Relevance to SPoRT objectives

1
Evaluation of WRF Using High-Resolution Soil
Initial Conditions from the NASA Land
Information System
SPoRT Science Advisory Committee, 13 June
2007 Presented by Jonathan L. Case

• Relevance to SPoRT objectives
• Work done since joining SPoRT team
• Project overview
• Hypothesis
• Background on Land Information System (LIS)
• Experiment design
• Results
• Land Information System vs. Eta comparison
• Impacts on short-term numerical forecasts
• Summary / Future Work

2
Relevance to SPoRT Program Objectives

• LIS work a natural extension of MODIS SST
  research
• Focus on 0?24 hour forecast problems
• Use of NASA systems and tools
• Land Information System developed by NASA Goddard
  Space Flight Center (GSFC)
• Software capable of incorporating EOS datasets
  derived from MODIS
• Fosters collaborations with external agencies
• GSFC Hydrological Sciences Branch
• NASA Center for Computational Sciences computing
  resources
• LIS capable of being transitioned to operations
• NCEP/AFWA operations

3
Activities Since Previous SAC Meeting

• Mr. Case joined SPoRT team in June 2006
• New SPoRT project since previous SAC meeting
• Gained familiarization with the LIS software
• Conducted preliminary modeling experiments using
  LIS with the Weather Research and Forecasting
  (WRF) model
• Built collaborations between SPoRT and the
  Hydrological Sciences Branch at GSFC
• Presented conference paper at annual AMS meeting
  (2007)

4
Project Overview

• Hypothesis Can short-term mesoscale numerical
  forecasts of sensible weather elements be
  improved by using optimally-tuned,
  high-resolution soil fields?
• Project Goals Investigate and evaluate the
  potential benefits of using high-resolution land
  surface data derived from NASA systems and tools
  on regional short-term numerical guidance (0?24
  hours)
• Use LIS software to initialize soil temperature
  and moisturein the WRF model
• Examine one month period with relatively benign
  weather
• Isolate influence of land-atmosphere interactions
• May 2004 over Florida peninsula

5
The Land Information System (LIS)

• Software that runs multiple Land Surface Models
  (LSMs) efficiently using high-performance
  computing
• Developed by GSFC
• LSMs Noah, Community Land Model, SiB, VIC,
  Mosaic
• Global, high-resolution datasets (down to 1 km)
• User configurable features
• Spin-up time for soil equilibrium
• Input datasets
• Atmospheric forcing data
• LIS has been coupled to the WRF model
• Can run WRF using LIS LSMs and land datasets not
  available in the standard WRF

6
Experiment Design

• LIS offline simulation using Noah LSM
• Nested 9-km/3-km grid domain over SE U.S.
• Simulation from 1 May 2002 to 1 June 2004
• Output every 12 hours during May 2004to
  initialize WRF runs
• Atmospheric forcing datasets
• North American Land Data Assimilation System
  (NLDAS hourly, 14 km)
• Global Data Assimilation System (GDAS 6-hourly,
  52 km)
• GDAS used where NLDAS forcing is missing
• Compare regional WRF simulations with
  high-resolutionLIS soil data to WRF runs with
  Eta model soil data
• Calculate verification statistics at 80 surface
  stations
• Plot fields to compare phenomenology differences

7
Control WRF and LIS/WRF Configuration

• Common characteristics
• Nested grids 9-km and 3-km spacing
• Noah LSM
• Daily 24-hour forecasts during May 2004
  initialized at 0000 UTC and 1200 UTC
• Atmospheric initial boundary conditions from
  NCEP Eta model on 40-km grid
• Differences
• Control WRF Initial soil data from Eta model
• LIS/WRF experiment Initial soil data from 2
  year LIS run on exact WRF grids

