of C4 grasses, due to physiological effects associated with lower concentrations of atmospheric carbon dioxide. The latter result is especially interesting in light of current large-scale forest decimation in the tropics, because it would make LGM climate in these regions a possible analogue to near future climate save for atmospheric CO2 should the deforestation continue uninterrupted. More recently, the findings of Crucifix et al. (2005) corroborated most of the above results, notably a disappearing Siberian taiga, increased shrub cover in Europe and expanded subtropical deserts. An analysis of bioclimatic relationships also revealed that the position of the boreal treeline was primarily constrained by water stress and soil properties (rather than summer temperature), and confirmed that a depletion of atmospheric CO2 (relative to pre-industrial values) would result in environmental conditions more favorable to grasses and shrubs by narrowing the climatic range where trees dominate the vegetation spectrum. All of the above papers made note of the profound impact of LGM climate on vegetation, especially at high latitudes, and their general success in using a coupled climate-vegetation model to reconcile simulated climate with observations would suggest some influence of

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

19

vegetation processes on the climate system during the same period. Crowley and Baum (1997) first tried to evaluate the potential role of vegetation changes by prescribing reconstructed LGM vegetation in a GCM, and found a significant impact on the terrestrial response to ice and SST changes. A more recent study (Crucifix and Hewitt, 2005) examined the regional and global impact of land cover changes in a series of LGM simulations involving the vegetation module TRIFFID. They found that the temperature and precipitation response on the continents was driven by regional interactions with vegetation, and the overall impact of vegetation dynamics (relative to present-day) resulted in additional surface cooling despite a warming trend in the tropics caused by the reduced tree cover. Furthermore, a strong correlation was found between enhanced glacial cooling in Siberia (caused by high surface albedo) and atmospheric dynamics in the tropics, suggesting a possible remote impact of high latitude vegetation changes on tropical climate. In an attempt to better quantify the impact of vegetation dynamics during the LGM, Jahn et al. (2005) used a factor separation technique (see previous section for other examples) in order to isolate the individual contributions of continental ice sheets, changes in CO2 concentrations, and vegetation feedbacks on the global climate signal. Their results highlighted previous

findings that the impact of vegetation changes would be mostly limited to regional-scale effects. Although a global cooling similar to that of Crucifix and Hewitt (2005) was found, further investigation revealed that the contribution from vegetation feedbacks to the temperature signal could have been indirect, for example by triggering a change in ocean circulation regime that would have caused further cooling.

2.3.3 Earlier periods


Among the earlier periods, ice age inception offers another potential interesting topic of study. Given the strong forest-albedo feedback mechanisms discussed above, the retreating boreal forest in favor of cold-adapted grasses (as observed in the palaeobotanical record during glacial inception) is sometimes credited with a major role in the expansion of continental ice

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

20

sheets by enhancing cold continental air masses and severely limiting warm-season snowmelt. This idea can be found, for example, in the work of Gallimore and Kutzbach (1996), as well as De Noblet et al. (1996). Both studies indicate that prescribing an expansion of tundra (or letting vegetation adapt naturally to decreasing summer insolation) results in a significant decrease in summer temperature in high-latitude North America and Eurasia, which increases the likelihood of summer snowfall and allows the snowpack to survive the warm season in a considerably larger area, a crucial part of initiating an ice age. Glacial inception has also been studied with the UVic ESCM (see Meissner et al., 2003). The model confirms a global decline in tree vegetation, occurring in both tropics and high latitude as an expansion of grasses or shrubs at the expense of forests, the consequence of which appear to double the effective atmospheric cooling during ice age inception, as well as reduce meridional overturning in the North Atlantic and significantly perturb precipitation patterns over the continents. Finally, it is always interesting to contemplate how vegetation dynamics could have impacted climate at various epochs of Earths history. For example, Kubatzki et al. (2000) underline several differences between the climate of the Holocene (present day) and that of the Eemian (last interglacial) which might be caused or amplified by climate-biosphere interactions. In particular, they note that ecological feedbacks amplify the orbitally induced warming in the Arctic (where temperature departures from present day conditions are greatest) and result in overall warmer conditions across the globe. An apparent expansion of subtropical vegetation also intensifies the monsoonal response to orbital forcing, while indirect interaction between vegetation and the ocean indirectly results in a reduced Atlantic meridional overturning circulation. Another example can be found in Schneck et al. (2012), who use an EMIC to evaluate the sensitivity of climate to vegetation changes during the Late Miocene, situated prior to the cycle of Quaternary glaciations in the geologic timeline and believed to have been warmer especially at the poles and more densely vegetated compared to the present day. In particular, one of their

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

21

simulations prescribes modern day high-latitude vegetation on Late Miocene climate. This causes a strong cooling effect extending up to the mid-latitudes, an expected result of increased surface albedo on boreal climate. Due to the geography of the time or perhaps a model shortcoming this apparent cooling does not lead to an intensification of Northern Hemisphere heat transport. However, the authors note that the inclusion of vegetation feedbacks eliminate a part of the discrepancy between simulations and paleorecords, bringing the research one step closer to explaining the weak equator-to-pole temperature gradient as suggested by data records.

2.3.4 Stability of the climate-vegetation system


One of the most intriguing aspects of the nonlinear climate system and its complex entourage of interacting components is its ability to produce several equilibrium states depending on the initial condition. The concept of dual equilibriums in the climate-biosphere system was first hinted at in early studies of the mid-Holocene climatic optimum as a possible explanation for the presence of a green Sahara in GCM simulations an alternate solution of climate system dynamics in place of the modern-day arid desert. For example, Claussen and Gayler (1997) found a northward expansion of savanna vegetation into the Sahara as well as generally wetter conditions in the northern half of Africa, especially in the west. These results, which were obtained with a coupled atmosphere-biome model, were much closer to paleogeological and palaeobotanical records than an atmosphere-only simulation, the latter tending to reproduce modern-day conditions even with mid-Holocene orbital forcing and sea surface temperatures. In other words, the single addition of vegetation dynamics would have provoked a change in tropical circulation in response to mid-Holocene orbital forcing and SST that would have drastically enhanced precipitation in the Sahel, hence the argument for a possibly crucial role of atmosphere-biosphere interactions in the emergence of the green state of the Sahara. The stability of the climate system in the Sahel has also been investigated for LGM (Kubatzki and Claussen, 1998) and present-day (Claussen, 1998) climates. Sensitivity tests in both of these periods revealed that initiating the ice-free land surface as a uniform forest, steppe

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

22

or dark (low albedo) desert would ultimately yield the greened state of the Sahara as described in mid-Holocene reconstructions and simulations; in order to obtain the actual (and LGM reconstructed) distribution of subtropical deserts, bright (high albedo) sand deserts had to be prescribed from the beginning of a simulation. These results were explained in the context of Charneys theory of self-perpetuating deserts through albedo enhancement, associated with the presence (or absence) of vegetation with a substantial, long-lasting impact on sub-tropical convection and monsoonal patterns. Another study (Claussen et al., 1998) confirmed the existence of a dual equilibrium for LGM and present-day climate, and revealed that the green Sahara was the only possible solution in the case of mid-Holocene boundary conditions. Using a conceptual bifurcation model, the authors argued that orbital forcing (through its effects on atmospheric circulation) could be responsible for locking the atmosphere-biome system into the green mode during the midHolocene, which would also explain the observed decrease in sub-tropical aridity during that period. However, some questions remain unanswered such as the subsequent shift back to desert mode and the authors insist that further investigations with fully a coupled (atmosphere-vegetation-ocean) model will likely be necessary in order to better understand the role of vegetation in climate system stability. Finally, the question of atmosphere-biome stability has also been raised regarding highlatitude vegetation. Since boreal forests tend to create warmer, moister environments, it was hypothesized that the sole presence of evergreen needleleaf trees over present-day tundra and polar deserts could modify the high-latitude climate sufficiently enough to make it adequate for their survival. This idea was quickly dismissed however, as experiments with coupled

atmosphere-biome models (Claussen, 1998; Levis et al., 1999) did not find multiple solutions of the Arctic climate-vegetation system; in a forested Arctic start, for example, the initial warming signal was insufficient to allow boreal evergreen trees to persist in higher latitudes, and the northern extent of the boreal forest gradually drifted back towards present-day values.

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

23

Brovkin et al. (2003) extended the stability analysis to several other climate models, none of which produced more than one steady state for high-latitude vegetation, even with doubled CO2 levels. However, they noted an increase in climate sensitivity for low vegetation cover, as well as an increased sensitivity with interactive ocean and sea ice, reiterating the key importance of ocean dynamics in assessing the influence of vegetation feedbacks on the climate system.

2.4 Present-day interactions between climate and the biosphere: analyzing vegetation response to climate change
2.4.1 Global vegetation feedback to increases in atmospheric CO2
The potential climatic impacts of anthropogenic increases in atmospheric CO2 have been a topical issue of the past decades, and the scientific community is only starting to grasp the full extent of its influence on various aspects of our environment. Because of its central role in the chemical equations that define plant life, changes in carbon dioxide have long been suspected to have a major impact on vegetation, in addition to a spatial redistribution of plant biomes because of the elevated surface temperatures. For example, Prentice et al. (1991) used a forest succession model to analyze its implication on forest composition and biomass dynamics. While some species reacted better than others, creating a rather complex spatial shift of vegetation boundaries, the model displayed an unequivocal northward shift of the boreal treeline as a direct consequence of anthropogenic changes in CO2. Another known impact of increased atmospheric CO2 is to cause physiological changes such as a reduction of plant stomatal conductance, which limits the loss of water through transpiration and thereby mitigates the cooling effect of tropical forests through latent heat release. However, climate model experiments with doubled CO2 levels (see for example, Betts et al., 1997; Foley et al., 1999; Levis et al., 2000) have shown that that, despite appreciable decreases in evapotranspiration on local scales, global physiological climate-vegetation feedbacks are mostly offset by a widespread increase in leaf area index (e.g. by expanding tree

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

24

cover). The main exception is in the continental mid-latitudes, where a reduction in plant transpiration leads to a depletion of soil moisture This further limits the availability atmospheric moisture through recycling, the regions main source of precipitation, therefore resulting in a noticeable aridification of the mid-latitudes. This effect is not as important in the high latitudes, where evapotranspiration is not a significant component of the hydrologic cycle; however, the region remains sensitive to climate change, mainly because of radiative feedbacks from the northward expansion of the boreal forest.

2.4.2 Climate response to high-latitude afforestation


As mentioned in the previous paragraphs, one of the major consequences of climate warming due to anthropogenic increases in CO2 hence a possible outcome for the mediumrange future is a northward expansion of the boreal tree line. In order to better understand the many climatic impacts of afforestation in the northern hemisphere, there have been a number of numerical experiments simulating increased greenness in both the mid (Swann et al., 2011) and high (Zhang et al., 2006; Swann et al., 2009) latitudes. In general, mid-latitude afforestation seems to have little impact on global temperature and CO2, but regional warming can occur due to the increased solar energy absorption, especially in regions where water limitation prevents compensation through latent heat release. However, these local effects can influence remote circulation patterns: for example, the model results from Swann et al. (2011) suggest that an anomalous heating redistribution through atmospheric circulation changes could alter the Hadley circulation, impacting precipitation patterns across tropical and sub-tropical latitudes. Of course, a better understanding of these patterns and the role of vegetation dynamics would be crucial in, for example, designing strategies for climate change mitigation. At higher latitudes, added forestry results in a warming and moistening of the atmosphere mostly driven by springtime increases in net surface radiation from the snow-albedo feedback. This combines with the projected climate warming due to increase in atmospheric CO2 to further exacerbate the warming trend (see above section). Furthermore, while not an important part of

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

25

the water cycle, plant transpiration nevertheless contributes to a net soil-to-atmosphere moisture transfer, creating a more unstable atmosphere that enhances convective cloud formation during summer. Interestingly, in the long term there are indications among model results from (Swann et al., 2009) that the increased cloud cover would provoke a top-of-atmosphere radiative imbalance of comparable effect to that caused by changes in albedo, overriding the warming trend established by the latter. Further investigation will likely be necessary in order to gain a better understanding of this matter.

2.4.3 Climate response to anthropogenic land cover change


Aside from injecting a large amount of carbon into the atmospheric reservoir, another known climatic impact of human activities comes from the large-scale alteration of land surface, chiefly for agricultural and, to a lesser extent, urbanization purposes. Recorded history of humankinds agricultural traditions suggest that significant influence on climate from land use changes may predate the industrial era by at least a few centuries, but a recent hypothesis (Ruddiman,, 2003) suggests that it could push back the beginning of the anthropogenic era by several thousand years. The impact of historical land cover change has been shown to be of comparable importance to the effect of CO2 emissions (mainly from land conversion) for at least a thousand years prior to the industrial era (Brovkin et al., 1999; Brovkin et al., 2006). In particular, Brovkin et al. (2001) prescribed a scenario of land use change over the past millennium with six different Earth models of intermediate complexity, all of which agreed on a crucial role of land surface changes as a climate forcing for several centuries. It is interesting to note that the cooling effect was strongest during the, 19th century, and lasted until the mid-20th century, when the trend was apparently reversed, likely in result to the rising atmospheric CO2 levels. The potential impact of modern-day land cover changes has also been studied extensively, both through data collection (e.g., see Lee et al., 2011) and numerical model studies (Claussen et

CHAPTER 2. CLIMATE-VEGETATION INTERACTIONS AND FEEDBACKS

26

al., 2001; Bathiany et al., 2010). In general, results are akin to those presented in large-scale deforestation experiments (see section 2.2.5), which is not surprising since most human-induced land use change consists of replacing forested areas with agricultural cropland. In particular, the sign and magnitude of the temperature response depends on vegetation type and latitude: removing tropical forests results in a temperature signal dominated by biogeochemical effects, with a CO2 increase of approximately 60 ppmv driving the global warming, while land cover changes in nontropical latitudes produce a temperature anomaly mostly driven by springtime biogeophysical feedbacks. These observations are critically important in light of future land cover changes, such as the use of reforestation as an option for the enhancement of carbon sequestration.

