Vous êtes sur la page 1sur 13

Atmospheric Research 99 (2011) 277–289

Contents lists available at ScienceDirect

Atmospheric Research
j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / a t m o s

Rainfall distribution in the Andes of southern Ecuador derived from blending


weather radar data and meteorological field observations
Rütger Rollenbeck ⁎, Jörg Bendix
Laboratory for Climatology and Remote Sensing (LCRS), Faculty of Geography, University of Marburg, Deutschhausstr.10, D-35032 Marburg, Germany

a r t i c l e i n f o a b s t r a c t

Article history: The Andes of Ecuador show an extreme heterogeneity of spatial and temporal distribution of
Received 22 September 2009 precipitation. The existing operational network of the national weather service is not capable of
Received in revised form 13 October 2010 reproducing these complex patterns. By using a cost-efficient rain radar and a network of high-
Accepted 16 October 2010
resolution rain gauges, the real complexity of the rainfall distribution and the meteorological
processes of rainfall formation can be assessed. A blending method encompassing geostatistical
Keywords: tools allows to derive a comprehensive rainfall climatology for the study area.
Precipitation
Precipitation is predominantly of the advective type, associated with humid air masses from
Radar
the Amazon basin transported by the tropical easterlies. The typical form is light to heavy
Andes
Calibration drizzle with long duration but lower rain rates.
Climatology However, in contrast to former knowledge there is no single mechanism of rain formation for
any given place. Several processes interact like small and large-scale convective cloud systems,
local and regional valley/mountain breeze systems and terrain-lines of preferred moisture
transport interact on various time scale. This leads to complex patterns of rainfall in space and
time.
Several types of characteristic weather situations are revealed by the study. They are
characterized by specific combinations of local and regional atmospheric processes and
interactions with the topographical configuration. They are modified by mesoscale and
continental circulation patterns as the annual shift of pressure cells, the east Andean low-level
Jet and katabatic flows.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction masses from the Amazon. Due to orographic lifting, most of


the moisture is potentially precipitated on the eastern
The Andes are the major climate divide of South America. escarpment of this mountain range, mainly in the form of
Close to the equator, they separate the humid lowland forests long lasting drizzle.
of the Amazon basin from dry coastal areas in south Ecuador Nevertheless, climate maps at hand do not reflect those
and Peru. In the south of Ecuador the mountain chain is rather higher precipitation amounts, but show a continuous decline
narrow and generally lower than in the adjacent regions to from the Amazon lowland forests to the dry western coast of
the north and south, hence, rather strong climatic gradient Ecuador. They generally rely on ground-based climate
occur in this region. The so called “Nudo de Loja”, close to the observation of the national weather services station network
provincial capital Loja forms the lowest passage for the (e.g. INAHMI, 1993), which is limited to the valley bottoms
prevailing tropical easterly flows, providing humid laden air and leaves out the higher mountain parts that receive the
most of the rainfall. Even worse, the national network has
been strongly reduced in the last two decades, particularly in
⁎ Corresponding author. Tel.: +49 6421 282 5979; fax: + 49 6421 282
8950.
the southeastern Andes of Ecuador. Moreover, current rainfall
E-mail addresses: rollenbeck@lcrs.de (R. Rollenbeck), maps do not explain the high temporal and spatial hetero-
bendix@staff.uni-marburg.de (J. Bendix). geneity of precipitation observed in this part of the Andes

0169-8095/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.atmosres.2010.10.018
278 R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289

