Projecting Climate-Driven Groundwater Variability in the Algarve using Deep Learning-Based Projections

Understanding how groundwater systems will respond to climate change is essential for water-scarce regions such as the Algarve, southern Portugal, where groundwater plays a central role in sustaining agriculture and ecosystems. Previous studies in Portugal have demonstrated that climate teleconnections influence aquifer recharge processes across interannual to decadal timescales, with NAO identified as the dominant driver in southern Portugal and EA and SCAND contributing to higher-frequency variability. However, most existing analyses have focused on historical observations, offering limited insight into future groundwater behavior under projected climate change.

This study integrates climate mode analysis with deep learning-based projections to assess future groundwater variability in the Algarve. Spectral analyses of historical piezometric and precipitation records were first conducted to characterize dominant variability regimes and classify aquifers into annual, mixed, and low-frequency dominated systems. These classifications were then incorporated into deep learning models trained using CMIP6 climate model outputs, namely precipitation and temperature. Groundwater levels were projected under multiple Shared Socioeconomic Pathway (SSP) scenarios for mid-century (2030–2050) and late-century (2050–2100) periods.

The preliminary results indicate a general decline in groundwater levels across Algarve aquifers under all future climate scenarios, with the magnitude and temporal structure of change varying by aquifer type. Aquifers characterized by strong low-frequency variability exhibited more pronounced long-term declines, suggesting increased vulnerability to persistent climate forcing. In contrast, systems dominated by annual variability showed greater short-term responsiveness but less pronounced long-term trends. Across scenarios, a reduction in low-frequency variability was observed, indicating a potential loss of groundwater system inertia and reduced buffering capacity against prolonged droughts.

The analysis further suggests that climate teleconnections will continue to play a significant role in shaping projected groundwater dynamics, with NAO remaining the primary large-scale driver and EA and SCAND influencing higher-frequency modulations. The findings offer valuable guidance for regional groundwater management and provide a transferable framework for assessing climate-driven groundwater variability in other Mediterranean and Atlantic coastal regions.

 

This work is supported by FCT, I.P./MCTES through national funds (PIDDAC): LA/P/0068/2020 - https://doi.org/10.54499/LA/P/0068/2020 , UID/50019/2025,  https://doi.org/10.54499/UID/PRR/50019/2025, UID/PRR2/50019/2025