Simulating rainfall in the Chehalis river basin using the WRF model

Presentation by Laura Read and Fernando SalasMay 5, 2009

Presentation Outline Motivation for modeling heavy rainfall Introduction to Weather Research &

Forecasting model The Chehalis River Basin and storm

event Experimental design and methods Model validation Results and conclusions Future work

Motivation: Places Flood Adjusted Flood Damage Costs in the U.S. from 1900-2007

Flood Forecasting System

Weather Research & Forecast Model

(WRF)

HEC-HMS

Flooding

Courtesy of John E. Strack et al.

Climate ChangeEvidence for more rain, stronger storms

IPCC report on regional climate conditions

National recognition among the NRC, NOAA

Uncertainty in current projections

Higher temperatures = More moisture flux and convergence in the atmosphere

A call to action

The Weather Research & Forecasting Model

Who, What, Where, When Simulation Process Developed by NCAR,

NOAA, NCEP, FAA, Air Force and Navy Research

Regional version of a GCM—finer resolution, set boundaries and initializations

Uses physical equations to conduct research, forecast weather (numeric weather prediction)

WRF Post-Processing

ARWpost

http://www.dtcenter.org/wrfnmm/users/docs/user_guide/V3/users_guide_nmm_chap3_files/image001.gif

Storm Event Justification Why Washington?

31 of 43 Disaster Declarations are flood related

NEXRAD storm database lists heavy precipitation events and its associated damage costs

http://www4.ncdc.noaa.gov/cgi-win/wwcgi.dll?wwEvent~Storms

The Chehalis River Basin Basin Area: 2,660 sq

mi Length: 115 mi Annual Average

Discharge: 11,208 cfs200 in/yr

45 in/yr

http://www.chehalislandtrust.org/_borders/MAP1.JPG

http://www.ngdc.noaa.gov/mgg/topo/img/wa.jpg

Chehalis River Basin USGS collects

continuous observational stream-flow data nationwide

Nov 3rd storm event eclipsed flood stage in Chehalis River

2007 and 2008 both had record breaking floods in the South-Central region…is this a trend? http://waterdata.usgs.gov/wa/nwis/dv/?

dd_cd=02_00065_00003&format=img_default&site_no=12027500&set_logscale_y=0&begin_date=20061103&end_date=20061109

Storm Event: Nov 3, 2006 to Nov 9, 2006

http://water.weather.gov/

Storm Event: Nov 3, 2006 to Nov 9, 2006

Storm Event: Nov 3, 2006 to Nov 9, 2006

Storm Event: Nov 3, 2006 to Nov 9, 2006

Storm Event: Nov 3, 2006 to Nov 9, 2006

Storm Event: Nov 3, 2006 to Nov 9, 2006

Storm Event: Nov 3, 2006 to Nov 9, 2006

WRF Domain

43º N

52º N

128º W 112º W

Courtesy of Google Earth

Experimental Design

ParameterizationsCode Cumulus (cu_physics) Code Microphysics (mp_physics)

1 Kain-Fritsch (new eta) 1 Kessler scheme

2 Betts-Miller-Janjic 2 Lin et al

3 Grell-Devenyi ensemble 3 WSM 3-class simple ice

5 New Grell 4 WSM 5-class

5 Ferrier (new Eta)

6 WSM 6-class graupel

Simulation Combinations

Grid Cell Resolution Time Step (sec) Microphysics Schemes Cumulus Schemes Simulation Duration

32 km 180 Kessler, Lin, WSM3, Eta KF, BMJ, GD, NG 5 days

16 km 30 Lin, WSM5, WSM6, Eta KF, BMJ 2.5 days

16 km 60 Lin, WSM5, WSM6, Eta KF, BMJ 2 days, 2.5 days, 3 days

Model Validation

Atmospheric – 32 km NARR Geo-potential height Specific Humidity/Water vapor mixing

ratio Wind and wind direction

Precipitation – 4 km NEXRAD Accumulated rainfall Temporal Distribution Spatial Distribution

Geo-Potential Height: 120 W Over TimeWRF: 01_02 Simulation NARR

Wind and Wind Speed: 700 mb

NARRWRF: 01_02 Simulation at 00z7nov2006

Specific Humidity: Constant Latitude01_02 Simulation:

00z7nov2006 NARR

Pres

sure

(mb)

Pres

sure

(mb)

Pres

sure

(mb)

NEXRAD ValidationProcess of Importing Our

Simulations

Model Validation with NEXRAD: Daily

Model Validation

Atmospheric – 32 km NARR Geo-potential height Specific Humidity/Water vapor mixing

ratio Wind and wind direction

Precipitation – 4 km NEXRAD Accumulated rainfall Temporal Distribution Spatial Distribution

