C. R.

Geoscience 338 (2006) 1098–1112

http://france.elsevier.com/direct/CRAS2A/

Internal Geophysics (Applied Geophysics)

Lake studies from satellite radar altimetry

Jean-François Crétaux a,∗ , Charon Birkett b
a CNES/LEGOS, 14, av. Édouard-Belin, 31400 Toulouse, France
b ESSIC, 2207 Computer and Space Sciences Building, University of Maryland, College Park, MD 20742-2465, USA

Received 6 October 2005; accepted after revision 10 August 2006

Available online 5 October 2006
Written on invitation of the Editorial Board

Abstract
For more than 10 years, satellite radar altimetry has been a successful technique for monitoring the variation in elevation of
continental surface water, such as inland seas, lakes, rivers, and more recently wetland zones. The surface water level is measured
within a terrestrial reference frame with a repeatability varying from 10 to 35 days depending on the orbit cycle of the satellite.
With several decades of technique refinement, current data processing can be fairly simple or complex depending on the mission
and the instrument tracking methods. Data acquisition is not affected by weather conditions, but the technique can have a number
of limitations. Rapidly varying topography or complex terrain may inhibit the retrieval of good elevation data. The instruments
can only operate in a profiling mode and do not have a true global view. Stage accuracies are also dependent on target size and
surface roughness, which will limit worldwide surveying and limnological applications. However, there is the scope for systematic
continental scale monitoring and the provision of new stage information where gauge data is absent. The technique is sufficiently
advanced to have enabled a number of inland water case studies. Focussing on the large lakes, the links between lake evolution and
the local climate cycle on seasonal to interannual timescales have been explored, and water storage balance for water management
has also been brought into focus. Typically altimetric stage measurements can range in accuracy from a few centimetres (e.g., Great
Lakes, USA) to tens of centimetres (e.g., Lake Chad, Africa), depending on size and wind conditions. In the case of the Aral Sea,
contributing E-P (1.5 km3 yr−1 ), river run-off (3.0 km3 yr−1 ) and altimetry (1.5 km3 yr−1 ) errors combine to give a ∼3.5 km3 yr−1
water mass balance error. Of particular interest are those lakes in arid and semi-arid regions where water is an essential economical
resource. Anthropogenic influences (irrigation and the construction of dams) may strongly affect lake and inland sea evolution or
affect the global water cycle in general, noting the worldwide development of large reservoirs over the last 50 years. With regards
to societal and economic issues then, this article reviews several of the case studies focussing on the Aral and Caspian Seas, and
Lakes Issykkul and Chad. To cite this article: J.-F. Crétaux, C. Birkett, C. R. Geoscience 338 (2006).
© 2006 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved.
Résumé
Altimétrie satellitaire sur les lacs. Depuis plus de 10 ans, l’altimétrie satellitaire est une technique utilisée pour le suivi des
variations de niveau des eaux continentales, telles que les lacs, les rivières, ou encore les zones inondées. Le niveau d’eau est
mesuré avec une répétitivité de 10 à 35 jours selon le type d’orbite utilisé, et il est exprimé dans un système de référence terrestre
ellipsoïdal. En fonction du type de la mission et du mode de fonctionnement des instruments en orbite, le traitement des mesures
altimétriques est plus ou moins compliqué. Si la technique présente de nombreux avantages, entre autres de ne pas être affectée par
les conditions météorologiques, comme dans le cas, par exemple, de l’imagerie satellitaire optique, elle peut toutefois présenter un

* Corresponding author.
E-mail addresses: Jean-francois.Cretaux@cnes.fr (J.-F. Crétaux), cbirkett@umd.edu (C. Birkett).

1631-0713/$ – see front matter © 2006 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.crte.2006.08.002
 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112 1099

certain nombre de limitations. Une topographie de surface aux alentours de la zone à étudier qui serait trop irrégulière peut, par
exemple, restreindre fortement le nombre de mesures de qualité satisfaisante. Par ailleurs, les altimètres fournissant une mesure
exactement au nadir du satellite, leur couverture spatiale dépend de l’espacement entre deux traces au sol, et ils ne permettent pas
une vue d’ensemble de la Terre. Pour ces raisons, la précision des mesures effectuées par le biais de l’altimétrie satellitaire n’est
pas homogène, et dépend, dans une large mesure, de la dimension des objets ciblés, de la rugosité de surface, limitant ainsi une
surveillance globale de la Terre. Toutefois, cette technique est suffisamment avancée aujourd’hui et offre un potentiel intéressant
pour le suivi systématique de l’eau de surface à l’échelle continentale, en particulier dans les nombreux cas d’objets non mesurés
par des instruments au sol. Par exemple, les liens entre l’évolution des grands lacs et les effets climatiques à différentes échelles
spatio-temporelles ont été étudiés, ainsi que des bilans hydrologiques dans certains cas spécifiques. À titre d’exemple, l’altimétrie
permet un suivi régulier des grands lacs américains, avec une précision de l’ordre de quelques centimètres. Pour d’autres lacs,
comme le lac Tchad, la précision obtenue à travers diverses études est de l’ordre de 10 cm. Dans un autre cas, celui de la mer
d’Aral, de telles précisions ont permis d’établir, de façon assez précise, le bilan hydrologique au cours de la dernière décennie, et
de faire un bilan d’erreur détaillé (en combinant les valeurs suivantes : évaporation moins précipitations, 1,5 km3 an−1 , débit des
fleuves, 3 km3 an−1 , altimétrie, 3 km3 an−1 , on obtient une précision de l’ordre de 3,5 km3 an−1 du bilan hydrologique total).
Pour des régions extrêmement arides, comme l’Asie centrale, ce type l’évaluation des ressources en eaux est d’une importance
capitale. Par ailleurs, les activités humaines ont une grande influence sur les eaux continentales par le biais de l’irrigation et de la
construction de grands barrages. Le cycle global de l’eau peut en être affecté, notamment avec le grand développement, au cours de
ces dernières années, de projets de barrages le long de tous les grands fleuves de la planète. Cet article fait le point sur l’utilisation
de l’altimétrie satellitaire pour les eaux continentales, dans le cadre de ces questions d’actualités liées à la gestion de l’eau et aux
changements climatiques, en se focalisant sur quelques cas caractéristiques, comme les mers d’Aral et Caspienne, les lacs Tchad
et Issykkul. Pour citer cet article : J.-F. Crétaux, C. Birkett, C. R. Geoscience 338 (2006).
© 2006 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved.

Keywords: Altimetry; Lakes; Climate change; Water management; Central Asia; Lake Chad

Mots-clés : Altimétrie ; Lacs ; Changements climatiques ; Gestion de l’eau ; Asie centrale ; Lac Tchad

1. Introduction: Lake evolution, climate change 1.1. Water-balance components

and ecological issues
Change in lake volume is a key parameter of the wa-
Observation of the spatial and temporal variability
ter budget (Eq. (1)).
of water resources is crucial for societal and scien-
tific issues. The volume of water stored within lakes dV /dt = R − A(E − P ) (1)
and reservoirs is a sensitive proxy for precipitation and
may be used to study the combined impact of climate V being the lake volume, R the runoff from the catch-
change and water-resource management. A lake system ment basin, E the evaporation rate over the lake per unit
is a complex interaction between the atmosphere, and area, P the precipitation rate per unit area over the lake,
surface and underground water, responding to climatic and A the area of the lake.
conditions, but tempered overall by upstream water uses It requires level variation measurements (available
for agriculture, industry or human consumption. The from gauge instruments installed on the lake shoreline
overall water volume (a function of surface area and or derived from radar altimetry observations), and hyp-
lake level) depends then on the balance between the sometric curves or bathymetry maps to transform level
water inputs and outputs. The inputs are the sum of di- into volume. One critical point is the precision of the
rect atmospheric precipitation over the lake and surface bathymetry map. Small deviations can lead to large er-
runoff. Underground seepage can also provide inflow. rors in the computation of volume variation [14,49].
Though this is usually a minor component of the wa- The establishment of a hydrological water budget ap-
ter budget, it can be an important source of chemical pears simple, but in practice is not so trivial. This needs
transport [49]. Volume outputs are the sum of the direct accurate and continuous measurements, and a broad un-
evaporation from the lake’s surface with river outflow derstanding of physical interactions between the lake
and groundwater seepage. Some of these water balance and the atmosphere. Lake evaporation is generally an
components can be easily measured (river runoff, pre- unknown but critical parameter and few methods of de-
cipitation, lake level), while others can only be deduced termination exist: the most accurate is the method of
from hydrological or climatic modelling (evaporation, energy-budget, which requires knowledge of numerous
underground water). physical parameters (water temperature, radiation, air
1100 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112