2.5

Summary
A number of climate modeling studies of atmosphere-biosphere feedbacks have

illustrated the important role of vegetation within the Earth system.

The dominant

biogeophysical effects vary depending on the location and time of the year; for example, hydrologic processes are predominant in the densely wooded tropical evergreen forests, while radiative effects are the main factor at mid to high latitudes during winter and spring. Globally, the latter have been shown to be the most important, such that the total effect of the worlds forests on climate is to increase global temperatures. Radiative effects are also crucially

important in assessing the influence of the boreal forest on high latitude climates, both in paleoclimate simulation and in scenarios of future climate change. Ultimately, one of the most striking aspects of climate-biosphere interactions is the natural tendency of vegetation to modify its environment in such a way as to increase its survivability. On either side of the climatic spectrum, plants act to mitigate temperature

extremes and enhance precipitation, all of which contribute in making the land more hospitable. This reflection suggests an interesting analogy between the above and the theory of selfperpetuating deserts (Charney, 1975).

Chapter 3
Model description
3.1 Earth system Models of Intermediate Complexity
Earth system models of intermediate complexity (EMICs) such as the UVic ESCM exist in order to bridge the gap between inductive models, which focus on a limited set of processes and mechanisms (for example, most box models), and the computationally expensive quasi-deductive models (for example, GCMs). Considered as middle-of-the-road regarding

model complexity and computational efficiency, EMICs maintain a large spectrum of interacting components typical of their comprehensive counterparts (atmosphere, ocean, sea ice, land surface / vegetation modules as well as biogeochemical cycles are common), albeit in a more simple form allowing for longer-term simulations of the climate system (Claussen et al., 2002). Unlike a general circulation model, each EMIC is characterized by its own field of specialization, making it more suited to a particular set of experiments than other EMICs. For example, the McGill Paleoclimate Model of Wang and Mysak (2000) was specifically designed for the study of ice age inception and millennial- to Milankovitch-scale climate variability during the Quaternary, and in order to simulate the long timescales several degrees of complexity were abandoned in favor of improved overall performance. According to the most recent table of EMICs (Claussen, 2005), the range of model expertise covers a large spectrum of possible research interests, such as atmospheric dynamics (CLIMBER-2), ocean circulation (Bern 2.5D), biogeochemical cycles (ISAM-2: terrestrial; UVic: oceanic), global environmental change (MoBiDiC: orbital-scale; MIT: anthropogenic), and extraterrestrial climate dynamics (Planet Simulator). Although computational performance varies greatly among EMICs, it is usually possible to simulate climate system evolution for periods spanning up to tens of thousands of model years within a reasonable lapse of computing time, making this brand of models particularly attractive for paleoclimate research.

27

CHAPTER 3. MODEL DESCRIPTION

28

3.2

General description of the UVic ESCM


The model used in this study is the University of Victoria Earth System Climate Model,

an intermediate complexity coupled atmosphere/ocean/sea-ice model introduced and described in great detail by Weaver et al. (2001). It consists of a three-dimensional ocean general circulation model (OGCM) with ocean chemistry, coupled to a thermodynamic/dynamic sea-ice model and an energy-moisture balance atmospheric model with parameterized dynamical feedbacks. The model was originally equipped with a thermomechanical land-ice model, but this approach has been abandoned in recent versions in favor of prescribed continental ice sheets. Because of the simplified atmospheric component in the UVic ESCM, the model is computationally efficient compared to a fully coupled atmosphere-ocean GCM. The land-sea configuration used in the UVic ESCM is coarse. The spatial domain is global, and features a spherical grid resolution of 3.6 (zonal) by 1.8 (meridional), which is comparable to most coupled coarse-resolution AOGCMs. The model once employed the Euler frame of reference with the North Pole shifted to Greenland in order to avoid grid convergence problems. In more recent versions of the model, this approach has been abandoned in favor of an artificial island at the North Pole. Below are descriptions of the major components of the original, 2001 model: atmosphere, ocean, and sea-ice. The vegetation module and its supporting land-surface scheme were added later to the UVic model, and will be discussed in the following sections of this chapter.

3.2.1 Atmosphere
The atmospheric module is a vertically-integrated energy-moisture balance model loosely based on Fanning and Weaver (1996), with two major simplifications. First, the conservation of momentum is achieved through a combination of specified wind data and dynamical wind feedbacks, removing the need for computationally demanding prognostic equations. Second, the fluxes of energy and moisture are parameterized by diffusive processes only, although heat and

CHAPTER 3. MODEL DESCRIPTION

29

moisture advection by winds is left as an option. In other words, the transport of heat and water vapor in the atmosphere is dictated mainly by meridional gradients (i.e., the inherent pole-toequator temperature and moisture gradients), while wind velocities are not explicitly prescribed. The main feature of this simplified atmosphere is the energy-balance equation, an evolution equation for the prediction of surface air temperature Ta:

where a is the surface air density, ht a representative scale height for temperature, and cpa the specific heat of air at constant pressure. Terms on the right-hand side represent the sources and sinks of heat which parameterize energy exchanges between the atmosphere and the underlying surface. The first term QT is the horizontal heat flux, which involves a combination of advective and Fickian diffusive processes. The term QLH denotes the transfer of energy through latent heat release, which is assumed to occur solely through precipitation, either as rain or snow. Energy exchanges with the outer space in the form of incoming shortwave and outgoing longwave radiation are represented by the terms QSSW and QPLW, respectively. The incoming solar radiation is written as:

where S is the solar constant, is the latitudinally- and time-dependant planetary albedo, and CA is a reduction parameter accounting for the absorption/scattering of about 30% of shortwave radiation in the atmosphere (from water vapor, dust, ozone and clouds, to name a few). Also, in its definition of top-of-atmosphere incoming radiation I, the model accounts for orbital configuration when establishing the annual cycle of solar insolation, as per the calculations of

CHAPTER 3. MODEL DESCRIPTION

30

Berger (1978). The parameterization of outgoing longwave radiation is based on Thompson and Warren (1982), and modified in order to depend on surface air temperature, relative humidity, and the atmospheric concentration of carbon dioxide. In particular, CO2 radiative forcing is applied in the model through a decrease in outgoing longwave radiation. Since the model does not permit the storage of heat or moisture on the land surface, the final two terms can only assume nonzero values over the ocean. One term, QLW, accounts for the strong longwave flux at the atmosphere-ocean interface (due to the oceanic heat reservoir), which is modeled according to a gray body version of the Stefan-Boltzmann law. The other term, QSH, denotes sensible heat exchanges between the surface and the atmosphere, which are evaluated using a bulk parameterization of surface variables. The model uses prescribed present-day winds in its climatology, and includes a set of dynamical wind feedbacks based on a latitudinally-dependent empirical relationship between air temperature and density. In order to account for the dynamic response of the atmosphere to variations in sea surface temperatures, wind stress anomalies are parameterized in terms of surface air temperature anomalies. The hydrologic cycle in Fanning and Weaver (1996) is parameterized by a simplified and vertically-integrated version of the balance equation for water vapor, in which the horizontal advection term is replaced by an eddy diffusive term. In the UVic model this setup is essentially untouched, although there is an option for moisture advection by vertically-integrated atmospheric winds specified from NCEP reanalysis data. The models equation for moisture balance also involves, among others, the use of a bulk parameterization to calculate evaporation and precipitation, the latter being assumed to occur whenever the relative humidity exceeds a certain threshold (usually 85%). A specified lapse rate is used to calculate temperature and precipitation anomalies due to orographic influences, allowing among other things a more realistic configuration of each of the 33 specified river basins.

CHAPTER 3. MODEL DESCRIPTION

31

3.2.2 Sea ice


The treatment of sea ice in the UVic model is done with a standard model involving simple two-category (sea ice, open water) thermodynamics and elastic-viscous plastic dynamics; however, several options are offered for a more sophisticated representation of sea-ice thermodynamics and ice-thickness distribution. The standard model evaluates ice thickness, areal fraction and ice surface temperature based on the zero-layer formulation of Semtner (1976) and the lateral growth and melt parameterization of Hibler (1979), while the momentum balance equation for ice dynamics is solved using the rheology developed by Hunke and Dukowicz (1997). Snow follows the same parameterization as other types of precipitation, and is assumed to occur when air temperature at the surface falls below a critical value (usually -5C). Snow accumulation on the ground is treated as another form of moisture storage, and it can be used as an elementary ice sheet model (in the sense increasing surface and planetary albedo). Over sea ice or ocean the accumulation of snow is treated as part of the surface energy balance. Several ice-thickness distribution options are incorporated in the UVic model as alternatives to the standard representation of sea ice. The most commonly used improvement, described in great detail in (Bitz et al., 2001), involves a multi-layer thermodynamic model with heat capacity (Bitz and Lipscomb, 1999) and a Lagrangian formulation of the sub-gridscale ice thickness distribution developed by Thorndike et al. (1975), allowing among others a better resolution of the vertical temperature profile. Continental ice in the UVic model consists of a simple prescription of the spatial coverage and height of ice sheets based on paleoclimate data records. In the version of the model used for this study, the land-ice configuration is updated every few thousand years based on data from the model ICE-5G (Peltier, 2004).

CHAPTER 3. MODEL DESCRIPTION

32

3.2.3 Ocean
The central piece of the UVic model is the Geophysical Fluid Dynamics Laboratory (GFDL) Modular Ocean Model (MOM), v2.2 (Pacanowski, 1995), a full-fledged 3-D OGCM based on the Navier-Stokes equations subject to Boussinesq and hydrostatic approximations. The horizontal grid resolution is the same as in the atmospheric and sea-ice components of the model, while the vertical grid consists of, 19 unequally spaced levels that vary gradually in size, from very small near the surface to very large near the bottom. Ocean bathymetry is included and taken from the Suarez and Takacs (1986) dataset. The density of seawater is given by a nonlinear function of potential temperature, salinity and pressure. The ocean top layer is driven by wind stresses and surface buoyancy forcing. In order to avoid subfreezing ocean

temperatures, the model calculates the maximum available heat in the top layer, which can then be redistributed to the atmosphere or sea ice. This allows for a (relatively) simple definition of the net heat flux into the ocean QH:

where Qto, Qb are adjusted downward heat fluxes from the atmosphere and sea ice, respectively, and Ai is the areal fraction of ice. Similarly, the implied surface salinity flux Qs is given by:

where 0, S*, Lf are representative constants for water density, salinity and latent heat, and R represents freshwater supplied from land runoff. The total heat flux from the atmosphere is distributed between the ocean (Qto) and ice (Qti) according to the areal fraction of ice. These transfer equations combine with ocean mixing by parameterized wind stresses and the primitive-equation ocean dynamics to form the backbone of this ocean component.

CHAPTER 3. MODEL DESCRIPTION

33

3.2.4 Coupling strategy


The atmosphere and sea-ice components are coupled using a time step of 30 hours. The ocean model time step is double that of the previous two, and coupling between the atmosphere/sea-ice models and ocean model is done every two ocean time steps. Each

component uses an intermittent Forward Euler time step with leapfrog time stepping scheme, which forces the use of a special calculation technique of flux exchanges between components in order to ensure the conservation of heat and salinity. The ocean model is first spun up for 5000 years under specified orbital forcing year and atmospheric carbon dioxide concentrations, and coupled to the other components when equilibrium is reached. Overall model efficiency varies between 50-150 years per CPU day depending on computer performance.

3.3

Recent additions and improvements to the model


The current version of the model is 2.9, which carries a number of differences from that

described by Weaver et al. (2001). These include an improved radiative transfer scheme, the inclusion of a land surface model, the addition of sulfates and aerosols as potential climate forcings, the introduction of a dynamic global vegetation model, and a coupling of the latters terrestrial carbon cycle and the oceans inorganic carbon cycle (Matthews et al., 2004). More recently, ocean biogeochemistry (Schmittner et al., 2008) and a sediment model (Eby et al., 2009) have been incorporated into the UVic model. The improved radiative scheme and new land surface model will be discussed below, while the new vegetation module will be the focus of section 3.4.

3.3.1 Enhanced radiative transfer model


In earlier versions of the UVic ESCM the top-of-atmosphere reflectivity was specified by a zonally averaged planetary albedo calculated in the single-layer atmosphere. The recent inclusion of land surface and vegetation schemes has prompted a modification to the radiative transfer scheme to provide an explicit representation of surface albedo as part of a two-

CHAPTER 3. MODEL DESCRIPTION

34

dimensional albedo field, which distinguishes it from atmospheric albedo. Based on the theory outlined in Haney (1971), planetary albedo p would then be defined as a function of surface albedo s, atmospheric albedo a and atmospheric absorption Aa:

In this new scheme surface albedo is taken from the land surface module, which evaluates it according to snow/ice cover and vegetation distribution, while atmospheric albedo is calculated as a sum of a background (clear-sky) albedo of 0.08 and cloud reflectivity. Because cloud cover is not explicitly represented in the UVic model, the cloud reflectivity is computed through a specified zonally-averaged combination of albedos from other inputs, including the original zonally-averaged planetary albedo. The net shortwave radiation at the surface is then given by: ( )

where IS is the incident shortwave radiation at the top of the atmosphere. While a definite improvement over its predecessor, the model is still lacking a dynamical treatment of clouds because of its zonally constant atmospheric albedo, and therefore a feedback between clouds and climate is still excluded.