because only monthly totals are published. Understanding of distribution and dynamics of rainfall pattern in the Andes of
the rainfall generating processes in this region would be southeast Ecuador is still poor. Most investigations hitherto
important to derive realistic climatic maps and improve the presented are based on limited data sets mainly analyzed to
implementation and validation of regional and global climate unveil rain-bringing processes in the Andes of southeastern
models. With regard to rainfall formation processes, one of Ecuador by synoptic observations or based on time series of
the greatest challenges is to cross-link local point observa- satellite data (e.g. GOES, NOAA), which are not suitable to
tions of rainfall and the modeled rainfall on continent-wide to properly represent the spatio-temporal rainfall distribution in
global scale, which requires a proper understanding of an appropriate quality and spatial resolution.
multiscale processes that modify rainfall formation particu- This deficit of knowledge related to the rainfall formation,
larly in complex terrain. its interaction with the complex terrain of the Andes and thus,
Previous studies about seasonal rainfall distribution in the spatio-temporal rainfall distribution can be overcome by
Andes of southern Ecuador have proven that a clear annual combining scanning instruments with traditional observation
cycle can be observed with the major rainy season occurring data and geostatistical tools. Especially radar is known as a
from May to August and a short but irregular dry season in useful tool at this scale, although it is difficult to operate in
October and November at the eastern slopes. However, the high mountain areas (Andrieu et al., 1997; Fornasiero et al.,
annual cycle varies on rather short horizontal distances: The 2006). Conventional weather radars are rather expensive and
intra-mountain basin of the city of Loja has its rain season demand large investments in infrastructure and operation
peaking in March while August is the driest month. The small costs. Since a few years, alternative low-cost systems like the
town of Catamayo only 30 km to the West already has a full- local area weather radar (LAWR: Jensen, 2002; Rollenbeck
arid climate with only a short rain season in April, mainly and Bendix, 2006) are available and already have been
caused by strong convection (Rollenbeck et al., 2007; Richter, successfully applied at the ecosystem level in several case
2003; Emck, 2007). studies (e.g. Bendix et al., 2009).
On shorter scales field observations showed that the Consequently, the combination of several recording and
spatio-temporal distribution of precipitation totals might totaling rain gauges with radar observations from the LAWR
vary for each valley, most likely depending on its exposure to shall shed light on the major factors determining the daily and
the main atmospheric flow and the neighboring topographic annual variations and the spatial distribution of precipitation
conditions. Current knowledge of topographic modifications in the high mountain area of the Andes of south Ecuador. To
of rainfall patterns in the Andes is mainly based on achieve this goal, a comprehensive data set of radar imagery is
hypothetical considerations like the thermal control of evaluated and new maps of area rainfall are presented.
convective cloud systems (Weischet, 1965) or satellite data Furthermore, the analysis aims at unveiling different weather
inspection (Bendix, 1997; Adler et al., 2000; Bendix, 2000; situations generating specific patterns of rainfall distribution
Bendix et al., 2004; Vuille et al., 2000; Giovannettone and in the high mountains of southern Ecuador.
Barros, 2008). The results will be discussed with regard to static
In the Andes of southeast Ecuador several influential approaches based on time series statistics (Bendix and
factors of spatial and temporal rainfall variability have to be Lauer, 1992; Emck, 2007), interpolation techniques (Buytaert
considered: Micro- and mesoscale katabatic drainage flows et al., 2006) and topographic models of rainfall distribution
(Lopez and Howell, 1967) and their influence on the regarding atmospheric moisture transport (Daly et al., 1994;
development of rain-bearing mesoscale cloud systems Baigorria, 2005; Hughes et al., 2009).
(Bendix et al., 2009; Trachte et al., 2010) and the role of the
east Andean Low level jet (Bendix, 2000). However, most 2. Study area materials and methods
important is the orographic rainout of moist air coming from
the Amazon under the influence of the tropical easterlies. This 2.1. Study area
will most likely cause the exposed mountain slopes to be the
wettest places. Diurnal oscillations of the topographic rainout The study region is located in southern Ecuador at an
are caused by the thermal control of convection in the altitudinal range of 800 to 3600 m asl (Fig. 1). The region has
lowland and local valley-winds in the mountains. On the an extent of 120 by 120 km and comprises the humid eastern
mesoscale, katabatic drainage flows through the larger Andean slopes, which are characterized by tropical rain
draining valleys towards the Amazon lowland seem to induce forests up to ~ 1800 m asl, and tropical evergreen cloud forests
convergence in the eastern foothills. The development of at altitudes above 1800 m asl. The relatively dry basins and
mesoscale cloud systems in the Amazon foreland causes valleys west of the main Cordillera Oriental de los Andes are
characteristic rain events in the study region, which contrib- partly covered by xerotropical vegetation (Richter, 2003). The
ute to unusual diurnal cycles of rainfall (Bendix et al., 2009; radar site is located within the Reserva Biólogica de San
Trachte et al., 2010). Probable interactions with the east Francisco, the central investigation area of a collaborative
Andean low-level jet are still under investigation for the ecological research group. At the research station Estacíon
equatorial part of the eastern Andes. It should be also stressed Científica San Francisco (ECSF, 3°58′ S / 79°04′W) data are
that medium and long-term oscillations of the Pacific sea compiled and archived (Fig. 1, right, refer also to Beck and
surface temperature as e.g. in the scope of the ENSO Müller-Hohenstein, 2001; Beck et al., 2008).
phenomenon can impose alterations of precipitation on the The LAWR radar was installed on the highest accessible
internal time scale (Vuille et al., 2000). mountain peak, about 15 km east of the provincial capital of
Despite the work conducted on the formation of rainfall Loja. The summit of the Cerro de Consuelo is situated at
cited above, the knowledge about exact spatio-temporal 3,180 m asl and provides the least obstructed view in this
R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289 279

Fig. 1. Study region and radar range. Climate stations are represented by a black dot. All following images show the area depicted in the right panel (Coordinates
are in UTM 17 s). The gray dotted lines indicate profiles used in Fig. 9.

