NOAA ESRL GSD Assimilation and Modeling Branch

1
NOAA ESRL GSDAssimilation and Modeling Branch
Hyperspectral soundings and the pre-storm
environment Assimilation of AIRS data into
the Rapid Refresh a little on satellite
Convective Initiation / Lightning DA
Steve Weygandt, Haidao Lin, Ming Hu, Jun Li,
Jinlong Li, Tim Schmit, Tracy Smith, Stan
Benjamin, Curtis Alexander, John Brown, David
Dowell, Brian Jamison, John Mecikalski
2
RAP Data assimilation engine for HRRR
Hourly cycling model
Data Assimilation cycle
Observations
HRRR
3
Use of GSI for Rapid Refresh

• ? NCEP, NASA GMAO supported full system
• Primary development by NCEP for operational DA
• Advanced satellite radiance assimilation
• GSI used by NCEP for GFS, NAM, and RTMA
• NASA GMAO work to create GSI-based 4DVAR
• Framework for hybrid ensemble system
• ? Community analysis system
• Many users and code contributors
• DTC work to make code available to research
  community
• Community-wide SVN code management

4
Rapid RefreshHourly Update Cycle
Hourly Observations RAP 2012 N. Amer
Rawinsonde (T,V,RH) 120
Profiler NOAA Network (V) 21
Profiler 915 MHz (V, Tv) 25
Radar VAD (V) 125
Radar reflectivity - CONUS 1km
Lightning (proxy reflectivity) NLDN, GLD360
Aircraft (V,T) 2-15K
Aircraft - WVSS (RH) 0-800
Surface/METAR (T,Td,V,ps,cloud, vis, wx) 2200- 2500
Buoys/ships (V, ps) 200-400
Mesonet (T, Td, V, ps) flagged
GOES AMVs (V) 2000- 4000
AMSU/HIRS/MHS radiances Used
GOES cloud-top press/temp 13km
GPS Precipitable water 260
WindSat scatterometer 2-10K
Nacelle/Tower/Sodar 20/100/10
Partial cycle atmospheric fields introduce GFS
information 2x/day Cycle hydrometeors Fully cycle
all land-sfc fields (soil temp, moisture, snow)
5
Challenges for regional, rapid updating satellite
assimilation

• Data availability
• -- Long data latency, short data cut-off, small
  domain
• ? Limited data availability
• Bias correction
• -- Cycled predictive bias correction in GSI
• -- Limited and non-uniform data coverage degrades
  BC
• Lower model top
• -- Many channels sense at levels near RAP model
  top (10 hPa)
• -- Use of these high peaking channels can degrade
  forecast

6
AIRS Data

• Launched May 2002 on NASA Earth Observing System
  (EOS) polar-orbiting Aqua platform
• Twice daily, global coverage
• 13.5 km horizontal resolution (Aumann et al.
  2003)
• 2378 spectral channels (3.7-15.4 µm)
• 281 channel subset is available for operational
  assimilation

AIRS Brightness Temperature (BT) simulated from
Community Radiative Transfer Model (CRTM)
7
AIRS Radiance Coverage in RAP

• 3 h time window (/- 1.5 h), in 3-h cycle RAP
  retro run

06Z
03Z
00Z
09Z
15Z
12Z
Brightness Temperature (BT) from AIRS channel 791
18Z
21Z
08 May 2010
8
AIRS SFOV Data

• Single Field of View (SFOV) soundings derived
  using CIMSS hyperspectral IR sounder retrieval
  (CHISR) algorithm (Li et al. 2000)
• Clear sky only soundings
• SFOV data from CIMSS