3.3.2 Land surface scheme


The current version of the UVic ESCM has integrated a single soil layer version of the Meteorological Office Surface Exchange Scheme version 2 (MOSES-2), which defines the state of the land surface in terms of surface temperature, soil temperature and moisture content, and snow cover. It features among others an interactive representation of plant photosynthesis and conductance, and a parameterization of evapotranspiration as a function of canopy resistance. MOSES-2 in its standard configuration recognizes the five TRIFFID vegetation types, in addition to four types of non-vegetation landcover (bare soil, land ice, inland water and urban areas). A new soil thermodynamic scheme is introduced to account for the melting and freezing

CHAPTER 3. MODEL DESCRIPTION

35

of soil water and the impact of frozen and unfrozen water on the soils thermal characteristics. Soil moisture is increased by precipitation and snow melt, and is decreased by evaporation and continental runoff. The size of the snowpack is updated according to snow accumulation, snow melt and the rate of sublimation. Solar radiation unto the surface is balanced by latent heat release due to phase changes, sensible heat fluxes, and direct accumulation of heat into the soil. For a complete description of the land surface scheme as well as the original formulation, see Cox et al. (1999).

3.4

Description of the vegetation module

3.4.1 Evolution of vegetation modeling


Before the appearance of vegetation models, many AGCMs employed simple transfer schemes involving a representation of short-term biophysical processes of energy, moisture, carbon and momentum exchanges between the land surface and the atmosphere. The BiosphereAtmosphere Transfer Scheme (BATS), developed for use within NCARs climate models (Dickinson et al., 1986), and the Simple Biosphere (SiB) model of Sellers et al. (1986) are examples commonly found in the literature. In such models, the land surface was parameterized by fixing the geographical distribution of vegetation in most cases based on the modern-day configuration and assigning each grid cell to one of several pre-defined biomes with specified leaf area index, albedo, rooting depth and roughness length (Foley et al., 1998). Due to their static vegetation distribution, it was impossible for simple land surface schemes to capture long-term feedback processes that would arise from changes in vegetation cover. Many preliminary attempts to introduce vegetation as an interactive component of the climate system involved the use of equilibrium biogeographical models to update the geographical distribution of vegetation. For example, Claussen (1994) linked the BIOME model of Prentice et al. (1992) to the atmospheric GCM ECHAM through an asynchronous coupling procedure using multi-year averages of the climate model simulation to drive changes in

CHAPTER 3. MODEL DESCRIPTION

36

vegetation cover, and bringing the coupled system to equilibrium through multiple iterations. The coupled behavior was found to be stable but heavily dependent on initial conditions (as discussed previously in section 2.3.4). Another class of early vegetation models, referred to as transient ecosystem models, are also found in the literature (Kittel et al., 2000). These models place a heavier focus on the transient dynamics of vegetation changes, by modeling a wide array of ecosystems and each possible vegetation transition in independent sub-modules. They are most useful when examining differences among ecosystems in terms of rates of succession, transition probabilities, and sensitivity to climate and environmental disturbances. While undoubtedly a step forward in vegetation modeling, the set of iteratively linked climate-vegetation models was found to have two major limitations due to the asynchronous coupling strategy, leaving room for further model development. First, the existing models could only simulate the equilibrium response of vegetation cover to changes in climate, without addressing the transient nature of atmosphere-biosphere response to climate variability. Second, the model sometimes required two independent treatments of physical processes (from both the vegetation model and the AGCM), leading to inconsistencies in land surface parameterization. This is notably the case for the coupled ECHAM-BIOME model, where energy and moisture exchanges at the surface must be defined in both the AGCM for the evaluation of landatmosphere processes, and the vegetation model in order to estimate the soil moisture requirements of plants.

3.4.2 An overview of Dynamic Global Vegetation Models (DGVMs)


The last decade saw the emergence of a new class of vegetation models, which were created specifically to address the issues outlined in the above paragraph by incorporating the latest advancements in plant geography, plant physiology and biogeochemistry, vegetation dynamics, and biophysics (Prentice et al., 2007). In particular, DGVMs feature a transient, more integrated and physically consistent simulation of vegetation structure, land surface and ecological processes when compared to earlier models, and they are designed to be directly

CHAPTER 3. MODEL DESCRIPTION

37

incorporated into AGCMs. Some of the most commonly used vegetation models in current research are DGVMs: IBIS (Foley et al., 1996), VECODE (Brovkin et al., 1997), LPJ (Smith et al., 2001; Sitch et al., 2001) and TRIFFID (Cox et al., 2001) are but a few examples in this new category which has come to play a dominant role in vegetation modeling. Many DGVMs have been developed from existing models using one of two approaches. The first consists in expanding an equilibrium biogeographical model to include vegetation dynamics, by coupling it to models that simulate rates of vegetation growth and disturbance rates; this method is usually referred to as the top-down approach. Conversely, it is also possible to build a DGVM from a regional model by bringing it up to the global scale and coupling it with a biogeochemistry model, a method also known as the bottom-up approach. One of the challenges in efficient dynamic vegetation modeling comes from the large range of time scales involved. For example, DGVMs need to account for short-term dynamics of photosynthesis and moisture/energy exchanges (seconds to minutes), seasonal patterns of carbon assimilation (weeks to months), and changes in vegetation structure due to competition, mortality and disturbance rates (years to decades). In general, the timescales associated with changes in ecosystem structure tend to be up to several orders of magnitude higher than for physiological processes. Kittel et al. (2000) identify some limitations of DGVMs, especially with regards to highlatitude climate modeling. They note that most DGVMs still lack an adequate representation of sub-gridscale processes associated with unusual biomes such as inundated landscapes (marshes and bogs), anaerobic soils, and permafrost. Another possible improvement to high-latitude vegetation modeling would come from better defining the role of small plant organisms such as moss and lichens in the biogeochemical dynamics of tundra and boreal ecosystems.

CHAPTER 3. MODEL DESCRIPTION

38

3.4.3 The Plant Functional Type (PFT) approach


The concept of plant functional types, which consists of classifying plants functionally rather than by evolutionary development, was introduced in order to reduce the complexity of global vegetation structure and diversity to a manageable level in the more expensive vegetation models (Woodward, 1987). Using this strategy implies several assumptions regarding the

terrestrial biosphere: (i) that plant species can indeed be grouped according to broad structural or functional characteristics; (ii) that parameterizations for each PFT can adequately represent the physiological properties of each individual species; (iii) that the definition of a PFT is independent of geographical location; and (iv) that most biomes can be recovered from the dominant PFT and climatic regime. In general, most PFT schemes differentiate between woody and herbaceous types, with further subdivisions based on attributes such as leaf longevity, temperature tolerance, and photosynthetic processes; ultimately, the PFT configuration is chosen according to the modeling framework and the desired level of complexity. The PFT approach is often used within DGVMs as an efficient way to simulate vegetation dynamics and evaluate land surface properties. Most DGVMs allow for multiple PFTs to coexist within a single grid cell (by defining the areal fraction of each PFT) in order to provide a transient representation of vegetation changes due to climate forcings, as well as a more realistic simulation of structural changes within a biome. The fractional distribution is determined by PFT competition for nutrients (in the form of net primary productivity) and space, which in turn is highly influenced by climate variability and natural disturbances.

3.4.4 General description of TRIFFID


The vegetation module TRIFFID (Top-down Representation of Interactive Foliage and Flora Including Dynamics) is a DGVM developed at the Hadley Centre for use in coupled climate-carbon cycle simulations, fully described in Cox et al. (2001). It describes the state of the terrestrial biosphere in terms of soil carbon content and vegetation distribution, which is expressed through the structure and coverage of five plant functional types (PFT): broadleaf tree,

CHAPTER 3. MODEL DESCRIPTION

39

needleleaf tree, C3 grass, C4 grass, and shrub. Plant distribution and soil carbon levels are updated based on a carbon balance approach, using land-atmosphere carbon fluxes (for example, plant photosynthesis and respiration) supplied by the land surface scheme MOSES-2 to drive vegetation changes. These fluxes are derived for each PFT using the photosynthesisstomatal approach of Cox et al. (1999). Areal coverage is determined by the net available carbon and interspecies competition, which is modeled using a Lotka-Volterra approach. The model also accounts for bud-burst, leaf-drop and large-scale vegetation disturbances that increase the soil carbon content.

3.4.5 Vegetation dynamics


The state of the terrestrial vegetation in TRIFFID depends on net primary productivity (NPP) i, which is provided for each plant functional type i by the MOSES-2 land surface scheme. A fraction i of this NPP is employed to increase the area of the particular PFT, while the remainder serves towards the growth of the existing vegetated area (in terms of leaf area index and canopy height). The evolution of its fractional coverage i is therefore governed by the following differential equation:

where

is the PFTs vegetation carbon and

. Here the first term on the

right-hand side denotes the expansion of the PFTs fractional cover in the grid cell, which is met however with a certain amount of resistance from other PFTs (as given by the term in brackets). The competition terms cij, which can range from zero to unity, represent the ability of vegetation type j to dominate over vegetation type i" and reduce the growth of i. They are determined through the Lotka-Volterra approach, which emphasizes the role of height in the vegetation dominance hierarchy. In addition to pressure from other vegetation types, each PFT experiences intraspecies competition (cii = 1) to prevent it from expanding into territory which it already

CHAPTER 3. MODEL DESCRIPTION

40

occupies. In order to allow a vegetation type to appear in a previously unoccupied grid cell (for example, when climate and competition levels become favorable), each PFT is seeded by never letting the effective fractional cover drop below a specified seed fraction. For the sake

of total carbon conservation, the fraction of NPP which cannot contribute to the expansion of this PFT due to competition is considered wasted and is returned to the soil. Finally, the second term on the right-hand side accounts for large-scale disturbance events, such as forest fires or insect swarms, which result in the loss of vegetated area at a prescribed rate . , combines

The total amount of vegetation carbon for a PFT, denoted by the variable

all of the carbon accumulated in the leaves, stems, and roots. Its evolution, which is coupled to that of areal fraction, is given by the relatively simple equation

In the first term on the right-hand side, all of the primary production not used to expand the fractional coverage of the PFT (or lost to PFT competition) goes towards increasing vegetation carbon. The second term accounts for the loss of vegetation carbon through litterfall, which is parameterized according to the turnover rates for leaves, roots and stems. There is an additional litter contribution from large-scale disturbances that destroy vegetation, but it is not explicitly included in the definition of because the phenomenon is already accounted for in equation 7.

3.4.6 Leaf phenology and soil carbon


The phenological state of the vegetation is calculated based on the maximum potential leaf area index Lb of trees and shrubs: , where L is the actual LAI of the canopy and p is

a fraction between zero and unity. Bud burst and leaf turnover rates are set to be equal under normal circumstances, but leaf mortality increases if the surface temperature drops below a critical threshold. In order to ensure conservation of carbon during phenological changes, the actual rate of leaf drop (used to calculate litterfall in equation 8) is computed separately. Trees

CHAPTER 3. MODEL DESCRIPTION and shrubs are allowed to grow towards full leaf status (

41 ) whenever the rate of leaf

turnover does not exceed twice that of bud burst; otherwise, p steadily decreases, leading to a decline in LAI. Overall, this parameterization of leaf phenology results in a seasonal variation of the canopy LAI of vegetation; this is limited only to trees and shrubs, however, as a similar approach has not yet been included for grasses. Soil carbon comprises all of the carbon which is stored on the land surface but not currently used by any plant functional type. It is increased by total plant litterfall, and a fraction of it is released on each timestep as CO2 to the atmosphere due to microbial respiration. The total litterfall, c, tallies all of the dead vegetation carbon accumulated from fallen leaves, largescale perturbation events, and wasted NPP due to PFT competition. The latter term implies that all of the NPP devoted to areal expansion will be converted to soil carbon once a PFT occupies all of the space available to it. The rate of respiration, RS, is given by a complex

parameterization based on soil temperature, volumetric soil moisture and soil carbon content. The temperature dependence is assumed to be weakly exponential (in a Q10 form), while moisture dependence takes a quasi-parabolic shape reaching a maximum upon a specified optimum moisture level.

3.4.7 Biophysical parameters in MOSES-2


In the biophysical feedback loop, TRIFFID employs several parameters supplied by the land surface scheme in its evaluation of vegetation changes, and then returns information on leaf area index and canopy height for each PFT that are used by MOSES-2 to recalculate its own biophysical parameters (while not explicitly computed in TRIFFID, canopy height is diagnosed directly from total stem biomass). Three such parameters are obtained in this manner:

aerodynamic roughness length, canopy catchment capacity, and surface albedo. Roughness length, which modifies the transport of heat, moisture, CO2 and momentum near the surface, is taken to be directly proportional to height. Canopy catchment, which affects the amount of moisture available for evaporation, has an assumed linear dependence on leaf area index.

CHAPTER 3. MODEL DESCRIPTION

42

More relevant to this study is surface albedo, which is calculated for each vegetation tile as the sum of soil albedo 00 and canopy albedo 0, weighted by leaf area index L:

where

represents the fraction of incoming light that passes through the vegetation

canopy and reaches soil level. In the case of snow-free land surface, canopy albedo is specified as for tree types, and for grasses and shrubs, and soil albedo takes the form

of a geographically-varying field as presented in (Wilson and Henderson-Sellers, 1985). When blanketed by snow both albedos become prescribed, PFT-dependent parameters. Snow albedo at the surface takes the value of canopy albedo is prescribed as for shrubs. for trees and for tree types, for grasses and shrubs, while for grass types and

3.4.8 Coupling with the UVic ESCM


All information required by the land surface scheme (radiation, heat fluxes and precipitation rates) are computed within the atmospheric model and passed to MOSES-2, which uses it to evaluate land-atmosphere heat and carbon fluxes and continental runoff. Net primary productivity is calculated in the land surface scheme, and passed to TRIFFID which distributes it into the growth and expansion of each PFT. The distribution and physical characteristics of the terrestrial vegetation (canopy height, leaf area index, etc), as well as their associated land surface parameters (albedo, roughness length, canopy catchment) are updated and returned to MOSES-2 every 30 days. The only exception concerns the phenological status of leaves, which is updated daily based on accumulated temperature-dependent leaf mortality rates. The typical coupling period between the atmosphere-ocean-sea ice system and the land surface scheme is 60 hours, while information concerning the net primary productivity of plants is sent to TRIFFID on a monthly basis (Meissner et al., 2003).