temperature, saturated and air vapour pressure, wind, 1.2. Behaviour of lakes under hydrological condition
etc.). Other ‘empirical’ methods (more than 30 such changes
methods exist in the literature) are more frequently used
[24,49]. Groundwater flux between lake and subsur- The volume of stored water will vary with time ac-
face is also difficult to measure, but can be acquired cording to changes in the hydrological budget. Under a
through the use of seepage minipiezometers [49] or constant climate scenario, the volume will tend towards
modelling the groundwater flow system. Operation of an equilibrium level over a given time period, display-
minipiezometers does not permit continuous monitor- ing a perfect balance between inflow and outflow [33].
ing, while modelling requires many tests and complex If one of the components varies, for example, the lake
geological investigations. As a consequence, ground- level will change to compensate and reach a steady state.
water fluxes are generally considered as negligible or If a further change in water balance occurs before the
defined as a constant value in the water budget equation. lake has reached its equilibrium level, then the lake will
Measurement of precipitation via rain gauge presents find a new equilibrium state. Lakes and reservoirs will
another large uncertainty. The problem is due to high thus exhibit seasonal changes in surface area and level
variability in space and time and the paucity of gauge due to proportional changes in precipitation and evapo-
locations. The problem is generally amplified over lakes ration. Mason et al. [33] stated that for the 200 largest
because of the inability to measure rain directly over its lakes, one can use lake level and area variations to moni-
surface area, and precipitation on the shoreline and on tor climate variation via an ‘aridity index’, which ranges
the lake can differ markedly for large lakes. In some in value from 0, for very arid lakes, to 1, for lakes sit-
cases, lake evaporation can return to the lake as precipi- uated in humid zones. The authors’ formalised sets of
tation. Several of the large African lakes have greater di- relations between the equilibrium return time and the
rect lake precipitation (Victoria by 30%, Tanganyika by lake geomorphology. Through these relations, the lake
20%) than over their catchments areas. Global precipi- level records alone could then be used to compute the
tation databases (constructed by meteorological centres aridity index variation and therefore note the chang-
such as the ECMWF or the NCEP) do exist and are ing climatic conditions in the lake basin. Note though
freely available on the Worldwide Web, though discrep- that there is scope for erroneous interpretation if human
ancies with published climatologies are apparent. For activities are the source of the original lake volume per-
example, comparison between NCEP and a climatology turbation.
published by New et al. [36] for Lake Issykkul (Kirgizs- Nonetheless, the assessment of lake water balance
tan) and Lake Titicaca (Peru) reveals a factor-2 differ- could provide improved knowledge of regional and
ence between the annual average precipitation. A large global climate change and a quantification of the hu-
literature database exists for the Aral Sea, but differ- man stress on water resources across all continents. The
ences of up to 100% exist between different sets of pre- choice of the lake is an important factor. Hostetler [24]
cipitation [14]. Satellite missions such as TRMM and noted that deep and steep side lakes are good proxies for
the future GPM may offer precipitation datasets in the high-amplitude/low-frequency changes, while shallow
future. water basins are better target for rapid low-amplitude
Surface water flow, measured by runoff gauge instru- changes. The latter are extremely sensitive for revealing
ments on many river systems is often not available for decreased water input and rising evaporation.
many rivers entering or leaving lakes. Maintaining such Observation of many lakes in different climatic and
a device is expensive and many rivers in the world are regional zones can be utilised to discriminate between
not monitored or the data not released into the public do- global and regional climate and anthropogenic phenom-
main. Braided streams are difficult to gauge, and uncer- ena occurring on different time and spatial scales. For
tainties can be added to the hydrological water balance example, some lakes located in the Andean Mountains
when gauges are installed too far upstream (in case of are sensitive to ENSO sequences being more sensi-
delta like in the Aral Sea [14] or when the gauges ceased tive to precipitation variations in a humid region. Other
to work operationally (Lake Chad [12]). Although no lakes are more affected by global warming in the form
satellite-based instruments can monitor changes in stor- of changing glacier feed. Examples include Lakes Ar-
age volume, there exist now many imaging and altimet- gentino and Viedma, located at the border of Chile and
ric instruments with capabilities to map a real extent and Argentina [13,15,51].
to monitor the variation in lake stage, respectively. This The East African Rift region, covered by a dense net-
presents new opportunities to better determine the water work of natural lakes and some regulated reservoirs is
mass balance, particularly for poorly gauged regions. also of interest. Here, lakes are not only influenced by
 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112 1101

global climatic conditions, but affect the regional cli- from the ecological consequences, problems associated
mate themselves due to their size. Nicholson and Yin with economic expenditure will also arise. Increasing
[37] have shown that precipitation over Lake Victoria is temperatures in mountainous regions will accelerate the
greater than over its catchment area by more than 30%. thaw of snow and glaciers, affecting both lake level and
However, the main components of water balance (evap- salinity, while in coastal regions variations in riverine
oration and precipitation) for the East African lakes are outflow could affect heat exchange impacting thermal
still very uncertain and assessment of water balance stratification and mixing dynamics [38,39]. The effects
and connection to regional climate variability is an on- of climate change are still uncertain however, but the
going debate [6,37]. Based on seven years of satellite sustainable preservation of lakes worldwide depends on
radar altimetry data from the TOPEX/Poseidon mission the extensive understanding of the processes that gov-
and the utilisation of precipitation data over East Africa ern their potential evolution and the monitoring of these
and the Indian Ocean, Mercier et al. [34] showed ev- bodies via remote sensing.
idences of hydrometeorological linkages between the The construction of reservoirs over the last 50 years
Indian Ocean sea-surface temperature (SST) and pre- and the potential effects on sea level rise have been dis-
cipitation effects over the East African continent, and cussed by Shiklomanov et al. [44] and Gornitz [20]. On
the changes in lake levels. New satellite radar altimetry a global scale, it could be assumed that reservoirs act as
missions (Jason, GFO, and Envisat) have recently in- simple retainers of water and thus do not contribute to a
creased the number of lakes observable over East Africa general increase in elevation. While Shiklomanov et al.
and climate studies continue. [44] estimated that since 1950 reservoir storage has con-
Arpe et al. [5] showed a strong linkage on a decadal tributed to a sea-level decrease of around 0.3 mm yr−1 ,
time scale between the level variation of the Caspian Gornitz [20] quoted a water loss of 0.56–0.81 mm yr−1
Sea and the El Niño southern oscillation (ENSO). Be- and stressed that seepage loss, evaporation and sedimen-
cause the Volga river discharge (VRD) and the sea evap- tation process are still unknown and associated errors
oration both dictate the evolution of the Caspian Sea can be high. However, the example of the Aral Sea (see
level, the authors showed that the equatorial Pacific SST next sections), which dramatically shrank after 1960
variability could be used to simulate the VRD and via due to the building of irrigation channels and reservoirs
validation exercises showed good correlation with the across its drainage basin, reveals a potential negative ef-
observed discharge. Other decadal climate fluctuations fect on global sea-level rise.
such as the Pacific Decadal Oscillation (PDO) and the
North Atlantic Oscillation (NAO) can also significantly 1.3. Present-day lake-level monitoring
change the water circulation in the atmosphere, and so
affect lake level changes across large regions. The NAO The systematic, global monitoring of these water
could prominently influence climate variability over a bodies is becoming a fundamental objective of the in-
continental scale affecting the freshwater ecosystems of ternational community (as faced by United Nations’ or-
Europe, Central Asia and the US. ganisations, numerous governments, and other public
General circulation models (GCM) that have been or private institutions), but the decline in ground-based
developed to assess the effect of the so-called ‘green- gauge information has dramatically increased during the
house effect’ have shown that one of the consequences last decade [2,3,48]. In many regions, the expense of
of global warming would be significant changes in pre- data collection is restricted due to economic reasons,
cipitation and surface temperature [16,50]. This will or the collected data is restricted, being considered as
have an impact on regional hydrology and ultimately sensitive national information. The physical removal of
affect lake and reservoir level. For example, the present- gauges from many lake domains is a common situation
day decrease in the Arctic Lake total surface has been in many parts of the world. Nevertheless observation of
highlighted from a set of satellite images over more than the status of lakes and reservoirs and calculation of their
10 000 lakes in Siberia [47]. For some other locations water balance can be improved with the help of satellite
(e.g., the North American Great Lakes), global warming measurements such as precipitation from the TRMM
could be accompanied by a drop in lake level or increas- mission, and inundated surface extent or lake area from
ing level depending on which GCM is used to determine missions such as AVHRR, Landsat, and MODIS. While
the effect of global warming [32]. Seasonal variation scientists are looking to the GRACE mission to enable
of the Great Lakes are also affected by global warm- estimates of the vertically integrated water mass change
ing, as precipitation and runoff increases in autumn and over large rivers basins, researchers continue to explore
winter, inducing higher extreme lake level [4]. Apart satellite radar altimetry data as a contributor of surface
1102 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112