Sample retrieved soundings compared to
radiosondes
Typical moisture and temperature biases for SFOV
SFOV
Raob
Less vertical structure in SFOV profiles
9
Diurnal aspects SFOV T innovations (O-B)
SFOV Temperature innovations horiz., daily avg.
dependence on height, and time of day
AK / Grnlnd
Sample SFOV profiles compared with raobs
Eastern NA
Western NA/AK
--- (north)
Eastern NA
Central NA
Western NA/AK
West of AK
SFOV assimilation 400 -- 800 mb
400 mb
800
00z 03z 06z 09z 12z 15z 18z 21z
10
Compare AIRS SFOV with Raobs
Conditions for matched profiles 3-h time
window, less than 15 km horizontal distance under
clear-sky
Temp RMS
Temp bias
3 SFOV data sets obtained from UW CIMSS V1
first set V2 improved V3 best set
Cool
Warm
Cool
Warm
Cool
Mixing Ratio bias
Mixing Ratio RMS
Improvements in SFOV retrievals ? All results
Shown from V3
11
Overall Temperature bias (vs. raobs)
Comparison of SFOV T to radiosonde data
3h fcst T bias (00z,12z)
V1 first set V2 improved V3 best set
T SFOV no bias correct
CNTL No SFOV
cooler
warmer
Cool
Warm
Cool
Warm

• Correspondence between raob comparison and fcst
  impact
• Overall average masks diurnal signal
• Model bias as well as observation bias

12
Histograms of Moisture Innovations (O-B)
Radiosonde vs. SFOV retrievals
AIRS SFOV
Gaussian distribution, small bias
Large dry bias, correction needed?
Dry
Dry
Moist
Moist
13
SFOV Moisture Bias Correction
RAP verification against raobs
Moisture innovations No bias correction
Analysis
SFOV NO BC
CNTL No SFOV
SFOV WITH BC
Dry
Moist
Moisture innovations 15 bias correction
12-h fcst
12-h forecast
SFOV NO BC
CNTL No SFOV
SFOV WITH BC
Dry
Moist
14
SFOV T Qv assimilation forecast impact
Wind
Relative Humidity
9-day retro average
12h fcst RMS
Combined assimilation of SFOV T and Q (400-800
hPa) with bias corrections reduces 12h forecast
RMS (relative to rawinsonde data) for all
variables, most levels
Temperature
15
SFOV T Qv assimilation forecast impact
Wind
Relative Humidity
9-day retro average
12h fcst RMS
Temperature
Normalize Errors EN (CNTL EXP)
CNTL
16
SFOV T Qv assim normalized RMS errors
Vertical average 400-800 mb
WITH BC
Better
NO BC
Better
Worse
Wind
Worse
Relative Humidity
3h 6h 9h 12h
Temp - erature
Combined assimilation of SFOV T and Q (400-800
hPa) with bias corrections Small positive impact
Better
17
HRRR case study initialized from RAP 2100 UTC 10
May 2010
0-hr 850-500 hPa mean relative humidity ()
NO SFOV
WITH SFOV
SFOV data
Analysis 850-500 hPa mean relative humidity ()
from RAP AIRS SFOV run
SFOV water vapor mixing ratio (g/kg) at 750 hPa
Analysis 850-500 hPa mean relative humidity ()
from RAP control run
18
HRRR case study initialized from RAP 2100 UTC 10
May 2010
2 hr Forecast Reflectivity
NO SFOV
WITH SFOV
Radar data
Observed radar composite reflectivity
HRRR forecast reflectivity initialized from AIRS
SFOV RAP run
HRRR forecast reflectivity initialized from
control RAP
19
Challenges for regional, rapid updating satellite
assimilation

• Data availability
• -- Long data latency, short data cut-off, small
  domain
• ? Very limited data availability
• Bias correction
• -- Cycled predictive bias correction in GSI
• -- Limited and non-uniform data coverage degrades
  BC
• Lower model top
• -- Many channels sense at levels near RAP model
  top (10 hPa)
• -- Use of these high peaking channels can degrade
  forecast

20
Two month time series bias coefficients
How long a period to adequately spin-up bias
coefficient corrections predictors?

