Francina Dominguez:
Uncertainty Quantification in Future Projections of Flood-Producing Precipitation Events
The changing frequency, spatial distribution and severity of flooding are some of the most societally relevant impacts of climate change. Decision makers traditionally use static flood hazard areas that fail to account for the physics of flood-generating processes as well as the expected consequences of climate change. Further, despite its crucial importance for decision-making, uncertainty quantification (UQ) is largely absent or incomplete, leading to improper risk assessment. As part of the Office of Naval Research, Dept. of Defense, Multidisciplinary University Research Initiative (MURI), our team is investigating historical and future flooding the U.S. Army Garrison Detroit Arsenal. Our team includes scientists in the Physical, Human and Decision Sciences. Our goal is to quantify current / future flooding challenges, understand and anticipate human behavior, and improve decision making with explicit UQ. Here we will describe the atmospheric sciences component, which blends weather forecasting and climate projections.
In this work, we incorporate UQ by generating and analyzing a large ensemble of potential future extreme precipitation fields while explicitly accounting for the physical mechanisms that generate flood-producing precipitation events. The proposed method combines: 1) detailed analysis of historical flood-generating meteorological conditions 2) analysis of CMIP6 GCM extreme precipitation intensity and frequency in historical and future simulations 3) large ensembles of WRF (Weather Research and Forecasting Model) simulations at the convection permitting (CP) kilometer scale and 4) pseudo-global warming CP simulations to evaluate the changes in these flood-producing storms in a future climate. This allows us to incorporate uncertainty associated with future greenhouse gas forcing and choice of GCM model as done in climate projections. In addition, the WRF-CP ensemble design incorporates different physics options, initial/lateral-boundary conditions, and stochastic processes like those used in operational weather forecasting. The result is a large ensemble of precipitation fields at the km-scale that can be used as input for the tRIBS (TIN-based Real-Time Integrated Basin Simulator) hydrological and hydraulic model.
Jeff Trapp:
Conspicuously absent from climate-change attribution studies are considerations of specific tornado, hailstorm, and “straight-line” wind events of high impact. To help remedy this, we have developed an attribution methodology based on the use of high-resolution weather-model simulations performed according to the event-level “pseudo-global warming” (PGW) climate modeling approach. In essence, the PGW approach involves numerical simulations of an event under its true 4D environment (the control; CTRL) as well as under its 4D environment modified by a climate-change perturbation (the PGW). A quantitative comparison of the CTRL – or, herein, factual – simulations with the PGW– or, herein, counter-factual – simulations yields an estimate of the potential impact of the imposed climate change. For this attribution application, the climate-change perturbations, or “deltas”, are based on the difference between a mean climate during a pre-industrial time period and a mean climate during a contemporary time period. The climate conditions are derived from GCM simulations following the piControl and historical experiments, respectively, under CMIP6.
An extreme hail event that occurred in Switzerland and its environs on 29 June 2021 has been chosen for the demonstration of this methodology. Of particular focus is the hail swath(s) in the vicinity of Lucerne and Zurich, which included individual hail stones of at least 9 cm and record numbers of insurance claims (over 20,000 damaged vehicles; over 12,000 damaged buildings). ERA5 reanalysis was used as initial and boundary conditions for the simulations, which were conducted with the weather research and forecasting (WRF) model. Results based on an ensemble indicate less intense, and less widespread hail in the Luzerne–Zurich region in the pre-industrial environment. These results are formalized using the fraction of attributable risk (FAR).