Climatic controls on streamflow and groundwater dynamics in a semi-arid catchment: Long-term trends and importance of episodic events

Groundwater is a vital water source in semi-arid and arid regions characterized by little and erratic precipitation and ephemeral river flow. In these regions, water scarcity is becoming more critical due to population growth, increasing irrigation demands, and climate change. Groundwater recharge here primarily occurs as episodic events and specifically as focused recharge during high river flow, but the controlling processes are poorly understood. Thus, understanding and quantifying the relationships between climate, streamflow and groundwater dynamics is crucial to assess the impact of climate change on future water availability. Historically, water resources assessments in Africa have relied on large-scale hydrological models, however, they often lack validation from groundwater observations. Here, we take an observation-based approach to identify climatic controls on streamflow and groundwater dynamics in the semi-arid Hout-Sand River catchment, Limpopo, South Africa. Using data spanning from 1940-2023, we analyze time series of precipitation, air temperature, stream discharge, and groundwater level and evaluate long term trends and relationships across a range of climatic indices and hydrologic and groundwater signatures. While we find no significant trends in long-term annual precipitation, the precipitation patterns are becoming increasingly extreme and exhibit higher intensities with longer dry periods. In response, streamflow patterns are changing towards longer no-flow periods although there is no significant trend in total annual flows. Long-term groundwater levels are not unanimously increasing or decreasing. However, we observe a dependence of streamflow and groundwater levels on multi-annual patterns of climate variability. Furthermore, preliminary results suggest that episodic precipitation and streamflow events contribute to the majority of total groundwater recharge, and that the recharge mainly occurs close to streams, i.e. as focused recharge. Finally, we aim to use machine-learning regression techniques to identify the most important controls on focused and diffuse recharge in order to perform spatial regionalization.