Joint MUSCATEN and NetICE Workshop http://muscaten.ut.ee/SNOW10

Modelling of snow-ice-atmosphere interactions

Kuopio, Finland, 24-26 March 2010

Notes of the working group on snow/ice data assimilation

draft - LR 7.4.2010

The group discussed various aspects of the present and near future snow and ice data assimilation, based on conventional and satellite observations. The discussion was somewhat biased to HIRLAM and CANARI, which seemed to be the only systems where snow data assimilation is operationally applied, represented in the working group. Participants of the group session: Ardi Loot, Eric Bazile, Eva-Stina Kerner, Irene Suomi, Jure Cedilnik, Kalle Eerola, Karoliina Ljungberg, Laura Rontu, Mariken Homleid, Marko Kaasik, Riinu Ots, Suleiman Mostamandy.

Satellite snow

In the workshop presentations by Jouni Pulliainen several possible new sources of satellite snow and ice information were presented. First results of implementing snow water equivalent (SWE) data by the Globsnow project (snow.fmi.fi) were reported by Eerola and Mostamandy. The working group suggests:

• To continue testing of the Globsnow data, using updated and corrected version of them. It seems reasonable to assimilate SWE directly, using the snow density information available in HIRLAM for possible conversions between snow depth and water equivalent. For this, error characteristics valid for satellite SWE should be obtained and applied. Kalle and Suleiman, FMI, will work with this within HIRLAM. The methods, error statistics and data should be easily applicable also in HARMONIE and other environments of snow data assimilation. It is expected that later the SWE data developed within Globsnow will be available to everyone within the EUMETSAT Hydro-SAF programme.

• To try EUMETSAT Land-SAF snow cover, especially to define the snow border when SYNOP observations do not necessarily report zero snow depths. Jure, Env.Agency of Slovenia, has been working with Land-SAF albedo and will continue with the snow cover within CANARI. FMI colleagues will find out if the second version level data are available for users, as it cannot yet be found at the Land-SAF web site in Portugal. Suleiman and Kalle, FMI are also interested in testing these data. Exchange of data and codes was agreed.

Practical aspects of snow analysis with optimal interpolation

Problems and suggested solutions to several practical questions of the optimal interpolation snow analysis applied in HIRLAM and in the CANARI surface assimilation were touched in the workshop presentations by Homleid, Cedilnik, Eerola and Mostamandy. The group discussed some of them, and suggested that the solutions suggested in the presentations should be tuned and applied.

• SYNOP snow depth observations are available unevenly in space and time. This may lead to unnecessary jumping of snow depth from one forecast-assimilation cycle to another when SYNOP observations present the only source of information.

• Relaxation of the first guess snow to climatology has been used as an insurance against missing observations, but in most cases it leads to problems. Bias correction of the observations has been applied but also mostly led to problems. Both of these should be avoided in their present form in HIRLAM and CANARI.

• A problem of not reporting zero snow depth may lead to erroneous snow distribution when somewhere within the influence radius snow is reported or found in the background (first guess) forecast. In the present code, the influence radius of the snow observations has been mostly too large.

• Quality control of the observations and the first guess needs improvements (relaxation) in order to avoid rejection of correct observations. It has been seen that too strict quality control tends to reject also most of the satellite data introduced into the system. Especially, in the start of forecast-assimilation cycles from so called cold start (first guess from climate information) quality control should be minimal to allow the observation information enter.

Validation and open questions

In the forecast model snow parametrizations, several (prognostic) variables are used whose observed values are not well known and whose definitions may vary in different models and according to model resolution: snow density, snow fraction (snow cover in a grid-square), snow albedo. Climatological information is used, but even a good snow climatology represents a long-time average that might be totally inapplicable during a particular forecast. In the model, snowfall

and snow melting are parametrized and influence the forecast directly and the snow data assimilation via the background (first guess). The working group suggests that some more validation of the snow-related variables could be done:

• Snow density is regularly measured within snow courses (lines of observations over selected locations e.g. in Finland, Russia and other countries, measurements made once a week or month or so). These measurements would allow checking of the climatological and forecast snow densities over some representative locations.

• Within a NWP model, climatology of the snow depth analysis increments could be studied and compared with the modelled and measured snowfall and snow melt.

• Connection between predicted (and analysed) screen-level temperature and humidity, melting of snow, soil moisture could be studied in the framework of the Kalman filter surface data assimilation, under development in SURFEX.

Ice issues

The working group shortly discussed ice cover and water surface temperature assimilation in NWP. Sources of data presently used e.g. in HIRLAM and CANARI include

• ECMWF ice cover and SST analysis, mostly based on satellite data such as EUMETSAT OSI-SAF products

• OSI-SAF products directly

• Local SST observations e.g. over Baltic sea

• Local climatological lake surface temperatures

Preliminary HIRLAM experiments have been run to test possibilities to assimilate in-situ and satellite observations on lake water temperature and ice cover. Over lakes, the prognostic Freshwater Lake model (FLake) has been applied as a parametrization scheme, in parallel of the diagnostic methods, to estimate lake surface temperature and ice thickness. Combination of lake temperature and ice data assimilation with the prognostic FLake is a challenge for the future.

• The working group suggests usage of ECMWF ice cover and SST analyses where possible. Also the direct usage of OSI-SAF data is already possible within HIRLAM and HARMONIE.

• Usage of climatological data over lakes should be in the future restricted to cold start conditions; for this good lake climatologies should be developed, instead of the present ones mostly interpolated to continental areas from the ocean temperatures.

• The working group encourages studies related to usage of satellite-based lake surface state observations.

• Lake-related issues will be discussed in another MUSCATEN workshop in Norrköping 15-17 September 2010, where all were wished welcome to participate.

Projects

In the workshop, the "European reanalysis - Euro4M" project was presented by Bazile. The working group suggests to study the project plan to understand its potential to create European data on surface climatology applicable in NWP, send suggestions and participate. Such climatological information combined with accurate surface description, provided e.g. by ECOCLIMAP and a fine-resolution digital elevation data set, would be most necessary in the cold start conditions.

In the presentations by Essery and Brun in the workshop, a possibility for snow data assimilation based on (precipitation) observations and a dedicated snow model was mentioned. This kind of approach might provide users e.g. in hydrology, perhaps also the NWP models with snow depth/SWE/snow fraction/other snow variables, instead of the direct snow depth/SWE assimilation presently done e.g. in HIRLAM and CANARI. Experience of a similar approach, where satellite snow data will be assimilated using a stand-alone snow model driven by atmospheric forcing from a NWP model, may be soon acquired within an European Space Agency project CoSDAS ( http://www.enveo.at/index.php?id=71&tx_ttnews%5Btt_news%5D=43&tx_ttnews%5BbackPid%5D=26&cHash=ed650174b5 )