No abstract

Water resource systems are under enormous pressures globally. To diagnose and quantify potential vulnerabilities, effective modeling tools are required to represent the interactions between water availability, water demands and their natural and anthropogenic drivers across a range of spatial and temporal scales. Despite significant progresses, system models often undergo various level of simplifications. For instance, several variables are represented within models as prescribed values; and therefore, their links with their natural and anthropogenic drives are not represented. Here we propose a data-driven emulation approach to represent the local dynamics of water resource systems through advising a set of interconnected functional mappings that not only learn and replicate input-output relationships of an existing model, but also link the prescribed variables to their corresponding natural and anthropogenic drivers. To demonstrate the practical utility of the suggested methodology, we consider representing the local dynamics at the Oldman Reservoir, which is a critical infrastructure for effective regional water resource management in southern Alberta, Canada. Using a rigorous setup/falsification procedure, we develop a set of alternative emulators to describe the local dynamics of irrigation demand and withdrawals along with reservoir release and evaporation. The nonfalsified emulators are then used to address the impact of changing climate on the local irrigation deficit. Our analysis shows that local irrigation deficit is more sensitive to changes in local temperature than those of local precipitation. In addition, the rate of change in irrigation deficit is much more significant under a unit degree of warming than a unit degree of cooling. Such local understandings are not attainable by the existing operational model.