MESO-NH cloud forecast verification with satellite observation

MESO-NH cloud forecast verification with satellite observation Jean-Pierre CHABOUREAU Laboratoire d Aérologie, University of Toulouse and CNRS, France http://mesonh.aero.obs-mip.fr/chaboureau/ DTC Verification Workshop 2010, Boulder, Colorado, 2 November 2010

Motivation CRM = cloud resolving model (Δx~1 km), it explicitly represents the circulation in a cloud system Simulated cloud fields over Brazil Identification of systematic errors in parameterizations e.g. microphysics 5 hydrom. species, 35 processes Measure for cloud forecasts as clouds affect the radiation budget and can produce rain

Outline Cloud verification with satellite observation A way to identify systematic errors in cloud models A too large ice content as seen in the thermal infrared Using a long statistics to evaluate a tuning of the microphysics A physical consistency obtained in the microwaves A measure for cloud forecasts Sensitivity of forecasts to initial conditions: example of a Mediterranean heavy rain event over southeastern France Sensitivity of forecasts to synoptic activity: example of one month of daily 48-h forecasts over West Africa

Méso-NH A research model at the state of the art Jointly developed by Météo-France and CNRS Non-hydrostatic (Δx = 100 km 10 m), anelastic system A large number of parameterizations Physics: radiation, turbulence, deep and shallow convection parameterization, statistical cloud scheme, one- and two-moment microphysical schemes Chemistry & aerosols in gas and aqueous phases Coupling with ocean, hydrology, electricity, etc. Run on real and idealized conditions: 1D, 2D, 3D, nesting Post-processing and diagnostics budgets, profilers, trajectories, satellite, radar, lidar, GPS Parallelized, vectorized (from PC cluster to NEC, IBM,...) More on http://mesonh.aero.obs-mip.fr/

model MESO-NH satellite Our approach: model to observation Radiative Transfer code Simulated Tb Quantitative comparison with scale similarity Observed Tb IR: RTTOV (parameterization) MW: ATM (int. size dist.) Active: Home-made (int. size dist.) High clouds (Tb 10.8 µm) Clouds/precip (183 to 37 GHz) Cirrus/dust (ΔTb 8.7, 10.8, 12 µm) Overshoots (ΔTb 6.2, 10.8 µm) 3D clouds/precip. (lidar/radar)

Verification of cloud cover An African squall line (15-17 Aug. 2004) Δx=40, 10, 2.5 km 2160 km MSG Observation Meso-NH Simulation

Control of cloud ice Key role of autoconversion of ice to snow for ice content control False similarity in µphysics with autoconversion droplets rain However most of the schemes use a Kessler-like formulation R iauts ( 0, r - r ) = kismax i i inverse time constant: k is =10-3 e 0.015 T C s -1 critical mixing ratio: r i *= 5.0 10-4 kg kg -1 literature : r i *= 1.0 10-5 kg kg -1 (Fowler et al. 1996) r i *= 1.8 10-4 kg kg -1 (Hong et al. 2004) Chaboureau et al., JGR 2002

Microphysical tuning Meteosat-5 r i *=5 10-4 kg kg -1 r i *=2 10-5 kg kg -1 r i *=2 10-5 kg kg -1 r i *=5 10-5 kg kg -1 r i *=5 10-4 kg kg -1 Chaboureau et al., JGR 2002

A refined tuning for cirrus Verification at t+21h with BTD 8.7-10.8 µm Autoconversion=f(T) (Ryan 2000) Δx=30 km ( -5 0,06 ( T-273,16) 3,5 2 10, ) ri = min 10 Statistics over 30 days Cirrus cover defined with BTD>1K Chaboureau and Pinty, GRL 2006

Convective overshoots in Brazil Model 4 (Δx=625 m) Geophysica flight track Altitude (km) 22 km Model 3 (Δx=2.5 km) Color: Total water content Black line: 6.2-10.8 µm BTD=3K Red line: Cloud limit Color: 10.8 µm BT Chaboureau et al., ACP, 2007

Verification of hydrometeors - I A tropical case: Hurricane Bret (1999) TMI MNH Δx=1.6 km An overall correct simulation of BTs Wiedner et al, JGR 2004

Verification of hydrometeors - II A midlatitude case: a front in Feb. 2000 Almost no scattering effect! Δx=10 km Better agreement when changing the optical properties from dry snow to soft parameter Meirold-Mautner et al, JAS 2007

