Download ppt - COSMO meeting, Zurich, 20-23 September 2005

A.Montani; COSMO-LEPS: operational aspects and verificationCOSMO meeting, Zurich, 20-23 September 2005

COSMO-LEPS: operational aspects and

verification

Andrea Montani, Chiara Marsigli

ARPA-SIM Hydrometeorological service, Bologna, Italy

COSMO meeting, Zurich, 20-23 September 2005


Outline

• Introduction• Present status:

– archiving of COSMO-LEPS products;– test of different clustering-selection techniques.

• Related projects:– SPITLAEF (case studies + tests);

– SPCOLEPS (Alpine suite);– SPCOWIND (related to wind forecast);– DEISA (related to hydrological forecasting).

• Verification results.• Future plans:

– COSMO-LEPS as an ECMWF “time-critical application”;– “respond” to ECMWF EPS upgrade?


Introduction

• What is it? It is a Limited-area Ensemble Prediction System (LEPS), based on Lokal Modell and developed within COSMO (COnsortium for Small-scale MOdelling, which includes Germany, Greece, Italy, Poland and Switzerland).

• Why? The horizontal resolution of global-model ensemble forecast systems is limited by computer time constraints and does not allow a detailed description of mesoscale and orographic-related processes.

COSMO-LEPS project: combine the advantages of global-model ensembles with the high-resolution details gained by the LAMs, so as to identify the possible occurrence of intense and localised weather events (heavy rainfall, strong winds, temperature anomalies, …);

COSMO-LEPS forecasts to improve the short to medium-range

forecast (48h < t < 120h) of the so-called “severe weather events”.


The COSMO-LEPS suite @ ECMWFThe COSMO-LEPS suite @ ECMWFsince June 2004since June 2004

d-1d-1 dd d+5d+5d+1d+1 d+2d+2 d+4d+4d+3d+3

middle EPSmiddle EPS

youngest EPSyoungest EPS

clustering clustering periodperiod

0000

1212

Cluster Analysis and RM identificationCluster Analysis and RM identification

4 variables4 variables

Z U V QZ U V Q

3 levels3 levels

500 700 850 hPa500 700 850 hPa

2 2 time time stepssteps

Cluster Analysis and RM identificationCluster Analysis and RM identification

European European areaarea

Complete Complete LinkageLinkage

COSMO-LEPS

Integration Domain

10 Representative 10 Representative Members driving the Members driving the

10 Lokal Modell 10 Lokal Modell integrations integrations

(weighted according (weighted according to the cluster to the cluster populations)populations)

employing either employing either Tiedtke or Kain-Tiedtke or Kain-

Fristch convection Fristch convection scheme (randomly scheme (randomly

choosen)choosen)

COSMO-LEPS

clustering area

• suite running every day at ECMWF managed by ARPA-SIM;

• Δx ~ 10 km; 32 ML;• LM_3.15 since 18/4/2005;• fc length: 132h;• Computer time provided

by the COSMO partners which are ECMWF member states.


Operational set-up

Core products: 10 perturbed LM runs (ICs and 6-hourly BCs

from 10 EPS members) to generate probabilistic output (start at 12UTC; t = 132h);

Additional products: 1 deterministic run (ICs and 3-hourly BCs from the

high-resolution deterministic ECMWF forecast) to assess the relative merits between deterministic and probabilistic approach (start at 12UTC; t = 132h);

1 proxy run (ICs and 3-hourly BCs from ECMWF analyses) to “downscale” ECMWF information (start at 00UTC; t = 36h).


Dissemination to the COSMO community (+

Hungary)


Archiving of C0SMO-LEPS products

From 1 July 2005, COSMO-LEPS forecasts are archived on MARS at ECMWF (class=co).

Deterministic run (fc+0h to fc+132h every 3h).

Ensemble Prediction System:10 perturbed forecasts (fc+0h to fc+132h every 3h):

PLEV (500, 700, 850 hPa): Z, RH, T.

