Projected Drought under Climate Change Using Deep Learning in a Semiarid Mediterranean Region (Medjerda, Northern Tunisia).

Water resource management is a huge challenge in climate change hotspots regions such as in the south Mediterranean region. Predicted climate change in these regions indicates precipitation decrease, temperature increase, and increase in recurrence, magnitude, and duration of extreme events. Climate hazards and natural disasters are therefore expected to increase, especially droughts. Monitoring, prediction, and risk management of drought are therefore of paramount importance for achieving the Sustainable Development Goals for Climate Change (SDG 13) in this region. In this study, standardized climate indices such as SPI, SRI and SPEI are used to predict climate change impacts on the hydrometeorological regime of semi-arid Mediterranean Medjerda catchment in Tunisia. Climate modeling using the statistical downscaling method was performed to obtain precipitation and temperature projections under RCP 4.5 and RCP 8.5 with the MIROC5 global climate model by 2100. Then, the conceptual hydrological model GR2m, developed under the AirGR package available for R language, was implemented for four subcatchments of the Medjerda to simulate projected hydrological behavior and the SCI package available for R language is used to calculate the standardized climate indices. Results show a consecutive decrease of predicted runoff and, indicating that high increase of extreme drought will be expected. This will impact available water resources of the Medjerda catchment. The results also confirm the need to implement urgently climate resilient water management strategies