Forum on the Caspian, Aral and Dead Seas-Perspective of Water Environmental Management and Politics


About UNEP	UNEP offices	News centre	Publications	Calendar	Awards	Milestones	UNEP Store

United Nations Environment Programme
Division of Technology, Industry and Economics

Newsletter and Technical Publications

A Straregy for Esimation of Water Budget in the Aral Sea Basin by Atmospheric Water Balance

Nobuhito Ohte, (Kyoto University, Department of Forestry),
Taikan Oki, (University of Tokyo, Institute of Industrial Science),
Yukihiro Morimoto, (Osaka Prefecture University, Department of Regional Environmental Science)

The water budget of the Amudarya-Syrdarya Aral Sea basin during the period from 1989 to 1992 were estimated by the atmospheric water balance method using the ECMWF (European Centre for Medium-Range Weather Forecasts) global objective analyzed data set and the precipitation data. When the atmospheric water balance can be evaluated successfully, the period mean-,_H!Q-dW/dt value, which is the net water vapor flux convergence, can be interpreted as precipitation minus evapotranspiration (P-E). Moreover, In the cases of the Aral Sea basin, when the groundwater runoff from these areas can be assumed to be negligible, the P-E values can also be interpreted as changes of basin water storage because there are no outflow rivers. The annual P-E values were estimated as nagative at for most of the calculation grids within the Amudarya and Syrdarya basins. The basin average monthly P-E valued were nagative during the irrigation period, i.e., from April to October. During the irrigation period. The annual basin deficit was estimated larger than the decrease rate of the Aral Sea water in last decade. This suggests that the decrease rate of the Aral Sea water in last decade. This suggests that the decrease in storage may have been occurring not only at the Aral Sea surface, but also at the other part of the Amudarya and Syrbarya basin, e.g., soil moisture and groundwater storage. The clustering procedure using the multi-temporal NDVI images could successfully provide appropriate land cover classification. By this land cover information, the spatial distribution of estimated evapotranspiration was verified.

Introduction

It is now widely known that huge and fairly extensive irrigation fields that stretch along Amudarya and Syrdarya has been pruducing large water deficits due to evapotranspiration, and this has led to shrinking of the Aral Sea during past thirty years. The desiccation of the Aral Sea has been causing not only water resources problems, but also various environmental problems such as social or political issues [Micklin, 1988]. The accurate estimation of the water budget, especially water consumption by evapotranspiration, is fairy important in formulating effective measures against the water resource problems in the Aral Sea basin and the Lake Balkhash basin, and also in formulating an effective plan for sustainable agricultural development in other arid regions. As a typical case of large scale agricultural development in the central Asian arid region, the purpose of this study is to develop the effective procedure for estimating the water budget of the Aral Sea and the Lake Balkhash region.

As the first step of this study, the large scale estimation of evapotranspiration in the Amudarya-Sydarya-Aral Sea basin (hereafter called the ASA basin) has been carried out using the atmospheric water balance method. The basin water deficit was comparatively examined with the decrease of water volume of the Aral Sea. Moreover, to examine the correspondence between the terrestrial surface condition and the spatial distribution of evapotranspiration, the land cover map was made using NDVI (Normalized Differrence Vegetation Index) images prepared from NOAA/GVI date sets.

Methods

Atmospheric Water Balance Method

This method is used to estimate terrestrial water balance in a large scale region from the information of the vapor flux convergence in the atmosphere. Its concept was firstly applied to estimate the terrestial water budget by Starr and Peixoto [1958]. Recently the "Object analyses data" is established for giving an initial value for numerical daily weather forecasting, using GCM and various observation data. Utilization of this kind of data set can overcome a difficulty to estimate the vapor flux that is caused by less atmospheric physical information. Moreover, by using the object analyses data, the water balance studies on a global scale has been available in recent five years [Oki et al., 1995].

The atmospheric water balance is described as:

Water vapor flux convergence is calculated using the central difference method to calculate the following equation, assuming that the earth is a sphere which has a radius of R_e:

zusiki(2)

Since it seems generally reasonable to state that water content of solid and liquid phases in the atmosphere is negligibly small, the equation (1) can be simplified as:

The mean evapotranspration of the arbitrarily defined area can be estimated if the corresponding precipitation data is available. Namely, the evapotranspiration can be computed as:

If both atmospheric and precipitaton data are available over a short time span such as monthly or daily, evapotranspiration can be estimated on corresponding time scale. Moreover, the estimate region of evapotranspiration is not required to be a closed basin, and the region depends just on the spatial scale of the atmospheric and precipitation data. It is resonable to suppose that the flexibile nature of estimation scale is the advantage of the atmospheric water balance method to the traditional river basin water balance.