8
Daily 0-10 cm initial soil moisture ()(0000 UTC
values during May 2004)
Eta soil moisture
LIS soil moisture
Difference (LIS Eta)
9
Daily 0-10 cm initial soil moisture ()(0000 UTC
values during May 2004)
Eta soil moisture
LIS soil moisture

• Much more detail in LIS (as expected)
• LIS drier, especially over N. FL S. GA
• LIS slightly more moist over Everglades

Difference (LIS Eta)
LIS Substantially Drier
10
Daily 0-10 cm initial soil temperature (C)(0000
UTC values during May 2004)
Eta soil temperature
LIS soil temperature

• LIS systematically cooler over most of domain

Difference (LIS Eta)
11
0-10 cm initial soil moisture ()(1200 UTC 6 May
2004)
Eta soil moisture
LIS soil moisture
Difference (LIS Eta)
12
Sample Sea Breeze Evolution Differences(9-hour
forecast valid 2100 UTC 6 May)
13
Sample Sea Breeze Evolution Differences(10-hour
forecast valid 2200 UTC 6 May)
14
Sample Sea Breeze Evolution Differences(11-hour
forecast valid 2300 UTC 6 May)
15
Sample Sea Breeze Evolution Differences(12-hour
forecast valid 0000 UTC 7 May)
16
Sample Sea Breeze Evolution Differences(Meteogram
plots at 40J and CTY)
17
Verification Stats 0000 UTC Cycle(29 forecasts
_at_ 80 surface stations)

• LIS/WRF runs reduced RMS errors by a few
  tenths of a degree over most forecast hours
• Nocturnal warm bias and daytime cold bias both
  improved

• Not much change in dewpoint verification stats
• LIS/WRF daytime dewpoints about 0.5C lower
  than control WRF
• Wind Speed (not shown) LIS/WRF improved
  nocturnal high bias

18
Summary / Preliminary Conclusions

• Configured and tested LIS/WRF on Florida case
• Initial soil fields generated on exact WRF grids
• LIS generated soil fields cooler and drier than
  Eta model
• Simulated atmosphere sensitive to changes in soil
  characteristics provided by LIS
• Demonstrated positive improvement in sea-breeze
  prediction on 6 May
• Improvements in diurnal prediction of 2-m
  temperatures during whole month (both 0000 and
  1200 UTC forecast cycles)

19
Proposed Future Activities with LIS/WRF

• Merge MODIS sea-surface temperatures with LIS
  soil data
• Study impacts of LIS soil data on convective
  initiation
• Different regional domains cases
• Varying weather regimes (e.g. supercells vs.
  air-mass storms)
• New case study period over Tennessee Valley
• Very warm March followed by killing freeze in
  early April 2007
• Use real-time MODIS greenness fraction products
  in LIS/WRF system
• Regional modeling ensembles
• Summertime forecast sensitivity to soil initial
  condition perturbations
• Run different LSMs within LIS/WRF for ensemble
  members
• Pathway to operational regional LIS/WRF runs

20
Alabama Freeze Case April 2007
21
Alabama Freeze Case April 2007

• Proposal
• Use real greenness fraction data in LIS/WRF
  simulations, derived from MODIS vegetation
  index composite products
• Measure impact on WRF forecasts compared to
  climo datasets

22
Backup Slides
23
Soil Moisture Grid-Wide Stats Land Points

• LIS is a few drier than Eta model in
  volumetric soil moisture
• Variation about mean is very similar to Eta
  model soil moisture

24
Soil Temp Grid-Wide Stats Land Points

• LIS 0-10 cm soil temperatures typically cooler
  than Eta at 00z
• LIS 0-10 cm soil temperatures about the same
  or slightly warmer at 12z

• LIS deeper soil temperatures consistently
  colder than Eta

25
Sample Sea Breeze Evolution Differences(Forecasts
from 1200 UTC 6 May Simulations)
26
Verification Stats 1200 UTC Cycle(Surface
station 40J)
27
Verification Stats 1200 UTC Cycle(Surface
station CTY)