The deterioration of the West African savannah in the last three decades is believed to be closely linked with about 0.5 C rise in temperature leading to evaporation losses and declining levels of the Volta Lake in Ghana. Although hydrological models can be used to predict climate change impacts on the regional hydrology, spatially-observed ground data needed for this purpose are largely unavailable. This thesis seeks to address this problem by developing improved methods for estimating energy and water fluxes (e.g. latent heat [ET]) from remotely sensed data and to demonstrate how these may be used to parameterize hydrological models. The first part of the thesis examines the potential of the Penman-Monteith method to estimate local-scale ET using groundbased hydrometeorological observations, vegetation coefficients and environmental data. The model results were compared with pan observations, scintillometer (eddy correlation) measurements and the Thomthwaite empirical method. The Penman- Monteith model produced better evaporation estimates (~3.90 mm day(^-1) for the Tamale district) than its counterpart methods. The Thomthwaite, for example, overestimated predictions by 5.0-11.0 mm day(^-1). Up-scaling on a monthly time scale and parameterization of the Grindley soil moisture balance model with the Thomthwaite and Penman-Monteith data, however, produced similar estimates of actual evaporation and soil moisture, which correlated strongly (R(^2) = 0.95) with water balance estimates. To improve ET estimation at the regional-scale, the second part of the thesis develops spatial models through energy balance modelling and data up-scaling methods, driven by radiometric measurements from recent satellite sensors such as the Landsat ETM+, MODIS and ENVISAT-AATSR. The results were validated using estimates from the Penman-Monteith method, field observations, detailed satellite measurements and published data. It was realised that the MODIS sensor is a more useful source of energy and water balance parameters than AA TSR. For example, stronger correlations were found between MODIS estimates of ET and other energy balance variables such as NDVI, surface temperature and net radiation (R(^2) = 0.67-0.73) compared with AATSR estimates (R(^2) = 0.31-0.40). There was also a good spatial correlation between MODIS and Landsat ETM+ results (R(^2) = 0.71), but poor correlations were found between AATSR and Landsat data (R(^2) = 0.0-0.13), which may be explained by differences in instrument calibration. The results further showed that ET may be underestimated with deviations of ~2.0 mm day 1 when MODIS/AATSR measurements are validated against point observations because of spatial mismatch. The final part of the thesis demonstrates the application of the ET model for predicting runoff (Q) using a simplified version of the regional water balance equation. This is followed byanalysis of flow sensitivity to declining scenarios of biomass volume. The results showed the absence of Q for >90% of the study area during the dry season due largely to crude model approximation and lack of rainfall data, which makes model testing during the wet season important. Runoff prediction may be improved if spatial estimates of rainfall, ET and geographical data (e.g. land-use/cover maps, soil & geology maps and DEM) could be routinely derived from satellite imagery.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:694789 |
| Date | January 2007 |
| Creators | Opoku-Duah, Stephen |
| Publisher | Durham University |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://etheses.dur.ac.uk/3181/ |
Page generated in 0.144 seconds