Hydrologicat Sciences~Journal-des Sciences

Hydrologiques, 47{ i ) February 2002 \ 9

Simulating the impacts of land-use and climate

change on water resource availability for a large

south Indian catchment

J. WILK

Department of Water and Environmental Studies,

Linkdping University. S-581 83 Linkdping,

Sweden

iulwi@tema.liii.se

D. A. HUGHES

Institute for Water Research, Rhodes University. PO

Box 94. Grahamstown, 6140 South Africa

Abstract A monthly rainfall-runoff model was

calibrated for a large tropical
catchment in southern India. Various land-use and
climatic change scenarios were

tested to assess their effects on mean annual runoff

and assured water yield at the

Bhavanisagar Reservoir in Tamil Nadu, India. The

largest increase in runoff (19%)

came from converting forest and savanna (the

indigenous control scenario) to

agriculture. Mean annual runoff decreased by 35%

after conversion to commercial

forest and 6% after partial conversion to tea

plantations. The predicted climate

scenarios of reduced dry season rainfall decreased

the annual runoff by 5% while

enhanced annual rainfall caused a 17% increase in

runoff. Even if land-use and

climate changes had relatively large effects on

runoff, the changes in reservoir yield
which can be assured every year, were often less
severe. This was probably due to the

buffering effect of the reservoir and variation in the

mean annual runoff.

Key words land-use change; climate change;

hydrology; water resources; runoff; reservoir

yield; India

Simulation des impacts des changements de climat

et d'occupation

du sol sur la disponibilité de la ressource en eau

dans un grand

bassin versant du sud de l'Inde

Résumé Un modèle pluie-débit mensuel a été calé

pour un grand bassin versant

tropical du sud de l'Inde. Différents scénarios de

changement du climat et de

l'occupation du sol ont été testés pour estimer leurs

effets sur l'écoulement annuel
moyen et sur le bilan du barrage Bhavanisagar, au
Tamil Nadu, en Inde. La plus

grande augmentation d'écoulement (19%) est

obtenue dans le cas de la conversion de

la forêt et de la savane (le scénario de contrôle

indigène) en agriculture. L'écoulement

annuel moyen décroît de 35% après conversion en

forêts plantées, et de 6% après

conversion partielle en plantations de thé. Le

scénario climatique de réduction de la

pluviosité en saison sèche réduit l'écoulement de

5%, tandis que celui de l'augmentation

de la pluviosité annuelle engendre une augmentation

de l'écoulement de 17%.

Même si les changements de climat et d'occupation

du sol ont des effets relativement

importants sur l'écoulement, les changements dans

la disponibilité annuelle du barrage
sont souvent moins sévères. Cela est probablement
dû à l'effet d'amortissement du

stockage dans le réservoir.

Mots clefs changement d'occupation du sol;

changement climatique; hydrologie;

ressources en eau; écoulement; bilan de barrage;

Inde

INTRODUCTION

Experimental studies have generally been in

agreement that a removal of forest cover

is associated with increased annual streamflow in

both temperate (Bosch & Hewlett,

Open for discussion until 1 August 2002

20 J. Wilk & D. A. Hughes

1982) and tropical environments (Bruijnzeel, 1990).

However, most studies were

performed on small experimental catchments of a

few square kilometres, and it is not
certain if similar results, such as streamflow
increases of more than 100% (ibid), will

emerge from larger river basins. Even on these small

catchments, the large streamflow

increases were mitigated and sometimes

disappeared altogether when alternative

vegetation had grown in. Qian (1983) could not

detect a change in streamflow on

catchments of 7-727 km" in southern China after a

30% loss of tall forest. Similar

results emerged from a catchment of 36 000 knr in

northern Thailand after a removal

of 50%) forest cover (Dyhr-Nielsen, 1986). No clear

trend of changes in évapotranspiration

or runoff could be found in a study of Swedish

catchments between 1400

and 25 000 km", which concluded that any possible

changes are overshadowed by the
uncertainty from insufficient input data and the
simplification of nature in the model

(Brandt, 1992). Results from a study in northeast

Thailand found no significant

changes in water balance after a decline in forest

cover from 80 to 27%. (Wilk &

Andersson, 2001). A large river basin is a mosaic of

different land uses and practices

with varying vegetative patterns in all stages of

growth, and with heterogeneous

geology, soils, and climate. It is apparent that this

spatial and temporal variability is

able to mask many of the cause-and-effect links

between land use and hydrology that

have been demonstrated in small experimental

catchments.