Chapter 4
Results of the transient simulations
4.1 An overview of the original study by Doughty et al. (2010)
The idea of associating the megafaunal extinction with climate change due to alterations in the vegetation cover was formulated by Doughty, Wolff and Field (2010, henceforth DWF2010), who used both an observational and a modeling approach to justify their assertion. In particular, they noted that the mass extinction coincided with a drastic change in vegetation, especially in Alaska and northern Asia, and sought to link the two events causally. Due to the general lack of paleoevidence, mostly in regards to megafauna remains (which makes it difficult to determine the exact date of extinction in a particular region), several assumptions were required, notably (1) that vegetation change followed the megafauna extinction rather that preceded it (Gill et al. 2009), and (2) that some of the larger species had a diet which would have involved the uprooting of a large number of trees (Owens-Smith 1988). In this regard, the case of the woolly mammoth is especially strong because their behavior can be directly related to that of their modern-day elephant cousins, which are known to play a determinant role in the maintenance of grassland and the expansion of trees in the African savanna (Caughley 1976). In their pioneering work, DWF2010 showed that pollen data records indicate a rapid increase in Betula over Siberia and Beringia (which encompasses territory within current-day Alaska and the Yukon) close to the time of the megafaunal extinction. Based on this information, they hypothesized that this increase in vegetation cover could not be entirely attributable to climate change and accordingly they estimated the part of the increase in Betula that would be caused by the extinction of the terrestrial megaherbivores. The pollen data were obtained from a compilation of the Global Pollen Database for the above regions between 10 and 20 ky BP, and these data were used to reconstruct vegetation cover during that time span and hence estimate the percentage cover of Betula. In addition, archaeological evidence for human and mammoth 43

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

44

presence was used to estimate the time of the mass extinction. In the modeling effort, several scenarios of elephant-tree interactions were examined using an extended Lotka-Volterra predator-prey model (Duffy et al. 1999) with a range of mature Siberian vegetation densities and a wide range of mammoth behavior scenarios (40-1200 trees uprooted/mammoth/year, following a Monte Carlo approach) in order to predict the impact of megafauna on vegetation cover and estimate the reduced Siberian and Beringian dwarf deciduous tree cover prior to the Holocene. A reduced percentage cover of dwarf deciduous trees was then prescribed in the NCAR CAM 3.0, a dynamic atmospheric model coupled to a slab ocean model, in order to evaluate the impact of this vegetation change on global temperatures. Finally, results from both the predator-prey and climate system models were combined to obtain a quantitative measure of temperature changes that would be directly attributable to the megafaunal extinction. In DWF2010, analysis of the pollen database revealed an average increase in Betula pollen of 26% over a span of roughly 850 years, corresponding in time with the archaeological evidence for the occupation of the land by humans and the extinction of mammoths in the area. Atmospheric temperature and carbon dioxide concentrations obtained from Greenland ice core temperatures and CO2 proxies confirmed that this time period also coincided with rapidly changing climate conditions, resulting in increasingly hospitable conditions for dwarf deciduous trees in northern high latitudes. Results from the predator-prey model suggested that on average 23% of the increase in Betula could be attributed to the mammoth extinction (up to 50% in regions of dense vegetation and large mammoth population), with the rest caused by natural climate change. Climate simulations indicated that each percent increase in high latitude

deciduous dwarf tree cover would results in a globally averaged 2-meter air temperature increase of 0.0043C (up to 0.021C locally); these numbers take into consideration both the decrease in surface albedo (positive feedback) and increased carbon sequestration (negative feedback) that result from an increase in tree cover. In their paper, DWF2010 combined these results to obtain an additional 6% increase in Betula (23% of 26%) due to the mammoth extinction, yielding an

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

45

additional regional warming of 0.13C (globally : 0.026C). These numbers were used to suggest that the Pleistocene megafaunal extinction had some impact on land cover.

4.2

Description of the present experiment

4.2.1 Differences with the original study


The objective of the study in this thesis is to extend the modeling effort of DWF2010 with a more detailed experimental approach. In DWF2010 the temperature response to changes in vegetation cover is obtained by comparing 100-year equilibrium (snapshot) simulations of the climate system with different areal coverage of deciduous dwarf trees in the high latitudes (for example, 20% cover vs. 40% cover). In a similar manner, feedbacks from carbon-cycle effects are determined by comparing the equilibrium global temperature for different values of prescribed CO2 levels. Other than these two factors, it is assumed that the climate is simulated within the context of pre-industrial boundary conditions (i.e. orbital parameters, extent of the ice sheets). In contrast, (1) this study examines the transient response (over 1000 years) of the climate system to changes in vegetation cover, and (2) uses an Earth system model of intermediate complexity with late Pleistocene boundary conditions (around 12-17 y BP) to simulate this climate change. (3) Another important difference lies in the treatment of vegetation: in DWF2010 different values of deciduous tree cover are prescribed within the same climatic context, and the effect on temperature can be directly calculated by comparing two simulations; in our study vegetation is constrained by the presence of mammoths, but allowed to evolve over time (in reaction to changes in climate) with the use of the dynamic vegetation model TRIFFID. As a modeling tool, the UVic ESCM is well suited to the project for a number of reasons. First, it is relatively inexpensive, allowing as much as a thousand model years to be computed in less than two weeks. This comes at the substantial cost of reducing the atmosphere to a

somewhat simplistic energy-moisture balance model with fixed winds, which severely limits the number of processes in the atmosphere's response to forcing in the land surface scheme (and thus

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

46

the energy exchanges with all other components of the climate system). The model therefore sacrifices short-term variability (i.e., weather) and the ability to produce a dynamical atmospheric response for the sake of computational efficiency. However, these limitations are not so detrimental to long-term simulations of the climate system (spanning an interval of time many orders of magnitude longer than the characteristic timescale for atmospheric response), which focus on climatic feedbacks that arise from energy imbalances in the atmosphere. Second, the UVic model incorporates a full ocean general circulation model, which is an important asset to this study because millennial-scale climate variability is mostly driven by ocean dynamics and by the very long timescale of oceanic response to external forcings (such as orbital cycles and CO2 fluctuations in the atmosphere). Finally, the land surface scheme accounts for a dynamical treatment of vegetation feedbacks, and the plant functional type (PFT) approach in TRIFFID allows a very simple parameterization of the megafaunal extinction (see below) to be used within climate model simulations.

4.2.2 Experimental approach


The first step in this experiment consists of adding a slight modification to the UVic model in order to simulate the mammoth extinction and its impact on global climate through an increase in high latitude tree cover (as per the hypothesis formulated in DWF2010). Since it would be unphysical to force the growth of trees beyond the models conceivable limits, a measurable change in climate can only be achieved by first removing a fraction of the tree vegetation and then letting it grow back. Therefore, our strategy for implementing the

megafaunal extinctions within the UVic model must first start by introducing a perturbation into the model in the form of reduced tree cover, and specifying an area of the worlds land surface (the mammoth habitat) in which to apply it. In the context of the PFT approach in TRIFFID, this perturbation amounts to limiting the growth of trees and shrubs in favor of C3 and C4 grasses, much in the same way as one would account for agricultural lands in the present-day configuration of vegetation in fact, the mammoth habitat is defined as croplands in the model

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

47

and overwrites the map of agricultural lands provided in the model package; this is not an issue, however, as there is little evidence for human agriculture during the late Pleistocene. There are two ways for which such a perturbation can be introduced in the climate model. One way is to spin up the model to equilibrium from rest with the reduced tree cover in the northern high latitudes (for boundary conditions corresponding to the approximate time of mammoth extinction), a lengthy process due to the long response time of the ocean (at least 5000 model years). In the context of this study we favored a less time-consuming alternative which consists of inserting the tree cover perturbation at some point of a transient model simulation (i.e., start the simulation with an already spun-up model), and letting the climate system evolve until a new, approximate equilibrium is reached (a few preliminary tests revealed that 500 years of climate model simulation were sufficient for the temperature signal to stabilize). The starting point for all of our simulations is an extensive model run spanning over 25 thousand model years (initiated in the context of another study), which was selected due to its time-dependent prescription of carbon dioxide levels in the atmosphere (necessary in order to isolate the effect of biogeophysical feedbacks. In the final step of this experimental strategy, we eliminate the perturbation to high latitude tree cover an event which symbolizes the extinction of the Pleistocene megafauna and we let the subsequent recovery of forest biomes act as the main driver of climate change for the rest of the simulation. Due to the strong dominance of tree and shrub PFT in the competition scheme, it takes only a few hundred years for boreal forests to fully recover from the perturbation. In order to isolate the warming signal due to biophysical feedbacks only, it is necessary to compare the output with that of a no extinction simulation, in which trees and shrubs are not allowed to grow back even after the supposed time of extinction.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

48

4.3

The maximum impact scenario

4.3.1 Short description and parameter tuning


The purpose of this experiment is to quantify the highest amount of warming that can be obtained in climate simulations with the UVic model as a result of high latitude vegetation changes. Most of the parameters related to the mammoth extinction are optimized to have the greatest possible impact on the climate system; this particular simulation is therefore not intended to offer a realistic portrait of climatic feedbacks to the megafaunal extinction. However, since these parameterizations essentially lead to a quick replacement of grassland by small trees (akin to several afforestation experiments), results from this experiment will also serve the second purpose of identifying general climatic feedbacks to high latitude vegetation change in the context of late Pleistocene boundary conditions, a novelty for this particular era. In this experiment we define the mammoth habitat as any land grid cell located north of the 30N latitude. This particular number was chosen because it represents the approximate southernmost limit of boreal forests in the northern hemisphere (needleleaf trees in TRIFFID), and also because most of the vegetation in North America is constrained between 30N and 45N due to the overwhelming presence of the Laurentide ice sheet at higher latitudes during the late Pleistocene. Additionally, mammoth are assumed to uproot every single tree within their habitat, which we represent my setting the perturbed tree and shrub fraction to zero everywhere on this large territory. Finally, the extinction of the Pleistocene mammals is assumed to be

instantaneous, in order to minimize the time required for forest biomes to regain their original status. The catastrophic extinction is taken to occur during the year 12000 BC (14 ky BP), as suggested by the evidence from various burial sites mentioned in DWF2010; this last parameter is not optimized, but its impact on the temperature signal will be examined in section 4.4. In the following sub-sections we will examine the evolution of various climate parameters over the following 500 years (from 14 ky BP to 13.5 ky BP).

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

49

4.3.1 Vegetation and surface albedo changes


The change in the fractions of trees, grasses and shrubs (in the mammoth habitat) is displayed in Figure 4.1. There are two striking features in the results shown here. First, we note that the major part of vegetation change occurs within the first 100 years of the simulation. While it is certainly unreasonable to expect the boreal forest to recover so quickly, this issue might be more associated with the model itself rather that the unrealistic scenario. In particular, the parameterization of competitiveness and the height-based dominance hierarchy in TRIFFID are likely to be responsible for this aggressive invasion of the boreal forest. Second, the forest recovery (which extends to the southern limit of 30N) is made up almost exclusively of shrubs, until needleleaf trees start appearing in Europe and southwestern North America during the last century of the simulation, in large part due to the natural increase in global

Figure 4.1: Change in vegetation fraction over the mammoth habitat (all land north of 30N) simulated by the UVic ESCM in the context of a maximum impact scenario. This figure and every subsequent one represent the difference between a simulation where mammoths go extinct, and a simulation where their extinction does not occur.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

50

temperatures. While this result does not contradict the conclusions of DWF2010 (based on its physical characteristics, and height in particular, the Betula species would be classified as shrub in the PFT scheme), it is surprising to find that needleleaf trees are almost non-existent north of 30N despite their high tolerance to cold (in TRIFFID they are prescribed to survive temperatures as low as -30C). Figure 4.2 displays the timeseries of surface albedo, which closely follows the change in vegetation. After 500 years of climate model simulations, the change in surface albedo that arises from biogeophysical feedbacks only amounts to -0.026 locally (averaged over mammoth habitat, see Fig. 4.2), and approximately -0.006 globally (not shown). The spatial distribution of this increase in albedo can be found in Figure 4.3a. We note that a large portion of the northern landmasses, especially in North America, is unavailable for tree growth due to the presence of massive ice sheets (see Figure 4.3b). Furthermore, several places in Asia are either too cold (in Northern Siberia) or too dry (all of the southern half, with the notable exception of the Himalayan mountain range) to support the growth of trees, limiting the appearance of shrubs to a (relatively) narrow strip of land stretching from Europe to the Pacific coast and western Alaska, as well as isolated blobs (the Himalayas, southwestern North America). Since shrubs are very similar to grasses in terms of snow-free surface reflectivity, a large part of this albedo decrease is due to the difference in snow-covered canopy albedo (0.6 for
Figure 4.2 : Change in surface albedo over the mammoth habitat (all land north of 30N) simulated by the UVic ESCM in the context of a maximum impact scenario.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

51

grasses versus 0.4 for shrubs). This fact is strongly evidenced in Figure 4.4, which displays the full annual cycle of surface albedo anomalies in the Northern Hemisphere. The largest departure in surface albedo are clearly located at the reforested latitudes, and the anomaly all but vanishes during summer and early fall when the ground is assumed to be snow free. In the context of late Pleistocene boundary conditions, the glacial climate regime in the Northern Hemisphere results in a much later melting of the snowpack, as characterized by albedo anomalies that persist until mid-June in these latitudes. This contrasts with earlier studies (notably, Thomas and Rowntree (1992), Chalita and Le Treut (1994)), in which present-day boundary conditions result in a March or April meltdown. It is also interesting to note that a minimum in surface albedo anomaly seems to occur during June, right before the anomalies vanish altogether at the NH midlatitudes. An analysis of surface air temperature anomalies (see Figure 4.7 below) and land surface temperature anomalies (not shown) also reveals that during the same time period this region is much warmer in the reforestation run than in the control run. These observations lead

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

52

us to conclude that the darker canopy of dwarf trees leads to an earlier spring, and snow melt is hastened by up to a few weeks due to the snow-albedo feedback.