height variability. The techniques are now so well val- over the altimeter footprint. Corrections must be applied
idated that lake and reservoir altimetric measurements to the range measurement, to take into consideration
can be found via several web sites: atmospheric refraction, tidal and electromagnetic bias
effects. No consideration is given to lake tides, which,
• http://www.legos.obs-mip.fr/soa/hydrologie/ with the exception of spring tides (e.g., ∼8 cm for Lake
hydroweb/, Superior), are generally ∼1–2 cm. Random rain events
• http://earth.esa.int/riverandlake/, can also affect the range measurement via attenuation of
• http://www.pecad.fas.usda.gov/cropexplorer/ the altimeter radar pulse. Estimates of the effect of very
global_reservoir/. heavy rain are in the order of some tens of centimetres.
A frozen lake surface may present an additional bias to
Because water storage variability is mainly driven the altimeter range measurement, since radiation pene-
by a combination of meteorological input (precipitation, tration into the ice surface will lead to a larger than true
evaporation) and human water use, there is the need value of the altimeter range and hence to a lower height
though to have both in-situ local and global datasets estimate.
coupled with remote sensing measurements. Radar al- With knowledge of the satellite orbit, the difference
timetry has been the dominant technique for lake-level between the corrected range and the orbit, reveals the
contributions so far although lidar (laser altimetry) also height of the surface above a given reference datum
shows promise [10,22,23]. A number of authors have (the reference ellipsoid). Due to the inhomogeneous
explored the radar application over inland waters, more mass distribution within the Earth, the altimetry mea-
generally [7,35] or to more specific lake basin case surements can be further corrected for the height of the
studies [1,8,10,11,14,34]. These studies show that radar geoid above the ellipsoid of reference [19]. For conti-
altimetry is an effective and powerful technique with nental water bodies, a low-resolution terrestrial geoid
many potential applications: assessment of hydrological like EGM96 can be utilised [31], the data being first av-
water balance [11,14], prediction of lake-level variation eraged across the lake over a given period to remove
[12], studies of anthropogenic impact on lake’s water all periodical and random fluctuations and so reveal a
storage [1], correlation of interannual fluctuation of lake more precise Mean Lake Surface (MLS). However, for
level on a regional scale with ocean–atmosphere inter- small lakes, this computation introduces additional er-
action [34]. rors due to lack of averaging data to average properly
Here, we will summarise the principle of the altime- for the geoid slope along and cross track.
try technique and illustrate its potential by reviewing The derivation of time series of surface-height vari-
several test case studies on a diverse selection of wa- ations involves the use of repeat track methods. This
ter bodies. methodology employs the use of a mean (reference)
lake height profile. This is derived from averaging all
2. Altimetry: principles and error budget height profiles across the lake expanse within a given
time interval, effectively smoothing out the varying ef-
A full discussion of the derivation of altimetric fects of tide and wind set-up. The height difference be-
heights and their associated errors can be found else- tween the reference pass and each repeat pass enables
where [7]. A summary is provided here. the time series of lake-height variation to be created.
The altimeter transmits a short microwave pulse in To date, a number of satellite radar altimeter mis-
the nadir direction, and the resulting echo reflected by sions have been launched. Table 1 lists their temporal
the surface is examined in the time domain. The time coverage, repeat periods, spatial coverage with respect
for the pulse to be reflected back to the altimeter cor- to large lakes and reservoirs. Overall, stage accuracies
responds to the distance (or ‘range’) between satellite range from a few to tens of centimetres. Minimum target
and earth surface [19]. The altimetric height H of a sizes are dependent on a number of factors, including
non-ocean surface is determined by the difference be- footprint size, topography complexity and instrument
tween the satellite orbit (Alt) and the altimeter range tracking logic. Table 1 gives approximations for each al-
Rcorr measurement, with a final correction for earth tide timetric mission, but note that ongoing altimetric range
TE effects. and radar echo interpretation studies are aiming to im-
prove expected elevation accuracy and minimum target
H = Alt − Rcorr − TE
size. Table 2 summarises the estimated rms error bud-
where H is considered with respect to a mathematical get on a single 10-Hz TOPEX/Poseidon (or T/P) height
reference ellipsoid. This surface height is a mean height measurement (taken from [7]). Summing the squares of
 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112 1103

Table 1
Summary of satellite altimetry general characteristics
Tableau 1
Résumé des caractéristiques générales de l’altimétrie satellitaire
Satellites Operation period Orbital cycle Accuracy Minimum target area and width
ERS2 1995–2002 35 days >9 cm rms >100 km2 > 500 m
ENVISAT >2002 35 days >9 cm rms >100 km2 > 500 m
T/P 1992–2005 10 days >3 cm rms >100 km2 > 500 m
Jason-1 >2002 10 days >3 cm rms >100 km2 > 500 m

