Dynamic Bayesian Networks as a Decision Support Tool for assessing Climate Change impacts on highly stressed groundwater systems

Abstract

Bayesian Networks (BNs) are powerful tools for assessing and predicting consequences of water management scenarios and uncertain drivers like climate change, integrating available scientific knowledge with the interests of the multiple stakeholders. However, among their major limitations, the non-transient treatment of the cause-effect relationship stands out. A Decision Support System (DSS) based on Dynamic Bayesian Networks (DBNs) is proposed here aimed to palliate that limitation through time slicing technique. The DSS comprises several classes (Object-Oriented BN networks), especially designed for future 5 years length time steps (time slices), covering a total control period of 30 years (2070-2100). The DSS has been developed for assessing impacts generated by different Climate Change (CC) scenarios (generated from several Regional Climatic Models (RCMs) under two emission scenarios, A1B and A2) in an aquifer system (Serral-Salinas) affected by intensive groundwater use over the last 30 years. A calibrated continuous water balance model was used to generate hydrological CC scenarios, and then a groundwater flow model (MODFLOW) was employed in order to analyze the aquifer behavior under CC conditions. Results obtained from both models were used as input for the DSS, considering rainfall, aquifer recharge, variation of piezometric levels and temporal evolution of aquifer storage as the main hydrological components of the aquifer system. Results show the evolution of the aquifer storage for each future time step under different climate change conditions and under controlled water management interventions. This type of applications would allow establishing potential adaptation strategies for aquifer systems as the CC comes into effect