University of Twente Student Theses

Login

Evaluation of Remote Sensing data sources for rainfall- runoff modeling : a case study of Nyabarongo Catchment Area, Rwanda

Nteziyaremye, E. (2024) Evaluation of Remote Sensing data sources for rainfall- runoff modeling : a case study of Nyabarongo Catchment Area, Rwanda.

[img] PDF
4MB
Abstract:Rainfall-runoff modelling is required for effective water resource management, flood risk assessment, and environmental planning, particularly in locations such as the Nyabarongo catchment area in Rwanda, where in-situ data are insufficient and of poor quality. Remote sensing data such as Digital Elevation Models (DEMs), Land Use Land Cover (LULC) maps, soil data, and Satellite Rainfall Estimates (SREs) such as CHIRPS, CMORPH, and GPM-IMERG, can be alternative sources of data for hydrologic modelling in such a context of limited data. The main objective of this study is to assess how remote sensing data sources affect the performance of rainfall-runoff modelling in the Nyabarongo catchment area. This study developed a semi-distributed HEC-HMS model using four model combinations of data sources: Local DEM (10m × 10m) with in-situ or SREs rainfall and Sentinel-2, Local DEM (10m × 10m) with in-situ or SREs rainfall and LandSat-8, SRTM (30m × 30m) with in-situ or SREs rainfall and Sentinel-2, and SRTM (30m × 30m) with in-situ or SREs rainfall and LandSat-8. The Local DEM (10m × 10m) with in-situ rainfall and Sentinel-2 served as a reference case because it showed high performance for the first run compared to the other models developed. The preliminary effect assessment of different data sources showed that the LULC map from Sentinel-2 provided a more detailed land cover type than LandSat-8. For DEMs, the Local DEM 10m × 10m provided more detailed topographic information and stream network delineation, and low volume-water storage compared to the SRTM DEM 30m×30m, which contributed to better runoff simulation and model performance within the Nyabarongo catchment area. The analysis of both in-situ and SRE products showed that CHIRPS and GPM-IMERG overestimated the in-situ rainfall, and CMORPH underestimated it. The effects of data sources on model performance, the HEC-HMS model that used high-resolution data of the Local DEM 10m × 10m, Sentinel-2 LULC map, and in-situ rainfall outperformed other model developed, achieving a Nash-Sutcliffe Efficient (NSE) of 0.89, and Relative Volume Error (RVE) of 2.9%. The study also evaluated and corrected errors in SREs, finding that the Power Transform (PT) bias correction technique was the most effective in reducing errors compared to other techniques (Time Space variant and Distribution Transformation). Additionally, the study assessed different time window sizes for bias correction using a Sequential Window approach and showed that a 7-day window is most effective. Furthermore, the effects of error propagation from these data sources on streamflow simulations were analysed. The uncorrected SREs showed an increase in error. However, applying bias correction effectively reduced these errors. Finally, a runoff coefficient evaluation showed that the highest coefficients occurred with the model using the Local DEM, Sentinel-2 LULC, and in-situ rainfall data, as well as the corrected SREs with the Power Transform. Keywords: Remote sensing data, Rainfall-runoff modelling, Nyabarongo catchment, HEC-HMS, SREs, LULC, Digital Elevation Model, In-situ data, bias correction techniques.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Subject:38 earth sciences
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:https://purl.utwente.nl/essays/103500
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page