Table 2 nitude higher than the total error budget. Ongoing re-
Total error budget for lake height search is examining the trade-off between minimum tar-
Tableau 2 get size observable and acceptable height accuracy for
Erreur totale budgétée pour la hauteur du lac the various hydrological applications.
Contribution Root-mean-square (cm)
NASA orbit 3.5* 3. Case studies
Range precision (variable)
Embias plus skewness 2 + 1.2
Earth tide 1 3.1. The lakes of Central Asia
Dry troposphere 0.7
Wet troposphere, ECMWF/TMR 3/1.2 Fig. 1 depicts the lakes and reservoirs of Central
Ionosphere, DORIS 1.7 Asia that are overpassed by satellite altimetry. Both
Minimum total (TMR/ECMWF) 4.85/5.57
the larger (Small and Big Aral, Issykkul, Balkhash)
Abbreviations are ECMWF, European Centre for Medium-Range
and smaller lakes and reservoirs (Sarykamish, Kap-
Weather Forecasts; TMR, TOPEX microwave radiometer; and DO-
RIS, Doppler orbitography and radiopositioning integrated by satel- shagoyskoye, Toktogul, Chardarya, Aydarkul, Karakul)
lite. can be observed with the altimeter.
* The removal of geographically correlated orbit error components
This figure illustrates the potential of altimetry in
via the use of repeat track techniques may reduce this value. large basins water monitoring, in particular in regions
where in-situ data are lacking, or not available. We are
the individual rms values derives the total height er- going to show how altimetry can be used to explore
ror. The range precision will be improved, as a mean various water-resource situations in this Central Asian
lake height (across the profile) is determined, effectively region.
averaging all valid height measurements (N) from coast-
line to coastline. 3.1.1. Aral Sea
The accuracy of a single lake’s height measurement The Aral Sea is located in an arid zone, characterised
then will vary depending on the knowledge of the range, by marked differences between summer and winter tem-
the orbit and the various corrections. Range precision peratures and has low precipitation all year round. Evap-
will vary according to the surface roughness and the se- oration is approximately ten times greater than precipi-
lected satellite mission, as more complex tracking logic tation, the sea being maintained at equilibrium by the in-
and algorithms try to maximize the quantity of data col- flowing waters of the Amu Darya and Syr Darya. These
lected over continental surfaces. In practice then, it is two rivers originally provided around 60 km3 yr−1 of
important to note that altimetric height measurements fresh water, roughly half of their total capacity. Around
are an average (across the effective footprint and across 1960, a decision was made to develop an intensive cot-
the lake profile) and in this way differ from a single- ton and rice economy. In such an arid zone, irrigation
spot height at a given location provided by a traditional provided the means to reach the planned agricultural
gauge. Validation exercises with ground-based gauge objectives of the Soviet Union government. Large-scale
data give some measure of the expected accuracy for development of ground infrastructure (irrigation chan-
a similar target type and size. Comparison of T/P and nels, reservoirs) began and the extent of the irrigated
Jason results with gauge data for the Great Lakes, for area increased from 4 billion to 8 billion hectares. The
example, shows an accuracy ∼3–5 cm rms [45]. Al- volume of water utilised for irrigation increased up to
though this accuracy will reduce for smaller lakes, the about 100 km3 yr−1 . The effect on the level of the Aral
derived level variations are generally an order of mag- Sea was drastic, dropping from +53 m (above the 0
1104 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112

Fig. 1. Central Asian lakes and reservoirs monitored by satellite altimetry (http://www.legos.obs-mip.fr/soa/hydrologie/hydroweb/; the image is
constituted from a mosaic of Landsat images taken in 1990).
Fig. 1. Lacs et réservoirs d’Asie centrale observés par altimétrie satellitaire (http://www.legos.obs-mip.fr/soa/hydrologie/hydroweb/ ; l’image est
constituée à partir d’une mosaïque d’images Landsat collectées dans les années 1990).

Baltic sea level) in 1960 to +40 m in 1989. This de- be further assessed by hydrogeological modelling and
crease led to the separation of the sea into two lakes, more accurate data on the evaporation and precipitation
the ‘Small Aral’ to the north and the ‘Big Aral’ to the rates.
south. From 1989 to 2005, Small and Big Aral were Even if this can be accurately established, the ground-
separated and evolved in different ways. With the T/P water flow will likely only slow the desiccation of the
altimeter, which overpasses both new distinct lakes, it Big Aral. To reverse the process or even to stabilise the
was possible to measure precisely the level variations sea level to the level of the mid 1990s would require
from 1992 until now. The elevation of the Big Aral more underground flow than it could be even supplied
reached a low of +30 m in 2004 [14]. The correspond- in the most optimistic scenarios [14].
ing decrease in surface extent and volume was 67 000 For Small Aral, the situation since 1989 is radically
km2 and 1083 km3 in 1960 [9] to 16 000 km2 and different. It has continued to be fed by the Syr Darya
100 km3 in 2004 [14]. The difference between evapo- River, and has dried up at a lower rate than its counter-
ration and precipitation for the Big Aral represents an part. There are two main explanations: first the area of
average loss of 25–30 km3 yr−1 over the last decade, the basin is much smaller, which diminishes the effect
while river discharge from Amu Darya have varied from of evaporation, and secondly, during the years 1992 to
0 to 15 km3 yr−1 in the 1990s. Overall, the Big Aral 1999 a dam was built in the Berg’s strait (see Fig. 3),
shrank during this period at a rate of 60 to 80 cm yr−1 . which stopped the loss of water from the Syr Darya into
Satellite remote sensing also allowed a method of the desert. This dam was destroyed (and rebuilt) three
exploring the complex new hydrology. Using a com- times during this period.
bination of precise digital bathymetry map (DBM) of Aladin et al. [1] demonstrated that during the pe-
the basin with level variation deduced from altimetry, riod (1993–1999), the existence of the dam allowed
Crétaux et al. [14] computed the resulting volume vari- some restoration of the Small Aral. They computed that
ation of the Big Aral Sea for the period 1993–2004. during periods of dam absence, only 20% of the river
They showed that the reduction of the lake volume as runoff as measured at Kazalinsk (Fig. 3) reached the
measured by T/P, GFO and Jason is smaller than that sea. The rest was lost by evaporation in the delta and
deduced from examination of the hydrological budget in the desert, as well as by underground infiltration and
(Fig. 2). There are errors within both methods, but the probably by some inflow to the Big Aral through the
possibility of an additional positive water inflow to Big Berg’s strait. This computation (with uncertainties ∼20
Aral (5 km3 yr−1 [14]) via significant underground wa- to 25%) was fully based on comparison of volume vari-
ter inflow cannot be ignored. This conjecture needs to ations deduced from a combination of altimetry and
 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112 1105

Fig. 2. Big Aral Sea volume variations, deduced from in-situ measurements and from altimetry data [14].
Fig. 2. Variations de volume de la grande mer d’Aral déduite de mesures in situ et de l’altimétrie [14].

3.1.2. Other Central Asian lakes

Water resources monitoring and management in Cen-
tral Asia have to take into consideration aspects of sus-
tainable development, ecological impact of water uses
and climate changes, and transboundary water policies.
Water management is indeed becoming a political and
societal issue that concerns different countries as the
drainage basins of the Amu Darya and the Syr Darya,
are shared by five different countries: Uzbekistan, Kaza-
khstan, Tajikistan, Kirgizstan and Turkmenistan. A po-
litical agreement was signed in the mid-1990s for a ra-
tional use of the water resources. Via the use of large
reservoirs, both Kirgizstan and Tajikistan agreed to re-
tain upstream water over the winter period for ultimate
use by Kazakhstan and Uzbekistan, who require a large
amount of water downstream for irrigation in the dry
season. As compensation, Kazakhstan and Uzbekistan
agreed to provide gas and electricity to Kirgizstan and
Tajikistan. The agreement is a source of contention and
Fig. 3. The Aral Sea. Location of the Aral Sea coastline in 1966 is still being contested. This transboundary water man-
(1), 1992 (2), 2002 (3), and TOPEX/Poseidon ground tracks (4)
agement issue is needed to reinforce the collaboration
(from [14]).
between the water management agencies, and to high-
Fig. 3. Localisation des cotes de la mer d’Aral en 1966 (1), 1992 (2),
2002 (3). Traces TOPEX/Poseidon (4) (d’après [14]).
light the maintenance of the water resources monitoring
and the sharing of the river runoff, reservoir storage, and
bathymetry data with volume variations deduced from hydroclimatologic data. In fact, UNESCO, NATO and
the hydrological budget. The same calculation, when the other international organisations aimed to strengthen
dam was in place, showed that the dam allows water re- this agreement by providing the framework to enhance
tention of 80% of the river runoff that enters via the Syr the collaboration between political managers, hydrolo-
Darya delta (Fig. 4). This computation determined the gists, and scientists in general. Both financial and tech-
correlation between the amount of water entering into nical tools have been developed to assess water storage
the Syr Darya delta and the level of Small Aral [1]. evolution in the whole Central Asia, and space technolo-
1106 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112

Fig. 4. Volume variations of the Small Aral Sea deduced from satellite altimetry (black stars) and from hydrological in-situ measurements (triangles).
Amount of runoff data from the Syr Darya River to the Small Aral Sea have been adjusted considering initial values measured in Kazalinsk
(from [1]).
Fig. 4. Variations de volume de la Petite Aral déduites de l’altimétrie (étoiles noires) et de mesures in situ (triangles). Les débits totaux du Syr
Darya dans la Petite Aral ont été ajustés en considérant les valeurs initiales mesurées à Kazalinsk (d’après [1]).