very mountainous region. The site is characterized by bleak temperatures. Nevertheless, cloud frequency and cloud density
weather conditions (fog/cloud frequency N70%, continuous derived from these images (Bendix et al., 2004) are a useful
heavy winds) and can only be accessed by a small footpath, indicator of the precipitation patterns. Due to all this limita-
which put high demands on the installation and mainte- tions only a combination of all available data sources allows to
nance. As far as we know, this is the second highest location of understand the complex precipitation patterns in this region.
a rain radar in the world.
2.3. Method
2.2. Data
Basic processing of the radar data and an experimental
The installation of the LAWR was initiated in December calibration approach have been presented in an earlier study
2001, and the first operational data were obtained in March (Rollenbeck and Bendix, 2006) and some work on the unique
2002. By March 2008, about 400,000 images have been characteristics of the LAWR-system has been published by the
recorded with an uptime of about 71%. manufacturer (Jensen, 2002). However, the strong damping of
Ancillary data sets are used to develop a comprehensive the X-band micro waves, the complex terrain and several
understanding of the precipitation dynamics. A network of operational issues required an extended processing-framework
8 automatic stations supplies hourly data since 1998 (Richter, of the images to acquire useful time series and integrated
2003), three principal stations of the INAMHI national products from the radar. With all clutter fields removed and the
weather service with daily data, and a set of 22 totaling rain strong dampening corrected, the final radar images only
gauges supply monthly values. Information on rainfall rate provide quantitative data for a part of the observed domain
and drop spectra taken at five-minute intervals is available and also suffer from a decaying trend in sensitivity.
from 2 present weather sensors based on scatterometer The major problem to be faced is the unequal spatial
technology (BIRAL VPF 730–130, Leroy et al., 1998) which distribution of calibration data and the varying sensitivity
provide samples each 5 min. Unfortunately, automatic cli- conditions of the radar. Known approaches of relating
mate stations are only available for a small part of the domain, measured reflectivity to rain rate do not apply, because the
while totaling gauges do not have the required temporal LAWR has a non-linear receiver and the output values are
resolution to depict diurnal processes and spatial patterns of uncalibrated raw counts. The relation to the actual reflectivity
single events. Furthermore, station data with good coverage, is unknown.
which can be merged, with radar data only covers well the Consequently, the calibration of the radar counts was
central and southwestern part of the radar range. To the done by constructing surfaces of the relation between radar
north, only three stations are available, and both eastern data and the available station data. By this, the precipitation
sectors are only equipped with one station in Paquisha, south amounts are obtained from the station network, but the
of Yanzatza (see Fig. 1). Unfortunately only in the core study spatial distribution follows the radar observations. Precipita-
area and in Catamayo and Paquisha a complete time series is tion in the core study area can reliably be based on the dense
available, all other stations operated only for a limited part of network of automatic stations available there.
the period considered here. The processing consists of three branches, starting with the
NOAA-AVHRR images are only available for certain time calibrated and rectified NOAA-AVHRR images. These are
steps and the derivation of actual precipitation is still an processed by a cloud classification scheme especially adapted
unsolved problem. This is especially true for the typical ad- to Ecuador (Bendix et al., 2004) and cloud frequency and cloud
vective type of rain occurring on the east Andean escarpment, top height is used to correct the viewing geometry of the
where rainout is mainly from low stratus clouds with higher LAWR. LAWR raw data is processed according to the scheme
280 R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289

described in Rollenbeck and Bendix (2006), then uncalibrated is interpolated from the relations between 30 points in the
monthly total maps are calculated. The climate station data is station data map and the normalized radar map. A window of
selected according to the uptime of the radar, values of 5 × 5 pixels is applied to calculate the relation to the ground
totaling gauges are reduced proportionally to neighboring truth data in each respective month, which results in 750
recording stations. All data are then detrended by the calibration points for each month.
altitudinal gradients observed in the data and an interpolated Formally the scaling for each radar pixel can be described by
precipitation map is produced with kriging to reflect hori-
zontal gradients in rainfall distribution. The DEM (digital
Ps
elevation model) is used to rescale this map to the actual kx;y =
rainfall map by reapplying the observed vertical gradients of Cr
0.05 mm/m of altitude, the average value for the whole study
region. The monthly total maps from the radar are normalized, with the averaged ground truth data (5× 5 moving window)
to remove any temporal sensitivity trends from the radar data,
caused by irregular hardware operation and anomalous pro- x;y + 2
pagation conditions. In the next step, the normalized maps is P sðx;yÞ = ∑ Psðx;yÞ = 25
related to the interpolated station data map and a scaling map x;y−2

Fig. 2. Processing chain to derive calibrated radar images. The broken line indicates the calibration approach published.
in Rollenbeck and Bendix (2006).
R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289 281

Fig. 3. Detrended interpolation of station data (left) and scaled LAWR totals map (right) for the period March 2002 to March 2008.

and finally radar precipitation rate is scaled by ment of storm cells on the basis of the actual 5-minute radar
images.
Cr
Pr = 3. Results and discussion
k
3.1. Evaluation
Where Ps are the precipitation values taken from the interpo-
lated station map, Cr are the raw counts of the radar, k denotes The average annual total maps from interpolated station
the spatially variable scaling factor and Pr is the final products data and the scaled radar maps are used to evaluate the
precipitation. developed method (Fig. 3). The distribution of precipitation
This procedure is used to build scaling maps for each fields is independent in both maps, only the scaling of the
month of radar operation. radar data depends on the ground truth data.
The scaling maps are then applied to each five-minute Fig. 3 shows the areas of agreement of the two measure-
radar image for the respective month. Because the final radar ments as well as some deviations: the eastern sector of the
images still show some areas of weak sensitivity especially in domain has the highest annual precipitation input, while the
the most interesting core study area, the dense network of western half shows the expected westward decline of rain
station data is used to fill this gap by blending the inter- fall. To the North and the South the radar suffers from
polated station map with the radar images. Blending is shadowing effects of higher mountains. However, the
weighted by a nearest distance function of the horizontal and interpolated maps more reflect the distribution of elevations
vertical station coordinates. By this, areas with high station
density and at similar altitudes like the stations are more
represented by the station data, while remote areas are more
represented by radar data.
With optimization runs, the parameters of view shed
correction and distance damping are adjusted to minimize
errors between a correlation of the station data Ps and the
final product Pr. The complete processing chain of the data is
depicted in Fig. 2.
The annual cycle of precipitation distribution is generated
by aggregating the 6-year average monthly totals as well as
the average annual total. The average monthly totals are then
combined to produce totals maps of the characteristic three
seasons of the study region. These data are compared to the
10-year average of interpolated station data and a published
precipitation map of the INAHMI.
Typical weather situations under the influence of different
controlling factors are explained by daily totals maps,
depicting the variability of distribution of the different days.
The underlying processes of rainfall formation are analyzed Fig. 4. Correlation of monthly station data vs. calibrated radar data for the
with the help of visually tracking development and move- maps shown in Fig. 3.
282 R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289