In a large catchment, it is not always easy to pick

out clear changes when
analysing time series data of land use derived from
different measuring authorities.

Many changes may be taking place, some more

relevant from a water resource point of

view, but these may not always be captured with the

available land-use data. The

quality and density of a certain vegetation type and

the methods used to classify land

use into groups are equally important as quantifying

the areal coverage of each group.

The degradation or thinning of natural forest can

affect interception and transpiration

rates even if the area of coverage is unaltered.

The aim of this paper is to assess the effects of land-

use changes on water yield in

the Upper Bhavani basin, India (4100 km"). The

Pitman monthly rainfall-runoff model

was calibrated for 21 years over the Upper Bhavani

River basin in Tamil Nadu, and is
described in the companion paper (Wilk & Hughes,
2002). The results of the

calibration study indicated that the simulated runoff

volumes were 21% higher than

observed, and part of this oversimulation is thought

to be largely due to

unrepresentative estimations of areal rainfall in part

of the catchment. Nevertheless, the

parameter values were considered representative

enough to be used to assess relative

differences between various land-use scenarios.

Parameters in the model that control

processes such as infiltration losses, interception

and évapotranspiration, were altered

to reflect chosen land-use patterns.

Changes to the flow regime, apart from the mean

annual runoff, may be important

but more difficult to detect in the runoff time series.

The use of a reservoir simulation
model (Hughes, 1992) to assess changes in the
assured yield of a catchment represents

an approach that can integrate a variety of changes

in the inflow time series into a

single impact on the yield. From a water resource

perspective, changes to the downstream

yield are also the most important effects of land-use

change. The scenarios in

this study represent the combined effect of applying

what is known about land-use

impacts on small catchment studies to a large

catchment. Hypothetical "worst case"

scenarios have been put into a wider perspective by

comparing them with more

Simulating the impacts of land-use and climate

change on water resource availability 21

heterogeneous land classifications from 1965 and

1994 and two predicted climate

change scenarios.
EFFECTS OF LAND USE ON HYDROLOGY

Forests generally influence local hydrology through

high transpiration, good infiltration

due to a rich litter and humus layer and the presence

of macropores, and high

interception because of their high canopy coverage

and tiered understory. However,

different climates influence the processes by which

trees affect the hydrological cycle

in different manners. The main regulating factors in

the wet tropical climates are

radiation, tree physiology and rain drop size (Calder,

1998). In the wet tropics, humid

rainforest is able to convert nearly all net radiation

into évapotranspiration, so that it is

unlikely that any other land use will have higher

évapotranspiration rates (ibid). In

South Africa, however, natural forests have been

found to be less efficient evaporators
than commercial forests, possibly because of the
better layering allowing better retention

of moisture in and under the forest canopy.

Transpiration rates from eucalyptus

species are likely to be similar to those from other

trees, except where species of

eucalyptus do not exercise stomatal regulation and

have access to freely available

water (Calder, 1992). In these situations, eucalyptus

can transpire at much higher rates.

In the wet tropics, evaporation from intercepted rain

makes up a significant

amount of annual forest évapotranspiration.

Interception loss for tea is lower than from

indigenous forest cover, because tea has an

aerodynamically smoother canopy

(Blackie, 1979a). Commercial forests, with their

smooth single layered canopies, also
have lower interception rates, though rates vary
depending on the spacing between

plants. Wet tropical forests in the Upper Bhavani

basin were found to intercept 38-

40% of rainfall, while the exotic species, acacia

intercepted 30-35% and eucalyptus

only 20-25% (Chinnamani & Sakthivadivel, 1981).

Another study, from the Nilgris

Plateau area, found that indigenous forest

intercepted 34% of rainfall, acacia 25% and

eucalyptus 22% (Samraj etal., 1982).

It is not only évapotranspiration rates that regulate

the amount of water loss;

moisture availability also plays a major role.

Although moisture availability is more of

a limiting factor in the dry tropics, it will still have

some effect in the study area during

the dry season. Trees have much deeper roots than

cultivated crops and grass, and,
therefore, can extract water from deeper layers even
when the topsoil is dry. Finger

millet (an annual crop in Karnataka, India) has a

rooting depth of about 2 m, while

forest sites had much deeper roots. One of the

eucalyptus sites had roots extending to a

depth of 8 m (Calder et al., 1997). In comparison,

anchor roots of tea plants can reach

depths of 2.5-3 m, if there is adequate soil depth

(Tea Research Institute, India,

personal communication), while they matched the

depth of montane rain forest at 6 m

in Kenya (Kerfoot, 1961, 1962).