Fig. 6. Chardarya reservoirs level variations deduced from in-situ

measurements (black triangles: http://water.freenet.uz/) and from
Fig. 5. Lake level variations from altimetry on multi-satellite mode
altimetry (grey square: http://www.legos.obs-mip.fr/soa/hydrologie/
(TOPEX/Poseidon, GFO, Envisat & Jason) from http://www.legos.
hydroweb/).
obs-mip.fr/soa/hydrologie/hydroweb/, versus in situ gauges provided
by the IWP (Institute of Water Problem) of Bishkek for the Issykkul Fig. 6. Variations du niveau du réservoir Chardarya déduits de mesures
Lake (Kirgizstan). in situ (triangles noirs : http://water.freenet.uz/) et de l’altimétrie (car-
rés gris : http://www.legos.obs-mip.fr/soa/hydrologie/hydroweb/).
Fig. 5. Variations de niveau du lac Issykkul (Kirgyzstan) par altimé-
trie satellitaire, calculées en mode multi-satellite (TOPEX/Poseidon,
GFO, Envisat et Jason) visibles sur le site http://www.legos.obs-mip.
fr/soa/hydrologie/hydroweb/, comparées aux mesures in situ fournies one of the storage reservoirs along the Syr Darya, and
par l’IWP (Institute of Water Problem) de Bishkek. used to release water into Kazakhstan for irrigation pur-
poses). In most cases, the radar altimeters continue to
gies have been widely utilised. Optical and radar remote provide monthly stage data.
sensing have been used for analysing soils and vegeta- Fig. 7 shows the case of the Toktogul reservoirs,
tion cover evolution, while radar altimetry could open downstream from Lake Issykkul and located in Kirgizs-
up a new mode of lake and reservoir monitoring. tan. In situ data are only available up to 2001, while
Several of these lakes are monitored by ground-based satellite radar altimetry can only provide measurements
gauges, though access to current information (post- from 2002.
2001) is still restricted. Where there are overlap periods No gauge data is available for the Sarykamish reser-
between gauge and altimeter observation, validation ex- voir, but altimetric measurements (Fig. 8) show the con-
ercises reveal the typical accuracy of the satellite mea- sequences of diverting the flow of the Amu Darya River.
surements. For example, accuracies of 5 cm rms (Fig. 5, This political decision, made by the Republic of Turk-
Lake Issykkul), 10 cm rms (Fig. 6, Lake Chardarya, menistan in the early 1990s has speeded up the shrink-
 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112 1107

works and can provide new information where gauges

are lacking. In this respect, it provides an additional tool
for decision-makers in the field of water management.
We have seen that the impact of water uses the amount
of released water from the big reservoirs, and diverting
of the inflowing river has impacts on terminal lakes like
the Aral Sea, and one can ignore neither the correlations
between stage variations in the Aral Sea or those in its
upstream reservoirs (Chardarya, Toktogul, Karakul...).
Monitoring the lake and reservoirs level in these re-
gions also helps to study the impact of climate change
Fig. 7. Reservoir level variations from Altimetry on multi-satellite on the lakes and reservoirs, though interpretation is
mode (TOPEX/Poseidon, GFO, Envisat & Jason) versus in situ gauges
complex. Lake levels may drop due to potential in-
for Toktogul (Kirgizstan) (from http://www.legos.obs-mip.fr/soa/
hydrologie/hydroweb/ and http://water.freenet.uz). (For interpretation creases in evaporation, but levels may increase due to
of the references to colour in this figure text, the reader is referred to the increase in thaw rates of the surrounding glacier and
the web version of this article.) mountain snow.
Fig. 7. Variations du niveau du réservoir Toktogul (Kirgyzstan) par The recent increase of the Lake Issykkul level
altimétrie en mode multi-satellite (TOPEX/Poseidon, GFO, Envisat (Fig. 5) with a rate of around 10 cm yr−1 has been con-
et Jason), comparées aux mesures in situ (issues des sites http://www. firmed by both in-situ and altimetry data. This abrupt
legos.obs-mip.fr/soa/hydrologie/hydroweb/ et http://water.freenet.uz).
change of level variation could be the consequence of
recent higher glacier melting [41].
Studies made by Small et al. [46] showed that global
warming has also accelerated the shrinking of the Aral
Sea by around 100 mm yr−1 from 1960 to 1990. With
altimetry monitoring, it would be possible to quantify
the impact of global warming on the Aral Sea during
the last 15 years.

3.2. Caspian Sea

The Caspian Sea is the largest inland water body of

Fig. 8. Reservoir level variations from altimetry on multi-satellite
mode (TOPEX/Poseidon, GFO, Envisat & Jason) for Sarykamish the world. With a mean salinity of ∼13 g l−1 , it is lo-
(Turkmenistan) from http://www.legos.obs-mip.fr/soa/hydrologie/ cated in a depression bordered by the Caucasian Moun-
hydroweb/. tains to the west, the Central Asian plateau and desert to
Fig. 8. Variations de niveaux du réservoir Sarykamish (Turkménis- the east, the Russian and Kazak plains in the north, and
tan) par altimétrie en mode multi-satellite (TOPEX/Poseidon, GFO, by the Elbruz Mountains to the south. The total basin
Envisat et Jason), issues du site http://www.legos.obs-mip.fr/soa/ catchments cover more than 3 billion square kilome-
hydrologie/hydroweb/.
tres, and the sea is fed by numerous rivers including the
Volga, Ural, Terek, and Kura rivers. The sea can be con-
ing of Big Aral as the water intake from Amu Darya to sidered as being comprised of three parts, the northern
fill in the Sarykamish Lake was initially feeding the Big Caspian, which is a shallow area with a mean depth of
Aral. 5 m, the central Caspian, with a mean depth of 190 m,
and the southern Caspian, with a mean depth of 350 m.
3.1.3. Discussion The South Caspian is deepest at 1025 m [44].
The results shown here (taken from http://www.legos. The Caspian Sea has been characterised by cyclic
obs-mip.fr/soa/hydrologie/hydroweb/ for altimetry and and high-amplitude water-level variations over histori-
http://water.freenet.uz/ for in-situ data of Central Asian cal time scales. For the last 2000 years, the sea level
reservoirs) highlight the fact that the altimetric accu- fluctuated around 15 m, but it has declined until ∼7 m
racy is not homogeneous, but depends on the quantity over the last five centuries. Fig. 8 shows more recent
and quality of the satellite data available on each lake or level variations; note the minimum in 1979 and the sub-
reservoir. One can see that altimetry is a good comple- sequent rise, with corresponding surface area variations
mentary source of information to the ground-based net- between 360 000 to 400 000 km2 [44]. They have been
1108 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112

Fig. 9. Caspian Sea Level from in-situ data.

