Applying Stochastic Weather Generation and Continuous Hydrologic Simulation for Critical Infrastructure Design

ABSTRACT ONLY - Applied Weather Associates teamed with Aterra Solutions to complete a stochastic weather modeling study to provide long term meteorological realization for hydrologic modeling, flood frequency analysis, and flood recurrence interval analyses. This utilized a multisite stochastic modeling approach using daily observations of precipitation, temperatures, and snow water equivalent (SWE) from 49 sites in the upper Midwest through the Multi-site Auto-regressive Weather GENerator (RMAWGEN)) framework. Stochastic weather generators are statistical models that simulate realistic or plausible random sequences of atmospheric variables. Resulting sequences provide meteorological realizations that can be used for risk evaluations and reliability assessments for various systems such as dams and nuclear generating facilities. Observed precipitation and temperature records were used to calibrate RMAWGEN for the 1949-2019 period. Validation was performed on the calibration period data. Results demonstrate that the model was able to capture spatiotemporal characteristics of observed precipitation and temperature. The model generated 12 iterations of 1,000-years of daily weather sequences of precipitation, temperatures, and SWE. Climate change projections were applied using RCP 4.5 and 8.5 to generate 12 iterations of 1,000-years of future sequences of precipitation, temperatures, and SWE.
Weather outputs were used in a continuous simulation hydrologic model built using HEC-HMS. This was calibrated against 3 different years of daily flow data at locations throughout an 88,000 mi2 basin. Normal, wet, and dry years were used for calibration. The final calibrated model was used to simulate runoff for each 12x1000-year simulations, including the three climate change projections. Uncertainty analyses, using a Monte-Carlo framework within HMS, bracketed potential outflow possibilities based on variability in hydrologic inputs identified in the calibration phase. Annual maximum flows were used to characterize probabilistic flood hazards (to as low as a 10-6 annual exceedance probability), considering a wide range of event parameters such as snow accumulation, spring melt patterns, and rainfall. Results will be used in safety assessments and seasonal flood operation planning.