The overall impact that various land uses have on

hydrology depends on the combined

effect of differences in interception, transpiration

and infiltration capacities. The

evidence from both model simulation and process

studies in Kenya suggests that total
water use by tea may be lower than under forest for
wet conditions, and similar, or

even higher, under dry conditions (Blackie, 1979b).

The balance between interception

and transpiration processes will depend on spatial

and temporal rainfall distribution.

When water is freely available, forests transpire high

amounts and intercepted water is

22 J. Wilk & D. A. Hughes

lost almost daily, while during dry periods,

interception plays a less dominant role in

the water balance.

THE UPPER BHAVANI BASIN

A description and map of the basin is given in the

companion paper (Wilk & Hughes,

2002).

LAND-USE ANALYSIS
To determine the patterns of land-use change in the
catchment over the past few

decades, aerial photographs and vegetation maps

were analysed with the aim of

quantifying the change in amount of forest cover

from 1965 to 1994. Over such a large

catchment, however, the term "forest" encompasses

many different vegetation types

that may have very different water uses. Even within

a defined vegetation group,

different densities and qualities (healthy vs

degraded) exist. Vegetation classes were

grouped according to their hydrological influences

(such as savanna, natural and

commercial forest and open land, including both

agricultural crops and grassland) as

far as this was possible with the available sources.

The classes included in open land
and indigenous forest categories are listed in Table
1. The sources on which the classes

are based are:

- 1965 aerial photographs, obtained from the US

Geological Survey;

- vegetation classifications based on 1973 Survey of

India topographical maps,

prepared by the French Institute, Pondicherry, India;

- a 1986 vegetation map prepared by Université Paul

Sabatier, Toulouse, France,

based on Landsat imageries between 1973-1979; and

- a 1995 vegetation map prepared by the French

Institute, Pondicherry, India, based

on a 1994 Survey of India map.

The vegetation analysis from 1965 indicates that

58% of the basin was covered by

indigenous forest, while in 1994 this figure is 74%,

obtained by combining savanna,
natural and commercial forest and tea. Although the
results are based on different

classifications and sources, making the actual

numbers uncertain, it does tend to

downplay the possibility of forest decline within this

period. The apparent decline in

open land (from 42% to 25%) may be a result of

placement of savanna in the classifications

(Fig. 1). Savanna ranges from open to more dense

tree and shrub savanna,

making it hard to distinguish as either open land or

forest.

Based on the land-use analysis, it was concluded

that the resolution of the

available data, compared to the nature of the

changes, precludes a clear definition of

land-use change. The changes have not been large

enough to overcome the potential
errors caused by classification and different
methods of interpretation. Hypothetical

scenarios have been created to assess the effects of

extreme land-use changes, using

model parameters based on an understanding of the

effects noted in small catchments.

Despite the uncertainties in the land-use

classifications, scenarios of land use as

defined by the 1965 and 1994 surveys have been

included to represent realistic

heterogeneous land-use patterns.

Simulating the impacts of land-use and climate

change on water resource availability 23

Table 1 Description of the vegetation classes from

different sources divided into open land and forest.

Year of

publication /

Scale
1973

1:100 000

1986

1:250 000

1995

1:100 000

Source

1973 Survey of

India

topographical

maps

Landsat

imageries from

1973-1979

1994 Survey of

India
topographical

Open land

Open scrub

Grass

Mosaic of degraded vegetation

and cultivated land, crops and

plantations

Various crops

Cultivated land

Irrigated crops

(sugar cane, paddy)

Predominating dry crops

(millet, groundnut)

Grassland

Irrigated agriculture

Dry land agriculture

Forest

Reserved forest

Unreserved forest

Dense mixed jungle

Fairly dense mixed jungle

Dense scrub

Fairly dense scrub

Open jungle

Wooded area

Wet evergreen forest

Moist semi-deciduous forest

Semi-deciduous or deciduous forest

Degraded moist forest

Low evergreen forest

Re-afforested areas

Thorny thicket
Shola forest

Dense evergreen forest

Semi-evergreen forest

maps Secondary moist deciduous woodland

Dense thicket

Discontinuous thicket to low

scattered shrubs

Woodland to savanna woodland

Scrub woodland

60

50

40

30

20

10

0
S 1965

D 1973

D 1973-1979

D1994

i r OH

Open Forest Tea

Fig. 1 Land-use cover (%) from 1965 to 1994.