above 2000 mm in the east from the drier parts in the west.
An important feature only visible in the combined map is the
extreme rainfall occurring in the Cordillera del Condor in
the southeastern sector and at the already mentioned slope
of the Yacuambi-valley. Annual totals here exceed 5000 mm.
The highest mountain chain, the Cordillera Real is already
slightly less wet, probably because atmospheric moisture has
precipitated on the eastward mountains.
The interpolated official map published by the Ecuadorian
Weather Service INAHMI (Fig. 6 right) is mainly based on 5
principal stations in Catamayo, Vilcabamba, Loja, Zamora and
Yanzatza. In comparison to the blended combination of ground
and radar data (Fig. 4, left), the spatial structures in the west
seem fairly represented. However to the east, strong deviations
from the combined product are visible. Generally, precipitation
is underestimated by the official map with maximum areas only
reaching 4000 mm and the main Cordillera Real is only shown
to have slightly above 1500 mm, which is in strong contrast to
Fig. 5. Blended average annual map of interpolated station data and scaled observations and measurements reporting values above
radar maps (2002–2008). 4000 mm.

than the dominant barrier effect of the mountain chains. The 3.2. Seasonal precipitation patterns
radar image shows the highest rain fall in the most exposed
parts to the south east and an extraordinarily steep slope in The maps of the seasonal variation (Fig. 7) in the research
the valley of Yacuambi, about 15 km northeast of the center of region exhibit the typical patterns which control the rain fall
the image, the station maps attribute highest rainfall to the regime: January to April is a transitional season, where ad-
Cordillera Real. The central parts are strongly underesti- vective and convective rainfalls are interspersed and precip-
mated, due to saturation of the radar receiver. itation is highly variable from day to day. The southern sector
The point-to-point calibration with a 5 × 5 pixel average of the domain receives the highest amounts of rain, because
window for the radar image shows a good agreement radiative heating of the high mountains enhances convection.
between the two data sets (Pearson r = 0.82, slope 0.92; From May to August the predominant rain generating
Fig. 4). mechanisms are the strong easterlies and the barrier effect
Generally the distribution in the radar image is more of the main mountain chains. The western parts go through
disperse and more bound to certain mountain peaks that are the driest month, while the eastern escarpment is almost
exposed to moisture transport. The blended map (Fig. 5 left) permanently wetted. Between September and December an
gives the most probable representation of precipitation. It increasing influence of westerlies with lower rainfall is
shows the isolated dry valley of Catamayo in the west, visible, especially in the southern sector. These westerlies
surrounded by slightly higher values, especially in the more also support the dry spells in the eastern sector, when Föhn
elevated parts. The role of the main mountain chains as effects inhibit precipitation and fine weather is observed
climate divide is clearly visible, as it separates areas of totals more frequently. Wind as the primary factor determining the

Fig. 6. Blended dataset from radar and interpolation for the complete observation period 1998 to 2009 (left) and published official precipitation map from the
national weather service INAHMI (various periods until 2000) (right).
R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289 283

Fig. 7. Seasonal variations of rainfall distribution (totals of the respective 4-month period).

spatial patterns of rainfall throughout the year follows a through the clear sky in the west. During the wet season in
similarly strict regime: March, convective cloud systems from the west often
Fig. 8 shows the relative frequency of wind directions per converge with easterly flows from the Amazon leading to
month, grouped by the four sectors NE, SE, SW and NW. The sometimes torrential rainfalls. This is enhanced by convection
different seasons are clearly marked by a change of the flow caused by the relative overheating of this basin.
regime in the boundary layer. Quite contrarily the three locations on the eastern
The different seasonality of the eastern and western sides escarpment are characterized by almost continuous light
of the Andes is depicted in Fig. 9 by some typical annual cycles drizzle, which amounts to rather high precipitation totals. At
of different locations in the radar range, covering the profile the margin of the Valle Yacuambi, about 15 km northeast of
lines given in Fig. 1. The profile North (N in Fig. 9) covers the the center of the images, a small rainfall maximum is visible
valley of Catamayo with its full-arid climate that is prone to a in the map. Here, a rather steep mountain slope is almost
very strong Föhn effect during times of easterlies. The driest perfectly exposed to the main flow coming up the Valley of
months are June to August with almost no rainfall at all. The the Rio Zamora. This valley guides moist air masses during
short wet season brings intensive showers, which often are periods of strong northeasterly winds. Even more extreme is
connected to humid air masses coming from southern di- the Cerro de Consuelo where the LAWR is installed. It is a
ections. Vilcabamba, well hidden from the easterlies, shows special location because it is the most exposed place to the
similar patterns, but the proximity to the main mountain flow of humidity stemming from the Amazon. Eastwards,
divide of the Cordillera Real and the slightly weaker Föhn there is no higher mountain in the radar domain and its flanks
effect lead to a semi-arid climate, with a similar seasonality. In are perpendicular to the channelized flow of air masses. Its
the Valley of Vilcabamba, conditions are only slightly eastern slope is therefore a place of rather intensive processes
different. The Föhn effect is not that strong, because it is not of forced uplift and consequent rainout. The diagram clearly
that deeply incised and receives some spillover of the wall shows the wind-regime of rainfall here, it almost perfectly
clouds above the main Cordillera. The intra-mountain basin of follows the frequency of easterlies during the year. The
the city of Loja is only a short distance away from the Cordillera de Condor is much lower, but receives almost the
locations of extreme precipitation in the Cordillera Real and same amount of rainfall per year.
the Cordillera de Consuelo. Hence, spillover of humid air The interaction of moisture transport, atmospheric flow
masses frequently occurs here. A frequent observation in this regime and topographical features leads to specific patterns as
city is a wall cloud above the mountains to the east, bringing later shown in Figs. 11 and 12. To understand the relevance of
sprays of drizzle, while at the same time the sun shines those patterns, sectoral totals of rain fall from the combined
dataset for the four sectors from 0° to 89°, 90° to 179°, 180° to
269° and 270° to 359° are extracted (Fig. 10). The time series
does not represent the flow situations, but only the totals of
rain in the four sectors, which depend on the interaction of the
different rain generating processes discussed in the next
chapter. Notably, the eastern sectors contribute the majority
(about 80%) of the precipitation to the study region, especially
the southeastern parts comprising of the Cordillera del Condor
and the exposed parts of the Cordillera de Corazon. The
southwestern and the northwestern sectors have about the
same fraction of 10% each of the annual precipitation, however
during the typical month of the rain season there, this value
Fig. 8. Monthly wind direction frequencies (relative) derived from the two can increase to more than 60%. Regarding annual and
most exposed climate station Paramo (3400 m, cordillera Real) and Antenas interannual variations, a strong heterogeneity of the annual
(3180 m, Cordillera de Consuelo) 1998–2009. cycle is visible. 2002 and 2003 exhibit normal rain peaks in
284 R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289