On the other hand, the water balance of the basin system is written as:

where S and R_o are basin storage and runoff, respectively. S includes not only soil moisture and groundwater storage, but also snow accumulation. R_o is defined as the net outflow rate from the basin which is considered. Namely, R_o is calculated as the outflow minus inflow from surrounding area. From the equation (3) and (5), the atmosphere-basin water balance equation can be obtained.

zusiki(6)

Study Area

The study area is shown in figure 1. The area stretches from the west coast of Aral Sea to the east tip of Lake Balkhash (55°E-80°E), and from the south end of Turkmenistan to the central part of a Kazakhstasn Plateau (35 °N-50°N).

"Desiccation of the Aral Sea " [Micklin, 1988] means nothing but the decrease of the basin storage. Therefore, the decrease of lake water, which has been widely published, can be examined by the atmospheric water balance.

To discuss the water budget and hydrologic cycle of the ASA basin, the basin area was defined as the group of 2.5(2.5 degree grids, which was necessary to match to the atomospheric data commented below (figure 1).

Atomospheric and Precipitation Data

The data used for the analysis of the atmospheric water balance is the objective analysis data set compiled by the European Centre for Medium-Range Weather Forecasts (ECMWF). It is made through the 4-dimensional data assimilation (4DDA) system. The data set consists of the variables, geopotencial height, wind vector, temperature, and relative humidity. These variables locate at each 2.5-degree grid points convering the globe by 144(73. The grid network has seven layers at 1000, 850, 700, 500, 300, 200 and 100 hPa height. Since 1985 the data set has been compiled twice a day (0000 GMT and 1200 GMT).

The climatological precipitation data was used for this study because of less information from the surface station within the study area, especially in Iran, Afghanistan, and Uzbekistan. The used data set was compiled by Legates and Willmott [1990]. The data set consists of gauge-corrected average monthly precipitation with its long term average values from 1920 to 1980. The data set was recompiled into 2.5-degree resolution to match the atmospheric data.

Classification of Land Cover Using the NOAA/ GVI Data

Land cover map of the study area was made from the multi-temporal NDVI (Normalized Difference Vegetation Index) images. The images were established from GVI (Global Vegetation Index) weekly composite data sets distributed by NOAA.

GVI data is composed by spatially sampling of AVHRR GAC (Grobal Area Coverage) dara. To avoid cloud cover on an image, seven-day maximum vegetation index composite are produced from the daily GAC data. The standard base projection is Plate Carreé. Resolution in the mapped data is 16 km at the equator [Kidwell, 1990]. First, hierarchical clustering to make ten categories was done focusing seasonal NDVI variation using 35 images composed for the period from March 4 to October 30, 1988. Secondly, these ten categories were grouped into five types of land cover based on the ground truth data prepared by Morimoto and Ogari [1995]. A similar technique has been used by Honda and Murai [1989] to compile the global vegetation and land use maps prepared by the Kazakhstan Academy of Science as powerful ground truths.

Results

Net Vapor Flux Convergence

The spatial distribution of 4-year (1989-1992) average annual net vapor flux convergence (=P-E) are shown in figure 2. In the disributions of all individual years, a fairly similar pattern was found. The areal and period (1989-1992) average annual net water vapor flux convergence was calculated as -206.8 mm.

Figure 3 shows the classification of the regions on which the deficit and surplus were evaluated. On almost all grids wihtin the ASA basin, the water deficit was evaluated. The net convergence on the grids corresponding to the Lake Balkhash basin was not smaller than those of the ASA basin. On the region close to the Pamir and Tien Shan mountainous area, including snow, compared to the low and flat plain that stretches over Kazakhstan, Uzbekistan, and Turkmenistan. It also suggests that the surplus generated in the mountainous region are consumed in the lower and flatter part of the ASA basin.

Average Evapotranspiration

To calculate evapotranspiration rate with net vapor flux convergence, a precipitation data set has to be properly prepared to match the resolution. Figure 4 shows the spatial distribution of annual precipitation of the study area which was prepared from the data by Legates and Willmot [1990]. There are several grids having over 400 mm annual precipitation along the Tien Shan Mountains. Besides these regions, namely, in the low and flat plains that mostly consists of the ASA basin, the annual precipitation ranges between 150 and 250 mm, and the significant spatial variation cannot be found. The areal average annual precipitaton of the study area is 256.2 mm.

