An assessment of the use of remote sensing to estimate catchment rainfall for use in hydrological modelling and design flood estimation.

Masters Degree. University of KwaZulu-Natal, Pietermaritzburg.

The accurate estimation of catchment rainfall is crucial, especially in hydrological modelling and flood hydrology which is used for the planning and design of hydrological infrastructures such as dams and bridges. Traditionally, catchment rainfall is estimated by making use of ground-based point rainfall measurements from rain gauges. The literature review conducted in this study supports that there is evidence of a decrease in the number of operational groundbased rainfall stations in South Africa which presents a challenge when estimating catchment rainfall for use in hydrological modelling and design flood estimation. Thus, innovative ways are required to estimate catchment rainfall and to improve the estimation of catchment design rainfall. This study investigated the use of remote sensing as an alternative way to estimate catchment design rainfall. To do this, a pilot study was first used to develop and test the methodology using a quaternary catchment that was selected based on the raingauge density. This was followed by the application of a refined methodology in another quaternary catchment which was used to verify the results that were obtained in the pilot study. After a comprehensive review of the literature, the remote sensing product selected for this study was the CHIRPS rainfall product. The methodology adopted first validated the remotely sensed rainfall data using the observed rainfall data and the estimated remotely sensed rainfall values were bias corrected using the observed rainfall data. The statistics that were used for validating are MAE, MBE, RMSE and D. The method that was used for bias correction was empirical quantile mapping Issues encountered, and as documented in the literature, include the unavailability of long periods of observed quality rainfall data and the limited and uneven spatial distribution of rainfall stations. Catchment rainfalls were estimated using observed rainfall, and this was assumed as the best estimate and was compared to the catchment rainfalls that were estimated using the biascorrected remotely sensed rainfalls. The performance of CHIRPS rainfall was varied among the approaches and the selected catchments. Nevertheless, the results from this study still show the potential of the use of remotely sensed rainfall to estimate catchment design rainfalls. At the daily timescale, satellite-derived and observed rainfall were poorly correlated and variable among locations. However, monthly and annual rainfall totals were in closer agreement with historical observations than the daily values. Despite the varied performance , the result of the study shows that CHIRPS rainfall product can be used to estimate catchment rainfall for hydrological modelling and flood frequency analysis. By acknowledging that the performance of remote sensing products is robust, it is of importance to note that the performance of the results presented is strictly for the catchments and stations selected for this project as well as the methods selected to validate and correct the bias in remotely sensed rainfall. The recommendations from the study are that a similar study is conducted in another region where there is even distribution of stations and a long record of quality observed rainfall beyond the year 2000 and consideration of the methods to identify outliers before making any meaningful estimations such as catchment rainfall from rainfall data.

No thesis submission form available.