Statistical assessment Joint BT distribution for AMSU and SSM/I 5 midlatitude rain events (SIMGEO cloud database) 150 GHz BT depression w. respect to 90 GHz due to ice scattering 85 vs. 37 GHz OK Polarization effect over sea reduced due to emission by rain Chaboureau et al, JAMC 2008

Outline Cloud verification with satellite observation A way to identify systematic errors in cloud models A too large ice content as seen in the thermal infrared Using a long statistics to evaluate a tuning of the microphysics A physical consistency obtained in the microwaves A measure for cloud forecasts Sensitivity of forecasts to initial conditions: example of a Mediterranean heavy rain event over southeastern France Sensitivity of forecasts to synoptic activity: example of one month of daily 48-h forecasts over West Africa

Sensitivity to initial conditions A 5-day episode resulting in 400 mm of rain MESO-NH initialized with ECMWF (ECM) ARPEGE (ARP) ALADIN (ALA) Acc. precipitation between 00 UTC 19 Nov and 00 UTC 24 Nov from raingauges and Meso-NH Time series of 3-h accumulated precipitation averaged over the CEV and ALP subdomains Clark and Chaboureau, JGR 2010

Heavy precipitation on Nov 2007 Rain conditions as seen from space (top) B3m5 and (bottom) 37V GHz BT from AMSU-B and SSM/I observations, at 1800 UTC 20, 21, 22 Nov. Black line is Δ37 at 40 and 50 K. Clark and Chaboureau, JGR 2010

Stratiform regime on 20 Nov 2007 Sensitivity of rainfall to surface wind speed Surface wind speed at 1900 UTC 20 Nov. from SSM/I observation and Meso-NH simulations. Black line is Δ37 at 40 and 50 K. Clark and Chaboureau, JGR 2010

Convective regime on 22 Nov 2007 Prediction of convective storms over the sea Time (hour) associated with 10.8 μm BT < 225 K from 0000 to 2300 UTC 22 Nov. Clark and Chaboureau, JGR 2010

Convective regime on 22 Nov 2007 Change in the prediction of precipitable water Precipitable water (mm) at 0700 UTC 22 Nov. from SSM/I observation and Meso-NH. Blue line is the CAPE at 200 and 400 J/kg Clark and Chaboureau, JGR 2010

A month of forecast verification Daily 48-h real-time runs from 23 July to 22 August 2006 over West Africa during the AMMA campaign initialized/coupled with ECMWF forecasts, Δx=32 km deep and shallow convection and subgrid cloud schemes An example of (not too bad) forecast MSG observation Meso-NH forecast t+36h Δx=32 km 12 UTC 14 Aug 2006 3840 km Söhne et al, MWR 2008

Characteristics for high clouds Right amount of clouds, but too patchy High clouds defined with 10.8 µm BT<160 K Cloud systems as groups of adjacent grid points with BT<260 K Patchiness = (# cloud systems + # clear-sky areas) / (# grid points) Söhne et al, MWR 2008

Categorial scores for high clouds Increase in score with larger-sized area Probability of Detection, False Alarm Ratio Heidke Skill Score Fractions Skill Score FSS uniform Söhne et al, MWR 2008

HSS change with synoptic activity MSG cloud fraction (shading, %) and 700-hPa meridional wind contours at -3 m/s (dashed) and at 3 m/s (solid). averaged between 10 and 15 N (a) MSG observations and ECMWF analyses and (b) Méso-NH forecasts at D + 1. The skill score increases with AEWs (c) HSS calculated point-by-point (line) absolute meridional wind intensity (shading, m/s) Söhne et al, MWR 2008

Summary Use of simple diagnostics taken from remote sensing BT space, in particular for high clouds (BT 10.8 µm) and BT difference for cirrus (8.7-10.8 µm), overshoots (6.2-10.8 µm), convection (183 1-183 7 GHz), A way to identify systematic errors in cloud models A too large ice content as seen in the thermal infrared Using a long statistics to evaluate a tuning of the microphysics A physical consistency obtained in the microwaves A measure for cloud forecasts Sensitivity of forecasts to initial conditions: sources of uncertainties identified with satellite observations Sensitivity of forecasts to synoptic activity: increase in skill score with forcing conditions related to African easterly waves