1

Application of Satellite Land Measurements in Improving NCEP Numerical Weather Prediction

Weizhong Zheng, Michael Ek, Helin Wei, Jesse Meng, Jiarui Dong and John Derber

WWOSC, Montreal, Canada16 – 21 August, 2014

Environmental Modeling Center (EMC)NOAA/NWS/NCEP, USA

Xiwu Zhan and Jicheng Liu

NOAA/NESDIS/STAR, USA

2

Outline

Requirements of satellite products for NCEP land model Utilization of satellite data sets (e.g. GVF)

Assimilation of satellite products (e.g. Soil moisture)

Satellite radiance assimilation

Summary and discussion

3

Uncoupled“NLDAS”(drought)

Air Quality

WRF NMM/ARWWorkstation WRF

WRF: ARW, NMMETA, RSM

Satellites99.9%radar?

Regional NAMWRF NMM

(including NARR)

Hurricane GFDLHWRF

GlobalForecastSystem

Dispersion

ARL/HYSPLIT

Forecast

Severe Weather

Rapid Updatefor Aviation (ARW-based)

ClimateCFS

3.5B Obs/Day

Short-RangeEnsemble Forecast

Noah Land Model Connections in NOAA’s NWS Model Production Suite

MOM32-Way Coupled Oceans

HYCOM

WaveWatch III

NAM/CMAQ

Regional DataAssimilation

Global DataAssimilation

North American Ensemble Forecast System

GFS, Canadian Global Model

NOAH Land Surface Model

NCEP,NCAR,

UT-Austin,U. Ariz.,

& others

4

• Close surface energy & water budgets,

• Determine heat, moisture, and momentum exchange between surface & atmosphere,

• Noah land model provides surface boundary conditions to parent atmospheric model, e.g. NAM, GFS, CFS.

Role of Noah Land Model

NCEP-NCAR unified Noah land model

5

To provide these proper boundary conditions, land model must have:

• Atmospheric forcing to drive land model,

• Appropriate physics to represent land-surface processes,

• Initial land states, e.g. soil moisture/ice and snow, analogous to initial atmospheric conditions, though land states may carry more “memory”, especially deep soil moisture, similar to SSTs,

• Land data sets e.g. land use/land cover (vegetation type), soil type, surface albedo and emissivity, and associated parameters, e.g. surface roughness, soil and vegetation properties.

Land Model Requirements

6

Precipitation Incoming solar Incoming Longwave

Wind speedAir temperature Specific humidity

+ Atmospheric Pressure Example from 18 UTC, 12 Feb 2011

Atmospheric Forcing

Remotely-sensed

• Global Land Data Assimilation System (GLDAS) used in the NCEP Climate Forecast System (CFS) relies on a “blended” precipitation product, a function of:

• Satellite-estimated precipitation (CMAP), heaviest weight in tropics where gauges sparse.

• Surface gauge network, heaviest in mid-latitudes.• High-latitudes: Model-estimated precipitation

based on Global Data Assimil. System (GDAS).

7

Atmospheric Forcing: Precipitation

CMAP Surface gauge GDAS

8

4-km Snow Cover (daily integratedNIC IMS product)

24-km Snow Depth(daily integratedAFWA product)

02 April 2012

Initial land states: Snow Products

9

SMOPS Blended DailySoil Moisture

Initial land states: Soil Moisture

04 March 2014

ESA SMOS

NASA SMAP(Oct 2014 launch)

• Testing assimilation of SMOPS into NCEP global model.

10

Green Vegetation Fraction (Multi-year, 1/8-deg, NESDIS/AVHRR)

Vegetation Type(1-degree, SiB)

Soil Type(1-degree, Zobler)

Land Data Sets (e.g. GFS T574 ~25-km)

Green Vegetation Fraction (Weekly real-time, 1/8-deg, NESDIS/VIIRS)

Mid-Jan Mid-July Mid-July 2013Mid-Jan 2013

• Fixed annual/monthly/weekly climatologies, or near real-time observations:

Veg_Type, Soil_Type, GVF, Max.-Snow Albedo, Snow-Free Albedo.

• Some quantities may be assimilated into Noah, e.g. soil moist., snow.

Land Conditions: Wildfire Effects

11

Wildfires affect weather/climate systems:(1) atmospheric circulations(2) aerosols and clouds(3) land surface states (green vegetation fraction, albedo and surface temperature etc.) --> impact on surface energy budget, boundary-layer evolution, clouds & convection.

Wu & Ek: “A parameterization for land surface physical characteristics of burning areas for weather and climate models” . Aug.19: 11:40 - 12:00, Room 520 F

12

• Highest temporal and spatial resolution available.

• “Blended” LEO and GEO products; include in situ.

• Global domain, consistent use between NCEP global and regional models.

• Full consistency between all products, e.g. a “burned area product” also reflected in green vegetation fraction (GVF), albedo, emissivity, surface temperature, soil moisture, etc.

• Future: account for lakes, wetlands, water bodies.

• Land data assimilation in Noah land model using NASA Land Information System (LIS), i.e. snow, soil moisture, GVF (future Noah with dynamic/growing vegetation).

