Fernerkundungsbasierte Modellierung und hydrologische Messungen zur Analyse und Bewertung der landwirtschaftlichen Wassernutzung in der Region Khorezm ( Usbekistan )

Abstract

The recently founded states of Middle Asia face serious economical and ecological problems in irrigated agriculture. Thus, the introduction of the Integrated Water Resource Management (IWRM) is one of the major aims of the Interstate Commission for Water Coordination (ICWC) of Middle Asia. Among others, important tasks addressed by the ICWC are to improve the transparency in water distribution, to control the land use and water withdrawals, and to optimize the performance of the extensive irrigation and drainage systems in the arid to semi-arid Aral Sea Basin. In the past years inconsistent and unreliable hydrological data limited the calculation of the water balances for the river basins of the Amu Darya and Syr Darya. Percolation losses in the rivers or canals and unapproved withdrawals caused tremendous uncertainties when accounting for water amounts between and within different irrigation systems. One of the basic needs of the IWRM is to establish a monitoring system of water flows to understand the functioning of the different irrigation and drainage systems. Objective, consistent, and reliable data are prerequisites to achieve this goal. This study focuses on the irrigation and drainage systems of Khorezm, located in the lower Amu Darya Basin. Serious ecological problems such as high groundwater levels, soil salinity, and non-sustainable use of land and water characterize the present situation in this region. The scientific gaols were (1) to generate objective and consistent data to measure agricultural land use and water consumption in irrigated areas of the Khorezm region and (2) to analyze the functioning of the irrigation system to assess the use of land and water. Remote sensing in combination with hydrological measurements and irrigation performance indicators were found suitable to achieve the aims of this study. A method was developed to classify agricultural land use for the entire Khorezm region by temporal segmentation of 8-day 250 m MODIS time series. Daily MODIS 1 km Land Surface Temperature (LST) data were used for modeling seasonal actual evapotranspiration (ETact) of the summer vegetation period. The Surface Energy Balance Algorithm for Land (SEBAL) was slightly modified to account for the coarse spatial resolution of MODIS data and for semi-operational purposes. Furthermore, flow measurements were recorded for 2005 to generate a hydrological data set for balancing. The water balance was achieved by integrating the remotely sensed evapotranspiration. Additionally, widely accepted irrigation performance indicators such as relative evapotranspiration, drainage over inflow ratio, and depleted fraction were calculated on a monthly base to investigate the functioning of the canal network in Khorezm on regional scale. The classification of agricultural land use from time series of MODIS Normalized Difference Vegetation Index (NDVI) comprises two important steps: (1) the generation of an enormous eature space by temporal segmentation of time series to consider the temporal variability within classes; (2) the application of the Recursive Partitioning And Regression Tree (RPART) algorithm to build a classification tree by selecting suitable features and generating threshold values. In Khorezm, cotton, rice, winter wheat, and crop rotations with winter wheat have the highest proportion in the regional crop production. Training and validation samples were extracted from high spatial resolution ASTER images and field data from 2004. The application of RPART on different feature sets enabled stable results and portability with 91% overall accuracy and a Kappa coefficient of 0.9 for both years, 2004 and 2005. MODIS 1 km land products (LST, emissivity, albedo, NDVI, and leaf area index), and meteorological data were combined for modeling ETact. The sensible heat flux was calculated according to the METRIC (Mapping EvapoTranspiration at High Resolution and Internalized Calibration) variant of SEBAL. Aggregated to MODIS 1 km scale, the land use classification was the determining parameter to select hot and cold anchor points needed to model sensible heat fluxes automatically. The probability to find completely dry or wet conditions within a 1 km grid is very low. Thus, classification results, NDVI, and ASCE-EWRI reference evapotranspiration (ETref) were used to adjust the estimations of the vertical temperature gradient at the best fitting anchor points (similar to METRIC). The validation was done by exhaustive plausibility analyses of the input data, the intermediate results, and the final modeling output using crop coefficients and literature references. Measurements of ETact over agricultural sites were not possible during field work. However, SEBAL has left the experimental stage and has been successfully applied and validated for ecozones and climatic regions including semi-arid environments. All input and