Figure 4.4 : Annual cycle of land surface albedo anomaly in the Northern Hemisphere during the last year of climate model simulations. Solid line represent positive contours, while dotted lines represent negative values. On the abscissa, months are displayed from January to December according to their numerical order.

4.3.3 Temperature
The impacts of this decrease in albedo on temperature are displayed in the form of a globally-averaged (Figure 4.5) and zonally-averaged (Figure 4.6 (a)) timeseries of temperature changes due to biogeophysical feedbacks only. The temperature trend is well-correlated with the decrease in albedo, and 100 years into the simulation the temperature anomaly is approximately 0.110C globally (in Figure 4.5) and an average 0.275C over the mammoth habitat (not shown). In a similar fashion to surface albedo, temperature keeps increasing over the following 400 years, albeit at a reduced rate, reaching 0.175C globally and 0.420C over the mammoth habitat. The additional warming is to be expected since natural deglacial climate change causes the Earth to become warmer and wetter and ice sheets to recede, leaving an ever increasing amount of space to be conquered by trees and shrubs, and thus further lowering the surface albedo compared to the no extinction simulation in which the tree cover is not allowed to expand north. Due to

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

53

Figure 4.5 : Globally-averaged temperature increase due to biogeophysical effects only, in the context of a maximum impact scenario.

this factor it is expected that the temperature departure between the two simulations will keep increasing beyond 500 years, perhaps even indefinitely. Not surprisingly, our numbers are several times higher than those obtained by DWF2010, even after only 100 years of model simulations. However, this difference can be explained by both the use of a much larger territory (although a large part of that area is covered by ice sheets during the late Pleistocene); and the fact that our maximum impact scenario effectively compares 0% tree cover against 100% tree cover (20% vs. 26% for DWF2010). The spatial distribution of temperature changes is displayed in Figure 4.6 (b). At first glance, it would seem that the warming pattern is directly related to changes in surface albedo, with areas that experience rapid reforestation following the megafaunal extinctions (see Fig. 4.3 (a)) also observing the largest increase in temperature. This is especially true for southwestern North America and the Himalayas. However, a few details cannot be fully explained by comparing with Fig. 4.3 (a) only, with two cases standing out in particular. First, we notice that the largest departure (0.6C) occurs over extreme northeastern Asia, which sees a change in albedo comparable to that experienced in central Europe, although the latter only experience half as much warming. Here, we suggest that the additional warming is caused by a strong snow-

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

54

Figure 4.6 : (a) Zonally-averaged temperature difference between the extinction and no-extinction runs; (b) spatial distribution of the temperature anomaly. The dotted lines represent 0.05C isotherms.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

55

albedo feedback, promoting warmer temperatures during the winter and spring (for a complete discussion, see Chapter 2). As mentioned above, there is compelling reason to believe that the warmer spring temperatures due to biogeophysical effects hasten the snowpack melt by a few weeks, creating a strong temperature anomaly during the late spring (see Fig. 4.7). A second case of interest comes from an unexpected area of (slightly) cooler temperatures off the coast of Antarctica, in the Weddell Sea. Since all of the forcing is

happening in the Northern Hemisphere, it makes sense that the distribution of temperature anomalies should also be heavily biased towards the latter. However, it is also known that temperature anomalies over the North Atlantic can lead to significant changes in the oceanic thermohaline circulation, impacting the rate of deep water formation near Antarctica. In our case an analysis of the
14

C content of the deep waters in the Weddell Sea reveals a strengthening

vertical gradient of 14C along with a reduced 14C ratio in the bottom waters (see Fig. 4.8), often indicative of stunted deep water formation and lower ocean temperatures (because sinking ocean

Figure 4.7 : Zonally-averaged, annual cycle of temperature anomalies over the northern Hemisphere. The contour interval of the isotherms is 0.1C. On the abscissa, months are displayed from January to December in their numerical order.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS water loses heat which it releases into the atmosphere).

56

These cooler waters can also lead to

more sea ice formation, triggering a local sea ice-albedo feedback with negative impacts on temperature. The combination of all of the above factors would then explain why this region could experience cooling despite a major warming in the Northern Hemisphere. As mentioned earlier, the annual cycle of temperature anomalies is displayed in Figure 4.7. Although the peak anomalies tend to occur later due to the cooler glacial climate, the seasonal cycle is very consistent with results from other similar studies (Thomas and Rowntree (1992); Bonan et al. (1992); Douville and Royer (1996)), successfully reproducing the strong temperature anomalies during the winter and spring.
Figure 4.8 : Variations in C anomaly as a function of depth. This particular snapshot is taken in the Weddell Sea, in the middle of the cold anomaly in Fig. 4.6(b), and averaged for the entire last year of the simulation.
14

4.3.4 Precipitation
Another important climatic factor, especially when considering the growth of vegetation, is the global distribution of precipitation. Figure 4.9 displays the change in total precipitation that would arise from biophysical feedbacks only. Contrary to our intuition and the results of several studies, notably Thomas and Rowntree (1992), and Bonan et al. (1995), the model output suggests that the increase in temperature brought by the change in vegetation cover is accompanied by a considerable decrease of precipitation rates over land. As shown in Figure 4.9, the total worldwide precipitation eventually recovers, but this is in large part thanks to a compensating increase of precipitation over the oceans.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

57

There are several arguments that could be used to explain the initial drop in precipitation rates that seem to coincide with the drop in surface albedo. Since the maxima of precipitation decrease are centered on the reforested areas (not shown) and occur mostly during the summer months (see Fig. 4.10), one possible explanation could come from parameterization of soil moisture in MOSES and TRIFFID. Looking at the annual cycle of precipitation anomalies can also offer some clues. In our case, it is interesting to note that during most of spring the precipitation anomaly is slightly positive, which would be in line with most studies associating an increase in boreal forest with warmer, wetter conditions. However, this anomaly takes a sudden reversal in June, and stays strongly negative throughout the summer. In this manner, our results are quite similar to those of Chalita and Le Treut (1994), who found that climate-vegetation interaction in Europe resulted in a wet spring followed by a warm, dry summer. However, their paper noted that soil moisture

Figure 4.9 : Change in total precipitation rates, shown for land only and land + ocean.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

58

levels were dramatically low in the early summer, likely because of the increased evaporation during spring, and dropped to a critical point below which precipitation was not possible. This is not the case in our model run, especially over the concerned regions, and so it is deemed unlikely that critically low soil moisture level would explain the anomalous precipitation rates. Similarly, changes in evaporation over land do not provide a convincing lead. Further experimentation with the UVic model will be necessary in order to gain a better understanding of this issue.

Figure 4.10 : Annual cycle of precipitation anomalies in the Northern Hemisphere during the last year of model simulations. -7 Solid lines represent positive contours, while dotted lines represent negative values. The contour interval is in units of 10 2 -1 kg m s . On the abscissa, months are displayed from January to December according to their numerical order.

4.3.5 Sea ice


The impacts on sea ice in the Northern Hemisphere are clearly visible in the model output. The timeseries of global sea ice volume anomalies, as well as the anomaly in sea ice thickness over the Arctic Ocean in late summer, are displayed in Figure 4.11. The change in ice volume follows the same pattern as surface albedo and temperature, and can therefore be directly associated with the decrease in surface albedo. Not surprisingly, the largest thickness change occurs on the Asian side of the Arctic, where most of the vegetation change occurs. Further

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

59

investigation will be required to determine whether this change in sea ice has a significant impact on temperature (through the sea ice-albedo feedback mechanism).

4.4

A set of more realistic experiments

4.4.1 Description of the experiments


The above is a hypothetical maximum impact scenario, with parameters set to unrealistically create a large perturbation. A criticism often heard when presenting the above results in seminars was that the mammoths could not survive if the very source of their diet (trees and shrubs) was completely removed. Or that their habitat could not possibly span such a large area as suggested in the maximum impact scenario. With these criticisms in mind, and given the large uncertainty regarding the mammoth diet and their time of extinction, we next designed a set of simulations to explore the parameter span more likely to have been encountered. The following subsection 4.4.2 deals with a set of sensitivity experiments concerning three circumstantial parameters of the mammoths presence and extinction: rate of tree clearing (referred to as herbivory in some papers), surface area of habitat, and timing of extinction. The first of these terms is parameterized as a reduction of the mammoths influence on the biosphere: instead of systematically causing the disappearance of forestry in their habitat, trees

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

60

and shrubs are allowed to grow at a rate equivalent to a certain fraction of their unperturbed state. The second factor is treated simply as a northward shift of the southernmost limit of the mammoths habitat. This is the only logical way to proceed since there are insufficient paleodata to reconstruct the exact surface area of their habitat; regardless, it would make but a small difference with the approach employed here, and would therefore be irrelevant to the purpose of this study. The last of these terms consists of starting the simulation at a different time (with different boundary conditions) in order for the extinction to occur in a different context of natural climate variability; in particular, we want to test whether any significant change can arise from the experiment being conducted at a different stage of deglacial climate change. Results from all three of these experiments are put together for the sake of comparison. In the final two subsections of this chapter, we relax two other assumption made in the maximum impact scenario. In subsection 4.4.3, we abandon the instantaneous extinction and look at the climate system evolution in the case of several different curves of gradual extinction : linear, sinusoidal, and exponential. In subsection 4.4.4, we question the use of prescribed CO 2 levels in the atmosphere, and try to determine the impact of biogeochemical effects on the total temperature signal in response to changes in vegetation cover.

4.4.2 Sensitivity tests


Out of many different model runs for this part of the project, we have selected six individual simulations to outline the results from the sensitivity study, two from each of the three parameters described in the above subsection. The details of each experiment are presented in Table 4.1, and these experimental parameterizations are compared with those used in the maximum impact scenario.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS


Parameters of sensitivity study # Date started (yrs after 0 AD) Length of simulation (yrs) Year of extinction -12000 -12000 Fraction of trees cleared 1.00 0.67 Southmost extent of habitat 30N 30N Notes

61

-1

-12500 -12500

1000 1000

Maximum impact scenario Tree clearance decreased by 1/3 Tree clearance decreased by 2/3 Extent of habitat further north Extent of habitat much further north Earlier time of extinction Later time of extinction

2 3 4 5 6

-12500 -12500 -12500 -15500 -10500

1000 1000 1000 1000 1000

-12000 -12000 -12000 -15000 -10000

0.33 1.00 1.00 1.00 1.00

30N 45N 60N 30N 30N

Table 4.1 : List of experiments used in the sensitivity study and their parameterizations. Results from entries in bold are shown in Figure 4.13 in the form of a world map of temperature anomalies 500 years after the prescribed extinction.

Figure 4.12 : Results of the sensitivity tests, presented here as a timeseries of temperature anomalies. The maximum impact scenario is shown in red for the sake of comparison.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS


1 3

62

Figure 4.13 : Spatial distribution of temperature anomalies for various simulations in the set of sensitivity experiments. The number besides each panel refers to the that of the specific experiment in Table 4.1. All of these figures are one-year averaged differences in temperature between the simulation and a related no extinction simulation with similar parameterizations. The year of averaging is 500 yrs after extinction.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS The results from the sensitivity study are shown in Figs. 4.12 and 4.13.

63 Overall these

results correspond well with our intuition. Decreasing the rate of tree clearing effectively reduces the forest recovery in the years following the instantaneous extinction, and therefore reduces the loss of albedo (when compared to the no extinction run) and the increase in temperature. Decreasing herbivory by 33% approximately halves the impact on global

temperatures, while a decrease of herbivory by 67% yields anomalies so insignificant they can hardly be distinguished from background noise. As can be seen in the top left panel of Figure 4.13, all of the main features of the maximum impact scenario (geographical distribution of anomalies, location of the largest departure in eastern Siberia, slight cooling in the Southern Ocean) can be retrieved for these experiments, albeit with diluted numbers. Results from the sensitivity to the area of habitat are also fairly straightforward in their interpretation. A reduction in the area of habitat effectively removes all potential input (in terms of albedo effects) from the areas which are excluded from the smaller habitat. Moving the southern border of the habitat to 45N reduces the impact by about half, while a displacement of this border to 60N essentially negates all possible effects, since there are very few ice-free locations north of this boundary that can support any kind of vegetation at all. The top right panel of Fig. 4.13 still shows several of the aspects of the maximum impact scenario, notably the maximum in Siberia and the negative anomaly near Antarctica. Experimenting on the time of extinction creates a little more interesting impact, since a different time period corresponds to a different stage of deglacial climate change, and a different potential extent of forest recovery. An earlier extinction by 3000 years (15 ky BP) yields a temperature anomaly slightly lower than the maximum impact scenario, but the difference becomes negligible with time. In the end-of-simulation output (bottom left panel of figure 4.13), we note large similarities between the two experiments, with the notable exception that the early extinction scenario does not reproduce the cold temperature anomaly in the Weddell Sea. We hypothesize that slightly lower global temperatures during that period (and therefore a lower

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

64

freshwater flux into the North Atlantic Ocean) prevented a reversal of the thermohaline circulation due to additional warming from the biogeophysical feedbacks. As expected, a later extinction by 2000 years (10 ky BP) results in an even larger increase in SAT anomalies than the maximum impact scenario. The spatial distribution of temperature anomalies (bottom right panel of Figure 4.13) is slightly different at the northern high latitudes due to a moderately large fraction of the continental ice sheets disappearing between 12 ky BP and 10 ky BP (a prescribed feature in the UVic ESCM), which opens up more room for the expansion of various vegetation types.