• Highly variable for different predictors and
  channels
• Some can take two months or more
• Problems due to big differences in data coverage
  for successive cycles (in contrast to global
  models)

AIRS channel 261 (CO2 channel, PWF 840 mb)
21
Temperature and Moisture Jacobians
Standard profile (0.01 hPa top)
RAP profile (10 hPa top)
Temperature
Artificial sensitivity due to low model top in RAP
dBT/dT (K/K)
Moisture
Artificial sensitivity due to low model top in RAP
(dBT/dq) q (K)
22
Radiance assimilation forecast impact
Wind
Relative Humidity
9-day retro average
6h fcst RMS
AIRS radiance assimilation with GSI bias
correction and channel selection reduces 6h
forecast RMS (relative to rawinsonde data) for
all variables, most levels
Temperature
23
RAP assimilation of lightning data

• Map lightning density to proxy reflectivity
• -- sum ground flashes per grid-box
• over 40 min period (-30 ? 10 min)
• REFLmax min 40, 15 (2.5)(LTG)
• Sin distribution in vertical

OLD specified relationship
NEW seasonally averaged empirical relationships
LTG and REFLmax REFLmax and vertical REFL profile
24
Lightning Flash Rate ? max reflectivity
NEW
Seasonally dependent empirical
Summer
OLD specification in RUC
Winter
25
SUMMER Reflectivity profile as a function of
column maximum reflectivity
Max dbz 40-45
Max dbz 30-35
Max dbz 35-40
Max dbz 45-50
26
WINTER Reflectivity profile as a function of
column maximum reflectivity
Max dbz 30-35
Max dbz 40-45
Max dbz 35-40
Max dbz 45-50
27
AVERAGE Reflectivity profile as a function of
column maximum reflectivity
Max dbz 30 - 35
Max dbz 40 - 55
Summer
Summer
40 dBz
30 dBz
Winter
Winter
Max dbz 35 - 40
Max dbz 45 - 50
Summer
Summer
44 dBz
36 dBz
Winter
Winter
28
Applications lightning DA technique Can apply
technique to lightning data and satellite-based
indicators of convective initiation ? GLD-360
lightning data -- good long-range
coverage Especially helpful for oceanic
convection ? SATCAST cloud top cooling rate data
-- good Convective Initiation (CI)
indicator Avoiding model delay in storm
development SATCAST ? work by Tracy Smith
using data provided by John Mecikalski
proxy flash rate - 2 x cloud-top cooling
rate (K/15 min)
29
Sat obs
Radar coverage
Rapid Refresh oceanic lightning assimilation
example
24 Apr 2012
Observed reflectivity
16z
No radar echo
Latent heating-based temper-ature tendency
Lightning flash rate
No radar coverage
16z
30
Sat obs
No radar echo
Rapid Refresh oceanic lightning assimilation
example
24 Apr 2012
No radar coverage
Observed reflectivity
16z
16z
16z 1h GSD RAP forecasts
17z
17z
LTG DA ? slight impact on RAP forecast storm
clusters
with LTG
NO LTG
31
Assimilation of satcast cloud-top cooling rate
CI-indicator data
Cloud-top cooling rate helpful for initializing
developing convection in GSD RAP retro tests
5 July 2012
18z
18z
5 July 2012
SATCAST cooling rate (K / 15 min)
IR image
17z
32
18z1h
19z
WITH satcast assim
Obs Reflect
Assimilation of Satcast cooling rates provides
more realistic short-range forecast of
convective initiation and development
18z1h
NO satcast assim
33
18z2h
20z
WITH satcast assim
Obs Reflect
Assimilation of Satcast cooling rates provides
more realistic short-range forecast of
convective initiation and development
18z2h
NO satcast assim
34
AIRS Assimilation Summary / Future Work

• Small positive impact in RAP forecasts obtained
  from assimilating of AIRS SFOV data with
  application of simple bias correction (competing
  with full mix of conventional
  observations)
• Assimilation of AIRS radiance data in RAP
  produces small positive impact for winds,
  temperature, relative humidity and heavy
  precipitation
• Work to address low model top issue
• (better channel selection, blend with GFS model,
  raise RAP top)
• Examination bias correction issues and cloud
  contamination, re-scripting RAP partial cycle to
  increase cutoff time
• Evaluate sensitivity AIRS data in conjunction
  with other satellite data types

35
LTG / satellite CI DA Summary

• Preliminary evaluation of impact from
  assimilation of two novel convection indicators
• GLD-360 lightning data
• -- good long-range coverage
• Helpful for oceanic convection
• ?Satcast cloud top cooling rate data
• -- good Convective Initiaation
• Avoid model delay in storm development
• Qualitative assessment ongoing

Plan HRRR runs from RAP w/ and w/o LTG, satcast