SURF: albedo, LCC, MCC, TCC, SW radiation flux, CAPE, hzerocl, snowlmt, mslp, T_2m, Td_2m, TMAX_2m, TMIN_2m, U_10m, V_10m, UVMAX_10m, large-scale rain, convective rain, large-scale snow, TP.

Forecast probability (various intervals and thresholds):SURF: CAPE, hzerocl, TMAX_2m, TMIN_2m, UVMAX_10m, TP, snowfall, showalter index.

Clustering information (population, clustering variable used, …).

back archiving of past runs (from 5/11/2002 onwards) will start soon.


Related projects

• SPCOLEPS (joint Italy and Switzerland ECMWF special project) to study possible modifications of the operational suite.

• SPCOWIND (joint Italy and Great Britain): new ECMWF special project so as to have the computer resources to run a limited-area ensemble system over North-Western Europe (within the EC project PREVIEW, WP: Windstorms).

• SPITLAEF (Italian ECMWF special project) to perform studies on limited-area ensemble size, clustering methodologies, model perturbations, EPS reruns, …

• DEISA (Distributed European Infrastructure for Supercomputer Applications): consortium of super-computing centres which can (hopefully!) provide computer time resources to rerun 1 full month (August 2002) of COSMO-LEPS forecasts for ensemble hydrological purposes (within the EC project PREVIEW, WP: Medium-range Plain-floods).


ALPINE suite (running on SPCOLEPS billing units)

Experimental suite running daily from 15 July 2005. Same configuration as the operational COSMO-LEPS (red),

but both clustering and integration domain are centred over the Alps (blue).

Products not yet disseminated, but saved at ECMWF.

On single events, differences between the 2 suites can be noticeble:

Prob maps of 3-day rainfall exceeding 50 and 100 mm (fc24-96h); forecast starting at 20050905 12UTC

In this case, more helpful maps from the Alpine suite

ope alp


COSMO–LEPS verification (SON 2004)

Verification grid


Verification of the distributions

Station observation

Grid point forecast

The verification has been made in terms of:

• Average value

• Maximum value

• 50th percentile (Median)

• 90th percentile

in a box


COSMO-LEPS

10-MEMBER EPS

tp > 20mm/24h

tp > 10mm/24h

tp > 10mm/24h

tp > 20mm/24h

tp > 30mm/24h

Average values

boxes 1.5x1.5 deg


COSMO-LEPS

10-MEMBER EPS

Average values

boxes 1.5x1.5 deg


COSMO-LEPS

10-MEMBER EPS

tp > 20mm/24h

tp > 10mm/24h

tp > 10mm/24h

tp > 20mm/24h

tp > 30mm/24h

Maximum values

boxes 1.5x1.5 deg


COSMO-LEPS

10-MEMBER EPS

Maximum values

boxes 1.5x1.5 deg


• COSMO-LEPS suite will soon become a “time-critical” application monitored by ECMWF:– involvement of ECMWF operators in the management of the suite;– file system dedicated and priority in job scheduling.

• Implement dissemination to the Czech Met. Service (if possible).

• Back archiving on MARS of past runs (from 5/11/2002 onwards).

• Modify (?) the operational configuration to “contrast” EPS upgrade:– EPS will increase the hor. res. from 80 to 50 km (TL255L40 TL399L62)

with 56-57 model levels between 40 hPa (approximate top-level height of LM) and the surface;

– EPS will have HIGHER vertical resolution than LM; is this a problem???– increase the COSMO-LEPS vertical resolution from 32 to 40 layers so

as to reduce the gap?– ask for 3-hour post processing of EPS until fc+144h.

• Study CAREFULLY the outcome of the Alpine suite.

Future plans


Thank you !