Fig. 10. Caspian Sea level variation from altimetry (grey circles) and
Fig. 9. Niveau de la mer Caspienne à partir de données in situ.
from in-situ measurements (black square).
Fig. 10. Variations de niveaux de la mer Caspienne à partir de l’alti-
correlated with height oscillations between −26 and métrie (cercles gris) et à partir de mesures in situ (carrés noirs).
−29 m (with respect to the zero ocean level) during the
last hundred years.
Coastal and delta areas are strongly affected by these
considerable sea-level changes with subsequent damage
on infrastructure and recurrent ecological disasters. The
events have led to many investigations of the links be-
tween level variability and potential impacts [30]. The
Caspian Sea has also provided a unique example to test
global sea-level rise models and the possible impacts on
coastal zones over a very short time of few dozen of
years [30].
Adaptation of human activities, efficient use of wa-
ter resources, sustainable development and ecosystem
protection fundamentally depend on the variations of Fig. 11. Caspian Sea level variations deduced from altimetry measure-
ments.
the Caspian Sea’s water level. Water balance investiga-
tions can utilise more than a hundred years of ground- Fig. 11. Variations de niveau de la mer Caspienne, déduites des me-
sures altimétriques.
based data based on the network of hydrometeorological
stations homogeneously scattered along the coast and
Islands of the Caspian Sea. Averaged data from four sta- level variations at a time when gauge data is sparse.
tions (Baku, Makhashkala, port Shevchenko, and Kras- Comparisons of recent altimetry measurement and in
novodsk) provide an ‘official’ Caspian Sea Level (CSL) situ gauges measurements are shown in Fig. 10 for
(Fig. 9), from which uncertainty in the water balance the period 1993 to 2001. This figure demonstrates
(mainly the evaporation rate) can be explored with a that the general trend of both time series are corre-
view to predicting the future evolution of the CSL [29, lated but some discrepancies, in particular the ∼7-cm
44]. level in 1995, can lead to a ∼30-km3 difference in
A detailed description of the water balance com- water volume. Fig. 11 shows the Caspian Sea level
ponents can be found in Shiklomanov et al. [44]. from radar altimetry up to 2005: http://www.legos.obs-
The Volga River provides more than 80% of the to- mip.fr/soa/hydrologie/hydroweb/. Note the 60-cm de-
tal inflow (400 km3 yr−1 ), and dictates the interan- crease observed during the seven years following the
nual variability of the Caspian Sea. Underground wa- peak in 1995, and ongoing rise after 2002. Seasonal
ter inflow is about 3–5 km3 yr−1 and the precipitation variations of the sea level (∼20 cm) are mainly driven
rate is ∼250 mm yr−1 , with high temporal variabil- by changes in river runoff, evaporation, and precipita-
ity. Evaporation represents the main outflow component tion.
(∼730 mm yr−1 ), without high temporal variability.
Discharge to the Kara Bogaz bay in the east has var- 3.3. Lake Chad, Africa – Seasonal prediction and
ied between 0 to 30 km3 yr−1 and depends on the level regional vulnerability
difference between the Caspian Sea and the bay.
Recent radar altimetry measurements have helped Lake Chad lies at an altitude of ∼282 m above
to explore the water balance of the Caspian providing sea level in the Sahel region of Central Africa. The
 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112 1109

catchment area is immense, covering an area of 2.5 × as one means of coping with the current situation. This
106 km2 . The climate in the vicinity of the lake is practice utilises the moist soil bed revealed by the re-
semi-arid in the south and arid in the north, with a dry treating floodwater to harvest a number of differing
season occurring between October and April. A short crops. As the waters rise, fishing replaces farming until
wet season (typically 550 mm yr−1 in the south and the waters once again decline, and the harvesting cycle
250 mm yr−1 in the north) occurs between May and commences [42]. Recent studies [8,43] have discussed
August. Approximately 90% of the lake’s water stems the plight of Lake Chad and have shown how remote
from the Chari/Logone river systems. These rivers have sensing technology can monitor the ever-changing in-
their origins in Cameroon and the Central African Re- undation conditions. The areal extent and duration of
public, where there are more humid zones with rain- the floodwaters were observed via image analysis using
fall from May to September. The remaining 10% stems the NOAA/AVHRR LAC dataset. The variation of the
from local precipitation, and from the El Beid and Ko- water level, within the permanent lake waters, the in-
madougou Yobé Rivers [17,18,28]. The lake is classed flowing rivers, and the surrounding marsh lands were
as ‘closed’, referring to zero water outflow, although observed over a 10-year period using data from the
[28] reports losses as high as 21% via ground water TOPEX/Poseidon mission (1992–2002). With an ac-
seepage, which occurs in the Southwest of the lake curacy 10 cm rms, the altimetry revealed seasonal
basin. With low humidity, high temperatures, and shal- water-level variations of the order of 0.5–6 m and there
low lake depths (mean ∼5 m), ∼80% of the lake’s water was a notable rise in minimum levels, 15–35 cm yr−1
is lost through evaporation [40], with associated annual within the lake basin (Fig. 13). Peak-level phase lags
water-level variations of ∼1–3 m and large fluctuations of 4–5 months were also observed between the head-
of the surface area [25,26]. waters of the Chari/Logone Rivers and the western lake
During the last 40 years, regional precipitation in the marshes. Image results reveal that the permanent lake
region has been reportedly low and there have been no- area is ∼1385 km2 , and that with an accuracy of ∼5%
table droughts between 1968 to 1974 and 1983 to 1985. the surrounding regions have been additionally flood-
This led to the drilling of many uncapped boreholes ing by up to ∼3600 km2 yr−1 (Fig. 12). Despite current
and a large-scale irrigation system was constructed in basin complexities, the synergistic area/level measure-
the south. Both schemes led to high evaporation losses ments corroborate the known hydraulic relationship at
and the irrigation system further failed due to the non- low water levels. Both measurement sets reflect the poor
implementation of contingency plans during the drought inundation years of 1993–1994 and 1997–1998 despite
periods [27]. The reduction in lake surface area, from the indication of a general recovery in regional precipi-
∼25 000 km2 in 1960 to ∼2500 km2 in the mid-1980s tation to the Southeast of the lake.
[21] clearly demonstrates the magnitude of the com- Coe and Birkett [12] developed the work of Birkett
bined effects. There has thus been a slow desiccation [8] by using a combination of gauge and radar altime-
of the lake from the north to the south, and exposed try data, not only to deduce inflowing river discharge,
ancient dunes now dominate the northern and eastern but also to predict in advance the downstream discharge
boundaries of the original coastline. The only remain- and lake level. In particular, the prediction of the wet
ing area of permanent open water exists in the south- season water height on the lake and marshes was of
ern region of the lake basin, where it is surrounded particular interest because the seasonal maximum level
by densely vegetated swamps and smaller pools. Each determines the success of the farming and fishing peo-
year, the Chari/Logone waters flow into the lake basin ple on the southern shore of the basin [43]. The tech-
from the south and gradually move northwards, supply- nique used upstream measurements of surface water
ing the permanent open water and seasonally inundating height from the TOPEX/Poseidon radar altimeter, cal-
the marsh regions. ibrated through simple empirical regression techniques
The diminishment of Lake Chad has had adverse ef- with ground-based gauge height and discharge data, to
fects on both food and water resources. The local pop- estimate the mean monthly river discharge of the Chari
ulation has learned to adapt to the reduced seasonal River at N’Djamena, Chad (about 500 km downstream
inundation of the lake basin, but is vulnerable to both of the altimetric measurements). It also aimed to pre-
droughts and excessive inflows. The lake however re- dict the wet season height of the surface waters within
mains an important source of water and protein (fish) the Lake Chad basin (more than 600 km downstream).
for both the sedentary and nomadic local people. With In the particular case of the Lake Chad Basin, a cal-
a current coastal population of ∼0.75 million, the in- culation of water height directly from satellite height
troduction of flood retreat farming has been employed was more instructive than a general analysis of volume
1110 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112