Plantations Savanna

THE PITMAN MODEL AND LAND-USE CHANGE

The details of the semi-distributed Pitman model are

provided in the companion paper

(Wilk & Hughes, 2002), but additional background

information on the way in which

the model can deal with land-use changes is

provided here. The main parameters of

24 J. Wilk & D. A. Hughes

concern in a land-use context are those that control
interception (PI), évapotranspiration

(R and FF) and infiltration or catchment absorption

(AI, ZMIN, ZA VE and

ZMAX). The original version of the model was

designed to simulate internal subcatchment

differences between forested and non-forested

areas. This was accomplished

by specifying a %Forest parameter, two PI

parameters (PIF and PIV), as well as

a FF parameter to allow the input potential

évapotranspiration to be scaled (up or

down) for the land use over the "forested" area. The
FF parameter essentially allows

the model to cater for more effective drying of the

soil moisture store by deeper

rooting vegetation. Although the model refers to

forest, the parameters can actually
apply to any two groups of land use; for example, tea
and natural grassland. The main

issue is that realistic values for PIV, PIF and FF need

to be established to represent the

effects of the two land uses. It is equally important

that the other relevant parameters

(R, AI, ZMIN, ZA VE and ZMAX) are set to values that

are realistic to reflect conditions

over the sub-catchment as a whole. The model also

allows for seasonal changes to the

values of some parameters (PIF, PIV ana ZMIN). The

basin has been divided into four

projects, which have been further divided into sub-

catchments as illustrated in Fig. 3 of

the companion paper (Wilk & Hughes, 2002) for

which the various parameters have

been applied.

Table 2 illustrates the sensitivity of the model to

some changes in the land-use
parameter values using an arbitrarily selected sub-
catchment (Project 2, sub-catchment

1) within the study area. Some of the input values

and the base parameter values

are given at the bottom of the table and the effects

on mean annual interception, actual

évapotranspiration, surface runoff and soil moisture

runoff (all in mm and as %

changes from the results using the base parameter

values) are tabulated. For each row

of the table the model has been run with the base
parameter values or the changes

listed in column 1. Changes to the interception

parameter values have the largest

impact on surface runoff, but do not influence the

water balance of the soil moisture

store as significantly. Over the full range of possible

values (0-1), R can have a 6%
effect on actual evaporation, which translates into a
9% (from -5 to +4) possible

impact on soil moisture runoff. Parameter ZMIN is

clearly quite sensitive in this

region, which experiences some very high monthly

rainfalls during both of the

Table 2 Parameter sensitivity analysis depicting

changes to various model results for sub-catchment
2.1.

Parameter chang

PIV'=0.5, PIF==

PIV= l.0,PIF =

PIV'=2.0, PIF ==

i? = 0

R=].0

FF = 1.0

FF= 1.5

ZMIN =350
ZMIN =600

;es

= 2.0

= 3.5

= 6.0

Interception

(mm)

120.7 (-

229.9 (-

-58%)

-19%)

423.9 (+48%)

285.5

285.5

285.5

285.5
285.5

285.5

Actual

évapotranspiration

(mm)

917.3 (-4%)

944.8 (-1%)

992.0 (+4%)

988.0 (+3%)

933.0 (-3%)

862.2 (-10%)

979.8 (+2%)

955.5

957.6

Surface

runoff
(mm)

60.9 (+25%)

52.3 (+8%)

40.4 (-17%)

48.5

48.5

48.5

48.5

77.9 (+61%)

38.0 (-22%)

Soil moisture

runoff

(mm)

651.1 (+4%)

635.2 (+1%)

605.2 (-4%)
597.6 (-5%)

652.0 (+4%)

726.5 (+16%)

605.5 (-4%)

600.5 (-4%)

638.2 (+2%)

Input data: area = 367 kirf ; annual potential

évapotranspiration = 1100 mm;

mean annual rainfall = 1638 mm.

Base parameters: %Forest = 50.0; AI= 0.0, ZMIN= 500

mm, ZMAX= 700 mm;

PIV = 1.5 mm, PIF = 4.0 mm; R == 0.6, FF = 1.4, 51 =

0.0 mm, ST = 500 mm, FT = 70 mm.

Simulating the impacts of land-use and climate

change on water resource availability 25

monsoon seasons. Parameter FF also has the

potential to cause quite large changes in
results where the %Forest parameter represents a
large proportion of the catchment.

