Snow-atmosphere interactions at Dome C, Antarctica Éric Brun, Vincent Vionnet CNRM/GAME (Météo-France and CNRS) Christophe Genthon, Delphine Six, Ghislain Picard LGGE (CNRS and UJF)... and many colleagues from both labs. Boulder, October 2011
2 Outline Boundary-Layer studies at Dome C : work from C. Genthon and colleagues Snow-atmosphere coupled simulations
LGGE - IPEV: Boundary layer on the Antarctic plateau
FROM LAST YEAR'S MEETING IN TOULOUSE: - Maintain instruments for CONCORDIASI SOP 2010: Yes - Compare ARPEGE / AROME CONCORDIASI analyses: No - Process sonic data => Turbulence (e.g. TKE) and energy balance analyses: To some extent - Expand from surface: - to full BL (tethered kites/balloons ; profiler) - to troposphere (clouds, satellites): To some extent
Air temperature C Maintain the observing system: now more than 2 years on record 2009 2010 very contrasted : 2009 / 2010 warmest / coldest winter in the decade
Model evaluation of the BL temperature ECMWF and MAR: - Agreement broadly good - Inversions under-stimated Comparison with ARPEGE / AROME still to be done
Sonic data: Direct turbulence evaluations
First comparisons betwween observed and simulated turbulence
Beyond 45 m? - Model / meteorological analyses - One radiosounding per day - Radiometers / profilers (AMSTRAD / P. Ricaud) - Flying devices
Dome C / Concordia : a very convenient site to simulate snow-atmosphere interactions Flat, horizontal and quasi-infinite terrain Relatively homogeneous snow surface (scale dependant) Moderate wind velocity Sophisticated instrumentation at Concordia base: BSRN radiation observations LGGE/IPEV and PNRA boundary layer and surface observations LGGE continuous snow temperature profiles LGGE detailed density, temperature and SSA profiles in Summer Radio-soundings... Differences with South Pole station: high diurnal cycle more challenging under IASI swath! 10
11 Boundary layer observation from a 45m tower (LGGE) for process studies and models evaluation
... assimilation of Infra Red radiances : a major challenge for most meteorological models Assimilation schemes rely on the capacity of meteorological models to simulate the observations (observation operators) IASI measures IR radiances over 8468 IR wave bands which are emitted by the surface and modulated by the atmosphere Measured radiances by IASI are very sensitive to surface temperature Snow temperature variations are poorly captured by simple snow schemes Concordiasi field campaign: an ideal framework to investigate the sensitivity of IASI data assimilation to the complexity of snow representation in meteorological models 12
13 Methodology of the present study Evaluation of a 11-day stand-alone simulation of Crocus snow model: Observed meteorological forcing: Downwards LW and SW observations from BSRN Air temperature, humidity and wind speed from LGGE No significant precipitation and no snow drift during the period Initialization from an observed temperature profile (L. Arnaud) Adjustment of initial top density (2cm hoar-like layer) and snow grains detailed evaluation of surface and near-surface temperature Fully coupled snow-atmosphere simulation using the meteorological model AROME and the same configuration of Crocus Evaluation of the coupled simulation both in the boundary-layer and in the near-surface snowcover
14 Main characteristics of Crocus : an ancient 1-D snow model now embedded in SURFEX LSM Physical processes: Snow /rain precipitation Radiative balance turbulent fluxes Liquid Water content Density Temperature Snow grains Ground thermal flux Water run-off - thermal diffusion - water flow, phase changes - spectral albedo, light penetration - compaction - metamorphism - snow/soil coupling - wind compaction - variable number of snow layers - no blowing snow - no interaction with vegetation - no air flow inside the snowpack
15 Snow surface temperature observations
Performance of Crocus in stand-alone mode: hourly snow surface temperature Stand-alone simulation ISBA_ES/Crocus Observation from emitted LW (BSRN) 2010 January 20th. to 31st January 2010 bias = +0.05 K rmse = 1.16 K correlation = 0.983 Input meteorological data from BSRN (ISAC-CNR) and LGGE 16
Simulation of the propagation of the diurnal heat waves inside the snowcover 2010 January 20th. to 31st Observed snow temperature profile from Laurent Arnaud 17
18 Simulated energy budget of snowcover Sensible heat released to the atmosphere during cloudy days
19 Characteristics of the coupled simulation Same configuration and initial state of the snow model 20 numerical layers z0 = 1 mm Richardson limit = 0.1 AROME : regional model for Numerical Weather Prediction 2.5km, non-hydrostatic, domain 625 x 625 km² centered around Dome C 60 vertical levels bottom 3 levels : 8.5, 27 and 51 m Turbulent Kinetic Energy as a pronostic variable turbulent fluxes boundary conditions and daily initial states from ARPEGE ARPEGE: global model stretched over Antartica, 4D-Var ARPEGE/IFS library : Météo-France, ECMWF, ALADIN/HIRLAM, Meso-NH Full coupling between snowcover and the atmosphere thanks to SURFEX (externalized land surface model) Cycling of the snowcover throughout the 11-day simulation
Coupled simulation results: surface temperature at the Dome C gridpoint bias = +1.29 K rms = 3.54 K correlation = 0.86 20
21 Surface Temperature ( C) Very encouraging improvement...... if we refer to operational ARPEGE forecasts Operational ARPEGE output Observation from emitted LW Operational ARPEGE output - too warm Observation from emitted LW - too weak diurnal cycle 2010 January 20th. to 31st
Coupled simulation results: near-surface snow temperature 22
Coupled simulation results at Dome C grid point: wind speed and boundary layer temperature 23
24 Poor performance of cloudiness simulation... bias = -0.2 W m 2 rmse = 28 W m 2 r = 0.05! => Is it due to a snow-atmosphere feedbacks?
25....same problems for most meteorological models! A very challenging issue: - deficencies in polar cloud micro-physics? - deficiencies in water vapor assimilation?
Snow surface temperature during the second half period (26-01 0h loc. time to 30-01 0h) bias = +0.69 K rmse = 1.8 K correlation = 0.97 26
27 Conclusions and perspectives Good performance of the detailed snow model in stand-alone mode: - diurnal cycle of surface temperature - diurnal heat-wave propagation through the snowpack -... but high sensitivity to the top snow layer characteristics (density, grain size) Performance of the coupled snow-atmosphere simulation: - reasonable simulation of surface temperature - good simulation of the propagation of diurnal heat-waves through the snow pack - good simulation of boundary-layer characteristics: profiles of wind speed and of the diurnal temperature amplitude - main errors due to a poor simulation of cloudiness Evolution of snow schemes in meteorological models: - Improvements to be expected from an evolution of snow schemes over the Antarctic Plateau in NWP and climate models - An extensive use of the driftsondes is planned for extending the present study to the assimilation process, to other areas and other seasons
28
29 Simulated boundary-layer conditions
30 Prevailing meteorological conditions
An example of density simulation from observed forcing data (Col de Porte, 1320 m a.s.l) Snow Depth (cm) Winter season 2005-2006 Nov. 1st Dec. 1st Jan. 1st Feb. 1st Mar. 1st Density(kg/m3) Apr. 1st May. 1st 31
First test on IASI assimilation in ARPEGE (V. Guidard) Surface temperature over Antarctica 32 Grid points frequency) Standard ARPEGE snow scheme Modified snow schemes Obs guess departure (K)