22nd General Assembly, Krakow, 21 Nov. 2017

Tour d’ALADIN

http://www.umr-cnrm.fr/aladin/

Algeria

1- First results of the implementation of surface data Assimilation system «CANARI» on the Algerian HPCModel: ALADINCycle : cy40t1Number of observations : 40

2.2. Roads weather warnings for Algeria

2m

te

mp

era

ture

(°C

)

2- Development of new products based on ALADIN, AROME and ALADIN_DUST outputs2-1. Weather warnings

Wind Rain Snow/Ice Extreme temperatures Thunderstorms Dust storms

Austria

„:

• Development of a parameter/process based stochastic physics scheme for the cLAEF

prototype ongoing (convection permitting EPS based on AROME)

• Continuation of work on AROME-RUC Nowcasting prototype: Tests with Austrian

Mode-S data, Tests with OPERA data

• Migration of ALARO and AROME to new HPC including upgrades (switch 12/2017)

• „Tornado show case“ Vienna Airport 20171017: nice AROME performance

top left: INCA analysistop right: AROME forecastbottom left: IFS forecast

source: www.kurier.at

Belgium- ALARO running 4 x daily at 1.3km resolution (cy40t1 + Alaro-1 + ISBA)- Combination of Alaro-1 with SURFEX available in cy43t2

Bulgaria

● cy41t1 is operational since 1st of November 2017

● There are two Models:● ALADIN BG 5000 m horizontal resolution,

105 vertical levels, 72 hours forecast range

● AROME BG 2500 m horizontal resolution, 60 vertical levels, forecast range 36 hours

● Comparative verification of the old operational model and cy41t1 was performed during the e-suite running

Croatia

2hPa

6 hours

- mainframe computer upgrade (288 cores total)- ensemble B matrix- testing quality of LBCs from different models- testing different LBC options- MESSI project (meteotsunami forecast)

Relative humidity

Wind speed

Wind direction

RMSE

RMSE

RMSEARP LBCs+TOUCANS

Mean sea level pressure measured (black) and two forecasts coupled to ARPEGE

MSLP standard deviation and bias starting from analyses computed using different B matrices

measured

forecasts

Czech Republic• Background error covariances were

derived for the BlendVar System;

• 3D VAR is forced to act mainly on the small scale part of the spectra;

• Details in Bučánek and Brožková, 2017.

• Assimilation of MODE-S MRAR data in operations since June 2017

• Future high resolution preparations on new CHMI supercomputer NEC LX to reach its full configuration in early 2018 – strong precipitation case 29/06/2017

2km resolution, ARPEGEanalysis dynamical adaptation,6h precipitation sum fcst at +12h

2km resolution, ALARO 4,7kmBlendVAR dynam. Adaptation,6h precipitation sum fcst at +12h

Radar & gauges verifyingobservations

Data Assimilation Algorithm

FranceAROME Overseas recent improvements

- New Sea Surface Temperature initialization, daily updated (see Figure below)- Coupling Arome with a 1D ocean mixing layer model- Ensure a smoother initial time stepping of dynamical adaptation (using “IAU”)- Specifically tuned diffusion of hydrometeors

Hungary

● Data Assimilation:

– Experiments with non-conventional observations in AROME: GNSS ZTD and RADAR reflectivity and radial wind

● Very-high resolution tests (500m-1km with AROME):

– Modification of the shallow convection parametrization in the grey zone

● Research in AROME microphysics:

– Experiments of low-stratus cases, underestimated by AROME – ICE3 scheme developments

– Tests of two-moment LIMA scheme

● Convection-allowing ensemble system (test AROME-EPS):

– Representation of model uncertainty with SPPT scheme

– Operational application of a time-lagged AROME-EPS

Morocco

21st ALADIN General Assembly

A paper in under review: GPS Meteorology: Validation and Comparisons with Radiosonde for the Moroccan GPS Permanent Stations

A paper in under review: A two-step radar-gauge merging method for quantitative precipitation estimate with C-Band radar in Morocco

AROME(2.5km,90L), 3DVAR CY41t1 every 3 hours, Conventional data

Work in progress: Verification using ECMWF analysis

Work in progress in collaboration with aeronautical service

Work in progress: impact of high vertical resolution within a nesting framework

Impact of wind on data assimilation system in multivariate mode and flow dependent B-Matrix

First work on separately detecting fog and low clouds using satellite data and ground-based observations

PolandPre-operational version of CROCUS snow model is ready and running every day, providing forecasts of snow depth and snow water equivalent. AROME forecasts are used as an input. Special products were prepared for hydrology – forecast for river basins, as well as verification scores.

Map of snow depth for Poland Map of snow depth for river Dunajec

Time series of mean snow depth for river Dunajec

E-suite of ALARO1-vB is running for several months now. Basic scores are better, compared to ALARO1-vA.

