Error or insight: Tracing how errors in dynamically downscaled hydrologic projections shape vulnerability inferences in complex water infrastructure systems

Abstract:

Water-resources planners often use basin-wide water management models (WMMs) to understand how systems behave under various climatic and anthropogenic stressors. Frequently, dynamically downscaled climate inputs are used in conjunction with land surface models (LSMs) to provide hydrologic streamflow projections, which serve as critical inputs for WMMs. Although there is an extensive body of research exploring LSMs’ performance relative to observed streamflow, it has not been well studied how their errors influence WMM-based assessments. In this study, we explore how LSM streamflow errors affect the way that we perceive water availability and financial stability for irrigation districts in the agriculturally important Central Valley, as represented in the California Food–Energy–Water Systems [CalFEWS] WMM. CalFEWS simulates 12 major surface water reservoirs that impact the operations of California’s two primary state-wide water transfer projects: the Central Valley Project and the State Water Project. Additionally, CalFEWS simulates the water allocation, groundwater banks, streamflow regulations, and financial risks of irrigation districts. Our analysis of streamflow prediction errors is based on the Weather Research and Forecasting (WRF) model and its coupled LSM (Noah-MP). Here, we show even modest LSM streamflow errors can strongly affect assessments of water availability and financial stability for irrigation districts in California’s Central Valley. Our results highlight that LSM errors in the analysis of flood and drought extremes can be highly interactive across timescales, are path dependent, and can be amplified by modeled infrastructure systems (e.g., misrepresenting banked groundwater). Our results also show that common strategies used to reduce errors in deterministic LSM hydrologic projections (e.g., bias correction) can themselves strongly distort projected climate vulnerabilities and misrepresent their inferred financial consequences. Overall, this work indicates the need to move beyond standard deterministic climate projection frameworks towards exploratory ensemble-based scenario methods that can bridge the process insights provided by LSMs with the broad suite of uncertainties that should be considered in WMM-based vulnerability assessments.