Fig. 9. Precipitation diagrams derived from the combined images of average monthly precipitation for the period 2002 to 2008. The upper line covers the northern
profile given in Fig. 1, the lower the southern profile across the Andes. Annual totals for each point are given as pTot.

April resp. May with the majority of rain falling in the eastern by mid 2007 annual rain totals more or less return to normal
sectors. In May 2003 95% of the rain falls occurred on the conditions, the SE-sector, normally the wettest of all,
eastern side of the Andean mountains. continues to show much lower peaks than during the years
In 2004 a first rain peak occurs in March, which also before. A possible explanation of the onset of this declining
included the eastern sectors. The following April is unusual precipitation trend may be the extraordinary drought, which
dry and rain fell primarily in the southwestern sector. Then, occurred in the Amazon basin in 2005 (Phillips et al., 2009).
the wettest month of the whole time series follow in May and However, in 2006 rainfalls in the Amazon returned to normal,
June, dominated by precipitation in the northeast sector. In while the years since then have been drier than normal in the
2005 the normal annual cycle is almost reverted. Peak study region. Further analysis of this unusual dry period is
rainfalls occur in the first third of the year, with a pronounced still pending.
wet season in the western parts in March (~55% of the month
total). Then after June 2005 monthly totals are declining for a 3.3. Characteristic weather types
long period. Between the unusual dry August 2005 (80 mm,
only 50% of the long year average for this month) and May The interaction of the different flow regimes with topo-
2007, no month exceeds a total of 200 mm. The rainy season graphy and the resulting precipitation patterns are illustrated
of 2005 and 2006 seem to be inhibited and the dominance of by four typical precipitation distributions (Fig. 11). Northwest
rain in the western sectors is clearly visible. From May 2005 situations are rather rare and hence, typical patterns cannot
until April 2006 wind data of the two most exposed climate be recognized from the few available radar images. The
stations show the unusual high frequency of westerlies of example from the 8th of February 2006 shows some isolated
almost 40%. Usually the maximum of westerlies occurs in rain fields above the eastern escarpment of the Cordillera Real,
December and reaches only 25% of the time. The annual totals possibly triggered by some convergence with the eastern flow
of the eastern sectors during these two years are conse- dominating the following days.
quently about 22% lower than the average of the rest of the The situation on the 7th of May 2005 is characterized by
years, reaching only 2100 mm instead of 2800 mm. Although widespread drizzle under a strong wind from the northeast-

Fig. 10. Monthly precipitation totals per wind sector (NE, SE, SW, and NW).
R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289 285

ern sector. The barrier effects of the higher mountaintops port. A special feature of that day is the intensive spill over of
mainly control the distribution of rainfall. The valley of the storm clouds over the central Cordillera Real, leading to strong
Rio Zamora serves as preferred way to transport moisture rain in the normally dry valley of Vilcabamba and further west
from the Amazon basin, because it is oriented from NE to SW. to Cariamanga. A similar phenomenon with lesser intensity
The Western half of the region remains almost free of rain, occurs in the north of the region close to Saraguro.
although some spillover occurs in the south. The dominant rain types occurring in the region are
In April southwestern flows are frequent, the example from explained by four examples from May 2002, October 2002,
the 10.4.2007 is quite typical of the precipitation distribution, June 2003 and November 2006 (Fig. 12). In November, when
where the inflow of rather warm air masses from the drier parts wind speeds are generally low, isolated convective showers
of the province of Loja cause intensive convection on the develop due to local overheating. The image shows several
western slopes of the Andes mountain chain. Strong rain storms shower cells above the Valley of Vilcabamba and the western
in the valleys of Catamayo, Vilcabamba and in the basin of Loja slopes of the Cordillera Real. The available climate stations
can be detected frequently around this date, almost in all the there show peak rain rates of 6.4 to 8 mm, falling in about 4 to
years observed. 5 h. Often these cells develop due to heat accumulations
Southeastern flows are the most frequent and show a wide above the intramontane basins but the peak rainfalls occur
range of precipitation patterns. The 14.3.2008 exhibits strong above the mountains due to upslope thermal winds.
forced uplift on the high mountains in the south of the During the main rain season, the dominant rain type is
Cordillera Real. The central parts of the region benefit from advective, as can be seen on the 20.6.2003. The strong wind
shadowing effects of the Cordillera del Condor and the mainly causes intensive rainout on the main Cordillera, while the
NE–SW oriented valleys prevent pronounced moisture trans- western part is free of rain because of the strong lee-effect.

