Use of Geostationary Satellite Observations for Dynamical Support of Model Cloud Fields

Arastoo Pour Biazar1, Richard T. McNider1, Kevin Doty1, Yun-Hee Park1, Maudood Khan2, Bright Dornblaser3

1. University of Alabama in Huntsville2. The Universities Space Research Association3. Texas Commission on Environmental Quality (TCEQ)

Presented at9th Annual CMAS Conference, October 11-13, 2010

Friday Center, UNC-Chapel Hill

Motivation

Use of Satellite Data to Improve the Prediction of Clouds in Air Quality Decision Models

1. The uncertainties in physical atmosphere such as temperature, moisture, winds, and cloud prediction greatly impact air quality model predictions. e.g., clouds have an important role in regulating the photochemical reaction rates, heterogeneous chemistry, and the evolution and partitioning of particulate matter.

2. Numerical meteorological models have traditionally had significant problems in creating clouds in the right place and time compared to observed clouds. This is especially the case during air pollution episodes when synoptic-scale forcing is weak.

3. The initiation and assimilation of clouds in weather forecast models has been the subject of many investigations. Yet, cloud prediction remains a pervasive problem and evidence suggests that there are still major errors in cloud placement. This is especially true at the spatial scale at which air quality models operate.

4. The problem in cloud prediction is particularly frustrating in air quality SIP modeling since they are retrospective simulations in which the observed cloud field is known from satellite observations but models have significant differences in cloud placement.

5. The overall purpose of the current effort is to improve model location and timing of clouds in the Weather Research and Forecast (WRF) meteorological model which is widely used in the air quality planning community.

Background

At UAH considerable attention has been given to the use of geostationary satellite observations to improve the representation of physical atmosphere in the air quality simulations.

One such effort has been replacing model cloud transmissivity with satellite observed transmissivity in air quality models. However, while this technique provided improvements in model performance, it produced a physical inconsistency in the model system. Insolation and photolysis fields were derived from satellite data, but deep convection or cloud venting of the boundary layer or the impact of cloud water on long wave radiation or chemistry could be in the wrong location according to model placement of clouds.

The new activity attempts to create an environment in the model conducive to creating clouds where there are observed clouds and remove clouds where the observations indicate clear sky.

Previous attempts at using satellite data to insert cloud water have met with limited success. Previous studies have also indicated that adjustment of the model dynamics and thermodynamics is necessary to fully support the insertion of cloud liquid water in models.

Our approach is to develop relationships between satellite-derived cloud properties and targeted variables internal to the model such as grid scale vertical velocity and to provide the dynamical and thermodynamical support needed to sustain or clear model clouds.

This technique was first implemented in MM5 and we are in the process of transitioning it to WRF. WE are also pursuing a simple alternative threshold approach.

0.65um VIS surface, cloud features

ASSIMILATION OF GOES-DERIVED CLOUD PRODUCTS IN MM5

Use satellite cloud top temperatures and cloud albedoes to determine a maximum vertical velocity (Wmax) in the cloud column (Multiple Linear Regression).

Adjust divergence to comply with Wmax in a way similar to O’Brien (1970). Nudge MM5 winds toward new horizontal wind field to sustain the vertical

motion. Remove erroneous model clouds by imposing subsidence and suppressing

convective initiation.

W<0

W>0

Underprediction

Overprediction

Satellite Model/Satellite comparison

Multiple Linear Regression Variables

• Total Cloud Depth • Number of Cloud Layers• Depth Wmax Layer• Wmax• Height of Wmax• Est. 1-h stable precip.• Est. 1-h convective precip.• Height of Max. Diabatic

Heating• Magnitude of Maximum

Diabatic Heating• Layer with Upward Motion

Satellite Products

• Cloud Albedo• Cloud Top Pressure /

Brightness Temperature• Short Wave Radiation• Surface Albedo• Skin Temperature

Surface incident shortwave radiation in W m-2 for July 2, 1999. Assimilation and satellite observation plots are @ 14:45 UTC, and the control run is @ 15:00 UTC. Control: Control run with no assimilation. Assimilation: Run with assimilation of satellite cloud information. Satellite Observation: Derived from GOES–8 satellite.