4.4.3 Gradual extinction experiment


It is clear that any kind of megafauna species did not disappear all at once; instead, their extinction is likely the end result of a slow decline due to a combination of climatic and anthropogenic stresses acting over thousands of years. In order to verify whether a gradual decline in population has a significant impact on the overall result, we designed a series of four simulations with slightly different patterns of population decline (in order to represent the decline in population, we set the fraction of trees cleared in a grid cell as equal to the fraction of the original population left). In the first two, we examine the basic linear pattern, with one being spread over a longer time than the other. The final two simulations both have an intermediate duration of 1000 years, and one simulates an exponential decay (fast early, slow late) of the population while the other follows a sine pattern (slow early and late, fast in between). In each of these simulations, the model is run for an additional 500 years after the population reaches zero. Results from these simulations are shown as a combined timeseries of temperature and albedo in Figure 4.14. Again, changes temperature seems to be very closely correlated to changes in surface albedo. Because of the length of these simulations, we can see some events that could not be witnessed in the 500 year-long maximum impact scenario. For example, in panels (a), (c) and (d), a sharp drop in albedo can be observed which is linked to a considerable

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

65

Figure 4.14 : Results of the gradual extinction experiments, presented in the form of temperature-albedo graphs. The four panels represent each of the individual simulations.

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

66

decrease in the prescribed land ice. The temperature anomaly is also significantly larger in each of these simulations (0.27 C for each of them, compared to 0.175 C), owing to the simulation lasting a longer time (more time for trees to grow and for slower components to respond) and extending to a later period than the maximum impact scenario (all of these end in 9500 YBP). A quick glance at the spatial distribution reveals little difference in the spatial pattern compared to the maximum impact scenario. This experiment was only intended to find out whether a non-instantaneous extinction pattern would result in a different temperature output, possibly due to long-term nonlinear interactions between various components of the climate system. There are numerous possible ways upon which this could be improved or extended: for example, one could subdivide the mammoth habitat into smaller regions, each with its own timing and pattern of mammoth extinction. Whether it would have a measurable impact on the overall result is still questionable, based on what was obtained here.

4.4.4 Free CO2 experiment


Since the main objective of this study was to quantify the climate response to biogeophysical feedbacks alone, it was only fitting to prescribe levels of CO2 in the atmosphere in order to avoid any interference from carbon cycle effects. In this final experiment, however, we turn off this prescription of CO2 and attempt to quantify the combined biogeophysical and biogeochemical effects on the climate response. Apart from this new element, all other

characteristics of the experimentation are left unchanged from section 4.3. Results from this simulation, which are shown in Figure 4.15, are rather counterintuitive. We would have expected the temperature effects from biogeophysical effects to be at least partly offset by interaction with the carbon cycle due to the increased carbon sequestration by trees and shrubs. Instead, they appear to enhance each other, resulting in a combined warming that exceeds double that from biogeophysical effects alone (Fig. 4.15 (a)). Naturally, this results

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

67

Figure 4.15 : A selection of results from the free CO2 experiment. (a) a comparison of the temperature anomaly between the free and prescribed CO2 experiments; (b) difference in atmospheric CO2 between the two simulations; (c) change in total soil carbon resulting from the vegetation change; (d) carbon flux from the atmosphere to the land (since it is mostly negative, it indicates a land-to-atmosphere flux.

from an opposite trend of atmospheric carbon dioxide: as displayed in Fig. 4.15 (b), CO 2 levels are increased by about 15 ppm in the first 150 years following the extinction, after which the trend is reversed, but the overall anomaly after 500 climate model years is still overwhelmingly positive. While the increase of the last 350 years can undoubtedly by attributed to the increase in carbon sequestration, the sudden increase early on cannot. An analysis of carbon fluxes between the land and atmosphere reveals that the initial increase in CO2 is closely related to the vegetation change. While this could be clear from the coincidence of both the increase in CO2 and the rapid afforestation (as represented by the decrease in surface albedo), it appears the UVic model, and TRIFFID in particular, defines

CHAPTER 4. RESULTS OF THE TRANSIENT SIMULATIONS

68

different limits of soil carbon content depending on the dominant PFT. Since this limit is lower for shrubs than C3 grasses, the replacement of the latter by the former in several grid cells north of 30N causes a massive release of soil carbon to the atmosphere. The change in soil carbon content is portrayed in Fig. 4.15 (c), and Fig. 4.15 (d) confirms that most of it is transferred into the atmosphere. As a side note, the blips that appear on Fig. 4.15 (d) are in fact caused by a decrease in the (prescribed) continental ice sheets. A bug in the model causes all of the land carbon in a grid cell to be trapped under the ice when the continental ice sheets are prescribed to appear; this carbon is then suddenly released into the atmosphere when the ice is removed. The changes visible in Fig. 4.15 (d) are very small (compared to those that can be seen in Fig. 4.14), and thus would not be easily observed in a temperature anomaly plot. It is difficult at this point to determine whether the carbon cycle response to the vegetation change is genuine or a product of some model artifact. Since logic would dictate that the increase in plant carbon sequestration should dominate over any other effects, a much more detailed experimental setup would be required in order to draw conclusions from these results.

Chapter 5
Conclusions
5.1 Summary
The objective of this thesis was to investigate biophysical feedbacks between the fauna, flora, and climate in the context of the Late Quaternary Extinctions, and provide a quantitative assessment of those feedbacks in terms of temperature and other climate parameters. To this end, we developed an experimental strategy that consisted of first prescribing a decrease in tree cover over a pre-defined mammoth habitat in the coupled UVic-TRIFFID ESCM, then allowing the model to reach an approximate equilibrium with the new environmental conditions, and finally terminating the perturbation in order to examine the transient climatic response to a subsequent recovery of tree fraction. This setup was used to explore several hypothetical cases of the mammoth extinction, including a catastrophic maximum impact scenario and a collection of more realistic variations of the former. Results from the maximum impact scenario were mixed, with some being rather unsurprising while others were more difficult to explain within a physical context. Due to a strong height-based plant dominance hierarchy in TRIFFID, shrub types were found to quickly recover their unperturbed territory following the mammoth extinction, perhaps too quickly for a natural reforestation. Tree types, on the other hand, did not conquer much terrain during the 500 simulation years, and thus had on their own a limited impact on the climate system. Since the experiment was set up specifically for changes in vegetation to drive the climate response, it follows that many other climate parameters observed a similar trend to that experienced by vegetation distribution. We noted a sharp decrease of surface albedo for the initial 100 years (after extinction), which then became very subtle for the rest of the simulation. Overall, after 500 years of climate model simulations, the albedo over land decreased by a little

69

CHAPTER 5. CONCLUSIONS

70

under 0.006, while strictly over the mammoth habitat the decrease was approximately 0.026. There was little change outside of the Eurasian subcontinent, mostly because of the Laurentide ice sheet, which was still quite large at the given time of simulation. Accordingly, it goes without saying that under warmer, ice-free conditions in North America the areal cover of albedo reduction would have been much larger. Another interesting consequence of late Pleistocene boundary conditions was found in the seasonal cycle of albedo anomalies, which revealed that the snow-masking effect of shrubs lasted several months longer into the spring season, suggesting that the change in vegetation might have had a greater effect than it would for present-day conditions. We found the change in temperature to be dominated by the albedo feedback, and at the surface the estimated warming was 0.175C globally and 0.420C over the mammoth habitat. We hypothesize that any warming happening after the initial 100-year pulse would be an indirect consequence of deglacial climate change, which over time makes more room for trees and shrubs to move in (as a result of melting ice sheets, as well as warmer and wetter conditions). As outlined in previous studies (Thomas and Rowntree (1992); Bonan et al. (1992); Douville and Royer (1996)) the warming is found to be greatest during winter and spring. Although neutral or positive temperature anomalies are observed almost everywhere on the globe, there is an area of slightly cooler temperatures in the Southern Ocean. After examining a depth profile of the 14C tracer, we suggest that a cooling at this location would likely be caused by a weakening of the overturning circulation, a result of which is to mitigate deep water formation in the Southern Ocean. This sporadically cooler anomalies could also enhance the formation of sea ice over these regions, a possible explanation to the albedo anomalies observed in that region. Still within the context of a maximum impact scenario, we examined the transient response of two other climate parameters, namely precipitation and sea ice. In regards to precipitation, we observed an unequivocal and unexpected drop in precipitation rates, especially during summer over the reforested latitudes. While these results can be related to at least one major study (Chalita and le Treut (1994)), they seem to contradict the notion that with warmer

CHAPTER 5. CONCLUSIONS

71

temperatures come more moisture availability for precipitation. More investigations with the UVic are likely necessary to get a satisfying answer. On the other hand, sea ice anomalies were found to go in the expected direction, and correspondingly the UVic model calculated a drop in sea ice volume during the melting season in the Arctic Ocean. In general, the range of sensitivity and gradual extinction studies were found to correspond well with our intuition. Decreasing the mammoth habitat or the tree clearance ratio resulted in a drop of the temperature response, while the timing of extinction did not seem to have a major impact (among the range tested, from 15 ky BP to 10 ky BP). The main conclusion from the gradual extinction tests was that the shape of the extinction pattern does not have a significant impact on the long-term temperature response. However, these simulations

confirmed that temperature anomalies continue increasing at least several thousands of years after the extinction. A simulation of free carbon exchanges with the atmosphere yielded

counterintuitive results, as the reforestation was found to coincide with a large CO2 anomaly in the atmosphere, most of it coming from the land surface reservoir. These results are likely not representative of reality, and further investigation with the UVic treatment of the land carbon cycle will be necessary before these results are taken into consideration. We therefore conclude that, with our experimental strategy, it was possible to reproduce and quantify climatic effects of the megafaunal extinctions within the UVic ESCM. Our results were of comparable significance to that obtained by several other studies which examined the impact of high-latitude vegetation change on either present-day or past (i.e., LGM or midHolocene) climate conditions. However, some of our simulations produced surprising, or at times unprecedented, results which could not be related to previous studies; for example, the decrease in precipitation observed in all simulations, or an increase in atmospheric carbon dioxide caused by the recovery of tree fraction, both of which could be caused by physical inconsistencies within the UVic model. Further experimentation with the UVic model will tell whether they are genuine physical processes or unwanted model artifacts.

CHAPTER 5. CONCLUSIONS

72

5.2

Future work
In general, many paleoclimate model studies involving vegetation feedbacks are focused

on either the mid-Holocene or the Last Glacial Maximum, both of which are peripheral to the period studied in this thesis. In a way, our work is a bit of an oddity for its temporal frame, which makes it a bit difficult to compare our numbers with other works. In that sense, our maximum impact scenario can also be looked at as a high-latitude (>30N) reforestation experiment within late Pleistocene conditions, which is a bit of a novelty for paleoclimate studies. It could be useful for future work to look at the climate response in other models (of intermediate complexity), since numerical results from one single model can hardly be taken without a grain of salt. An element of the Doughty et al. (2010) study that we overlooked throughout the entire project is the predator-prey relationship between the mammoth and the tree vegetation. Whereas the original paper used a Lotka-Volterra approach to simulate a range of realistic tree-mammoth scenarios, in this project we simply wrote of this relationship in terms of a pre-defined tree grazing rate, whose influence we examine in Section 4.4.2. Hence, a further extension could be made to our work by including the results of a predator-prey model as a dynamical component impacting vegetation cover. Of course, this could be rather difficult with the current version of TRIFFID, as it does not clearly define any measure of tree density which would be used in such a case; an altogether different model could be the best choice in order to apply this suggestion. Finally, as it was mentioned several times in the thesis, a thorough investigation of land surface processes would be required in order to shed some light on peculiar results obtained in the context of our study. In our case, both the counterintuitive precipitation and carbon dioxide trends seem to happen during a sudden change in the dominant PFT (and then, it could be argued that the abruptness itself is also caused by an irregularity in the competition scheme). A similar experiment with a different vegetation model might also enlighten us as to whether or not the processes suggested by TRIFFID are physically meaningful.