Fig. 11. Examples of the four different flow regimes.


286 R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289

The general rain rate derived from the radar image is On 21.5.2002 a characteristic situation developed when
significantly lower, but certain climate stations show the strong easterlies spilled moisture over the mountains and
intensity of these type of events: The mountain station at the brought widespread advective rainfalls for almost the whole
Cerro de Consuelo (Radar site) accumulates 53 mm on that domain. The constant strong rainfall during the whole day
day with rain rates of up to 9.5 mm/h, the whole event lasted results in a high rain rate of 2.1 mm/h. All stations show
almost 72 h. similar totals of 30 to 50 mm and maximum rain rates be-
Typically, convective patterns are more frequent in October, tween 9 and 12 mm/h.
November, March and April and advective patterns dominate Rain types and spatial patterns are not strictly bound to
the rain season from May to August (see also Fig. 11). certain seasons or locations: There are frequent occasions of
However, both types of precipitation can be observed strong easterlies with advective rain in the drier month of the
throughout the year and sometimes occurring simultaneously year, but also longer dry spells occurring during the main
on a single day. The example from the 17.10.2002 shows rainy season. Advective rainfalls are not limited to the eastern
advective precipitation in the eastern half of the image, while escarpment, but frequent events of spillover occur which
the west develops several storm cells bound to the higher cause enhanced precipitation especially in the more elevated
mountains. The highest parts of the Cordillera Real remain parts (see also Fig. 10). Convection frequently increases
largely dry. The totals and rain rates below the advective parts rainfalls inside large stratus clouds, which precipitate on the
to the east are significantly lower, reaching about 3 mm daily eastern escarpment, which probably contributes to the high
with a maximum rain rate of 2.1 mm/h. In Vilcabamba, 7.3 mm rain rate on the 21.5.2002. Convection is more bound to
fall in only 2 h. mountain crests and peaks: The rather rare thunderstorms

Fig. 12. Four examples for typical rain types. Rain rates (RR) are given as captions.
R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289 287

observed on the eastern side of the Cordillera Real are almost shorter but intense showers frequently occurring during
always spilling over from the Loja-basin under westerly these periods. Visual inspection of the 5-minute radar images
flows, following the main ridge lines and dissipating, when show widely dispersed small showers in the whole radar
they reach the eastern lowlands. range for extended time periods, especially from September
to November 2006. Apparently the drier phase in 2005 and
3.4. Rain generating processes 2006 has not only diminished the influence of the easterlies
(see Fig. 10), but also weakened convection resulting in a
The three most important processes of rain formation are considerable rainfall anomaly. Comprehensive climatological
further analyzed statistically by using data of the scatterom- explanations for this period require further analyses of the
eter instruments. The topographical control is clearly detect- data collected by the ongoing research initiative in south
able in the images shown above and emphasizes the role of Ecuador.
the easterlies as the primary factor for rainfall distribution.
The second influencing factor is local atmospheric stability 4. Conclusion
conditions, which affect the development of the rain type.
Two variables are analyzed for this matter: the rain rate and The analysis of almost 6 years of radar monitoring of
the average drop size of precipitation (effective radius reff as rainfall processes in the Andes of south Ecuador confirms
defined by Wyser, 1998), as observed by the scatterometer at several well-known features of the rainfall dynamics but adds
the ECSF research station (Fig. 13). Although under normal some important new insights into the amount of spatial
circumstances reff is not necessarily larger in convective heterogeneity and temporal variability of the rainfall pat-
showers (Tokay et al., 1999) it is assumed, that the special terns. First of all the, data products used so far have strongly
atmospheric environment of the tropical mountains here and underestimated the annual totals as well as the spatial
the capabilities of the scatterometer instrument cause variability of monthly and annual totals actually occurring
convective showers to show a higher average drop size and in this region. The dominant role of the strong easterlies and
higher rain rates. The frequent and widespread light drizzle the barrier effect of the Andean mountain chain are re-
occurring here is typically associated with advective rain sponsible for the long-term average spatial distribution.
cause by the forced uplift on the east Andean slopes. Mountains with slopes perpendicular to the most frequent
Both parameters exhibit a similar temporal behavior wind direction receive the highest amounts of rain, given that
(Pearson r² = 0.42), higher rain rates are almost always specific pathways for moisture transport exist. Daly et al.
caused by larger drop sizes. Generally the average drop size is (1994) proposed to model precipitation as a function of
rather low, because the majority of the rain falling is of the terrain elevation and statistically derived horizontal gradi-
advective type and the thin air of these mountain sites cause ents, a technique applied here too. However the detrended
higher terminal velocities of raindrops with earlier breakup interpolation clearly shows limitation, when compared to the
into smaller droplets (Foote & Du Toit, 1969; Straka, 2009). radar maps. Small-scale structures are not represented and the
Higher values of reff up to 0.35 mm and peak rain rates of individual exposure of a mountain and the influence of
10 mm/h mainly occur in JFMA and SOND, the main rain surrounding terrain are not considered. Moreover, the influ-
season is characterized by values with drop sizes 10% and rain ence of topographic pathways for moisture transport is totally
rates 25% lower. Most interesting are times when both values left out as an influencing factor. To model all these parameters
deviate. The peak rain rates in late 2002 and early 2003 and under constantly changing boundary conditions of wind di-
the larger differences in all rainy seasons point to a higher rection, wind speed, moisture transport and radiative forcing of
fraction of small droplets and consequently a more advective convection as proposed by Kyriakidis et al. (2001) appears
regime. As expected, the transitional JFMA seasons of 2004 almost impossible in such a remote region, where there are
and 2005 are marked by larger drop sizes and hence more very few climate data available. Hence, the application of re-
convective rain falls, as already concluded for the drier mote sensing techniques is indispensable.
season. Lower intensities and larger droplets mark the The second most important process is local heating with
extended dry period in 2005 and 2006 already mentioned resulting convective showers. This process is responsible for
above. This contradictory behavior may be attributed to the majority of rain falls in the western sectors, although a

