Designing Statistical Prediction Systems to Capture Seasonal Lake Water Balance Dynamics
Date:
Presentation: American Geophysical Union (AGU) Fall Meeting, San Francisco, CA
VanDeWeghe, A., Luo, Y., and Gronewold, A. 2023. “Designing Statistical Prediction Systems to Capture Seasonal Lake Water Balance Dynamics.” American Geophysical Union (AGU) Fall Meeting. (Presentation)
Abstract
Large lakes around the world store a majority of globally accessible freshwater and support immensely valuable ecological and economical systems. Dual pressures of anthropogenic water demand and a changing climate underscore the need to better understand the dynamics of water availability in these lakes while simultaneously complicating this very endeavor. Mechanistic representations of the movement of water through large lake systems via coupled hydrologic, atmospheric, and lake models can be costly, computationally intensive, and challenging to calibrate to observed conditions. Here, we present a statistical alternative to represent plausible sequences of hydroclimate variables that influence lake water supply. Specifically, we examine the implementation of a regular vine copula over the Laurentian Great Lakes system to simulate physically plausible monthly sequences of precipitation, overlake evaporation, and surface runoff. However, selecting the structure of this copula depends on the goals of its implementation, including length of the forecasting horizon, importance of capturing seasonal dynamics, and timing of recurring use. We present a framework for identifying the advantages and disadvantages of variable selection based on user-informed operational goals, and further suggest that statistical water supply prediction systems should: 1) capture spatiotemporal correlations between variables over the length of the seasonal forecasting horizon or at least 12 months, 2) prioritize representation of highly correlated variables and those that are most influential to the water balance during calibration, and 3) define temporal model boundaries to minimize the influence of statistical discontinuities on the magnitude and dynamics of lake water supply. The applicability and flexibility of this system to capture unique water supply dynamics of large lakes across North America will be discussed.
Citation
VanDeWeghe, A., Luo, Y., and Gronewold, A. (2023). Designing Statistical Prediction Systems to Capture Seasonal Lake Water Balance Dynamics. AGU Fall Meeting Abstracts, 2023, H53E-06.