sport real-time vegetation dataset and impact on land surface and numerical models

SPoRT Real-time Vegetation Dataset and Impact on Land Surface and Numerical Models

Sixth Meeting of the Science Advisory Committee28 February to 1 March 2012

transitioning unique NASA data and research technologies to operations

National Space Science and Technology Center, Huntsville, AL

Relevance / accomplishments since 2009 SAC Importance of vegetation in models SPoRT real-time vegetation product Modeling sensitivity methodology Case study results Summary and future


Relevance to NASA/SPoRT

• Addresses NWS forecast challenge– Improved local model forecasts

needed, especially for warm-season processes

• Use of NASA EOS datasets– MODIS Aqua and Terra

• Data can be incorporated into both Land Information System (LIS) and Weather Research and Forecasting (WRF) models– Improve current SPoRT modeling capabilities– Expand upon data denial experiments with SPoRT partner WFOs– Supports the NASA-Unified WRF (NU-WRF) project:

A larger-scale NASA project hosted by GSFC


Accomplishments since 2009 SAC Meeting• 2009 SAC Recommendation: “Broaden the use of this [data denial]

or similar concepts to help quantify the impact of unique NASA data sets.”

• Developed vegetation dataset that can be used by SPoRT partners for real-time modeling– Wrote module to incorporate vegetation data into LIS and NU-WRF– Transitioned data for use by SPoRT partners in WRF model and

WRF-EMS without any required source code changes– Instructions document for WFOs to use SPoRT vegetation in WRF-EMS

• Impact investigations– 2011 intern student

• Publications and presentations– Annual AMS meetings (2011, 2012)– SPoRT virtual workshop (31 Aug 2011)

Importance of Vegetation in Models

• Evapotranspiration from Healthy Vegetation– Significant contribution of moisture transport into

boundary layer during warm season– Important to represent vegetation accurately in models

• Horizontal and Vertical Density of Vegetation– Greenness Vegetation Fraction (GVF), horizontal density– Leaf Area Index (LAI), vertical density– In Noah Land Surface Model (LSM), LAI is prescribed while

GVF varies spatially by model grid cell• Other parameters depend on vegetation (e.g. albedo)


SPoRT Daily Real-Time Vegetation Product• Continental-U.S. NDVI/GVF grid

– NDVI from real-time MODIS swaths, mapped to CONUS grid– Inverse time-weighted NDVI composites

• Query up to 6 NDVI values in previous 20 days at each pixel• Similar to original MODIS SST algorithm

– Daily composites generated since 1 June 2010– Calculate GVF on 0.01° grid for use in LIS/Noah LSM

• Following Zeng et al. (2000) and Miller et al. (2006)• Based on distributions of NDVImax as a function of land use class


minimax

minii NDVINDVI

NDVINDVIGVF

,

NDVImax,iNDVImin

GVF 1-year Diff on 15 July: 2010 to 2011

Big 1-yr diff in GVF– TX was very wet in

early summer 2010– Virtually no rain in

TX/OK in 2011– 1-yr reduction in GVF,

up to 40%+– Shows how much

GVF can change from year to year

– Climo doesn’t cut it!

2011 MODIS GVF actually compares better to NCEP climo Texas summers

should be dry!

2010 2011


Approach and Methods

• Assess impacts of real-time MODIS vegetation data on LIS/Noah and model forecasts of severe weather– Document model sensitivity to real-time versus

climatological greenness vegetation fraction (GVF)– Can incorporating real-time GVF improve simulations of

severe weather events?• Employ unique NASA assets in modeling study

– NASA Unified-Weather Research and Forecasting (NU-WRF)– Combination of several NASA capabilities and physical

parameterizations into a single modeling software package


Approach and Methods, cont.

Offline LIS run:Noah Land Surface ModelAtmospheric analyses from GFS Data Assim. System

....

1 June2008

1 June2010

Offline LIS/Noah run(spin up w/ NCEP GVF)

NU-WRF runs: 36-h fcsts

LIS snapshot output to initialize

WRF runs

Severe events simulated:• 10, 11 June 2010 (CO supercells)• 15 June 2010 (SE U.S. convective winds)• 17 July 2010 (Upper Midwest tornadoes/wind)• 27 April 2011 (Super tornado outbreak)• 22 May 2011 (Joplin, MO EF-5)• 24-25 May 2011 (OK/MS Valley tornadoes/severe)

Results: General Observations

• Substantial changes to sensible/latent heat fluxes, and 2-m temperatures/dewpoints in places

• However, most WRF simulations look similar to each other– Similar model errors in onset/timing of convection– Good/poor control simulations good/poor experimental forecasts

• Greatest impact occurs with maximum surface heating– 17 July 2010: “Clean” case (little pre-existing clouds/precip) with

convection firing late in the day


17 July 2010 WRF Case Study

• Urban areas can be resolved much better by the SPoRT GVF.• Higher GVFs prevail from NE to ND.

