Surface water bodies provide vital support to the society and fundamentally affect ecosystems in various manners. Precise knowledge of the spatial extent of surface water bodies (e.g. reservoirs) as well as of the quantity of water they store is necessary for efficient water deployment and understanding of the local hydrology. Remote sensing provides broad opportunities for surface water mapping. The main objectives of this thesis are: 1) delineating surface water area of partly vegetated water bodies only from remote sensing data without field data input; 2) obtaining the surface water storage, and 3) analyzing its spatio-temporal variations for northeastern (NE) Brazil as a representative for a densely dammed semi-arid region. At first, I investigated the potential of digital elevation models (DEMs) generated from TanDEM-X data, which were acquired during the low water level stage, for reservoirs’ bathymetry derivation. I found that the accuracy of such DEMs can reach one meter, both in the absolute and relative respects. It has shown that DEMs derived from TanDEM-X data have great potentials for representing the reservoirs’ bathymetry of temporally dried-out reservoirs. Subsequently, I targeted at developing a method for mapping the water surface beneath canopy independent of field data for further delineation of the effective water surface. Instead of the commonly used backscattering coefficients, I investigated the capability of the Gray-Level Co-Occurrence Matrix (GLCM) texture index to distinguish different types of Radar backscattering taking place in (partly) vegetated reservoirs. This experiment demonstrated that different types of backscattering at the vegetated water surface show distinct statistical characteristics on GLCM variance derived from TerraSAR-X satellite time series data. Furthermore, with the threshold established based on the statistics of the sub-populations dominated by different types of backscattering, the vegetated water surfaces were effectively mapped, and the effective water surface areas were further delineated with an accuracy of 77% to 95%. ii Based on the investigation of the DEMs generated from TanDEM-X data, I derived the formerly unknown bathymetry for 2 105 reservoirs of various sizes in four representative regions of an overall area of 10 000 km2. The spatial distributions of surface water storage capacities in the four regions were subsequently extracted from the combination of the reservoir bathymetry and the water surface extents provided by RapidEye satellite time series. Furthermore, the spatio-temporal variations of surface water storage were derived for the four representative regions on an annual basis in the period of 2009-2017. This study showed that 1) The density of reservoirs in NE Brazil amounts to 0.04-0.23 reservoirs per km2, the corresponding water surface and surface water storage are 1.18-4.13 ha/km2 and 0.01-0.04 hm3 m/km², respectively; 2) On the spatial unit of 5×5 km2, the surface water storage in the region constantly decreased due to a prolonged drought with a rate of 105 m3/year from 2009 to 2017, with a slight increase from 2016 to 2017 in a few reservoirs; 3) Local precipitation deficit controls the variation of the overall surface water storage in the region. In this thesis I demonstrated the great potential of the great potential of SAR and optical satellite time series data for hydrological applications. The method I developed for delineating the effective water extent from the vegetated reservoirs has shown high potential transferability for other similar regions. The data gaps of bathymetry and surface waters storage capacity were filled for 2 105 reservoirs in NE Brazil. The results of the spatio-temporal variations of surface water storage in four representative regions from 2009-2016 can support future water management and improve hydrological prediction in NE Brazil.
