Abstract: Although in situ rainfall data remains the most accurate, the gauge network density in East Africa is sparse. It lacks continuity, thus making it inadequate to accurately assess the spatial and long-term rainfall trend and variability. This study evaluates the capabilities and limitations of remote sensing data compared with ground-based observations in Tanzania's Pangani Basin in assessing the seasonal and annual rainfall trends and variability. Data from the Climate Hazards Group Infrared Precipitation with Stations (CHIRPS) and twenty-three ground stations were analyzed, comprising a time series from 1990 to 2021. Trend analyses were developed by applying the Mann-Kendall test. The spatialization of the increase and decrease of precipitation for the state was performed using the slope method of Sen. The CHIRPS rainfall stations show only two out of twenty-three had overestimation of above 50%. None of the stations had an underestimation of lower than -50%. Spatial analysis revealed disparities between datasets, showcasing moderate annual rainfall variation from both sources. However, ground station data of at least ten stations shows an increased trend for yearly rainfall in agreement with remote sensing data. The seasonal rainfall trend for Vuli and Masika shows that about 50% of the stations show disparities with remote sensing data. While remote sensing shows an increased trend, the ground station shows a decreasing trend. The study highlights the necessity for calibration and validation to avert misinterpretations in climate trend analyses, especially at the basin level. Overall, this study provides a critical assessment of the effectiveness of remote sensing data in monitoring rainfall variability, emphasizing the importance of aligning diverse data sources for accurate climate trend assessments and informed decision-making.
Keywords: Precipitation; Water Resources; CHIRPS; Remote Sensing; Tropical Catchment
Manuscript submitted to Physics and Chemistry of the Earth | This study is financially supported by a Carnegie Corporation of New York grant. The authors gratefully acknowledge support from the Future Africa Research Leader Fellowship (FAR-LeaF) Programme at the University of Pretoria.
