Transcript Set Title of Presentation here

Evaluation of the Latent heat nudging scheme
for the rainfall assimilation at the mesogamma scale
Andrea Rossa* and Daniel Leuenberger
MeteoSwiss
*current affiliation: ARPA Veneto , Centro Meteo Teolo
Motivation
QPF poorest performance area of NWP, especially in summer
(convection)
Rainfall assimilation to mitigate spinup effect
Characteristics of LHN
1
Radar observations for high-resolution NWP models
Buoyancy driven
4DDA, timely assimilation of high-frequency observations
This study: reevaluation of Latent Heat Nudging (LHN) within
high-res model using idealised and real radar data
[email protected]
WSN05 Toulouse, 5.-9. September 2005
Computational efficiency
Idealised Experiments: Setup and Strategy
Model: Local Model (LM) in idealised configuration
Unstable environment supportive for supercell storms
Reference simulation
4D fields for validation
2
Surface precipitation for assimilation
Assimilation simulations
Sensitivity to uncertainty in observations and environment
[email protected]
WSN05 Toulouse, 5.-9. September 2005
Proof of concept with identical twin simulation (CTRL)
Reference and CTRL simulation
z
x
total sfc rain
CTRL y
y
x
RH, W and qe
z
x
z , W and cold pool
y
+5%
x
x
x
3
y
[email protected]
WSN05 Toulouse, 5.-9. September 2005
REF
Delay: 12 min
[email protected]
WSN05 Toulouse, 5.-9. September 2005
4
Extrema of w
5
Transient phase: 45 min
[email protected]
WSN05 Toulouse, 5.-9. September 2005
Assimilation increments
Observation Uncertainty
Amplitude Error
Factor 0.5 (2) results in 88% (133%) of total precipitation
Environment and model dynamics damps error
Temporal resolution
Strong sensitivity
6
Combined amplitude and structure error
Potential strong sensitivity
[email protected]
WSN05 Toulouse, 5.-9. September 2005
Non-Rain Echoes
Non-Rain Echoes: convective conditions
7
y
6h Sum of Model Precipitation
y
x
x
[email protected]
WSN05 Toulouse, 5.-9. September 2005
6h Sum of Radar
Non-Rain Echoes: AP conditions
8
Vertical mixing due to strong, spurious updrafts
t
x
[email protected]
WSN05 Toulouse, 5.-9. September 2005
w
z
Environment Uncertainty
Perfect observations, degraded environment
Low-level Humidity
-4% (4%) error in PBL humidity  -14% (11%) precip. deviation
For slightly drier conditions: convection suppressed
Environmental Wind
9
For slightly moister conditions: multi-cell development
[email protected]
WSN05 Toulouse, 5.-9. September 2005
Can lead to distorted system dynamics
Uncertainty in the Environment: Error in Wind
Bias in environmental wind: +2m/s; assimilation only
10
y
x
x
x
Surface precipitation
DTLHN, cloud, precip
z , W and cold pool
[email protected]
WSN05 Toulouse, 5.-9. September 2005
z
y
[email protected]
WSN05 Toulouse, 5.-9. September 2005
11
Case study of 8. May 2003 21 UTC
200 km
Simulations with Dx = 2.2km, no CPS
17-18 UTC
Analysis
21-22 UTC
4h Forecast
RADAR
12
LHN
[email protected]
WSN05 Toulouse, 5.-9. September 2005
CTRL
13
1h Forecast
LHN
Analysis
[email protected]
WSN05 Toulouse, 5.-9. September 2005
8. May 2003 Storm
RADAR
Findings
LHN has good potential for triggering convection
Excellent results in identical twin simulation
Scheme sensitive to observation and environment errors
LHN triggers storm at right location and intensity
Radar data clearly improves QPF in the first hours
Outflow located too far downstream  position error
Storm kept in model for more than 4h
Storm environment important, particularly PBL! Use mesoscale
data assimilation to complement rainfall assimilation (AWS,
VAD/radial winds, GPS tomography, PBL profilers)
[email protected]
WSN05 Toulouse, 5.-9. September 2005
Case study of real storm
14
Non-rain echoes pose serious problem  Data quality
[email protected]
WSN05 Toulouse, 5.-9. September 2005
15
Thank you for
your attention !