Monitoring arid-land groundwater abstraction through optimization of a land surface model with remote sensing-based evaporation

Abstract

The increase in irrigated agriculture in Saudi Arabia is having a large impact on its limited groundwater resources. While large-scale water storage changes can be estimated using satellite data, monitoring groundwater abstraction rates is largely non-existent at either farm or regional level, so water management decisions remain ill-informed. Although determining water use from space at high spatiotemporal resolutions remains challenging, a number of approaches have shown promise, particularly in the retrieval of crop water use via evaporation. Apart from satellite-based estimates, land surface models offer a continuous spatial-temporal evolution of full land-atmosphere water and energy exchanges. In this study, we first examine recent trends in terrestrial water storage depletion within the Arabian Peninsula and explore its relation to increased agricultural activity in the region using satellite data. Next, we evaluate a number of large-scale remote sensing-based evaporation models, giving insight into the challenges of evaporation retrieval in arid environments. Finally, we present a novel method aimed to retrieve groundwater abstraction rates used in irrigated fields by constraining a land surface model with remote sensing-based evaporation observations. The approach is used to reproduce reported irrigation rates over 41 center-pivot irrigation fields presenting a range of crop dynamics over the course of one year. The results of this application are promising, with mean absolute errors below 3 mm:day-1, bias of -1.6 mm:day-1, and a first rough estimate of total annual abstractions of 65.8 Mm3 (close to the estimated value using reported farm data, 69.42 Mm3). However, further efforts to address the overestimation of bare soil evaporation in the model are required. The uneven coverage of satellite data within the study site allowed us to evaluate its impact on the optimization, with a better match between observed and obtained irrigation rates on fields with higher frequency of available data. The inclusion of novel remote sensing sources (e.g. CubeSats) that offer higher frequencies and higher resolution can also be explored to improve the methodology, although further validation of these systems is needed. The developed framework has the potential to be used as a water management tool to monitor groundwater losses over large remote regions.