Assimilation of Globsnow data into HIRLAM Mostamandy Suleiman Sodankyla August 2011
Motivation Small number of observations (especially in eastern Europe and Russia) 00 cm of snow problem Non-uniform in time distribution of SYNOP snow observations
00 cm of snow problem Many stations do not report 00 cm snow Information spreads from existing stations into the area of no observations Snow edge is difficult to analyze
Distribution of SYNOP snow observations 00UTC 12UTC 06UTC 18UTC
Main Goal To improve data coverage To solve 0cm problem (Snow edge) To make distribution of snow observations more uniform in time To improve quality of forecast (T2M, RH2M, )
Globsnow www.globsnow.info
Globsnow Spaceborne passive microwave radiometer data combined with ground-based synop observations 25km spatial resolution Time resolution once per day Temporal coverage 1979-2011 NRT (near-real-time) service ongoing www.globsnow.info Reference: Pulliainen J. 2006. Mapping of snow water equivalent and snow depth in boreal and subarctic zones by assimilating space-borne microwave radiometer data and ground-based observations. Remote Sensing of Environment, Vol. 101, pp 257269.)
Globsnow Globsnow products Snow Water Equivalent (SWE) Snow Extent (SE) Output Formats 1. HDF (only data without coordinate) 2. NetCDF ( data + coordinate) 3. PNG ( Quick look )
Quality of data 2009
Quality of data 2009
Quality of data 2010
Quality of data 2010
Quality of data 2010
Quality of data 2010
Technical aspects Reading globsnow data Converting hdf to ascii Converting netcdf to ascii ( not yet ) Preparing file for SPAN Converting the ascii file into specob format Converting SWE to SD using (ρsnow =0.24) Modyfing SPAN Code cntrol.f getobs.f seterr.f setlim.f setoip.f siodev.f specob.f, Span script
Design of the experiments Hirlam version Latest 7.3 version New surface scheme ( newsnow ) Resolution 0.15 deg., 60 levels 3DVAR Time 15.Oct 24.Dec 2009 27. Mar 27. Apr 2010 20 Nov 20 Dec 2010 Globsnow observations Use them when available, some days with missing Globsnow file If exist, use the same file for every cycle of the day for winter 2010-2011 they are available near real time (only one day is missed) Resolution of the globsnow data: ~25km, comparable to the grid box size 28.01.11 16
Different data scenarios More than 30 exp. have been done Reference: only SYNOP observations Globsnow data, elevation from gtopo (high resolution) Globsnow data using HIRLAM orography Globsnow data in addition to the reference Globsnow data without SYNOP data Globsnow data only at 00 UTC without synop Globsnow thinned 4 times (to 75 km) Results from last experiments (2010) are shown Reference Globsnow without synop Globsnow + synop Reduced resolution of globsnow Globsnow assimilated once per day 28.01.11 17
Results 27 March 2010
results 10 April 2010
Results 25 April 2010
Results 20 November 2010
Results 15 December 2010
Results 20 December 2010
Results
Results Forecast verification Dec.2010
Results Forecast verification April.2010
Results Forecast verification April.2010 SNOW_73_SWGO TYPE LL BIAS RMSE STDV TT 0-0.200 2.074 2.065 TT 6-0.143 2.537 2.533 TT 12-0.005 2.667 2.667 TT 18 0.066 2.520 2.519 TT 24 0.034 2.723 2.723 TT All -0.084 2.459 2.458 SNOW_73_SWSG N BIAS RMSE STDV 32829-0.209 2.059 2.048 32821-0.179 2.526 2.519 16256-0.038 2.654 2.654 16577 0.036 2.521 2.521 16263 0.012 2.704 2.704 114746-0.109 2.447 2.445 SNOW_73_ref_SPRING_2010 N BIAS RMSE STDV N 32829-0.209 2.046 2.036 32829 32821-0.203 2.512 2.504 32821 16256-0.059 2.641 2.640 16256 16577 0.021 2.491 2.491 16577 16263 0.005 2.672 2.672 16263 114746-0.123 2.429 2.426 11474
Results Forecast verification April.2010
Conclusions, problems and plans Data from Globsnow to SPAN technical problems are solved The autumn and spring experiments has been examined The snow extant is reasonable in Globsnow Snow depth underestimated in some area No snow on mountains No independent observation to verify against Globsnow data can be used together with synop data or separately? Globsnow can be used only once The quality of forecast not much depend on resolution of Globsnow data (25 -> 75) Changing R(obs. err.) matrix for Globsnow data Quality control for Globsnow data. Writing fortran interface to converting globsnow data form netcdf to ASCII Defining the new correlation function for Globsnow data