Downscaling Rainfall at a Small Watershed Level in a Data Scarce Dryland Woodland Using the Regression Model

Abstract

Using a regression model, this study proposes a statistical strategy for downscaling rainfall data at a small watershed level in data-scarce dryland woods. Because dryland ecosystems are especially vulnerable to changes in precipitation patterns, accurate rainfall estimation is critical for successful water resource management and ecosystem conservation. Obtaining trustworthy rainfall data at small scales in a data-scarce environment, on the other hand, remains a difficulty. To overcome this issue, we offer a downscaling strategy that uses regression models' power to estimate rainfall at the local watershed level with readily available parameters. The study region is in dryland woods with few rainfall monitoring stations. Global Precipitation Climatology Centre (GPCC) data were downloaded from 1986 – 2005 to test the validity of the regression model in downscaling the rainfall. The regression model implemented in a stepwise operation uses auxiliary data such as locational and topographic features (altitude, latitude, and longitude) extracted from 90 m Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM), to establish a statistical link between large-scale rainfall measurements and local-scale factors. The trained regression model performance was then assessed using cross-validation procedures. The results show that the suggested downscaling method is excellent at capturing the spatiotemporal variability of rainfall within the small watershed. The predicted rainfall data were validated with the original gridded GPCC points, and the observed result shows that there was a very strong correlation for all the years at p > 0.05 significant level. The regression model successfully converts large-scale rainfall patterns into localized estimates, allowing for more accurate precipitation representation at a finer resolution. This data can help decision-makers in the dryland woodland with water distribution, land management, and ecological planning. This research advances the subject of hydrological modeling in dryland ecosystems by presenting a realistic method for downscaling rainfall data at the watershed level. The suggested regression model is a useful tool for filling data gaps, especially among conservationist who may not have the luxury of using complex tools for estimating rainfall