The streamflow output from the Pitman model can be

used as input to a water

balance simulation model of the Bhavanisagar

Reservoir to determine the overall

impact of the various land-use scenarios on the yield

of this reservoir. The

Bhavanisagar Reservoir has a capacity of 986 Mm'

and supplies irrigation water to

downstream agricultural users. The annual runoff

volume is 410 mm with peaks

occurring in July and November corresponding with

the monsoons. Year-to-year

variability is large, ranging from 240 to 930 mm

during the 21-year study period,

resulting in a coefficient of variation of 0.39. The

storage volume relative to inflow is
60%. The assured yield of the dam, given various
scenarios of upstream inflow, can be

determined by iteratively changing the draft

requirement until negligible shortfalls in

supply are experienced. This approach is a useful

method of putting the effects of landuse

change into a water resource management context.

LAND-USE AND CLIMATE CHANGE SCENARIOS

The indigenous scenario was considered the control

against which others were

compared. It was predominantly a mixture of wet

and dry deciduous tropical forest

(35%), tree or scrub savanna (59%) in the most

northern and southern parts of the

basin and grassland (6%) on the high plateau. This

was created by transforming all

agriculture, tea and commercial forest areas in the

1973 maps to the nearest indigenous
land classification, based on knowledge of the basin
climate, soils and topography. The

land-use scenarios tested were: an indigenous

control scenario, large-scale conversion

to tea, commercial forest and agriculture and land

cover as in 1965 and 1994.

The most extreme land-use scenarios were created

by giving them maximum

coverage in the sub-catchments, where this was

physically and climatically plausible

Table 3 The area of each sub-catchment and the

percentage of coverage by different land-use classes
for

the hypothetical land-use scenarios.

Project

3
4

Basin

Subcatchment

1.1

2.1

2.2

2.3

2.4

2.5

2.6

3.1

3.2

3.3

3.4

3.5

3.6
4.1

4.2

Area (km2)

37

366

122

264

256

117

356

554

90

107

206

216

180
670

563

4100

Indigenous cover (%):

Forest Savanna Grassland

30

50

100

50

30

40

100

20

20

40

30
30

35

70

60

30

100

50

70

60

80

40

50

50

70

70

59
30

10

20

20

40

30

10

Commercial

forest (%)

100

100

100

100

100

100
100

100

100

100

100

100

100

100

100

100

Agriculture

(%)

100

100

100

100
100

100

100

100

100

100

100

100

100

100

100

100

Tea

(%)

100

40
70

30

50

100

30

50

100

100

50

50

50

50

30

49

26 J. Wilk & D. A. Hughes

(Table 3). Agriculture was given 100% coverage;
although it would not be assumed to

occur in very steep mountain areas, these areas are

very small. Commercial forest was

given 100% basin coverage, as species such as

eucalyptus can be grown in both wet

and more arid environments. Tea was placed on all

high elevation areas with cooler,

humid climate and covered 47% of the basin. The

chosen land-use scenarios are not

considered to be a prediction of future conditions,

but represent the most extreme

effects that changes could cause, if reflected by

suitable parameter values.

Land-use cover in 1965 was based on the division of

land into open land and forest

cover from aerial photographs. Forest covered 42%

of the basin. Open areas that were
classified as savanna in 1973-1979 were considered
as such in 1965 (22%) while the

remainder of the open area was considered to be

agriculture (20%). The area of tea was

considered to be half of that in 1973-1979 (5%).

Commercial forests were considered

to cover approximately 7% of the basin, the same

area as in 1973-1979. Land use for

1994 can be seen in Fig. 2.

To compare the maximal effect on streamflow

generation from land-use changes

with that which can be expected from altered rainfall

generation as predicted from

general circulation models (GCMs), two climate

change scenarios were included in the

study. Results were used from climate scenarios

based on data from four GCMs: the

Canadian Climate Centre model (CCCJ1), the United

Kingdom Met Office model
(UKMOH), the Geophysical Fluid Dynamics model
(GFDLA) and the Australian

CSIR09 model. These are described in Whetton

(1994). For the year 2070, the models

(a)

(b)

50

20

-40

3500

3000

2500

2000

! 1500

1000

500

DMAR
H Assured Yield

à, CV of MAR

&S»!% CdJ I hi D

1° Agr(2%) Agr(1%) Agr 1965 Inc. P 1995 Dec. P Tea

Part

DMAR

Q Assured Yield

0-

__

Agr (2%) Agr(1%) Agr 1965 Inc. P 1995 Dec. P Tea

Plantation Indigenous

Fig. 2 (a) Mean annual runoff (MAR), assured yield

(MmJ) and coefficient of variation (CK) of

MAR (%) for the scenario groups; and (b) change in

MAR and assured yield (%) as compared to

the indigenous scenario.

