1. Key note: Short range NWP - SRNWP Thibaut Montmerle1,2, Jean-François Mahfouf 1 1 Météo-France (CNRM/GMAP) 2 WMO ET-EGOS’s PoC for HR-NWP

2. OUTLINES Introduction Example of a SRNWP system : AROME-France
SRNWP vs. Nowcasting
Towards the use of future instruments
Conclusions and challenges

3. SEVIRI 10.8 m simulated by AROME-France (zoom)
Context : Short Range NWP at convective scale
Non-hydrostatic models (in the 1-4 km horizontal resolution range) allow realistic representation of convection, clouds, precipitation, turbulence, surface interactions
Specific features :
Observations linked to clouds and precipitation can be considered (e.g radars)
Need detailed surface conditions, and coupling models to provide Lateral Boundry Conditions (LBCs)
Analyses must be performed frequently
Forecasts are very expensive in computational time
SEVIRI 10.8 m simulated by AROME-France (zoom)

4. Example of SEVIRI active radiances
Satellite data for SRNWP
Example of SEVIRI active radiances
SRNWP benefits from DA of satellite data at Global scale through LBCs
Data Assimilation of satellite data inherited from Global scale NWP systems : same obs operators, generally used with a finer thinning
Because of their high temporal resolutions, GEO satellites are essential, but LEO play an important role at high latitudes

5. Example of SRNWP system : AROME-France
Operational since 2008
Spectral limited area non-hydrostatic model with explicit moist convection
Horizontal resolution : 1.3 km
90 vertical levels (up to 10 hPa)
3D-Var assimilation (1-h window)
Coupling files : hourly forecasts from global model ARPEGE
Forecast range : up to 42 hours, times a day

6. Example of SRNWP system : AROME-France Assimilated observations
Same obs as ARPEGE (+) 5 SEVIRI/MSG radiances (with Ts inversion) (+) radar DOW and Z (RH) (–) GNSS RO, IR and MW sounders with a different set of channels

7. + model top lowered at 10 hPa
Example of SRNWP system : AROME-France
Time evolution of monthly averaged number of assimilated obs :
April 2015 : 1.3 km + 1h cycle
+ model top lowered at 10 hPa
Since 2015 :
more obs with high temporal resolution : SEVIRI, CONV, RADAR
no more IASI channels above 10 hPa
Radar data with a higher spatial resolution
Sept 2008 – 2.5 km
Oper with Radar DOW
April 2010
RADAR Z / IASI added 7

8. Example of SRNWP system : AROME-France 1h cycled 3DVar assimilation :
3DVar analyses use observations in a [-30min, +30 min] assimilation window
Analyses
Forecasts
Guess
Update of the forecast by IAU
Production cycle
Assimilation cycle
A SRNWP system needs to wait for his LBCs
Cut-off times are not a real issue (contrarily to Nowcasting Systems)
Scheduling of the operational tasks

9. Example of SRNWP system : AROME-France
Temperature
Time evolution of analysis error reduction by obs type
GEO - LEO
Radar Z strongly impact q (and wind) in low to mid troposphere in rainy conditions
Brousseau et al., 2013
Humidity
dry
rainy
dry
rainy

10. Cloud and precipitations analysis in NOAA’s RAP/HRRR
Observations
Map to cloud field
No cloud
Cloud
Unknown
Merge cloud field
Update hydrometeors
based on the cloud field
Stan Benjamin

11. SRNWP vs. Nowcasting Main goals of Nowcasting :
provide to forecasters analyses and very short term (6h) forecast asap
provide a better alternative to radar image extrapolation
AROME-PI
AROME-oper
Hydrométéores
Surface
Set up of a degraded version of AROME: AROME-PI
Very short cut-off times (10 min), excluding a lot of observations
Hourly analyses
No cycling: the most recent AROME forecast is used as a guess, asynchronous coupling files 11

12. SRNWP vs. Nowcasting Much shorter cut-off times
IASI/20
Averaged daily number of assimilated satellite data :
SEVIRI/20
AMSUA
AROME /20
AROME-PI /20
AROME
AROME/PI
SSMIS
MHS
Much shorter cut-off times
Much less satellite data assimilated
Very short timelines needed !

