Transitioning Land Surface Skin Temperature and Snow Improvement to Operation at NCEP/EMC

1

Transitioning Land Surface Skin Temperature and Snow Improvement to

Operation at NCEP/EMC

Xubin Zeng (University of Arizona, Tucson) Zhuo Wang (NESDIS)

Mike Barlage and Fei Chen (NCAR) Helin Wei, Weizhong Zheng, Mike Ek (NCEP/EMC)

11 October 2012 [email protected]

2

Assimilated NOAA-18 AMSU-A channel 15 (89 GHz) data in operational GDAS in July 2007(Zheng et al. 2012)

Q: For our JCSDA project, why do we focus on the improvement of NCEP land model (Noah)?

Less data assimilated over land

Even less over arid regions in daytime

..partly because of Noah deficiencies

3

Q: For our JCSDA project, why do we focus on the improvement of NCEP land model (Noah)?

Snow data are assimilated, but the initial information is lost too fast.

..partly because of Noah deficiencies

GFS/Noah 7-day forecastingof snow over one grid cellin western U.S. in April 2010

1. Skin temperature over semi-arid regions

July 2007

Zheng and Mitchell (2008) Zheng et al. (2012)

2

TsDiff

GVF

4

CLM has a similar problem

2

5

Obs

Zeng et al. (2012)

Question

Can we develop unified formulations to improve both Noah and CLM in the simulation of the Ts diurnal cycle over arid regions?

Yes (Zeng et al. 2012; Zheng et al. 2012). Main ideas:

1. To improve the formulations of roughness lengths 2. for momentum and heat; 3. To improve the treatment of stable turbulence in 4. the atmospheric boundary layer through the 5. interplay between sensible and ground heat fluxes

2

6

Noah CLM

37

New

CON

Results: daytime and nighttime improvements

Desert Rock

Gaize

New-10 layers

Mean absolute deviation (K)

Desert Rock GaizeNoah (Con) 2.8 5.8 Noah (New) 0.5 1.6

CLM (Con) 1.9 4.6CLM (New) 0.7 1.8

118

9

Tb bias in satellite pixels used in GFS GSI (NOAA-17 HIRS-3 Ch. 8) (Zheng et al. 2012)

Con New

Case: 18Z, 20070702

W CONUS: 243

Veg_7: 66

W CONUS: 483

Veg_7: 274

CON NEW

Tb from NOAA-18 AMSU-A Ch. 15

10

Tb simulation for NOAA-18 AMSU-A Ch. 1

GFS CON CON + improved surface emissivity

CON + improved surface emissivity +improved Noah

11

1212

2. Major snow deficiency of Noah over forest areas

Snow melts too early, too fast

Offline Noah Simulations of SWE over Niwot Ridge, CO Obs SWE Noah

GFS/Noah 7-day forecasting

This deficiency has been known for a few years for Noah and some other land models

13

Our solution (Wang et al. 2010):

Main ideas : Vegetation shading effect; Snow density adjustment; Under-canopy resistance; Revised Z0m under snow condition

Using existing Noah model structure (for easy operational implementation)

1414

Niwot Ridge forest (40.03N, 105.55W, 3050 meters)

July 2006 –June 2007

Results are also improved over grassland

Snapshot 20110317 120h Forecast(Wei et al. 2012)

Improvement

Con

New

Obs

15

Time series over 100-120W, 50-60N

Operational GFS

NewObs

3/17 3/2516

1-month (11 Mar – 10 Apr 2011)

Q: Can better snowpack forecast lead to better prediction of other fields?

17

1-month (20110311-20110410)

Obs

New

Impacts on other forecast scores (e.g., 500 hPa height, precipitation) are small 18

Our Noah improvements (UA) perform as well as the explicit canopy model (MP) at maintaining snowpack in spring.

They will be released in next WRF release in Spring 2013

Barlage et al. (2012)

Evergreen Needleleaf WRF Snow in an idealized 6-mon simulation

Impact of Noah improvements in WRF v3.4.1

CON MP NEW

19

Summary

Improvements in Noah daytime Ts were implemented in GFS in May 2011 - a successful R2O transition;

they also increase the number of GSI-assimilated satellite Tb data from surface-sensitive IR and MW channels;

Nighttime Ts improvements are ready for testing in GFS;Diurnal Ts improvements have been tested in CLM and are

ready for implementation.

Noah snow improvements have been tested in GFS and are ready for implementation;

They have also been tested in WRF and will be available to the community in the next WRF release in Spring 2013.

2

20

21

• Noah Ts improvements don’t necessarily improve T2m or troposphere temperature

Better forecast of snowpack doesn’t necessarily result

in better surface temperature forecast,

because GFS atmospheric model had been tuned to partially compensate for Noah land model biases.

• We have to improve both land and atmospheric models together (so that GFS GSI could assimilate more satellite data over land and have a positive impact on NCEP forecasting).

Critical Issues