Simulating the impacts of land-use and climate

change on water resource availability 27
considered suggested a -10 to 10% decrease in
rainfall during the dry season and a

0-10% increase during the wet season. In this study,

two scenarios were created to

represent the most extreme changes (the driest and

the wettest) in seasonal precipitation:

a 10% decrease in dry season rainfall and a 10%

increase in annual rainfall. The

climate change scenarios were run with the

indigenous control parameters while only

changing the precipitation input equally for all

rainfall events.

The range of model parameters chosen for various

land uses is shown in Table 4,

while the range of parameter values for each project

for the different scenarios is

shown in Table 5. The parameter ZMIN (controlling

infiltration) was set to the lowest
value (Table 4) for the agricultural scenario, to
reflect reduced infiltration. The model

was run with 1 and 2% impervious areas (AI

parameter) to represent activities that

compact the land surface and decrease infiltration

capabilities, such as plough pans,

Table 4 The range of model parameter values for

different land-use classes.

ZMIN

(mm month"1)

PIV

(mm)

PIF

(mm)

FF

Indigenous forest

Agriculture
Tea

Commercial forest

Tree savanna

Scrub savanna

600

400

450

550

500

510

1.5

1.5

1.5
4-6

3.5

3.8

3.8

0.2

0.3

0.6

0.4

1.3

1.4

1.5
1

1.1

Table 5 The range of model parameter values used in

each project for the land-use scenarios.

ZMIN

(mm month"1)

PIV

(mm)

PIF

(mm)

FF

Project 1

Project 2

Project _

Project 4

Control (indigenous)
Agriculture AI= 0

Tea

Commercial forest

1965

1995

Control (indigenous)

Agriculture AI = 0

Tea

Commercial forest

1965

1995

Control (indigenous)

Agriculture AI = 0

Tea

Commercial forest

1965
1995

Control (indigenous)

Agriculture AI = 0

Tea

Commercial forest

1965

1995

500

350

350

500

400

400

450-900

300-600

400-850
400-800

350-750

300-700

500-1100

300-700

350-1000

450-1000

500-1100

400-1150

800

500

650-700

700

900

900

1-1.5
0

0.5-1

1-1.5

1.5-2

0.5-1

1-2

1-1.3

1-1.5

0.5-1.5

0-1.5

1.1-2

1-1.2

1-1.5

1.4-2
1-1.9

1-1.6

1-1.3

1-1.5

1.4-1.8

1.4-1.8

4.5-5

1.5-2

3-3.5

4.5-5

5.5-6

2.5-3

3.5-6

1.5-2

3.3-3.7
4.5-5

2.5-5

2.5-5

4-6

1.5-2

3.5-3.6

4.5-5

4.5-6

4-5

3.3-4.8

1.5-2

3.5-3.7

4.5-5

4-5

4.5-5

0.3
1

0.3

0.2

0.7

0.2-0.4

0.3-0.4

0.15-0.7

0.15-0.9

0.2-0.4

0.3-0.45

0.2-0.4
0.15-1.2

0.35-0.4

0.4-0.45

0.3-0.35

0.3-0.35

1.1

1.4

1.5

1.2

1.15

1.1-1.2

1.2-1.4
1.5

1.05-1.25

1.05-1.15

1.1-1.2

1.2-1.4

1.5

1.1-1.5

1-1.2

1.1

1.2

1.5

1.1-1.2

1.1-1.2

28 J. Wilk & D. A. Hughes

road building, increased urbanization and poor land
management. For those scenarios

that include indigenous forest, interception rates

(PIF) were lowered in the dry season,

as the majority of trees are deciduous. Natural forest

was characterized by the highest

interception rate of all scenarios, but was assigned a

lower FF than both tea and commercial

forest, to suggest lower évapotranspiration to be

consistent with more

conservative water use. To account for relative root

depth, R values for natural forest

were set lower than for agriculture, tea and savanna,

yet higher than for commercial

forest.

Tea, being perennial, has the same interception

values for both seasons. To

simulate a large-scale conversion to tea from the

indigenous condition, R and FF were
increased slightly from the indigenous scenario,
while interception was set lower than

both indigenous and commercial forest. Commercial

forests are expected to have

extreme values for both rooting depth and

évapotranspiration, since both acacia and

eucalyptus can readily exploit even limited water

availability by tapping groundwater

reserves at relatively great depths.

RESULTS

The mean annual inflow to the Bhavanisagar

Reservoir under the indigenous scenario

has been simulated as 2060 Mm' and the assured

yield as approximately 1300 Mm3.