Fig. 14. Comparison of observed versus predicted water height for

the open lake region. The observed (local) height is obtained directly
from the T/P radar altimeter observations. The predicted height (with
39-day phase lag) is obtained via observation and correlation of water
heights at the Chari location (from [12]).
Fig. 14. Comparaison des hauteurs prédites et observées pour les eaux
permanentes du lac. Les données observées (locales) de hauteur sont
obtenues directement par les observations du radar altimètre de TO-
PEX/Poseidon. Les hauteurs prédites (déphasées de 39 jours) sont
Fig. 12. AVHRR LAC channel-2 noontime images (∼1 km per pixel)
obtenues par corrélation avec les observations faites à Chari (issues
of Lake Chad. The permanent lake waters can be clearly seen in the
de [12]).
centre of the southern basin. Inundated marsh, revealed by dark grey
hues, can be seen on the January, 1996 image marking the dry (but
high Chari River inflow) season.
ence could be used to estimate river discharge about
Fig. 12. Image AVHRR (canal 2, résolution de 1 km par pixel) à
mi-journée du lac Tchad. Les eaux permanentes du lac sont clairement 500 km downstream at N’Djamena and do so 10-day
identifiées dans la partie centrale du bassin sud. Les zones inondées in advance (r 2 = 0.9611). Via simple linear correla-
en marge (indiquées en grisé) sont visibles sur l’image prise en jan- tion methods, the stage measurements were also used
vier 1996, qui correspond à la saison sèche, mais avec un fort débit de to estimate the height of the permanent waters of the
la rivière Chari.
lake (600-km downstream) 39 days in advance (r 2 =
0.9297, Fig. 14). Predicting the water height on the
western marshes of the lake-bed though was poorer
(r 2 = 0.7958), due to a change in response time of the
local stage to the seasonal floods. This change may be
due to a variation in the hydraulic connectivity of the
marshes with the open lake, which coincides with an
observed increase in mean water level in the latter half
of the 1990s.
Fig. 13. Time series of altimetric lake level variations for the perma- The excellent river discharge and lake water predic-
nent waters of Lake Chad. Data is taken from the TOPEX/Poseidon tions show that the altimetry is an additional useful tool
mission. Time axis is 4-month intervals. where rapid water resource assessment is desirable. In
Fig. 13. Séries temporelles des variations de niveau pour la partie comparison to limited available gauge data, the pre-
permanente du lac Tchad. Les données sont issues de la mission diction study found the satellite observations advanta-
TOPEX/Poseidon. L’axe des abscisses représente le temps avec une geous. Not only the remote sensing data are received
période de 4 mois.
continuously, but they are also potentially available to
the user within a few days after measurement. This
change, because the local population of the lake basin accessibility together with spatial coverage across the
is more dependent on lake level fluctuations for their basin (i.e. the favourable time-lag between upstream
livelihood and because the conversion of lake basin vol- and downstream points) provides an opportunity to pre-
ume changes into water height changes is extremely dict, as much as 1–2 months in advance, the downstream
difficult due to the enormous seasonal fluctuations in the water resources. Such a general development technique
complex marsh area. for water resource quantification and prediction from
Coe and Birkett [12] found that the altimetric stage satellite altimetric data may be potentially useful in
measurements, upstream at the Chari/Ohuam conflu- other regions of the globe.
 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112 1111

4. Conclusions Victoria and Tanganyika, Int. J. Climatol. 24 (13) (2004) 1613–

1624.
Both lake elevation and storage volume are impor- [7] S. Birkett, Contribution of TOPEX/Poseidon to the global moni-
toring of climatically sensitive lakes, J. Geophys. Res. 100 (C12)
tant parameters for routine monitoring and for study- (1995) 25179–25204.
ing water balance variations across the catchment basin. [8] C.M. Birkett, Synergistic remote sensing of lake Chad: variabil-
Applications can be interdisciplinary, ranging from haz- ity of basin inundation, Remote Sens. Environ. 72 (2000) 218–
ard prediction to environmental and social concerns. 236.
Threading through all themes is the influence, control [9] V.N. Bortnik, Alteration of water level and salinity of the Aral
and effect of climate variability. Radar altimetry can Sea, in: Creeping Environmental Problems and Sustainable De-
velopment in the Aral Sea Basin, Cambridge University Press,
contribute to the measurement of lake height variabil- 1999, pp. 47–65.
ity over the lifetime of the satellite. The measurements [10] S. Calmant, F. Seyler, A. Cazenave, F. Frappart, Amazon River
can be acquired along the position of the satellite ground Stages by ENVISAT vs. Other Satellite Altimeters, Poster Ses-
track with accuracy of the order of a few centimetres sion 4P12–02 ‘Lake Levels’, Abstract 671, 2004 ERS and EN-
to tens of centimetres rms. Such measurements can be VISAT Symposium, Salzburg, Austria, 2004.
[11] A. Cazenave, P. Bonnefond, K. Dominh, Caspian Sea level from
combined with bathymetry and imaging-derived area
TOPEX/Poseidon altimetry: level now falling, Geophys. Res.
extent to further deduce changes in storage volume. Lett. 24 (1997) 881–884.
Case studies of the Aral Sea have shown that estimates [12] M.T. Coe, C.M. Birkett, Water Resources in the Lake Chad
of the water mass balance are accurate to ∼3 km3 yr−1 , Basin: Prediction of river discharge and lake height from
which contributed to solve the issue of underground wa- satellite radar altimetry, Water Resour. Res. 40 (10) (2005),
ter flow. For remote lakes, altimetry can be the unique doi:10.1029/2003WR002543.
[13] A. Coudrain, B. Francou, Z.W. Kundzewicz, Glacier shrinkage
system to determine an accurate water balance. in the Andes and consequences for water resources, Hydrol.
There are sampling limitations to the technique, but Sci. J. 50 (6) (2005) 925–932.
combining datasets from different altimetry missions [14] J.-F. Crétaux, A.V. Kouraev, F. Papa, M. Bergé Nguyen, A. Ca-
provides higher temporal and spatial resolution. Com- zenave, N.V. Aladin, I.S. Plotnikov, Water balance of the Big
pilation of data from the Jason-1 (10-day repeat), GFO Aral Sea from satellite remote sensing and in situ observations,
J. Great Lakes Res. 31 (4) (2005).
(17-day) and ENVISAT (35-day) missions can provide
[15] P.J. Depetris, A.I. Pasquini, The hydrological signal of the Perito
a repeat measurement at a frequency better than a sin- Moreno Glacier damming of Lake Argentino (Southern Andean
gle mission alone. While Jason-1 may only view ∼300 Patagonia): the connection to climate anomalies, Global Planet.
lakes, ENVISAT gains an additional ∼700 targets. Fu- Change 26 (4) (2000) 367–374.
ture satellite missions based on a combination of altime- [16] M. De Wit, J. Stankiewicz, Changes in surface water supply
try and interferometry (see [3] for discussions) could across africa with predicted climate change, Science 311 (5769)
(2006) 1917–1921.
provide a promising future for true global lake studies
[17] Famine Early Warning System (FEWS), Lake Chad untapped
with height and width observation of all targets with potential, FEWS Spec. Rep. 97-4, 1997.
centimetre accuracy. [18] Famine Early Warning System (FEWS), Lake Chad polders,
FEWS Bull. AFR98-3, 1998.
References [19] L.L. Fu, A. Cazenave, Satellite Altimetry and Earth Science,
A Handbook of Techniques and Applications, International Geo-
[1] N.V. Aladin, J.-F. Crétaux, I.S. Plotnikov, A.V. Kouraev, A.O. physics Series, vol. 69, Academic Press, 2001.
Smurov, A. Cazenave, A.N. Egorov, F. Papa, Modern hydro- [20] V. Gornitz, Impoundment, groundwater mining, and other hydro-
biological state of the Small Aral Sea, Environmetric 16 (2005) logical transformations: impacts on global sea level rise, in: B.C.
375–392. Douglas, M.S. Kearney, S.P. Leatherman (Eds.), Sea Level Rise,
[2] D.E. Alsdorf, D.P. Lettenmaier, Tracking fresh water from space, History and Consequences, Academic Press, San Diego, 2001,
Science 301 (2003) 1491–1494. pp. 97–119.
[3] D. Alsdorf, D. Lettenmaier, C. Vörösmarty, et al., The need for [21] A.T. Grove, African river discharges and lake levels in the twen-
global, satellite-based observations of terrestrial surface waters, tieth century, in: T.C. Johnson, E. Odada (Eds.), The Limnology,
EOS Trans. 84 (29) (2003) 269–276. Climatology and Palaeoclimatology of the East African Lakes,
[4] E.P. Argyilan, S.L. Forman, Lake-level response to seasonal cli- Proc. IDEAL Conference, Jinja, Uganda, Gordon and Breach,
matic variability in the Lake Michigan–Huron system from 1920 Amsterdam, 1996, pp. 95–100.
to 1995, J. Great Lake Res. 29 (3) (2003) 488–500. [22] D.L. Harding, M.F. Jasinski, ICESat Observations of inland sur-
[5] K. Arpe, L. Bengtsson, G.S. Golitsyn, I.I. Mokhov, V.A. Se- face water stage, slope, and extent: A new method for hydrolog-
menov, P.V. Sporyshev, Connection between Caspian Sea level ical monitoring, Fall AGU, 2004, C21B-05 (abstract).
variability and ENSO, Geophys. Res. Lett. 27 (17) (2000) 2693– [23] M.A. Hofton, J.B. Blair, J.B. Minster, J.R. Ridgway, N.P.
2696. Williams, J.L. Bufton, D.L. Rabine, An airborne scanning laser
[6] L. Bergonzini, Y. Richard, L. Petit, P. Camberlin, Zonal circula- altimetry survey of Long Valley, California, Int. J. Remote
tions over the Indian and Pacific Oceans and the level of Lakes Sens. 21 (12) (2000) 2413–2437.
1112 J.-F. Crétaux, C. Birkett / C. R. Geoscience 338 (2006) 1098–1112