Example of 6-Hour Error Analysis

6 Hours MAX (mm) Error RANGE (mm) Error MEAN (mm) Error STD (mm) Error SUM (mm) Error Score

NEXRAD 95.25000 N/A 95.25000 N/A 37.40030 N/A 15.35030 N/A 24871.20000 N/A  N/A

hi_01_02 73.53860 22.79% 68.93870 27.62% 23.34100 37.59% 12.23360 20.30% 3897.95000 84.33% 11

hi_01_04 60.39740 36.59% 49.22840 48.32% 22.54510 39.72% 9.65613 37.09% 3765.03000 84.86% 16

hi_01_05 59.63400 37.39% 54.94480 42.32% 21.75210 41.84% 11.86330 22.72% 3632.60000 85.39% 16

hi_01_06 62.77250 34.10% 51.44550 45.99% 22.77840 39.10% 9.96768 35.07% 3803.98000 84.71% 16

hi_02_02 76.73080 19.44% 74.95240 21.31% 23.60430 36.89% 13.81840 9.98% 3941.92000 84.15% 4

hi_02_04 60.28520 36.71% 51.83900 45.58% 23.33370 37.61% 10.83900 29.39% 3896.73000 84.33% 16

hi_02_05 60.97900 35.98% 56.89180 40.27% 23.57740 36.96% 13.09710 14.68% 3937.42000 84.17% 12

hi_02_06 63.77300 33.05% 54.76460 42.50% 23.58740 36.93% 11.19840 27.05% 3939.10000 84.16% 13

Legend1st

2nd

3rd

Other

Model Validation with NEXRAD: 6 Hours

Parameterization Results Summary 02_02: Betts-

Miller-Janjic, Lin et al.

01_02: Kain-Fritsch, Lin et al.

02_05: Kain-Fritsch, new Eta

Simulations Overall Score

01_02 47

01_04 78

01_05 77

01_06 70

02_02 39

02_04 79

02_05 49

02_06 76

Lin et al. microphysics scheme produced lowest percent errors compared with NEXRAD

Histogram of Precipitation Distribution

Testing for Statistical SignificanceNEXRAD 01_02 Simulation

Test for Homogeneity and T-Stats

F-Test: Datasets are not homogeneous

T-test: reject the null hypothesis (small p)

Model Validation

Atmospheric – 32 km NARR Geo-potential height Specific Humidity/Water vapor mixing

ratio Wind and wind direction

Precipitation – 4 km NEXRAD Accumulated rainfall Temporal Distribution Spatial Distribution

Temporal Distribution Results Source Day 18z6nov200

6 Error 00z7nov2006 Error 06z7nov200

6 Error 12z7nov2006 Error Score

NEXRAD 100.00% 31.27%  N/A 32.60%  N/A 17.43% N/A 18.70% N/A N/A

hi_01_02 82.68% 32.77% -4.81% 26.11% 19.90% 23.08% -32.41% 18.03% 3.57% 12

hi_01_04 83.53% 33.54% -7.27% 26.95% 17.35% 22.06% -26.55% 17.45% 6.67% 15

hi_01_05 79.02% 33.60% -7.46% 28.91% 11.32% 21.76% -24.84% 15.73% 15.90% 13

hi_01_06 84.63% 33.42% -6.91% 26.88% 17.55% 21.97% -26.02% 17.73% 5.21% 14

hi_02_02 86.89% 34.67% -10.88% 25.59% 21.50% 21.57% -23.73% 18.17% 2.83% 13

hi_02_04 83.63% 35.01% -11.97% 27.59% 15.38% 20.94% -20.10% 16.79% 10.22% 12

hi_02_05 87.59% 33.56% -7.35% 27.09% 16.89% 21.26% -21.95% 18.08% 3.31% 13

hi_02_06 85.06% 34.90% -11.61% 27.20% 16.57% 20.90% -19.90% 17.00% 9.08% 12

Legend1st

2nd

3rd

Other

Time period with the lowest amount of accumulated rainfall (06z7nov2006) has the highest associated percent error

Total Precipitation: 12z6nov2006 to 12z7nov2006

The simulations do not capture the magnitude or the trend of the rainfall over the 24 hours

**Note scale difference

Model Validation

Atmospheric – 32 km NARR Geo-potential height Specific Humidity/Water vapor mixing

ratio Wind and wind direction

Precipitation – 4 km NEXRAD Accumulated rainfall Temporal Distribution Spatial Distribution

Spatial Results: Percent Error

0102 Simulation 0205 Simulation

Spatial Error Analysis: Percent Error6 Hour 12

Hour

18 Hour

24 Hour

Conclusions

We need to improve our model WRF underestimates rainfall for all

parameterizations tested BMJ and Lin et al. model physics

work best Spatial conclusions Need to learn about the dynamics of

the atmosphere

Complete spatial analysis—error maps

Run simulations at higher resolution

With better results, look into feeding modeled precipitation into a river basin model (HEC-HMS)

How do WRF errors translate into stream-flow errors for flood forecasting?

Questions? Comments?

No Comments?

“Do you like statisticians?” **”Probably”