The mean annual runoff and assured reservoir yields

under the various land-use covers,

based on a 21-year modelling period, are shown in

Fig. 2 (as absolute values and
relative to the indigenous yield). It should be noted
that an increased mean annual

inflow will not necessarily result in a higher reservoir

yield. The yield is determined by

the combination of storage, long-term inflow volume,

seasonality and long-term

variations in inflow volumes (as measured by the

coefficient of variation, CV, for

example).

The agriculture scenario represented a total

conversion of indigenous cover to

crops such as vegetable, banana, coconut and

spices. This resulted in the largest

increase in MAR, 19% greater than the indigenous

scenario. In terms of assured yield,

this scenario resulted in an increase of 6%.

Increasing the impervious area of the

catchment by 1 and 2%, caused MAR increases of 20

and 21%, respectively, over the
indigenous scenario.

The 1965 scenario showed a modest decrease in

MAR (2%), but a 3% increase in

assured yield. The low variation in the annual runoff

could account for the small

change in yield. Land use in 1994 was a mosaic of

agriculture, savanna, natural and

commercial forest, and tea. Despite the 1994 land

use having extensive agricultural

areas, the MAR for the 1994 scenario was of a

similar order to the indigenous scenario.

This could be because the indigenous scenario had

large areas of savanna (59%). Both

the 1965 and 1994 scenarios represent a more

realistic picture of large catchments,

with mixed land use, and the changes from the

indigenous scenario are much more

modest than the extreme conversions to agriculture

and commercial forest. This
supports findings from large heterogeneous
catchment studies, where little or no

changes in annual runoff can be detected after

removal of large tracts of forests. The

tea scenario, covering 47% of the basin, had a

slightly reduced MAR and reservoir

yield.

Simulating the impacts of land-use and climate

change on water resource availability 29

3=

-J

QC

Zi

c:

<
o

5?

18

15

12

3I

0'-—

Jan March May July Sept Nov

Fig. 3 Seasonal distribution of runoff for the chosen

land-use scenarios.

Commercial forest over the entire basin resulted, not

surprisingly, in the lowest

MAR and yield. Despite a 33% reduction in MAR, the

reduction in yield was only 21%,
partly due to the reservoir's storage capacity, which
acts as a buffer between inflow

and water use. The seasonal distributions of runoff

show a similar pattern for almost

all land-use groups except that of commercial forest

(Fig. 3). During the drier periods,

commercial forest species have a greater ability to

extract water than other vegetation

types, thus a smaller percentage of the annual runoff

occurs during these periods as

compared to other vegetation groups. However,

these drier seasons are not critical

periods of reservoir recharge, so reduced runoff is

not equally reflected in the assured

yield.

The climate change scenarios showed modest

changes in assured reservoir yield.

The increased annual precipitation scenario

generated a similar MAR ( 17% increase) to
that of converting the entire basin to agriculture
(19% increase). However, the annual

CV of inflows was higher (38% as compared to 35%

for the agriculture scenario),

resulting in a lower assured yield. Decreasing the dry

season rainfall by 10% decreases

MAR by 5% and has a marginal effect on the yield.

CONCLUSION

Only moderate land-use changes have been

experienced in the catchment of the Upper

Bhavani basin over the period for which data are

available. Although there has been

little overall change in terms of area, there have

been large local changes, as tea,

commercial forests and agricultural crops have been

established and then replaced by

alternative vegetation types on different tracts of

land. However, there is little indication
that the area of indigenous forest has decreased in
the last few decades.

Of the scenarios, total conversion to agriculture in

the basin generated the largest

assured reservoir yield as compared to the

indigenous control scenario, while the

greatest reduction came from total conversion to

commercial forest. The other

scenarios, including those related to climate change,

showed relatively negligible yield

changes. The changes for the agricultural and

commercial forest scenarios are quite

substantial, indicating that large-scale conversion to

different land uses will affect the

- Forest

- Plantation

Tea

Agriculture
//

//

-'/

30 J. Wilk & D. A. Hughes

volume of runoff from basins in the region. In terms

of water resource utilization, the

effects can be much reduced where reservoir

storage is available to buffer the changes.

