Mapping steady-state groundwater levels in the Mediterranean region: The Iberian Peninsula as a benchmark

Abstract

Groundwater is a strategic freshwater resource that is over-exploited in the Mediterranean region mainly for agricultural uses, leading to severe groundwater depletion in many aquifers. Uncertainties in groundwater availability are further exacerbated by climate change and its associated impacts. To address these challenges and ensure the longevity of this resource, sustainable groundwater management is essential. Unfortunately, our knowledge of groundwater status at the regional Mediterranean scale is limited due to the lack of consistent in-situ monitoring and data sharing. Groundwater modelling at the global scale offers a tool to evaluate the status of data-scarce regions. This study aims to assess steady-state water table depth in the Mediterranean region and its uncertainty by examining the simulations of three global gradient-based groundwater models. To examine these models’ agreement with observed data, we focused for demonstration purposes on the Iberian Peninsula for its climatic diversity similar to that of the Mediterranean region and for its relatively high density of in-situ data. Results showed that the models represented reasonably well the observed groundwater heads of the Iberian Peninsula (R2 = 0.70–0.74). The models performed better at low elevations than in mountainous areas. To overcome the limitations of the regional models, a geostatistical approach is proposed to downscale the average of the three models for different subsets of the observed data (i.e., 10, 30, 50, and 70 %). Results revealed that when the average simulated groundwater depth was conditioned with at least 50 % of observations (equivalent to about three wells per 1000 km2), the spatial groundwater patterns in the Iberian Peninsula were well reproduced (R2 ≥ 0.65). Overall, this study shows that despite their underlying assumptions, global models can be used to map regional groundwater resources as long as they are conditioned on observed data.