A Meta-Analysis to Disentangle the Impacts of Climate and Land Use Changes on Streamflow

Abstract

Climate change and land use/land cover (LULC) changes influence streamflow by altering precipitation patterns, evaporation, and hydrological processes. Disentangling their combined effects is critical for effective water management under increasing climatic and environmental pressures. This meta-analysis integrates quantitative and qualitative approaches to assess the impacts of precipitation, temperature, and LULC changes on streamflow using published data sets, multiple linear regression, and Random Forest models. Precipitation emerges as the dominant driver, showing significant variability and a direct linear correlation with streamflow. Temperature impacts are inconsistent, while LULC changes demonstrate nuanced effects. Conversions to agriculture generally increase streamflow, whereas transitions to forests reduce it. Multiple linear regression revealed that precipitation alone explains nearly half of the variance in streamflow, with LULC changes contributing an additional but smaller percentage. In contrast, temperature changes have minimal influence. Variability in LULC conversions correlates with residuals, underscoring diverse impacts across land use types. The Random Forest model, which allows the consideration of non-linear dependencies, achieved R2 values of 0.7, confirming precipitation as the most critical predictor, followed by temperature and LULC changes. Including catchment area and climate zone added no significant improvement. These findings highlight the combined importance of precipitation, temperature and LULC changes in shaping streamflow dynamics. While comprehensive, the meta-analysis may overlook local factors such as micro-climate variations or land management practices. The variability in model predictive power underscores the challenge of modeling nonlinear relationships between climate, LULC changes, and streamflow. The results offer critical insights for sustainable water resource management and predictive hydrological modeling.