Motivation

Progress in Radar Assimilation at MeteoSwiss

Daniel Leuenberger1, Marco Stoll2 and Andrea Rossa3

1 MeteoSwiss2 Geographisches Institut, University of Bern3 Centro meteorologico di Teolo, ARPA Veneto, Italy

2

Motivation

Convection often missed in the model

model deficiencies

improper initial conditions

Prerequisites for convection

Prefrontal environment (instability,wind)

Trigger (frontal pressure disturbance,local low-level convergence)

Radar rainfall assimilation provides trigger at the right time and location

3

Latent Heat Nudging refresher

Simple, economic 4DDA scheme for radar rainfall

Forcing via buoyancy

Temperature adjustment given by ratio of radar and model precipitation

Vertical distribution given by model

Scale nearby or idealised profile if no suitable model profile is available

RadarModel

Rain

rate

Diabatic Heating

z

4

LHN Experiments

aLMo with 7km grid size, diagnostic precipitation

6 summer convection cases over Switzerland of airmass (2), prefrontal (2) and frontal (2) type

focus to role of low-level environment and response of model dynamics to radar forcing

mostly missed convection in CTRL runs, but one case was well captured

3-6h assimilation duration

Best radar estimate of surface precipitation from 3 Swiss radar stations (clutter reduction, vertical profile correction), measurements 5min apart.

5

Observation weight w(x,y,t)

Quality function based on visibility of radar

Extendable (e.g. clutter maps…)

6

22.7.2003 Case: Missed frontal convection

Free forecast

Assimilation

23

2

2

2

1

2

0

1

9

1

8CTRL LHN RADAR

7

Role of low-level Environment

OBS

CTRL from aLMo ANA 12UTC

LHN from aLMo ANA 12UTC

LHN from aLMo ANA 15UTC

Free forecast

8

Impact of improved low-level environment

3h sums (+1 to +4 h free forecast)

Additional three hours of conventional aLMo assimilation improve environment and thus precip forecast started from LHN!

LHN from12 UTCaLMo ANA

LHN from15 UTC aLMo ANA

9

Response of model dynamics to forcing

OBS CTRL

10

Findings

LHN is an effective convection trigger

Positive impact in QPF up to 5 hours

General improvement of postconvective environment (though sometimes locally too strong forcing during assimilation)

Weak overestimated precipitation is not sufficiently removed

Rapid loss of precipitation signals may be caused by wrong thermodynamical/dynamical PBL structure

Need to improve low-level atmosphere, particularly humidity

11

Errors in Radar Data can be a Problem !

6h cumulated clear sky echo6h cumulated model response

6h Assimilation of Clear-Sky Echos (CAPE = 800 J/kg)

12

Anaprop

Stable stratification (strong inversion) and no rain

assimilation of clear-sky echos (6h)

no model response (0% rain!)

updrafts of 6m/s (for PJC) and 12m/s (for OMC) are induced

no errorneous rain, but updrafts could possibly influence larger environment

13

Findings

Non-meteorological echos can be drastically amplified by LHN in unstable, moist situations

Area of echo seems to be as important as amplitude

Wind can drift rain out of forcing area

Problem can be reduced by quality control of data and by filtering the input data in the model

Effect is reduced in drier or more stable situations

14

Towards operational application

LHN promising for very short-range forecasts (up to 12h)rapid update cycle (aLMo/2, 18h forecasts per day, started every 3h)use in concert with other observations, particularly surface observations

Extended testsLong periods including different weather situationsaLMo/7km and aLMo/2.2km configurations

Sensitivity tests Radar quality (ground clutter)Composite size (Swiss Composite vs. Eurocomposit)