Control Run AssimilationSatellite Observation

A

B

C

Downward shortwave radiation in W m-2 at 2200 UTC 6 July 1999.

(A) Derived from GOES–8 satellite. (B) Control run with no assimilation. (C) Run with assimilation of satellite cloud information.

MODEL

ASSIMILATION

MODEL

CNTRL

Satellite OBSERVED

Insolation

SATELLITE DATA IS UTILIZED TO CORRECT MODEL CLOUD FIELDS IN A DYNAMICALLY CONSISTENT MANNER

BIAS = (Total model predicted cloud) / (Total observed cloud)

Series 1: control simulation; Series 2: satellite assimilation

CALL PROGRAM

WADJ

READ MM5,SATELLITE,SFC DATA

INTERPOLATETO SIGMA-H

GRID

CLUSTER ANALYSES

CALCULATETOTAL CLOUD

DEPTH

REMOVE SHALLOW

CLOUDS

CALCULATEZHMAX

ESTIMATEPRECIPRATES

CALCULATEZWMAX

CHECK FORPRIORWMAX

CALCULATEWMAX

CALCULATEWMAX CLOUD

DEPTH

CALCULATENUMBER OF

CLOUD LAYERS

CALCULATEMAXIMUMHEATING

DETERMINESTABLE/CONV.

CATEGORY

DO STABLEADJUSTMENT

DOCONVECTIVEADJUSTMENT

REMOVEERRONEOUS

CLOUDS

DETERMINENUDGING

TIME SCALES

CALCULATEUPWARD

DEPTH

INTERPOLATE/AVERAGE BACK TO

MM5 GRID

ADD NEWDIVERGENT

COMPONENTS

CALCULATENEW

DIVERGENTCOMPONENTS

ADJUSTDIVERGENCE

FOR WMAX

SUBTRACTORIGINAL

DIVERGENTCOMPONENTS

CALCULATEORIGINAL

DIVERGENTCOMPONENTS

CALCULATEORIGINAL

DIVERGENCE

WRITE OUTNUDGING

FILES

MM5CONTINUEWay to complicated and computationally expensive for operational use

Flowchart showing the flow of the processes needed for cloud

adjustment at each hour

MM5 PAUSE:WRITE SPECIAL

HISTORY

The Way Forward

Two different tracks are followed:

Streamline the current technique and implement it in WRF.

o Clearing erroneous clouds are more difficult in WRF. WRF’s response to suppressing the convective parameterization is different from MM5 (WRF compensate by creating grid resolved clouds).

Revisit the problem and develop a simpler approach.

o Focusing on daytime clouds, revisit the relationship between internal model cloud variables and relate them to what satellite can observe.

Case study: summer of 2006; WRF configuration: 36-km grid spacing, CONUS with 42 vertical layers; SW radiation: Dudhia; LW radiation: RRTM; Monin_Obukhov similarity with NOAH LSM; PBL scheme: YSU; Microphysics: Lin; Cumulus parameterization: Kain-Fritsch; IC/BC/nudging: EDAS.