RMSE for September 2017, 00 UTC runs temperature at 2 mALARO-v1A (black) and ALARO-v1B (yellow)

RMSE for September 2017, 00 UTC runsrelative humidity at 2 mALARO-v1A (black) and ALARO-v1B (yellow)

Scatterplot for snow depth Verification of swe for station Kasprowy Wierch

Verification of snow depth for station Kasprowy Wierch

PortugalOperations

Increasing the number of levels(46→60)

Increasing the number of daily runs (2→4)

DA Local Progress

Positive scores on 24-hour screen level forecasts by moving from a 6-hour to a 3-hour OI_MAIN cycling in AROME-Portugal with Iberian WMO BUFR data

Model Development

Successful validation of Alqueva Lake physiography in AROME-Portugal for ECOCLIMAP-II

DA Project Activities

Hosting “ALADIN Data Assimilation basic kit Working Days” in Lisbon

New subroutine ‘amdarWMO’ for processing WMO BUFR E-AMDAR observations over Europe in collaboration with OPLACE (CHMI)

T2M HR2M V10M

--- OPER__ 3-hour DA__ 6-hour DA

Romania

increasing the integration domain and the horizontal resolution (including the Black Sea) => better representation of precipitation pattern

the amount of precipitation in south-eastern part of the country is not accurately simulated

http://sharaku.eorc.jaxa.jp/GSMaP/

Observations

24 h cumulated precipitation

Slovakia

After discovering and correcting a serious bug in quadratic coupling (Fig.left) the ALARO-1vB with 4.5km/L63 on CY40t1bf07 has been declared operational on new HPC on 20/03/2017. It clearly outperforms previous operational version in most of the parameters (Fig. right).

Good performance of new operational system in cases of high impact weather. Example of strong wind event on 14/09/2017: correct indication of extreme wind gust. However in general the timing, intensity and localization of the extreme weather patterns is not always excellent.

wind gust by ALARO-1vB 4.5km wind gust INCA analysis33.2m/s

MSLP evolution: bugged and reference run T2m scores: old and new version

Meteogram Nitra

Slovenia● Upgrade of the operational suites and HPC

– cy40t2bf7 with ALARO-1vB physics

– 11 additional compute nodes, additional memory, robot tape library

STD

T2m

BIAS

NEW

specific humidity increment due to assimilated radar reflectivities

drying

moistening

OLD

● First steps towards radar data assimilation

Tunisia

Models installed on the HP research computer

Configuration of AROME-TunisiaAROME-Tunisia 1.3 1.3 kmkm(CY40, Coupled to ARPEGE 10Km, Time step 45s,90 vertical levels Installation of the cycle 40t1_ bf.05 on HP computer

ALADIN operational

AROME HARMONIE

Model version CY38t1_bf.03 CY38t1op2 CY38t1_bf.03

Spatial Resolution 7.5 km 2.5 km 2.5 km

Vertical Levels 70 60 65

Boundaries ARPEGE 10km ARPEGE 10km

ALADIN 7.5km

Time step 45 0s 45 s 45 s

Fig1: Vertical profile of the standard deviation of specific humidity (q), temperature (t), vorticity (v) and divergence (d) for AROME-TUNISIE during winter (blue line), inter-season (cyan line) and summer (red line) periods; AROME–Tunisie (mean of the 3 periods) (blue dot) and ALADIN-TUNISIE (green dot).

The catchement of Medjerda, located in the north-western region of Tunisia, is an area at risk that suffers from severe flooding every year.As it holds the biggest river in Tunisia and several dams around it, Medjerda watershed represents an important hydrometeorological study area.

AROME- MEDJERDA

Model version CY40

Resolution 1.3 km

Vertical Levels 90

Boundaries ARPEGE 10km

Time step 45 s

Center (8.75,36.5)

N° Points 128*85

AROME-MEDJERDA: a small domain over Medjerda catchment

Background error covariance matrix computation and

assessment for AROME 2.5 km & ALADIN 7.5 km

24H Accumulated precipitations of 25/02/2015 - from left to right : Observation, AROME-Medjerda 1,3km and ALADIN 12.5 km

Case study 09/02/2017 North-West system

Case Study 29/09/2016 East system

Model Inter-comparison: Motivation of migrating from Aladin 7.5km to Arome 1.3 on the new Calculator

From left to right : Observation, AROME 1,3km and ALADIN 7.5 km

A sample of case studies comparing AROME 1.3 km and ALADIN 7.5 km was made in order to justify the cost of the new HPC as AROME 1.3 is much more expensive in term of computation cost.

The following illustrations, North-West system and East system, present two atmospheric systems that characterize the main perturbations over Tunisia where we can see that regarding the localization and the intensity, AROME 1.3 km is much more accurate.

3DVAR Data Assimilation Implementation Operational & Parallel Suites

Turkey

Towards Operational 3DVar at TSMS (cy40)

- Calculation of B Matrix with NMC method - Observation Pre-Processing:

• Using all available BUFR conventional data on GTS.• Non-GTS-TR synop data converted from TAC to BUFR.• Validation of high resolution data included in BUFR TEMPs.• Operational use of SEVIRI radiances in GRIB.• AMSU-A Brightness Temperature tested.• Developing obs. monitoring tool in Python

- Surface Analysis:• CANARI configuration was run on the soil in cy40.

- Coupling:• ECMWF LBCs were successfully tested in the assimilation in cy40

Updates of HPC Resource - New HPC installation and acceptance tests have been finished on July 2017 (SGI ICE XA 4032 core, E5-2690 v4, 2.60GHz, 192GB DDR4 RAM/node, ~168Tflops peak)