Estimating Water Balance Components of Lakes and Reservoirs Using Various Open Access Satellite Databases

Abstract

There are millions of lakes and ten thousands of reservoirs in the world. The number of reservoirs is still increasing through the construction of large dams to meet the growing demand for water resources, hydroelectricity and economic development. Accurate information on the water balance components of lakes and reservoirs is deemed necessary for managing demand (i.e. of the various water user communities) and supply (gauged and ungauged inflow from surrounding catchments). Information on storage and availability of fresh water is a national security issue in many countries. In-situ hydrological measurements of reservoirs are usually not publically available. Satellite measurements and the application of specific interpretation algorithms are alternative sources of information which are publicly accessible and can help communities to better utilize scarce water resources. Ultimately, a rich, readily available and accessible data source will lead to better management decisions, with various benefits and services for all stakeholders. This thesis explores the use of satellite measurements to estimate the various key water balance components of lakes and reservoirs, ultimately leading to predictions of releases from reservoirs. Various open-access satellite databases have been explored. This thesis places emphasis on the following aspects: (i) improving existing satellite products linked to lakes and reservoirs; (ii) integrating and customizing various satellite products; (iii) evaluating and comparing alternative satellite products and (iv) developing remote sensing algorithms for the generation of new products for lakes and reservoirs. A summary of some existing open-access databases pertaining to water balance components of lakes and reservoirs is presented in the Appendix of this thesis. A pre-requisite for estimating the inflow to lakes and reservoirs from the surrounding catchments is to have access to accurate precipitation data at high spatial resolution and for the mountains where gauges are often absent. A new integrated downscaling-calibration procedure is described in Chapter 2. It is based on an integration of the TRMM (Tropical Rainfall Measuring Mission) 3B43 precipitation product (at 0.25° resolution) and NDVI (Normalized Difference Vegetation Index) from SPOT-Vegetation satellite data. This procedure creates an improved product of monthly pixel-based precipitation data at 1 km resolution. Only limited rain gauge datasets are required for calibration. This new procedure has been successfully tested in two different basins: Lake Tana Basin in Ethiopia with a humid climate and the Caspian Sea Region in Iran with a semi-arid climate. Chapter 3 describes the quantification of time series of water level, lake surface area and lake water volume from open-access databases based on satellite measurements. The four currently available satellite altimetry databases providing lake levels are firstly evaluated for three different lakes: Lake Mead (U.S.A), Lake Tana (Ethiopia) and Lake IJssel (The Netherlands). The four databases are: (i) Global Reservoir and Lake Monitoring (GRLM), (ii) River Lake Hydrology (RLH), (iii) Hydroweb and (iv) ICESat-GLAS level 2 Global Land Surface Altimetry data (ICESat-GLAS). Time series of lake surface areas were determined from Landsat TM/ETM images using the Modified Normalized Difference Water Index (MNDWI) method. A new method has been developed to estimate lake water volume, using purely satellite data, by combining satellite altimetry databases and Landsat TM/ETM images. The new method makes it possible to estimate the volume of water above the historical minimum level identified from satellite altimetry databases. The validation shows that all estimated water volumes agreed well with in-situ water volume measurements derived from bathymetric surveys for both Lake Mead and Lake Tana, with R2>0.95 and RMSE ranging between 4.6 and 13.1% of corresponding mean values of in-situ measurements. Similar successful application of the proposed method has also been shown for Lake Nasser (Egypt-Sudan) and Roseires Reservoir (Sudan). Evaporation is a major component of the water and energy balance of lakes and reservoirs. The heat storage changes (Qt) term is in turn an essential part of the energy balance of lakes. Qt should be quantified and incorporated into surface energy balance combination models (including the Penman equation) to estimate lake evaporation. Qt is often ignored in both scientific literature and practical engineering type studies mainly due to the lack of routine data of vertical water temperature profiles. In Chapter 4, a first experimental dataset of heat storage changes in 22 lakes has been created from selected international publications. A hysteresis model (Qt=a*Rn+b+c*dRn/dt) was developed that fits the 22 independently gathered bi-weekly and monthly datasets satisfactory when the three coefficients (a, b and c) were determined using local measured flux data. Predictive models for the three coefficients using net radiation (Rn) and water surface temperature (T0) time series have been developed. The simple Slob’s equation that computes Rn from shortwave solar radiation (Rs) was evaluated and showed to perform well for lakes. Since Rs and T0 can be accurately estimated from various operational satellite measurements, the proposed procedure can be applied to estimate Qt for various lakes and reservoirs for which no local measurements are available. The performance of evaporation models, using as input Qt estimated by the developed hysteresis model, is described in Chapter 5. Three energy balance-based evaporation models were evaluated: the De Bruin-Keijman (DK), Penman, and Duan-Bastiaanssen (DB) models. The DB model is a non-published energy balance residual model based on an Ohm-type parameterization of sensible heat flux H with standard empirical coefficients for surface roughness, and the difference of surface water temperature and air temperature. A thorough analysis was conducted on the performance of three different evaporation models for five different lakes with different geographical conditions. Using an estimate of Qt, based on the theories outlined in Chapter 4, all three model’s evaporation estimates agreed well with measurements, and considerably better than cases with Qt excluded. The DK model with its minimum data requirements generally performed best for the five lakes. For two lakes, the new DB model ranked second, followed by the classical Penman model. Therefore, the widely used Penman model should be evaluated more critically, as it appears not always to be the most accurate method. One integrated case study on Lake Tana was conducted by combining the methods proposed in the individual chapters. Chapter 6 quantifies the water balance components of the poorly gauged Lake Tana in a wet year 2006. The specific question is whether the outflow from Lake Tana can be estimated accurately from the residual of the lake water balance. The lake inflow was estimated as the total runoff from the surrounding catchments. The runoff was approximated as the residual of the land-based catchment water balance. The estimated runoff and lake outflow were satisfactory at annual time scales but not at monthly time steps. The discrepancy of the estimated runoff into the lake at monthly time scale can be ascribed to the poor quantification of underground storage changes (changes in soil moisture and groundwater storages). The GLDAS (Global Land Data Assimilation System) open-access database was used for soil moisture storage. Given the coarse resolution of GLDAS and a similar limitation of the current GRACE satellite gravity mission (Gravity Recovery and Climate Experiment), it is preferred to apply a local rainfall-runoff model that takes into account soil moisture and groundwater storage changes at monthly and shorter time scales. Hence, the daily, weekly and monthly water balance of lakes and reservoirs requires the involvement of a hydrological model, and cannot be solved solely on the basis of earth observation data. The estimates of precipitation, evaporation from land and water surface, and the changes of open water storage are however crucial for describing the hydrological system behavior adequately, including the outflow from reservoirs and the determination of deep percolation losses. In summary, the following aspects of lakes and reservoirs have been investigated: (1) improving existing satellite products; An integrated procedure to generate local rainfall at high spatial resolution and with improved accuracy has been developed. (2) integrating and customizing various satellite products; Water volume above a reference level in lakes and reservoirs can now be calculated by integrating levels and surface areas from satellite products. (3) evaluating and comparing alternative satellite products; Different products on water levels, solar radiation, surface temperature and net radiation were tested. (4) developing remote sensing algorithms for the generation of new products; A new procedure for estimating heat storage changes Qt from operational satellite products was developed which is essential for calculating open water evaporation.