Fig. 13. Average monthly drop size of rain fall at the ECSF (rEff, gray area, left axis) and average rain rate (black line, right axis). Times of no rain are excluded from
the statistics. (measurement of drop spectra started 6 month later than radar observations).
288 R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289

contribution of advective spillover and a certain amount of B1.4 (RI370/18-1). We also would like to thank the foun-
orographic rain is still present there. However, the altitudinal dation Nature & Culture International (NCI) Loja and San
range of orographic rains jumps from about 1500 m on the Diego for logistic support.
eastern escarpment to elevations above 3000 m, after air
masses have crossed the main mountain chains. Apparently, References
the rainout causes a significant horizontal gradient.
Convective enhancement of rain rates also occurs on the Adler, R.F., Huffman, G.J., Bolvin, D.T., Curtis, S., Nelkin, E.J., 2000. Tropical
rainfall distributions determined using TRMM combined with other
eastern sides, where normally advective processes dominate,
satellite and rain gauge information. Journal of Applied Meteorology
as can be seen in the examples presented in this study. vol. 39, 2007–2023.
Convective processes, normally well reproduced by current Andrieu, H., Creutin, J.D., Delrieu, G., Faure, D., 1997. Use of weather radar for
climate models, are strongly overestimated, if downscaled to the hydrology of a mountainous area. Part I: radar measurement
interpretation. Journal of Hydrology 193, 1–25.
the resolution necessary to represent the strong gradients Baigorria, G.A. 2005. Climate interpolation for land resources and land use
given here, especially, when orographic enhancement is in- in mountainous regions. Ph.D. dissertation. Wageningen University,
volved (McGregor, 1996, p. 112). Laboratory of Soil Science and Geology, Wageningen, the Netherlands.
168 p.
The traditional distinction of a primarily convective Beck, E., Müller-Hohenstein, U., 2001. Analysis of undisturbed and disturbed
regime in tropical lowlands and the absolute dominance of tropical mountain forest ecosystems in southern Ecuador. Die Erde;
advective processes in tropical mountains obviously are too Themenheft: Tropische Wald-Ökosysteme 132, 1–8.
Beck, E., Makeschin, F., Haubrich, F., Richter, M., Bendix, J., Valerezo, C., 2008.
simple. Bendix and Lauer (1992) already identified the un- The Ecosystem (Reserva Biológica San Francisco). In: Beck, E., Bendix, J.,
usual patterns of the variations in the annual cycle occurring Kottke, I., Makeschin, F., Mosandl, F. (Eds.), Gradients in a Tropical
on the east Andean escarpment, but due to a lack of Mountain Ecosystem of Ecuador, Ecological Studies, vol. 198. Springer,
Berlin, pp. 1–14.
observations could not identify the complexity of the processes
Bendix, J., 1997. Adjustment of the convective stratiform technique (CST) to
involved. Emck (2007) analyzed high-resolution point data estimate 1991/92 El Niño rainfall distribution in Ecuador and Peru by
along two profiles crossing the Andes in the study region and means of METEOSAT 3. International Journal of Remote Sensing 18,
1387–1394.
identified the influence of exposure and moisture pathways,
Bendix, J., 2000. Precipitation dynamics in Ecuador and northern Peru during
but also could not observe process dynamics, because the the 1991/92 El Niño: a remote sensing perspective. International Journal
remote sensing perspective was missing. Several attempts have of Remote Sensing 21, 533–548.
been made, to improve interpolation techniques by applying Bendix, J., Lauer, W., 1992. Die Niederschlagsjahreszeiten in Ecuador und ihre
klimadynamische interpretation. Erdkunde Bd. 46, 118–134.
external drift variables (Buytaert et al., 2006) and incorporating Bendix, J., Rollenbeck, R., Palacios, E., 2004. Cloud classification in the Tropics
wind dynamics, derived from cloud tracking (Giovannettone —a suitable tool for climate ecological studies in the high mountains of
and Barros, 2008). However, all these geostatistical models Ecuador. International Journal of Remote Sensing vol. 25 (21),
4521–4540.
remain specifically suited to certain types of mountains with Bendix, J., Trachte, K., Cermak, J., Rollenbeck, R., Nauss, T., 2009. Formation of
specific characteristics. The strong interannual variations, convective clouds at the foothills of the tropical eastern Andes (South
which occurred during the only 6 years of radar observation, Ecuador). Journal of Applied Meteorology and Climatology 48, 1–17.
Buytaert, W., Celleri, R., Willems, P., de Bievre, B., Wyseure, G., 2006. Spatial
show, that only real time observations can capture the dynamic and temporal rainfall variability in mountainous areas: a case study from
of precipitation distribution. Remote sensing instruments are the south Ecuadorian Andes. Journal of Hydrology 329, 413–421.
the only mean to cover remote regions with a scarcity of Daly, C., Neilson, R.P., Phillips, D.L., 1994. A statistical–topographical model
for mapping climatological precipitation over mountainous terrain.
observation. The long period of below-normal precipitation
Journal of Applied Meteorology 33, 140–158.
from 2005 till 2007 was registered by some stations, but the Emck, P., 2007. A Climatology of South Ecuador. Diss. Univ, Erlangen.
extent of this event and the spatial response of the rainfall fields Foote, G.B., Du Toit, P.S., 1969. Terminal velocity of raindrops aloft. Journal of
Applied Meteorology 8, 249–253.
would have gone unnoticed without the radar data. The dry
Fornasiero, A., Bech, J., Alberoni, P.P., 2006. Enhanced radar precipitation
period in the Nudo de Loja was certainly triggered by the estimates using a combined clutter and beam blockage correction
drought which occurred in the Amazon basin in 2005, however, technique. Nat. Hazards Earth Syst. Sci. 6, 697–710.
there is still no clear concept, why this period lasted so much Giovannettone, J.P., Barros, A.P., 2008. Probing regional orographic controls
of precipitation and cloudiness in the Central Andes using satellite data.
longer in the Andes mountains of south Ecuador. Journal of Hydrometeorology vol. 10 (1), 167–182.
Regarding the methodological advances, an advantage of Hughes, M., Hall, A., Fovell, R.G., 2009. Blocking in areas of complex
the straightforward blending technique presented here is, topography and its influence on rainfall distribution: 508. Journal of
Atmospheric Sciences 66 (2), 508–516.
that it can be used with a very limited set of ground truth INAHMI (Instituto Nacional de Hídrologia y meteorologia), 1993. Anuario
data, which is normally the case in mountain regions with Meteorologíco. . Quito.
extreme rainfall. It also is rather robust against variations in Jensen, N.E., 2002. X-band local area weather radar—preliminary calibration
results. Water Science and Technology vol. 45, 135–138.
radar sensitivity and propagation conditions, normally the Kyriakidis, P.C., Kim, J., Miller, N.L., 2001. Geostatistical mapping of
biggest problems for radar calibration. precipitation from rain gauge data using atmospheric and terrain
characteristics. Journal of Applied Meteorology vol. 40, 1855–1877.
Leroy, M., Bellevaux, C., Jacobs, J.P., 1998. WMO Intercomparison of present
Acknowledgments
weather sensors. Instruments & Observing Methods report WWW/OSY
No. 73. World Meteorological Organization, Geneva.
The authors are indebted to the German Research Lopez, E.L., Howell, W.E., 1967. Katabatic winds in the equatorial Andes.
Journal of Atmospheric Sciences 24, 29–35.
Foundation (DFG) for the funding of the work in the scope
McGregor, J.L., 1996. Regional climate modelling. Meteorology and Atmospheric
of the Research Unit RU816 ‘Biodiversity and Sustainable Physics 63, 105–117.
Management of a Megadiverse Mountain Ecosystem in South Phillips, Oliver L., co-Authors, 2009. Drought sensitivity of the Amazon
Ecuador’, sub project B3.1 (BE 1780/15-1). Many thanks to rainforest. Science vol. 323 (5919), 1344–1347. doi:10.1126/science.
1164033. 6 March 2009.
Michael Richter and T. Peters for providing the data of the Richter, M., 2003. Using epiphytes and soil temperatures for eco-climatic
ECSF and Cerro met stations collected within the sub project interpretations in Southern Ecuador. Erdkunde Band 57, 161–181. Heft 3.
R. Rollenbeck, J. Bendix / Atmospheric Research 99 (2011) 277–289 289