COSMO-LEPS

10-MEMBER EPS

50th percentile values

boxes 1.5x1.5 degtp > 10mm/24h

tp > 20mm/24h

tp > 10mm/24h

tp > 20mm/24h

tp > 30mm/24h


COSMO-LEPS

10-MEMBER EPS


boxes 1.5x1.5 deg


COSMO-LEPS

10-MEMBER EPS


boxes 1.5x1.5 degtp > 10mm/24h

tp > 20mm/24h

tp > 10mm/24h

tp > 20mm/24h

tp > 30mm/24h


COSMO-LEPS

10-MEMBER EPS


boxes 1.5x1.5 deg


Clustering-selection technique

EVALUATION OF THE METHODOLOGY

with respect to clustering settings:

• clustering intervals (SON 2003)

• ensemble size (JJASON 2004)

• clustering area (JJASON 2004)

Verification against “proxy-rain” (ECMWF deterministic forecast starting

at 00 UTC and cumulated from fc+6h to fc+30h).


CLUSTERING INTERVALS

Consider a fixed configuration in terms of ensemble size (10 RMs selected out of 2 EPS sets, 2eps-10rm) and the properties of the “reduced” (10-member) global ensemble in 4 different cases:

OPE: the 10 members are selected like in the operational set-up (clustering variables: z,u,v,q; clustering levels: 500, 700, 850 hPa; clustering times: fc+96h, fc+120h);

D2: like OPE, but clustering times: fc+24h, fc+48h;D3: like OPE, but clustering times: fc+48h, fc+72h;D4: like OPE, but clustering times: fc+72h, fc+96h.BS

S

Brier Skill Score: OPE has slightly better scores at all verification ranges (less evident for ROC area .. not shown);

Outliers percentage: results heavily depend on the verification range.

outliers


ENSEMBLE SIZE and CLUSTERING AREA (1) Evaluation of the properties of the reduced global ensemble for 3 different ensemble sizes

(either 10, or 15, or 20 RMs selected out of 2 EPS sets, 2eps-10rm, 2eps-15rm, 2eps-20rm) and 2 different clustering areas.

2eps-10rm (ope): the 10 members are selected like in the operational set-up (clustering variables: z,u,v,q; clustering levels: 500, 700, 850 hPa; clustering times: fc+96h, fc+120h);

10 alp: like OPE, but the clustering area is centred over the Alps (43-49N, 4-16E);

2eps-15rm: like OPE, but 15 members are selected;2eps-20rm: like OPE, but 20 members are selected.BSS (fulldom)

Brier Skill Score: 2eps-20rm has better scores at all verification ranges; larger impact from 10 to 15 members rather than from 15 to 20 members (the same for ROC area .. not shown);

Outliers percentage: the same story.

Outliers (fulldom)


ENSEMBLE SIZE and CLUSTERING AREA (2) Evaluation of the sensitivity to the clustering area used to select the RMs:2eps-10rm (ope): 10 members selected with the operational clustering area (30-60N,

10W-30E);10 alp: like OPE, but the clustering area is centred over the Alps (43-49N, 4-

16E);2eps-15rm: like OPE, but 15 members are selected;2eps-20rm: like OPE, but 20 members are selected.ROC

(alpdom)

ROC area: due to the smaller verification area, there are very few occurrences at high thresholds; this gives lesser statistical significance to the scores, despite the long verification period. No positive impact of reducing the clustering area (actually “ope” is slightly better than “10 alp”!); small impact in terms of BSS (at fc+66h).

Outliers percentage: the same story as for the ROC area.

Outliers (alpdom)

BSS (alpdom)


NW Europe suite (running on SPCOWIND billing

units)

Generation of a new limited-area ensemble forecast system for the prediction of wind and wind gusts over North-Western Europe.

Aim: provision of site-specific wind forecasts to be post-processed and combined with other products by UK Met. Office.

Same configuration as the operational COSMO-LEPS (red), but both clustering and integration domain are shifted northwards (blue).

Experimental suite will start on June 2006.