Remotely-Sensed Land Data Sets We Need:

Comparison of GVF between VIIRS and Climatology

Climatology data: (Gutman & Ignatov, 1998)

➢ 5 year mean monthly climatology from AVHRR;

➢ Resolution: 0.144 degrees (~16km);

➢ Period: April 1985 – March 1991 (1988 excluded)

VIIRS weekly real-time data:

By Marco Vargas, Zhangyan Jiang & Junchang Ju

➢ Weekly real-time;

➢ Resolution: 0.036 degrees (~4km, global) & 0.009 degrees (~1km, regional)

Algorithm: Uses VIIRS red (I1), near-infrared (I2) and blue (M3) bands centered at

0.640 μm, 0.865 μm and 0.490 μm, respectively, to calculate the Enhanced Vegetation

Index (EVI) and derive GVF from EVI.

Utilization of Satellite Data Sets in the Model Utilization of Satellite Data Sets in the Model

14

Analysis on Anomaly Correlation at PMSL: Day5

+0.007 ↑

Good improvement of the PMSL scores in NH.

15

Temperature fits to RAOBS: f24 and f48

Bias reduced

in the lower

atmosphere;

RMSE slightly

reduced.

Reduction of bias & rmse.

16

Precipitation Skill Scores over CONUS: f60-f84

Improved scores for medium & heavy precipitation but slightly degraded biases.

17

Snow Data Assimilation (temporal comp.)

Comparison of snow cover fraction between the MODIS (blue circles), the open loop simulation (black line) and the assimilation simulation (green line).

Comparison of snow water equivalent between the open loop simulation (green), the assimilation simulation (red) and the in-situ measurement (black) averaged over all SNOTEL sites in the study region.

Snow Cover Fraction

Snow Water Equivalent

* Jiarui Dong (NCEP/EMC)

Obs

DA No DA

Assimilation of Satellite ProductsAssimilation of Satellite Products

18

Schematic representation of assimilating satellite soil moisture products from NESDIS/SMOPS into NCEP Global Forecast System (GFS)

Cooperators: Xiwu Zhan & Jicheng Liu (NESDIS/STAR)

19

f36-f60: Improved slightly the scores and reduced the bias;f60-f84: Improved the scores for light and medium precipitation but not for heavy precipitation. Good improvement for the bias.

f36 - f60 f60 - f84

SMOPS DA in GFS: Precipitation Skill Scores over CONUS36-60hr and 60-84hr forecast for 2 April through 5 May 2012

Threat Scores: Black: w/o SMOPS Red: w/ SMOPS

Scores increased with SMOPS (>0).

Bias: Black: w/o SMOPS Red: w/ SMOPS

Error reduced with SMOPS (<0).

It requires a forward radiative transfer model (RTM) to calculate Tb with input of model atm profiles and sfc parameters. (GSI/CRTM) Surface-sensitive channels Tb simulation:

(a) Sfc parameters such as LST and soil moisture; GFS LST daytime cold bias: new Zot (Ops GFS in May, 2011 ) DICE (Diurnal Land/Atmosphere Coupling Experiment )

(b) Sfc emissivity (IR/MW)

Improvement of MW land surface emissivity model (Implemented in the latest release CRTM v2.1). Sub-grid scale land surface: sea, land, ice and snow.

Satellite Radiance AssimilationSatellite Radiance Assimilation

▶ Problems in surface sensitive channels data assimilation:

Much less satellite data (IR/MW) is assimilated over land than over ocean. e.g. in GFS/GSI, the large Tb bias can be seen over the CONUS from the GDAS radiance assimilation monitoring .

▻ West CONUS:

(a) Substantial cold bias of land surface skin temperature (LST) in GFS, resulting in the large simulated Tb bias (IR/MW) in the GSI; (b) In desert area, errors of emissivity calculation for MW; ▻ East CONUS: errors of emissivity calculation for MW.

▶ Approaches of improvement:

(a) Reduction of GFS LST bias with new formulations (Zom and Zot) (Zeng et al., U. of Arizona; Zeng et al. 2012; Zheng et al. 2012). (Operational GFS in May, 2011) (b) Improvement of MW emissivity calculation in the MW land emissivity model (Weng et al., 2001). (Implemented in the latest release CRTM v2.1).

New zot and z

om implemented in NCEP Ops GFS on May 9, 2011.

Large cold bias Improved !July 2010 July 2011

Diurnal variation of Tsfc

and T2m

Verified with SURFRAD Observation Network

http://www.srrb.noaa.gov/surfrad

Comparisons of Tb bias and rmse and the number of obs assimilated in GSI

Red: CTLBlue: SEN

24

Several satellite data sets developed recently were tested in the NCEP models and the results show good improvements, compared with the current data sets;

The weekly real-time VIRRS GVF data shows reduction of errors of temperature, humidity and wind speed forecasts, and improvement of precipitation scores. The SMOPS soil moisture product was assimilated in the GFS and it shows some positive impacts on NWP;

We have been continuing our efforts and working with many research teams to improve satellite data utilization over land in NCEP data assimilation system and then improve the GFS numerical weather prediction (NWP).

Summary and DiscussionSummary and Discussion

THANK YOU!

25