Decision-Making with Uncertainty for Mission Assurance of DoD Installations under Climate Change Impacts on Flood Risks
ATMOSPHERIC SCIENCES TEAM - University of Illinois
Francina Dominguez – Hydrometeorology, WRF Modeling
Brian Jewett– WRF Modeling
Deffi Putri – MS student, Meteorological analysis
Kevin Gray – Postdoc, WRF modeling
Our project has three key Thrusts: T1–Physical sciences; T2–Human and social sciences; T3–Decision science
The atmospheric sciences team is within Thrust 1 and encompasses the global-to-regional scales.
MOTIVATION: Climate change is altering the frequency, intensity, and location of storms, contributing to vulnerability and compounding risks such as flooding.
“Since the 1950s, there has been an upward trend in heavy precipitation across the contiguous US…These changes have contributed to increases in river and stream flooding in these regions” – National Climate Assessment
‘Recurrent Flooding’ emerged as the primary concern for nearly two thirds of analyzed DoD installations. Floodwaters can damage DoD infrastructure, equipment, vehicles and aircraft, off-base housing and cause land degradation.
June 2016, nine U.S. Army soldiers drowned at Fort Hood (now Fort Cavazos), TX when their tactical vehicle was swept downstream, while attempting to cross an at-grade low water crossing during heavy flooding
We will focus on two DoD locations for our analysis:
Southeastern Michigan Region
Fort Cavazos
Our group has experience in three critical areas:
a. Probabilistic weather forecasts to drive hydrology-hydraulic models.
b. Dynamical Downscaling using WRF to storm-resolving scale for flooding applications.
c. Climate Change projections using very high resolution WRF simulations.
a. Our group worked on precipitation and flooding predictions at Fort Cavazos (Matus et al. 2020, Journal of Hydrometeorology)
We used an ensemble of model-simulated precipitation forecasts to drive hydrologic-hydraulic models.
NCAR Ensemble Precipitation System
48-hour forecast
3km spatial resolution
This adds a probabilistic perspective to flow velocity and depth forecasts and communicates uncertainty.
b. Our group has worked on dynamical downscaling to improve rainfall forecasts over Chicago.
Stage IV observations
Different lead times.
(one source of uncertainty)
Different spatial resolutions
Driving model
WRF (Weather Research and Forecast) is currently the best model for downscaling global model data to the regional scale.
GCMs cannot explicitly represent small-scale physical processes such as convective initiation, organized convection, or mesoscale land-atmosphere interactions, which are critical for extreme precipitation.
3. Our group has significant experience developing future climate projections using the Pseudo Global Warming method.
MURI ATMOSPHERIC TEAM GOAL: To analyze historical and future flood-producing storms and generate an ensemble of regional simulations at the convection permitting scale (grid spacing < 4 km) that is required to capture critical physical processes impacting deep convection and extreme precipitation.
CRITICAL QUESTION:
- Are extreme flooding events associated with frontal systems, organized convection, extra-tropical cyclone..etc.?
Can GCMs capture the most important types of flood-producing storms?
How is the large-scale environment going to change in the future?
How is the frequency going to change in the future?
2) Use historical simulations of GCMs to understand how they represent the intensity and frequency of flood-producing storms.
3) Simulate the historical flood-producing storms using WRF at a resolution less than 4km.
We will select about several flooding events per case study. We will analyze the events using daily average surface pressure, temperature and surface wind maps created using ERA5, NOAA U.S. Daily Weather Maps and Stage IV Quantitative Precipitation Estimate.
We will have ensemble simulations using different initial conditions and different spatial resolutions. We will force the model using ERA5.
9- and 3-km domains
Our goal for the historical simulations is to provide the optimal WRF configuration that best represents precipitation**.
**Optimal representation of precipitation includes total precipitation, precipitation intensity, timing, spatial location.
18
9 km, 2/3/2
9 km, 2/2/2
9 km, 1/2/1
9 km, 2/2/2-2
3 km, 2/3/2
3 km, 1/2/1
Integrated cloud�& precipitation (mm)
CRITICAL QUESTION:
- Because different flooding events will be associated with different meteorological conditions. The “optimal WRF configuration” will depend on the specific event.
- There will be a suite of physics options that we would select.
- The value added by WRF downscaling will likely depend on the type of meteorological conditions.
- WRF will likely be more useful for representing organized convection.
-Regardless of the type of meteorological conditions, WRF will provide information about precipitation (intensity/timing) that would not be available from GCMs.
ERA5-derived initial and boundary conditions for the historical period are perturbed with climate change signal.
temperature, relative humidity, sea surface temperature, soil temperature, sea ice, and sea level pressure. Greenhouse gasses are also changed.
Historical: WRFinput = GCMhistorical
Future: WRFinput = GCMfuture
The future PGW simulation can be interpreted as how the historical climate would be different if it had occurred in a future climate, with changes in the mean-state climate corresponding to what occurred in the GCM.
Historical: WRFinput = ERA5
Future: WRFinput = ERA5 + DGCM
DGCM = GCMfuture - GCMhistorical
Direct (traditional) Downscaling
Pseudo Global Warming
DGCM can come from an ensemble of multiple models/ensemble members.
4) Simulate these storms in a warmer climate using Pseudo Global Warming (PGW).
Direct (traditional) Downscaling | Pseudo Global Warming |
Cons:
| Pros:
|
Pros:
| Cons:
|
1. Global Climate Model uncertainty
2. Internal Variability (ENSO)
3. Emission Scenarios (SSPs)
We will sample different sources of uncertainty related to future climate projections.
4. Downscaling methods (compare with other existing downscaling methods such as statistical and “direct” dynamical).
We must also include the uncertainty related to downscaling precipitation to the regional scale.
CRITICAL QUESTION:
- How can we provide useful precipitation projections into the future while considering all these different sources of uncertainty?
- Can we leverage the physical realism of PGW simulations within the ecosystem of existing (less physically realistic) GCM, statistical and dynamical downscaled products?