Calculation of the evapotranspiration was done using the 1989-1992 average net convergence and the precipitation data. The distribution of the calculated evapotranspiration was shown in figure 5. A similar pattern of the net convergence can be found also in the distribution of evapotranspiration. It can be found that the estimated annual evapotranspiration in the ASA basin was quite high, and that of the mountainous region around Tien Shan is much lower than in the low region. The areal average of annual evapotranspiration is 463.0 mm.

Land Cover Classification Using NOAA/GVI Data

Ten categories of land covers were objectively classified using 35 multi-temporal NDVI images using a clustering technique. These ten categories were then identified by what type of surface conditions they were, and grouped into next five types of land covers using the ground truth information: 1) Water surface, 2) Desert, 3) Irrigated land, 4) Steppe (vegetation is active mainly in spring), 5) Forest.

The classified land cover image is presented as figure 6. The classified irrigated area has a nice match with a same kind of map presented by Micklin [1988]. The ratio of each land cover pixel in the study area is shown in table 1. Although the pixel ratio is not directly equal to the areal ratio because of the projection method, the values are efficient to know how the irrigated land is large. The irrigated area was estimated larger than the water surface. The value in parenthesis indicates the frequency of the ASA basin. The irrigated land is almost two times as large as the water surface, which is mainly the Aral Sea. It is a fairly reasonable estimation referring to the report by Zhu et al. [1991].

Discussion

Water Budget of the ASA Basin and the Decrease of the Aral Sea Water

First, let us discuss the ASA areal average water budget. The estimated seasonal change of evapotranspiration is shown in figure 7 and compared with that of the forest basin in the Asian warm humid region [Suzuki, 1980]. Although the precipitation amount is dramatically different, during May, June and July, the evapotranspiration rate is the ASA basin represents the same level as that of the humid forest basin. This apparently suggests that irrigated water from the Amudarya and Sydarya is mainly consumed by the evapotranspiration, and the volume of the irrigated water is much larger than the precipitation input.

The annual precipitation and estimated evapotranspiration of the ASA basin in 235.8 mm and 534.5 mm, respectively. Namely, the annual water deficit is 298.7 mm. Figure 8 shows the water volume and sea area decreases of the Aral Sea between 1960 and 1987 estimated by Micklin [1988] using the NOAA/ AVHRR image and the data book compiled by the USSR Academy of Science. This figure is telling that the average annual water loss from the Aral Sea was approximately 800~830 mm. This annual water deficit can be converted to about 40~45 mm in all the ASA basin area, which is much smaller than the estimated basin average deficit above. This suggests that the decrease of the water storage on the ASA basin has been occurring not only at the Aral Sea surface, but also in the other parts of the basin, namely in the soil moisture and the groundwater along the two rivers.

Secondly, in order to discuss in more datail the hydrologic cycle in the ASA basin, the basin is divided into two parts according to the distribution of the estimated net vapor flux convergence in figure 3; the water surplus region and the deficit region. Here, the grids which have over 100 mm water surplus was defined as the surplus region. The seasonal variation of the average water convergence in each region is shown in Fig.9. The annual convergence of the surplus region and the deficit region is 793.5 mm and -578.2 mm, respectively. As we mentioned in figure 3, the grids having the water surplus is mostly located at the mountainous region around Tien Shan and Pamir. The difference in annual convergence between the two regions, thus, obviously suggests that the surplus generated in the mountainous region has been consumed in lower and flatter plains along the Amudarya and Syrdarya. The areal ratio of two regions is approximately estimated as 4:15 from the numbers of grids within each region. Considering the areal ratio, the ratio of the surplus volume to the deficit volume is 793.5(4:578.2(15, which is nearly equal to 1: 2.7. This clearly suggests that the evapotranspiration consumes the surplus from the mountainous region, in addition to it, it causes in significant decrease in the basin storage.

Correspondence Between the Spatial Distribution of Evapotranspiration and Land Cover

Based on the classified land cover in figure 6, the areal ratio of the irrigated land to the vegetation land (irrigated land + steppe) of the surplus region and the deficit region were calculated as 14 % and 41 % respectively. This proves that the irrigated land expanded in the plain region of the ASA basin plays a main role in making the huge water deficit.