Rollenbeck, R., Bendix, J., 2006. Experimental calibration of a cost-effective Trachte, K., Nauss, T., Bendix, J., 2010. The impact of different terrain
X-band weather radar for climate ecological studies in southern configurations on the formation and dynamics of katabatic flows –
Ecuador. Atmospheric Research 79, 296–316. idealized case studies. Boundary-Layer Meteorology 134 (2), 307–325.
Rollenbeck, R., Bendix, J., Fabian, P., Boy, J., Dalitz, H., Emck, P., Oesker, M., Vuille, M., Bradley, R.S., Keimig, F., 2000. Climate variability in the Andes of
Wilcke, W., 2007. Comparison of different techniques for the measurement Ecuador and its relation to tropical Pacific and Atlantic sea surface
of precipitation in tropical montane rain forest regions. Journal of temperature anomalies. Journal of Climate 13, 2520–2535.
Atmospheric and Oceanic Technology vol. 24 (2), 156–168. Weischet, W., 1965. Der tropisch-konvektive und der außertropisch-
Straka, Jerry M., 2009. Cloud and Precipitation Microphysics. Cambridge advektive Typ der vertikalen Niederschlagsverteilung. Erdkunde 19,
University Press. Cambridge Books Online. Cambridge University Press. pp. 6–14.
Tokay, A., Short, D.A., Williams, C.R., Ecklund, W.L., Gage, K.S., 1999. Tropical Wyser, K., 1998. The effective radius in ice clouds. Journal of Climate 11,
rainfall associated with convective and stratiform clouds: intercomparison 1793–1802.
of disdrometer and profiler measurements. J. Appl. Meteor. 38, 302–320.

Vous aimerez peut-être aussi