13. Use of future instruments
METOP-SG A+B ( )
IASI-NG, SCA
MWS, MWI + ICI, RO
Heritage : IASI, ASCAT, MHS, SSM/I
New instruments : ICI (ice clouds in the MW at high resolution) – SCA (improved info on strong winds)
MTG-I MTG-S 2021
Main challenge :
IRS => PC or L2
New instruments :
FCI and LI, how to use them for SRNWP ?
ADM-AEOLUS 2017
Need for local processing to get L2B HLOS winds in SRNWP
Usefulness of data from Sentinel ? SAR winds ? LST ?

14. Set up of future instruments
Impact of a MW GEO sounder onto SRNWP using Observation System Simulation Experiment (OSSE)
DA of simulated radiances from 6 channels around 183 GHz from a MW GEO sounder, on the top of a control AROME run
Impact assessed with respect to the know truth (“Nature Run”)
Relative impact of an accurate MW GEO sounder
on Relative Humidity AROME forecasts
Results suggests that
a MW GEO sounder, as accurate as LEO sounders (e.g. σo ≈ 2K) could lead to some NWP improvements.
Impacts would be much more limited if such an accuracy cannot be reached due to technology limitations.
(P. Chambon, F. Duruisseau
ESA funding) 14

15. Conclusions Compared to Global NWP, SRNWP :
is highly impacted by Conventional and Radar data (when raining)
makes more use of GEO (raw radiances, AMVs, cloud analysis) than LEO, except at high latitude when multiple LEOs available (> 60°)
do not necessarily requires data with very short timelines (contrarily to nowcasting systems)
High spatial and temporal resolutions are essential to capture rapid- growing Mesoscale Convective Systems and pre-convective conditions
GEO and multiple LEOs provide essential informations in clear air areas, esp. in mid to high troposphere
The continuity of these observations is essential

16. Challenges For satellite observations :
Need to describe horizontal and vertical gradients of wind, temperature and humidity
Detailed cloud information should be extracted (macro-structure and micro-physics)
Very stringent latency requirements for short assimilation cut-off times (esp. Nowcasting)
For a better use of satellite data in NWP :
• Huge number of spectral radiances to proceed : compression of information (PCs) and inter-channel correlation errors
• Data at high temporal and spatial densities: explicit observation error correlations / scaling issues in the observation operators

17. Challenges In terms of Data Assimilation schemes :
4D DA to better exploit informations in consecutive images
Ensemble DA : new variables could be initialized with suitable multivariate background errors (e.g. hydrometeors, aerosol concentrations, ... )
Cloudy radiances : needs to better simulate observations from new instruments (e.g. LI/MTG) and MW radiances (model+obs operators)
Coupling with ocean, chemistry and improved land surface schemes : new information to be considered (SST, aerosols, skin T, soil moisture, vegetation parameters, ...)
Need to define suitable metrics for the evaluation the impact of satellite data in SRNWP (e.g “convective scale FSOI”)

18. Thanks for your attention !
RADAR
AROME
26th of August h, 12h forecast

19. Satellite data for SRNWP
Observation type
Satellites/Instruments
Analyzed Model Variable
AMVs
MSG
U, V
Clear IR Radiances
MSG/SEVIRI
NOAA’s & MetOp-A/B HIRS
T, q
Clear and “cloudy” radiances
MSG/SEVIRI MW channels
MetOp-A/B IASI
AQUA AIRS
NPP ATMS & CrIS
Clear MW Radiances
NOAA’s AMSUA/AMSUB
MetOp-A/B AMSUA/MHS
AQUA AMSUA
DMSP SSMIS
Scat. winds
MetOp-A/B ASCAT
ISS RapidSCAT
GPS ZTD
Ground based GPS Stations

20. Example of SRNWP system : AROME-France
Average amount of assimilated obs :
AmsuA/B
All Obs
Sat
Surface
IASI
GPS ZTD
TEMP
AMDAR
SEVIRI
RADAR
Satellite data : 16 %
(10 % GEO + 6 % LEO)
SEVIRI : 44 %
IASI : 22 %
AMSU/MHS : ≈10 %
Much more radar data in rainy conditions !
(One day without strong rainy events)