Figure 10 represents the relationship between areal ratio of the irrigated land and annual evapotranspiration of each grid within the plain (deficit) region. Except three desert grids and the Aral Sea grid, a positive correlation can be found between the irrigated land ratio and annual transpiration, although the evapotranspiration rate also depends on the ratio of the steppe area. These facts indicate that there is significant correspondence between the estimation of the evapotranspiration by the atmospheric water balance method and the terrestrial condition.

Some contradictions are, however, found at the grids that are mainly covered by desert. For both the atomospheric and precipitation data in such part, the decrease of reliability should be considerable because less density of observation. The improvement of the resolution and reliability of the objective analysis data is required.

Conclusion

The water budget of the Amudarya-Syrdarya-Aral sea basin in recent years was discussed using the atmospheric water balance method. The estimated spatial and temporal distribution of net vapor flux convergence and evapotranspiration leads to the next conclusions by comparative examination with the decrease of the Aral Sea water and the terrestrial information obtained by NOAA/GVI images.

The annual water deficit, that is, the decrease of the basin water storage of the ASA basin was clearly detected by the atmospheric water balance method. The annual basin deficit was estimated larger than the decrease rate of the Aral Sea water in last decade. This suggests that the storage decrease may have been occuring not only at the Aral Sea surface but also at the other parts of the ASA basin, eg., soil moisture and groundwater storage.

The correspondence between the estimated grid average evapotranspiration and its areal ratio of the irrigated land was clearly found in the lower and flatter parts of the ASA basin. This represents that the clustering procedure using the multi-temporal NDVI images could successfully provide appropriate land cover classification, and the spatial distribution of estimated evapotranspiration was varified by the land cover information.

Reference:

Honda, Y. and S. Murai (1989): Vegetation mapping using global vegetation index and weather data, Proc. on the 10th Asian Conference on Remote Sensing, A-2-4-1-A-2-4-6.

Kidwell, K. B. (1990): Global vegetation index user's guide, NOAA, Washington D.C., pp.30.

Legates, D. R. and C. J. Willmott (1990): Mean seasonal and spatial variability in gauge-corrected precipitation, International Journal of Climatology, 10, 117-127.

Micklin, P.P. (1988): Disiccation of the Aral Sea: A Water Management Disaster in the Soviet Union, Science, 241, 1170-1176.

Morimoto, Y. and N. P. Ogari (1994): Vegetation monitoring of arid and semi-arid regions in central Asia using remote sensing -Landscape classification using NOAA/GVI data- (unpublished).

Oki, T., K. Musiake, H. Matsuyama, and K. Masuda (1994): Global atmospheric water balance and runoff from large river basin, Hydrological Processes (on submitting).

Starr, V.B. and J.Peixoto (1958): On the Global Balance of Water Vapor and the Hydrology of Desert, Tellus, 10, 189-194.

Suzuki, M (1980): Evapotranspiration from a small catchment in hilly mountains (I) -Seasonal variation in evapotranspiration, rainfall interception and transpiration-, J.Jap.For.Soc.,62,46-53.

Zhu, Z., P. Raskin and D. Stavisky (1991): Water Development Strategies for the Aral Sea Region, Proc. IWRA 7th World Congress on Water Resources, P.27.

Table 1. Ratio of Pixel Numbers of the Five Types of Land Cover of the Study Area

Land cover type	Ratio
Water	0.026 (0.036)
Desert	0.669 (0.652)
Irrigated land	0.059 (0.088)
Steppe	0.181 (0.161)
Forest	0.064 (0.063)

The value in parenthesis indicates the ASA basin average .

(larger image)

Figure 1: Study Area: The Amudarya-Syrdarya-Aral Sea Basin (enclosed area) as Defined for the Atmospheric Water Balance Analysis

Figure 2: Calculated Annual Net Vapor Flux Convergence

Figure 3: The Classification of the Deficit and Surplus Regions in Net Vapor Flux Convergence

Figure 4: The Spatial Distribution of the Annual Precipitation

Figure 5: The Spatial Distribution of the Annual Evapotranspiration

Fugure 6: The Land Cover Map Made from the Multi-Temporal NDVI Images

(larger image)
Figure 7: The Seasonable Changes of Evapotranspiration and Precipitation Comparing with
Those of the Humid Forest Basin, Kirya, Japan (Suzuki, 1980)