References
ALROY, J. (2001). A multispecies overkill simulation of the end-Pleistocene megafaunal mass extinction. Science, 292, 1893-1896. BALA, G., CALDEIRA, K., WICKETT, M., PHILLIPS, T. J., LOBELL, D. B., DELIRE, C., ET AL. (2007). Combined climate and carbon-cycle effects of large-scale deforestation. Proc. Natl. Acad. Sci. USA, 104, 6550-6555. BARNOSKY, A. D., KOCH, P. L., FERANEC, R. S., WING, S. L., & SHABEL, A. B. (2004). Assessing the causes of late Pleistocene extinctions on the continents. Science, 306, 70-75. BATHIANY, S., CLAUSSEN, M., BROVKIN, V., RADDATZ, T., & GAYLER, V. (2010). Combined biogeophysical and biogeochemical effects of large-scale forest cover changes in the MPI earth system model. Biogeosciences Discuss., 7, 387-428. BERGER, A. L. (1978). Long-term variations of daily insolation and Quaternary climatic changes. J. Atmos. Sci., 35, 2362-2367. BERINGER, J., TAPPER, N. J., MCHUGH, I., CHAPIN III, F. S., LYNCH, A. H., SERREZE, M. C., ET AL. (2001). Impact of Arctic treeline on synoptic climate. Geophys. Res. Lett., 28, 4247-4250. BETTS, A. K., & BALL, J. H. (1997). Albedo over the boreal forest. J. Geophys. Res., 102, 28,90128,909. BETTS, A. K., BALL, J. H., & MCCAUGHEY, J. H. (2001). Near-surface climate in the boreal forest. J. Geophys. Res., 106, 33,529-22,541. BETTS, R. A., COX, P. M., LEE, S. E., & WOODWARD, F. I. (1997). Contrasting physiological and structural vegetation feedbacks in climate change simulations. Nature, 387, 796-799. BITZ, C. M., & LIPSCOMB, W. H. (1999). An energy-conserving thermodynamic model of sea ice. J. Geophys. Res., 104, 15,669-15,677. BITZ, C. M., HOLLAND, M. M., WEAVER, A. J., & EBY, M. (2001). Simulating the ice-thickness distribution in a coupled climate model. J. Geophys. Res., 106, 2441-2463. BONAN, G. B. (2008). Forests and climate change: forcings, feedbacks, and the climate benefits of forests. Science, 320, 1444-1449. 73

REFERENCES

74

BONAN, G. B., CHAPIN III, F. S., & THOMPSON, S. L. (1995). Boreal forests and tundra ecosystems as components of the climate system. Clim. Change, 29, 145-167. BONAN, G. B., POLLARD, D., & THOMPSON, S. L. (1992). Effects of boreal forest vegetation on global climate. Nature, 359, 716-718. BROVKIN, V., CLAUSSEN, M., DRIESSCHAERT, E., FICHEFET, T., KICKLIGHTER, D., LOUTRE, M.-F.,
ET AL.

(2006). Biogeophysical effects of historical land cover changes simulated by six

Earth system models of intermediate complexity. Clim. Dyn., 26, 587-600. BROVKIN, V., GANOPOLSKI, A., & SVIREZHEV, Y. (1997). A continuous climate-vegetation classification for use in climate-biosphere studies. Ecol. Model., 101, 251-261. BROVKIN, V., GANOPOLSKI, A., CLAUSSEN, M., KUBATZKI, C., & PETOUKHOV, V. (1999). Modelling climate response to historical land cover change. Glob. Ecol. Biogeogr., 8, 509-517. BROVKIN, V., LEVIS, S., LOUTRE, M.-F., CRUCIFIX, M., CLAUSSEN, M., GANOPOLSKI, A., ET AL. (2003). Stability analysis of the climate-vegetation system in the northern high latitudes. Clim. Change, 57, 119-138. BROVKIN, V., RADDATZ, T., REICK, C. H., CLAUSSEN, M., & GAYLER, V. (2009). Global biogeophysical interactions between forest and climate. Geophys. Res. Lett., 36, L07405. CALDEIRA, K. (2006). Forests, climate, and silicate rock weathering. J. Geochem. Expl., 88, 419422. CAUGHLEY, G. (1976). The elephant problem-an alternative hypothesis. J. East Afr. Wildlife, 14, 265-283. CHALITA, S., & LE TREUT, H. (1994). The albedo of temperate and boreal forest and the Northern Hemisphere climate: a sensitivity experiment using the LMD GCM. Clim. Dyn., 10, 231240. CHARNEY, J., STONE, P. H., & QUIRK, W. J. (1975). Drought in the Sahara: A biogeophysical feedback mechanism. Science, 187, 434-435. CLAUSSEN, M. (1994). On coupling global biome models with climate models. Clim. Res., 4, 203-221.

REFERENCES

75

CLAUSSEN, M. (1998). On multiple solutions of the atmosphere-vegetation system in present-day climate. Glob. Chang. Biol., 4, 549-559. CLAUSSEN, M. (2005). Table of EMICs (Earth System Models of Intermediate Complexity). Potsdam, Germany: Potsdam Institute for Climate Impact Research. CLAUSSEN, M., & GAYLER, V. (1997). The greening of the Sahara during the mid-Holocene: results of an interactive atmosphere-biome model. Global Ecol. Biogeog. Lett., 6, 369377. CLAUSSEN, M., BROVKIN, V., & GANOPOLSKI, A. (2001). Biogeophysical versus biogeochemical feedbacks of large-scale land cover change. Geophys. Res. Lett., 28, 1011-1014. CLAUSSEN, M., BROVKIN, V., GANOPOLSKI, A., KUBATZKI, C., & PETOUKHOV, V. (1998). Modelling global terrestrial vegetation-climate interaction. Phil. Trans. R. Soc. Lond. B, 353, 53-63. CLAUSSEN, M., MYSAK, L. A., WEAVER, A. J., CRUCIFIX, M., FICHEFET, T., LOUTRE, M.-F., ET AL. (2002). Earth system models of intermediate complexity: closing the gap in the spectrum of climate system models. Clim. Dyn., 18, 579-586. COX, P. (2001). Description of the "TRIFFID" Dynamic Global Vegetation Model. Met Office: Hadley Centre technical note 24. COX, P. M., BETTS, R. A., BUNTON, C. B., ESSERY, R. L., ROWNTREE, P. R., & SMITH, J. (1999). The impact of new land surface physics on the GCM simulation of climate and climate sensitivity. Clim. Dyn., 15, 183-203. COX, P. M., BETTS, R. A., JONES, C. D., SPALL, S. A., & TOTTERDELL, I. J. (2001). Modelling vegetation and the carbon cycle as interactive elements of the climate system. Met Office: Hadley Centre technical note 23. CROWLEY, T. J., & BAUM, S. K. (1997). Effect of vegetation on an ice-age climate model simulation. J. Geophys. Res., 102, 16,463-16,480. CRUCIFIX, M., & HEWITT, C. D. (2005). Impact of vegetation changes on the dynamics of the atmosphere at the Last Glacial Maximum. Clim. Dyn., 25, 447-459. CRUCIFIX, M., BETTS, R. A., & COX, P. M. (2005). Vegetation and climate variability: a GCM modelling study. Clim. Dyn., 24, 457-467.

REFERENCES

76

CRUCIFIX, M., BETTS, R. A., & HEWITT, C. D. (2005). Pre-industrial-potential and Last Glacial Maximum global vegetation simulated with a coupled climate-biosphere model: Diagnosis of bioclimatic relationships. Glob. Planet. Change, 45, 295-312. CRUCIFIX, M., LOUTRE, M.-F., TULKENS, P., FICHEFET, T., & BERGER, A. (2002). Climate evolution during the Holocene: a study with an Earth system model of intermediate complexity. Clim. Dyn., 19, 43-60. DAVIN, E. L., & DE NOBLET-DUCOUDR, N. (2010). Climatic impact of global-scale deforestation: radiative versus nonradiative processes. J. Clim., 23, 97-112.
DE

NOBLET, N. I., PRENTICE, I. C., JOUSSAUME, S., TEXIER, D., BOTTA, A., & HAXELTINE, A. (1996). Possible role of atmosphere-biosphere interactions in triggering the last glaciation. Geophys. Res. Lett., 23, 3191-3194.

DICKINSON, R. E., & HENDERSON-SELLERS, A. (1988). Modelling tropical deforestation: A study of GCM land-surface parametrizations. Q. J. R. Meteorol. Soc., 114, 439-462. DICKINSON, R. E., HENDERSON-SELLERS, A., KENNEDY, P. J., & WILSON, M. F. (1986). Biosphere-Atmosphere Transfer Scheme (BATS) for the NCAR Community Climate Model. Boulder Colorade: National Center for Atmospheric Research technical note. DOUGHTY, C. E., WOLF, A., & FIELD, C. B. (2010). Biophysical feedbacks between the Pleistocene megafauna extinction and climate: The first humaninduced global warming? Geophys. Res. Lett., 37, L15703. DOUVILLE, H., & ROYER, J.-F. (1997). Influence of the temperate and boreal forests on the Northern Hemisphere climate in the Mto-France climate model. Clim. Dyn., 13, 57-74. DUFFY, K. J., PAGE, R. B., SWART, J. H., & BAJIC, V. B. (1999). Realistic parameter assessment for a well known elephant-tree ecosystem model reveals that limit cycles are unlikely. Ecol. Model., 121, 115-125. EBY, M., ZICKFIELD, K., MONTENEGRO, A., ARCHER, D., MEISSNER, K. J., & WEAVER, A. J. (2009). Lifetime of anthropogenic climate change: millenial time scales of potential CO2 and surface temperature perturbations. J. Clim., 22, 2501-2511.

REFERENCES

77

FANNING, A. F., & WEAVER, A. J. (1996). An atmospheric energy-moisture balance model: climatology, interpentadal climate change, and coupling to an ocean general circulation model. J. Geophys. Res., 128, 15,111-15,128. FOLEY, J. (1994). The sensitivity of the terrestrial biosphere to climatic change: A simulation of the middle Holocene. Global Biogeochem. Cycles, 8, 505-525. FOLEY, J. A., COSTA, M. H., DELIRE, C., RAMANKUTTY, N., & SNYDER, P. (2003). Green surprise? How terrestrial ecosystems could affect earth's climate. Front. Ecol. Environ., 1, 38-44. FOLEY, J. A., KUTZBACH, J. E., COE, M. T., & LEVIS, S. (1994). Feedbacks between climate and boreal forests during the Holocene epoch. Nature, 371, 52-54. FOLEY, J. A., LEVIS, S., PRENTICE, I. C., POLLARD, D., & THOMPSON, S. L. (1998). Coupling dynamic models of climate and vegetation. Glob. Chang. Biol., 4, 561-579. FOLEY, J. A., PRENTICE, I. C., RAMANKUTTY, N., LEVIS, S., POLLARD, D., SITCH, S., ET AL. (1996). An integrated biosphere model of land surface processes, terrestrial carbon balance, and vegetation dynamics. Glob. Biogeochem. Cycles, 10, 603-628. FRAEDRICH, K., JANSEN, H., KIRK, E., & LUNKEIT, F. (2005). The Planet Simulator: Green planet and desert world. Meteorologische Zeitschrift, 14, 305-314. FRAEDRICH, K., KLEIDON, A., & LUNKEIT, F. (1999). A green planet versus a desert world: estimating the effect of vegetation extremes on the atmosphere. J. Clim., 12, 3156-3163. GALLIMORE, R. E., & KUTZBACH, J. E. (1996). Role of orbitally induced changes in tundra area in the onset of glaciation. Nature, 381, 503-505. GALLIMORE, R., JACOB, R., & KUTZBACH, J. (2005). Coupled atmosphere-ocean-vegetation simulations for modern and mid-Holocene climates: role of extratropical vegetation cover feedbacks. Clim. Dyn., 25, 755-776. GANOPOLSKI, A., KUBATZKI, C., CLAUSSEN, M., BROVKIN, V., & PETOUKHOV, V. (1998). The influence of vegetation-atmosphere-ocean interaction on climate during the midHolocene. Science, 280, 1916-1919. GIBBARD, S., CALDEIRA, K., BALA, G., PHILLIPS, T. J., & WICKETT, M. (2005). Climate effects of global land cover change. Geophys. Res. Lett., 32, L23705.

REFERENCES

78

GILL, J. L., WILLIAMS, J. W., JACKSON, S. T., LININGER, K. B., & ROBINSON, G. S. (2009). Pleistocene megafaunal collapse, novel plant communities, and enhanced fire regimes in North America. Science, 326, 1100-1103. GUTMAN, G., OHRING, G., & JOSEPH, J. H. (1984). Interaction between the geobotanic state and climate: a suggected approach and a test with a zonal model. J. Atmos. Sci., 41, 26632678. HANEY, R. L. (1971). Surface thermal boundary condition for ocean circulation models. J. Phys. Ocean., 1, 241-248. HARVEY, L. D. (1988). On the role of high latitude ice, snow, and vegetation feedbacks in the climatic response to external forcing changes. Clim. Change, 13, 191-224. HAYS, J. D., IMBRIE, J., & SHAKLETON, N. J. (1976). Variations in the Earth's Orbit: Pacemaker of the Ice Ages. Science, 194, 1121-1132. HECK, P., LTHI, D., WERNLI, H., & SCHR, C. (2001). Climate impacts of European-scale anthropogenic vegetation changes: A sensitivity study using a regional climate model. J. Geophys. Res., 106, 7817-7835. HENDERSON-SELLERS, A. (1993). Continental vegetation as a dynamic component of a global climate model: a preliminary assessment. Clim. Change, 23, 337-377. HENDERSON-SELLERS, A., DICKINSON, R. E., DURBIDGE, T. B., KENNEDY, P. J., MCGUFFIE, K., & PITMAN, A. J. (1993). Tropical deforestation: modeling local- to regional-scale climate change. J. Geophys. Res., 98, 7289-7315. HIBLER III, W. D. (1979). A dynamic thermodynamic sea ice model. J. Phys. Ocean., 9, 815-846. IMBRIE, J., & IMBRIE, J. Z. (1980). Modeling the climatic response to orbital variations. Science, 207, 943-953. JAHN, A., CLAUSSEN, M., GANOPOLSKI, A., & BROVKIN, V. (2005). Quantifying the effect of vegetation dynamics on the climate of the Last Glacial Maximum. Clim. Past, 1, 1-7. KABAT, P., CLAUSSEN, M., DIRMEYER, P. A., GASH, J. H., DE GUENNI, L. B., MEYBECK, M., ET AL. (2004). Vegetation, water, humans and the climate: A new perspective on an interactive system. In Global Change: The IGBP Series (p. 566). Berlin: Springer.