10.7um IR sfc/cloud top temperature0.65um VIS surface, cloud features

Use of Daytime Cloud Albedo/Cloud Top Temperature for Model Evaluation

Model Cloud Albedo Satellite Cloud Albedo

Underprediction

Overprediction

Areas of disagreement between model and satellite observation

Model cloud albedo Model cloud top temperature

Areas of Underprediction/Overprediction can be identified for Correction

Kain-Fritsch scheme

0

0.05

0.1

0.15

0.2

0.25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

days in August

WRF-C,GOES-NC WRF-NC,GOES-C

Overprediction

Underprediction

Agreement between the model and OBS

0.6

0.62

0.64

0.66

0.68

0.7

0.72

0.74

0.76

0.78

0.8

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

days in August

Kain-Fritsch Grell-Devenyi New Grell

Agreement Index (AI) =(Clear/Cloudy agreements) / (Total Number of Grids)

Increased cloudiness over the domain

Weak Synoptic-scale forcing

Evaluating Model Cloud Prediction During August 2006

cloud albedo

RH < 95% RH > 95%

positive W

Functional relationships between cloud water and/or cloud albedo with model vertical motion was not clear. Thus, threshold relationships with vertical motion and relative humidity were examined. A contingency probability approach, where the coincidence of clouds/clear occurring with positive/negative vertical motion were examined.

Model Cloud, W, and Relative Humidity

 

CLEAR CLOUD

OCEAN LAND OCEAN LAND

height (m) w (m/s) RH (%) w (m/s) RH (%) w (m/s) RH (%) w (m/s) RH (%)

sfc 1000 -0.00253 72.08438 0.00377 39.52232 0.004865 99.12765 0.01269 99.6

1000 2000 -0.00588 59.14449 -0.00278 51.23995 0.034022 97.07111 0.02132 99.9

2000 4000 -0.00499 49.06997 -0.00745 41.42338 0.045954 95.62551 0.04551 100

4000 7000 -0.00608 40.36083 -0.01002 31.64465 0.054684 101.8438 0.06112 100

7000 10000 -0.01260 44.54638 -0.01433 36.94441 0.058007 99.79606 0.05639 98.96

10000 13000 -0.01579 47.13423 -0.01054 33.53775 0.065545 97.62615 0.05350 96.8

13000 ~top 0.00018 33.25936 0.00067 19.85797 0.044565 94.18938 0.03255 93.2

Threshold Table for target W (August 2006 Simulation)

Alternative Simple Approach for Creating Dynamical Support for Clouds

Obtain threshold vertical velocities and moisture needed to support cloud formation from WRF.

From GOES observations identify the areas of cloud under-/over-prediction and use the threshold information to obtain the needed vertical velocity in the model to achieve agreement with observations.

Having the threshold vertical velocity as the target, use one dimensional variational technique to calculate new divergence fields and target horizontal winds.

Use the new horizontal winds and threshold moisture fields as nudging fields in WRF to sustain the target vertical velocity.

Original wind field @ 25th layer

New wind field

A Snapshot of the Modified Wind Field For August 19.

Color filled contours represent vertical velocity. Horizontal wind fields were

modified to sustain the target vertical velocity.

Preliminary Results From

1D-Variational Scheme

Control WRF simulation: August 19, 2006, 20 GMT

Preliminary Results

Satellite assimilation: August 19, 2006, 20 GMT

CONCLUSION & FUTURE WORK

• Statistical relationships between clouds and WRF model variables were examined. While functional relationships were not clear, an examination of coincident relations showed that threshold relations between vertical motion and relative humidity were very robust.

• 98% of the model cloudy grids were associated with positive vertical motions and over 65% of the grids with clear condition were associated with negative vertical motions. This largely confirms the working hypothesis that in a GOES black and white image, white areas are associated with lifting and negative areas with subsidence.

• The statistics suggested that imposing positive/negative vertical velocity can create/clear clouds where there are inconsistencies with GOES observation.

• Preliminary results from the use of threshold vertical velocities are encouraging.

• The results presented here are preliminary and the technique needs further refinements. Coincidently, relative humidity will be adjusted below clouds to be consistent with model statistics.

Acknowledgment

The findings presented here were accomplished under partial support from NASA Science Mission Directorate Applied Sciences Program and the Texas Commission on Environmental Quality (TCEQ).

Note the results in this study do not necessarily reflect policy or science positions by the funding agencies.