| Summary: | The tropical Andes are characterised by highly variable precipitation patterns in both space and time, combined with low coverage of ground-based meteorological observation. Satellite precipitation products are increasingly used to fi ll this gap, providing high-resolution precipitation estimation in ungauged catchments. However, accuracy of space-borne sensors is limited by errors associated with rain-rate retrieval, temporal sampling, spatial resolution and ground-correction. This PhD thesis provides a detailed ground-validation of the Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar (TPR) over the tropical Andes. The TPR is the single space-borne active precipitation sensor with long-term historical observations (1998 - 2014) of tropical rainfall at a high spatial resolution (5 km). TPR accuracy was found to vary substantially across distinct climatic regions, demonstrating a dependence on the local rainfall regime. Satellite errors due to low-frequency sampling and rainfall retrieval showed a wide spatial spread, generally exceeding gauge errors. However, the TPR is an accurate estimator of mean climatological precipitation, which formed the basis of generating a high-resolution mean monthly precipitation dataset by satellite-gauge merging. Furthermore, this research focussed on the spatial scaling of rainfall at daily time-scales, showing that scaling behaviour varies substantially between different precipitation variables. Lastly, high-resolution climatological precipitation variables from the TPR were used to develop a spatial disaggregation model for regular-gridded satellite precipitation. Probabilistic disaggregated rainfall fields at 1 km resolution demonstrated better estimation of the statistical distribution than rainfall fields obtained by conditional simulation of gauge observations. The generic disaggregation approach is modular in design, supplies uncertainty estimates via probabilistic simulation and is independent of local gauge data. Thus, it is a useful tool for disaggregation of any gridded precipitation product with potential applications from extremes estimation to distributed hydrological modelling. Future work is required to improve the spatial consistency at di fferent spatial scales, possibly through the integration with geostatistical methods.
|
|---|