REFERENCES

79

KITTEL, T. G., STEFFEN, W. L., & CHAPIN III, F. S. (2000). Global and regional modelling of Arctic-boreal vegetation distribution and its sensitivity to altered forcing. Glob. Chang. Biol., 6 (Suppl. 1), 1-18. KLEIDON, A., FRAEDRICH, K., & HEIMANN, M. (2000). A green planet versus a desert world: estimating the maximum effect of vegetation on the land surface climate. Clim. Change, 44, 471-493. KOCH, P. L., & BARNOSKY, A. D. (2006). Late Quaternary Extinctions: state of the debate. Annu. Rev. Ecol. Evol. Syst., 37, 215-250. KPPEN, W. (1936). Das geographische System der Klimate. In W. Kppen, & R. Geiger (Eds.), Handbuch der Klimatologie (p. 46). Berlin: Borntraeger. KUBATZKI, C., & CLAUSSEN, M. (1998). Simulation of the global bio-geophysical interactions during the Last Glacial Maximum. Clim. Dyn., 14, 461-471. KUBATZKI, C., MONTOYA, M., RAHMSTORF, S., GANOPOLSKI, A., & CLAUSSEN, M. (2000). Comparison of the last interglacial climate simulated by a coupled global model of intermediate complexity and an AOGCM. Clim. Dyn., 16, 799-814. LEVIS, S., & FOLEY, J. A. (1999). Potential high-latitude vegetation feedbacks on CO2-induced climate change. Geophys. Res. Lett., 26, 747-750. LEVIS, S., FOLEY, J. A., & POLLARD, D. (1999). CO2, climate, and vegetation feedbacks at the Last Glacial Maximum. J. Geophys. Res., 104, 31,191-31,198. LEVIS, S., FOLEY, J. A., BROVKIN, V., & POLLARD, D. (1999). On the stability of the high-latitude climate-vegetation system in a coupled atmosphere-biosphere model. Global Ecol. Biogeog., 8, 489-500. LEVIS, S., FOLEY, J., & POLLARD, D. (2000). Large-scale vegetation feedbacks on a doubled CO2 climate. J. Clim., 13, 1313-1325. LORENZ, S., GRIEGER, B., HELBIG, P., & HERTERICH, K. (1996). Investigating the sensitivity of the Atmospheric General Circulation Model ECHAM 3 to paleoclimatic boundary conditions. Geol. Rundsch., 85, 513-524. MATTHEWS, H. D., WEAVER, A. J., & MEISSNER, K. J. (2005). Terrestrial carbon cycle dynamics under recent and future climate change. J. Clim., 18, 1609-1628.

REFERENCES

80

MATTHEWS, H. D., WEAVER, A. J., MEISSNER, K. J., GILLETT, N. P., & EBY, M. (2004). Natural and anthropogenic climate change: incorporating historical land cover change, vegetation dynamics and the global carbon cycle. Clim. Dyn., 22, 461-479. MEIR, P., COX, P., & GRACE, J. (2006). The influence of terrestrial ecosystems on climate. Trends Ecol. Evol., 21, 254-260. MEISSNER, K. J., WEAVER, A. J., MATTHEWS, H. D., & COX, P. M. (2003). The role of land surface dynamics in glacial inception: a study with the UVic Earth System Model. Clim. Dyn., 21, 515-537. MYSAK, L. A. (2008). Glacial inceptions: Past and future. Atmosphere-Ocean, 46, 317-341. NOBRE, C. A., SELLERS, P. J., & SHUKLA, J. (1991). Amazonian deforestation and regional climate change. J. Clim., 4, 957-988. NOTARO, M., & LIU, Z. (2008). Statistical and dynamical assessment of vegetation feedbacks on climate over the boreal forest. Clim. Dyn., 31, 691-712. O'HALLORAN, T. L., LAW, B. E., GOULDEN, M. L., WANG, Z., BARR, J. G., SCHAAF, C.,
ET AL.

(2012). Radiative forcing of natural forest disturbances. Glob. Chang. Biol., 18, 555-565. OTTERMAN, J., CHOU, M.-D., & ARKING, A. (1984). Effects of nontropical forest cover on climate. J. Clim. Appl. Meteorol., 23, 762-767. OTTO, J., RADDATZ, T., & CLAUSSEN, M. (2011). Strength of forest-albedo feedback in midHolocene climate simulations. Clim. Past, 7, 1027-1039. OTTO, J., RADDATZ, T., CLAUSSEN, M., BROVKIN, V., & GAYLER, V. (2009). Separation of atmosphere-ocean-vegetation feedbacks and synergies for mid-Holocene climate. Geophys. Res. Lett., 36, L09701. OWEN-SMITH, N. (1987). Pleistocene extinctions: the pivotal role of megaherbivores. Paleobiol., 13, 351-362. OWEN-SMITH, R. N. (1988). Megaherbivores: The Influence of Very Large Body Size on Ecology. London: Cambridge University Press. PACANOWSKI, R. C. (1995). MOM 2 documentation, user's guide and reference manual. Princeton: GFDL Ocean Group Technical Report, NOAA, GFDL. 232 pp.

REFERENCES

81

PELTIER, W. R. (2004). Global Glacial Isostasy and the surface of the ice-age Earth: The ICE-5G (VM2) Model and GRACE. Annu. Rev. Earth Planet. Sci., 32, 111-149. PIELKE, R. A., & VIDALE, P. L. (1995). The boreal forest and the polar front. J. Geophys. Res., 100, 25,755-25,758. PIELKE, R. A., AVISSAR, R., RAUPACH, M., DOLMAN, A. J., ZENG, X., & DENNING, A. S. (1998). Interactions between the atmosphere and terrestrial ecosystems: influence on weather and climate. Glob. Chang. Biol., 4, 461-475. POTTER, G. L., ELLSAESSER, H. W., MACCRACKEN, M. C., & LUTHER, F. M. (1975). Possible climatic impact of tropical deforestation. Nature, 258, 697-698. PRENTICE, I. C., BONDEAU, A., CRAMER, W., HARRISON, S. P., HICKLER, T., LUCHT, W.,
ET AL.

(2007). Dynamic global vegetation modeling: Quantifying terrestrial ecosystem responses to large-scale environmental change. In Global Change: The IGBP Series (pp. 175-192). Berlin: Springer. PRENTICE, I. C., CRAMER, W., HARRISON, S. P., LEEMANS, R., MONSERUD, R. A., & SOLOMON, A. M. (1992). A global biome model based on plant physiology and dominance, soil properties and climate. J. Biogeogr., 19, 117-134. PRENTICE, I. C., SYKES, M. T., & CRAMER, W. (1991). The possible dynamic response of northern forests to global warming. Glob. Ecol. Biogeogr. Lett., 1, 129-135. RIVERS, A. R., & LYNCH, A. H. (2004). On the influence of land cover on early Holocene climate in northern latitudes. J. Geophys. Res., 109, D21114. RUDDIMAN, W. F. (2003). The anthropogenic greenhouse era began thousands of years ago. Clim. Change, 61, 261-293. SCHMITTNER, A., OSCHLIES, A., MATTHEWS, H. D., & GALBRAITH, E. D. (2008). Future changes in climate, ocean circulation, ecosystems, and biogeochemical cycling simulated for a business-as-usual CO2 emission scenario until year 4000 AD. Glob. Biogeochem. Cycles, 22, GB1013. SCHNECK, R., MICHEELS, A., & MOSBRUGGER, V. (2012). Climate impact of high northern vegetation: Late Miocene and present. Int. J. Earth Sci. (Geol Rundsch), 101, 323-338.

REFERENCES

82

SELLERS, P. J., MINTZ, Y., SUD, Y. C., & DALCHER, A. (1986). A Simple Biosphere model (SiB) for use within general circulation models. J. Atmos. Sci., 43, 505-531. SEMTNER, JR., A. J. (1976). A model for the thermodynamic growth of sea ice in numerical investigations of climate. J. Phys. Ocean., 6, 379-389. SHUKLA, J., NOBRE, C., & SELLERS, P. J. (1990). Amazon deforestation and climate change. Science, 247, 1322-1325. SITCH, S., SMITH, B., PRENTICE, I. C., ARNETH, A., BONDEAU, A., CRAMER, W., LPJ dynamic global vegetation model. Glob. Change Biol., 9, 161-185. SMITH, B., PRENTICE, I. C., & SYKES, M. T. (2001). Representation of vegetation dynamics in the modelling of terrestrial ecosystems: comparing two contrasting approaches within European climate space. Glo. Ecol. Biogeogr., 10, 621-637. SMITH, T. M., SHUGART, H. H., & WOODWARD, F. I. (1997). Plant Functional Types: their relevance to ecosystem properties and global change (5th ed.). Cambridge, UK: Cambridge University Press. SNYDER, P. K., DELIRE, C., & FOLEY, J. A. (2004). Evaluating the influence of different vegetation biomes on the global climate. Clim. Dyn., 23, 279-302. SPRACKLEN, D. V., BONN, B., & CARSLAW, K. S. (2008). Boreal forests, aerosols and the impacts on clouds and climate. Phil. Trans. R. Soc. A., 366, 4613-4626. SUAREZ, M. J., & TAKACS, L. L. (1986). Global 5'5' depth elevation. Boulder CO 80303: Technical report, National Geophysical Data Centre, NOAA, U.S. Dept. of Commerce, Code E/GC3. SWANN, A. L., FUNG, I. Y., & CHIANG, J. C. (2012). Mid-latitude afforestation shifts general circulation and tropical precipitation. Proc. Natl. Acad. Sci. USA, 109, 712-716. SWANN, A. L., FUNG, I. Y., LEVIS, S., BONAN, G. B., & DONEY, S. C. (2010). Changes in Arctic vegetation amplify high-latitude warming through the greenhouse effect. Proc. Natl. Acad. Sci. USA, 107, 1295-1300.
ET AL.

(2003).

Evaluation of ecosystem dynamics, plant geography and terrestrial carbon cycling in the

REFERENCES TEMPO
AUTHORS.

83 (1996). Potential role of vegetation feedback in the climate sensitivity of

high-latitude regions: A case study at 6000 years B.P. Glob. Biogeochem. Cycles, 10, 727-736. TEXIER, D., DE NOBLET, N., HARRISON, S. P., HAXELTINE, A., JOLLY, D., JOUSSAUME, S., ET AL. (1997). Quantifying the role of biosphere-atmosphere feedbacks in climate change: coupled model simulations for 6000 years BP and comparison with palaeodata for northern Eurasia and northern Africa. Clim. Dyn., 13, 865-882. THOMAS, C. D., CAMERON, A., GREEN, R. E., BAKKENES, M., BEAUMONT, L. J., COLLINGHAM, Y. C., ET AL. (2004). Extinction risk from climate change. Nature, 427, 145-148. THOMAS, G., & ROWNTREE, P. R. (1992). The boreal forests and climate. Q. J. R. Meteorol. Soc., 118, 469-497. THOMPSON, S. L., & WARREN, S. G. (1982). Parameterization of outgoing infrared radiation derived from detailed radiative calculations. J. Atmos. Sci., 39, 2667-2680. THORNDIKE, A. S., ROTHROCK, D. A., MAYKUT, G. A., & COLONY, R. (1975). The thickness distribution of sea ice. J. Geophys. Res., 80, 4501-4513. WANG, Y., MYSAK, L. A., WANG, Z., & BROVKIN, V. (2005). The greening of the McGill Paleoclimate Model. Part II: Simulation of Holocene millenial-scale natural climate changes. Clim. Dyn., 24, 481-496. WANG, Z., & MYSAK, L. A. (2000). A simple coupled atmosphere-ocean-sea ice-land surface model for climate and paleoclimate studies. J. Clim., 13, 1150-1172. WEAVER, A. J., EBY, M., WIEBE, E. C., BITZ, C. M., DUFFY, P. B., EWEN, T. L., ET AL. (2001). The UVic Earth System Climate Model: model description, climatology, and applications to past, present and future climates. Atmos. Ocean, 39, 361-428. WEBB, S. D. (1989). Faunal dynamics of Pleistocene mammals. Ann. Rev. Earth Planet. Sci., 17, 413-438. WILSON, M. F., & HENDERSON-SELLERS, A. (1985). A global archive of land cover and soils data for use in general circulation climate models. J. Climatol., 5, 119-143.

REFERENCES

84

WILSON, M. F., HENDERSON-SELLERS, A., DICKINSON, R. E., & KENNEDY, P. J. (1987). Investigation of the sensitivity of the land-surface parameterization of the NCAR Community Climate Model in regions of tundra vegetation. J. Climatol., 7, 319-343. WOHLFAHRT, J., HARRISON, S. P., & BRACONNOT, P. (2004). Synergistic feedbacks between ocean and vegetation on mid- and high-latitude climates during the mid-Holocene. Clim. Dyn., 22, 223-238. WOODWARD, F. I. (1987). Climate and plant distribution. Cambridge, UK: Cambridge University Press. WROE, S., FIELD, J., FULLAGAR, R., & JERMIN, L. S. (2004). Megafaunal extinction in the late Quaternary and the global overkill hypothesis. In Alcheringa: An Australasian Journal of Palaeontology (pp. 291-331). London: Taylor & Francis. ZHANG, J., & WALSH, J. E. (2006). Thermodynamic and hydrological impacts of increasing greenness in northern high latitudes. J. Hydrometeorol., 7, 1147-1163.

Vous aimerez peut-être aussi