Decision Making under uncertainty in groundwater management

Managing water resources is one of the most pressing environmental chal-
lenge of the coming decades. With a changing climate, growing popula-
tion and change of land use, the stress on water resources has dramati-
cally increased (De Marsily & Abarca-del Rio, 2016). Whereas policies are
emerging on the use of surface water, groundwater is still lacking sustain-
able exploitation strategies. Designing optimal policies for groundwater
extraction is challenging. This is especially due to the uncertainties on the
climate, subsurface and human behavior. This study is focusing on the
importance of data in a decision-making framework to solve groundwater
management problems. Two approaches of groundwater management are
considered. In a first time, the formalism of Markov Decision Processes is
adopted. The problem of a single decision-maker trying to optimize the
groundwater extraction under an uncertain recharge is modeled and solved
using a reinforcement learning algorithm to produce optimal groundwater
usage policies. A Q-Learning algorithm successfully calculate the optimal
policy in that case. In a second time, the focus is directed on human be-
havior and in particular multiple groundwater user interactions. A Game
Theoretical framework is adopted and a two-players is successfully created.
The study is focusing on the sensitivity analysis of the Nash Equilibrium,
in particular on the different hydrogeological parameters that influence it.
A Value of Information problem is tackled, showing the importance of ac-
curate geological data in such a framework.