However, these are extreme changes where the

entire basin has been converted to one

land-use type—a very unlikely transition. The

scenarios based on mapped 1965 and

1994 land-use distributions, which are more realistic

mixtures of different land uses,

indicated relatively unchanged runoff as compared

to the control scenario. This

suggests that the mosaic of land-use changes that

has occurred in the recent past is
unlikely to be detectable in runoff records, even if it
were very accurate and accounted

for all artificial transfers. Perhaps the most

important implication of this study is that

changes that are likely to occur (rather than the

extreme scenarios referred to above)

will have a negligible impact on the yield of the

Bhavanisagar Reservoir.

Acknowledgements Financing for the project was

generously provided by SAREC

(Swedish Agency for Research Co-operation with

developing countries). The Swedish

Institute also provided funding to facilitate research

co-operation between Linkôping

and Rhodes Universities.

REFERENCES

Blackie, J. R. ( 1979a) The water balance of the

Kericho catchments. East Africa Agrie. For. J. 43,
Special Issue, 55-84.
Blackie, J. R. ( 1979b) Summary of results of the
Kericho experiments. East Africa Agrie. For- ./. 43,
Special Issue, 139-

141.

Bosch, J. M. & Hewlett, .1. D. (1982) A review of

catchment experiments to determine the effect of
vegetation changes on

water yield and évapotranspiration. J.Hydrol. 55,

323.

Brandt, M. (1992) Skogens inverkan pâ

vattenbalansen (The impact of forests on the water
balance, in Swedish). Report

no. 37, Swedish Meteorological and 1 lydrological

Institute, Norrkôping, Sweden.

Bruijnzeel, L. A. (1990) Hydrology of moist tropical

forests and effects of conversion: a state of
knowledge review.

Faculty of Ivarth Sciences, Free I iniversity,

Amsterdam, The Netherlands.
Calder, I. R. (1998) Water use by forests, limits and
controls. Tree Physiol. 18, 625-631.

Calder, 1. R. (1992) Water use of eucalvpts—a

review. In: Growth and Water Vse of Forest
Plantations (ed. by

I. R. Calder, R. !.. Hail & P. G. Adlard). John Wiley &

Sons Fid, Chichester, West Sussex, UK.

Calder, F R., Rosier, P. T. W., Prasanna, K. T. &

Parameswarappa, S. (1997) Eucalyptus water use
greater than rainfall

input—a possible explanation from southern India.

Hydro!. Earth System Sci. 1, 249-256.

Chinnamani, S. & Sakthivadivel, R. (1981) An

integrated study of the hydrology of the Bhavani
basin. College of

Engineering, Anna University of Technology, Madras,

India.

Dyhr-Nielsen, M. (1986) 1 lydrological effect of

deforestation in the Chao Phraya basin in Thailand.
Paper presented at the
international Symposium on tropical Forest
Hydrology and Application (Chiang Mai. Thailand, 1 1
-14 June, 1986).

Hughes, D. A. (3992) A monthly time step, multiple

reservoir water balance simulation model. H'ater SA,
18(4), 279-286.

Kerfoot. O. ( 1961 ) lea root systems. In: Proc.

Seventh Conf., Tea Research Institute of ERA,
Pamphlet no. 19.

Kerfoot, (). ( 1962) Roots systems of forest trees,

shade trees and tea bushes. East Africa Agrie. For.
J. 27, Special Issue.

Qian, W. (1983) Effects of deforestation on Hood

characteristics with particular reference to Hainan
Island. China. In

Hydrology of Humid Tropical Regions (ed. bv R.

Keller) (Proc. Hambure Svmp., August 1983). 249-
257. IAHS

Publ.no''. 140.
Samraj, P.. llaldorai. B. & Henry, C. (1982)
Conservation forestry. In: 25 Years Research on Soil
and Water Conservation

in Southern Hilly High Rainfall Regions. Monograph

no. 4. Central Soil and Water Conservation Research
and

Training Institute, Dehra Dun. India. 153-199.

Whetton, P. ( 1994) Constructing climate scenarios:

the practice. In: Climate Impact Assesstnent
Methods for Asia and the

Pacific (ed. by A. J. jakeman & A. B. Pittock) (Proc.

Regional Symp. organized by ANUTECT I Pty Ftd on
behalf of

the Australian International Development Assistance

Bureau. 10-12 March 1993, Canberra, Australia), 21-
27.

Wilk, J. & Andersson, F. (2001) Hvdrologieal impacts

of forest conversion to agriculture in a larae river
basin in Northeast

Thailand. Hydrol. Processes "15, 2729-2748.

Wilk, .1. & Flushes, D. A. (2002) Calibrating a rainfall-
runoff model catchment with limited data. Hydro!
Sci. J. 47(1).

3-18. Received 10 April 2000; accepted 16 July 2001