output data showed plausible values. Cloud free conditions in Khorzem enabled to execute SEBAL more than 130 times within 215 days of each vegetation period in 2004 and 2005. Interpolation and compositing resulted in daily, ten-day, monthly and seasonal products of ETact. Hydrological discharge measurements at 21 measurement sites covered 82 % of the Khorezm irrigation system. The Palvan-Gazavat-Subsystem located in the distant portion of the canal system was studied in more detail. Automated gauge readings and stream-flow measurements applying the velocity-area method were used to measure the discharge throughout the vegetation period in 2005. The maximum errors of flow measurements were approximately 7 %. The results show intensive agricultural land use in entire Khorezm dominated by cotton (58 % of the entire agricultural land, Uzbek state order). Rice (25 %) grows mainly close to the Amu Darya River and in vicinity to numerous lakes in the South. Furthermore, rice is traditionally the dominating land use in Gurlen Rayon. Crop rotations with winter wheat can be found in central Khorezm and in the downstream parts of the irrigation system. 2004 and 2005 depict similar land use patterns with a slight increase of wheat rotations and rice growing in 2005. The results are suitable for regional water managers to analyze seasonal land use patterns of different Water Users Associations (WUAs) rather than for exact statistics due to the heterogeneity within the 250 m MODIS pixels. Between April and October 2004 average values of ETact exceeded 775 mm (2005: 750 mm). Homogeneous 1 km rice (>80 % proportion of rice) pixels achieved 1007 mm (2005: 877 mm). In 2004 and 2005, the modeled ETact over cotton fields was 831 mm and 833 mm, respectively. Spatial and temporal variations of crop water consumption appeared throughout Khorezm. A negative tendency of ETact towards the distal parts of the irrigation system indicated decreasing water availability on the field level. Relative evpotranspiration indicated an oversupply and therefore potentials for water savings in the upstream part. In the downstream part the water demand was not met. For agricultural use, withdrawals of 5.38 km3 were measured in the vegetation period 2005. The values were on average 37% higher than the official data of the ICWC. Within the system boundaries water amounts of 22,782 m3/ha were available for irrigation. Comparisons between subsystems showed regional disparities of withdrawals ranging from 17,000 m3/ha to 30,000 m3/ha. The upstream-downstream gradient of irrigation water supply expected from the remote sensing modeling results could not be found at the regional water distribution level. In comparison with the remote sensing results it can be summarized that water consumption at the field level (MODIS pixel) or WUA level does not reflect the water intake at the upstream distribution nodes. High transportation losses and increasing groundwater levels during the vegetation period were concluded. Monthly water balances and performance indicators highlighted similar results. During the leaching and the main irrigation period in 2005, an increase of soil moisture and groundwater was recorded. The discharge of groundwater followed the irrigation phase in September. However, even in the main irrigation season (July and August), the average drainage over intake ratio is 45% and in the upper part of the irrigation system almost reaches 60%. This concludes a high potential for water saving. Although high discharges in the regional drainage system were found poor drainage systems are reported at the field level. Evidently the main drainage canals of the region work as large scale groundwater collectors rather than fulfill their designated use to collect saline water from the field level. Further measurements including years with less water availability, and detailed analysis of groundwater levels in relation to the drainage canals are required to quantify and localize weak points of the collector system. There were enormous discrepancies of system functioning inside the irrigated areas of Khorezm in 2005. Especially the Palvan-Gazavat-Subsystem showed high withdrawals and depleted fractions far below the critical value of 0.5. The latter clearly indicates the risk of groundwater and soil salinity in this distant part of the Khorezm region. One important contribution to the situation is an improper irrigation practise in the beginning of the cotton growth period, which was shown by former studies on farm level. Drainage is blocked to rise the groundwater tables. This practise is also used in times of water scarcity. The role of the lakes located South of the Palvan-Gazavat-System remained unclear. Further investigations have to show if these lakes also drain the groundwater body of Khorezm (lateral outflow, in this study assumed to be zero) or if connections to the Ozerny Collector exist (flow to sinks, unaccounted in this study). In summary, demand-oriented supply in Khorezm is a major challenge for regional water management. However, the