[24] S.W. Hostetler, Hydrological and Thermal Response of Lakes to East African great lakes: limnology, paleolimnology and biodi-
Climate: Description and Modelling, Physics and Chemistry of versity, Advance in Global Change Research, vol. 12, Kluwer
Lakes, Springer-Verlag, 1995. Academic Publishers, 2002.
[25] C.F. Hutchinson, P. Warshall, E.J. Arnould, J. Kindler, Develop- [38] W.H. Nowlin, J.M. Davies, R.N. Nordin, A. Mazumder, Ef-
ment in Arid Lands, Environment 34 (1992) 16–43. fects of water level fluctuation and short-term climate varia-
[26] ILEC, International Lake Environment Committee Survey of the tion on thermal and stratification regimes of a British Columbia
State of the World’s Lakes, Data Book of the World Lake Envi- reservoir and Lake, Lake Reservoir Manage. 20 (2) (2004) 91–
ronments, Otsu, Japan, 1988, AFR-2, 1, 1–7. 109.
[27] S.A. Isiorho, J. Njock-Libii, Sustainable water resources man- [39] A. Rimmer, M. Boger, Y. Aota, M. Kumagai, A lake as a nat-
agement practice, in: South Chad, Proc. Eco-Informa ’96, Lake ural integrator of linear processes: Application to Lake Kinneret
Buena Vista Florida, Global Networks for Environmental Infor- (Israel) and Lake Biwa (Japan), J. Hydrol. 319 (1–4) (2006)
mation 11 (1996) 855–860. 163–175.
[28] S.A. Isiorho, G. Matisoff, K.S. When, Seepage relationships be- [40] M.A. Roche, Tacage hydrochimique naturel du mouvement des
tween Lake Chad and the Chad aquifer, Ground Water 34 (1996) eaux dans le lac Tchad, in : Hydrology of Lakes Symposium,
819–826. Helsinki, IAHS-AISH 109, 1973, pp. 18–27.
[29] A.N. Kosarev, E.A. Yablonskaya, The Caspian Sea, SPB Acad- [41] V.V. Romanovski, Water level variations and water balance of
emic Publishing, The Hague, 1994. Lake Issykkul, in: Lake Issyk-kul: Its Natural Environment,
[30] S.B. Kroonenberg, E.N. Badyukova, J.E.A. Storms, E.I. Ignatov, Kluwer Academic Publisher, 2002, pp. 45–57.
N.K. Kasimov, A full sea level cycle in 65 years: barrier dynam- [42] M.-T. Sarch, Fishing and farming at Lake Chad: Overcapital-
ics along Caspian Sea shores, Sediment. Geol. 134 (3–4) (2000) ization, opportunities and fisheries management, J. Environ.
257–274. Manage. 48 (1996) 305–320.
[31] F.G. Lemoine, S.C. Kenyon, J.K. Factor, R.G. Trimmer, N.K. [43] M.T. Sarch, C.M. Birkett, Fishing and farming at Lake Chad:
Pavlis, D.S. Chinn, C.M. Cox, S.M. Klosko, S.B. Luthcke, M.H. Responses to lake level fluctuations, Geogr. J. 166 (2) (2000)
Torrence, Y.M. Wang, R.G. Williamson, E.C. Pavlis, R.H. Rapp, 156–172.
T.R. Olson, The Development of the Joint NASA GSFC and the [44] I.A. Shiklomanov, V.Yu. Georgievsky, Z.D. Kopaliani, Water
National Imagery and Mapping Agency (NIMA) Geopotential balance of the Caspian Sea and reasons of water level rise in the
Model EGM96, NASA tech. publ. 1998-206861, Greenbelt, MD, Caspian Sea, in: IOC Workshop report No. 108, Paris, 1995, pp.
USA, 1998. 1–28.
[32] B.M. Lofgren, F.H. Quinn, A.H. Clites, R.A. Assel, A.J. Eber- [45] C.K. Shum, Y. Yi, K. Cheng, C. Kino, A. Braun, S. Calmant,
hardt, C.L. Luukkonen, Evaluation of potential impacts on Great D. Chambers, Calibration of Jason-1 altimeter over Lake Erie,
Lakes water resources based on climate scenarios of two GCMs, Mar. Geod. 26 (3–4) (2003).
J. Great Lakes Res. 28 (4) (2002) 537–554. [46] E.E. Small, F. Giorgi, L.C. Sloan, S. Hostetler, The effects of
[33] I.M. Mason, M.A.J. Guzkowska, C.G. Rapley, F.A. Street-Perrot, desiccation and climate change on the hydrology of the Aral
The response of lake levels and areas to climate change, Clim. Sea, J. Clim. 14 (2001) 300–322.
Change 27 (1994) 161–197. [47] L.C. Smith, Y. Sheng, G.M. MacDonald, L.D. Hinzman,
[34] F. Mercier, A. Cazenave, C. Maheu, Interannual lake level fluc- Disappearing Arctic Lakes, Science 308 (5727) (2005) 1429.
tuations (1993–1999) in Africa from TOPEX/Poseidon: Con- [48] C. Vörösmarty, A. Askew, W. Grabs, R.G. Barry, et al., Global
nection with ocean–atmosphere interactions over Indian Ocean, water data: A newly endangered species, EOS Trans. 82 (5)
Global Planet. Change 32 (2002) 141–163. (2001) 54–58.
[35] C.S. Morris, S.K. Gill, Evaluation of the TOPEX/Poseidon al- [49] T.C. Winter, Hydrological processes and the water budget of
timeter system over the Great lakes, J. Geophys. Res. 99 (C12) lakes, Physics and Chemistry of Lakes, Springer-Verlag, 1995.
(1994) 24527–24539. [50] C.Y. Xu, E. Widen, S. Halldin, Modelling hydrological con-
[36] M. New, D. Lister, M. Hulme, I. Makin, A high resolution data sequences of Climate change – Progress and challenges, Adv.
set of surface climate over global land areas, Clim. Res. 21 Atmos. Sci. 22 (6) (2005) 789–797.
(2002) 1–25. [51] R.P. Zola, L. Bengtsson, Long-term and extreme water level
[37] S.E. Nicholson, X. Yin, Mesoscale patterns of rainfall, cloudi- variations of the shallow Lake Poopo Bolivia, Hydrol. Sci.
ness and evaporation over the Great lakes of East Africa, The J. 51 (1) (2006) 98–114.