DSMOMA

MSP

17 Jul 2010, WRF 21-h fcst: 2-m Temp

• Higher SPoRT GVFs in the western portion of the focus area correctly led to lower forecast 2-m temperatures


17 July 2010, WRF 21-h fcst: 2-m Dewp/CAPE

• Higher GVF values generally led to higher 2-m dewpoints• Net result was increase in CAPE up to 1000 J kg-1


17 July 2010, WRF Fcst 1-h precip: 27/33 h

• Both model runs close on placement with initial development and movement

• Some intensity variations

• CNTL run moves convection southward through Iowa too quickly

• EXP run correctly re-develops convection along IA/NE border in region of higher residual CAPE

CNTL CNTLEXP EXP

DIFF DIFFOBS OBS



Summary and Conclusions

• Developed a CONUS GVF dataset– 1 June 2010 to date– Can be ingested into LIS, WRF, NU-WRF, and WRF-EMS

• Significant GVF changes from summer 2010 to 2011– Anomalously green over Texas in July 2010– Year-to-year variations cannot be represented by a climatology dataset

• Using SPoRT GVF showed some improvement on the 17 July 2010 severe case

• Real-time vegetation datasets have potential to enhance accuracy of forecast models


Future Work• Short-term

– Contribute results to COMET module– Additional case studies, verification statistics– Study impact of GVF differences over

Texas related to 2011 drought– Improve albedo in LIS/WRF using new

look-up table method based on input GVF• Long-term

– Collaborate with NCEP/EMC: NOAA Global Vegetation Index

• Enhanced climo (weekly vs. monthly)• Initial development of real-time product

documented in Jiang et al. (2010, JGR)

Original seasonal Albedo

Albedo: Look-up table

Backup Slides


Default GVF Dataset in WRF/LIS

• Five-Year Monthly Global Climatology– Derived from AVHRR Normalized Difference Vegetation Index

(NDVI) data from 19851991– 0.144° resolution, valid at mid-point of each month– Currently in Noah LSM within NCEP/NAM and WRF models– Operational at NCEP since 1997 (Ek et al. 2003)

• Cannot account for variations in GVF due to:– Weather/climate anomalies (e.g. drought, excessive rain)– Land-use changes since the early 1990s (e.g. urban sprawl)– Wildfires and prescribed burn regions



NASA-Unified WRF (NU-WRF)

• Demonstrate NASA model capabilities at satellite-resolved scales

• Based on Advanced Research WRF dynamical core• Incorporates numerous capabilities & NASA assets

– Physics schemes (Goddard radiation, microphysics)– Goddard satellite data simulator unit (GSDSU)– Land Information System (LIS) and LIS Verification Toolkit– Goddard Chemistry Aerosol Radiation and Transport (GOCART)– Model verification and numerous post-processing options– Coupling between WRF model, LIS/GOCART, GSDSU

GVF Comparison: 17 July 2010

• Improved resolution– Ability to resolve vegetation

variation in complex terrain • Greener in Western U.S.

– Nearly 20-40% over High Plains– High Plains rainfall well above

average in previous 3 months

NW NE

SW SE

• SPoRT GVFs:– Consistently higher in

western U.S.– Have more day-to-day

variability– NE: Slightly lower in

mid-summer, then higher in early Fall

– SE: Closest to NCEP– Warmer than normal

temperatures out East in Fall 2010

GVF Quadrant Comparison: SPoRT vs. NCEP; 1 June to 31 October


Land Surface Model Stats: NW Quadrant


• Higher SPoRT GVFs lead to:– Higher latent heat fluxes

due to increased evapo-transpiration

– Lower sensible heat flux initially

– More rapid drying of soil moisture

• By late Summer/Fall:– Both latent and sensible

heat flux increase– This is due to drier soils

that cause more rapid surface heating

GVF Diff (SPoRT – NCEP)

Latent Heat Flux Diff:Time of Peak Heating

Sensible Heat Flux Diff:Time of Peak Heating

Soil Moisture Diff

0 6 12 18 24 30 36Forecast Hour

2-m Dewpoint

-4

-2

0

2

4

0 6 12 18 24 30 36

Bias

(°C)

Forecast Hour

2-m Temperature

NCEP_NPLSPoRT_NPLNCEP_MDWSPoRT_MDW

(a) (b)


17 July 2010: Verification Stats Bias

Error Standard Deviation

0

2

4

6

0 6 12 18 24 30 36

Std

Dev

(°C)

Forecast Hour

2-m Temperature

NCEP_NPLSPoRT_NPLNCEP_MDWSPoRT_MDW

0 6 12 18 24 30 36

Forecast Hour

2-m Dewpoint

(a) (b)

22 May 2011: Joplin, MO tornado day

22 May 2011, WRF fcst 1-h precip: 23/27 h

• More intense 1-h rain rates in sportgvf runjust prior to tornadic event

• Both runs have too much false alarm in AR

• After event, sportgvf run better handles squall line evolution into Arkansas

• Reduced false alarm in central Arkansas

CNTL EXP

DIFF OBS

CNTL EXP

DIFF OBS


Daily GVFs: 1 June to 30 Oct 2010


sport real-time vegetation dataset and impact on land surface and numerical models

Documents

sport vegetation

research technologies

sport partners

sac importance of vegetation

developed vegetation

modis gvf

data denial experiments

nuwrftransitioned data