Priority Project "Sat-Cloud"
Assimilation of satellite data with clouds and over land

Last updated: 2009

Project leader: Reinhold Hess (DWD)

Description

The aim of the project is the increased use of satellite data for regional applications of the COSMO-model in order to improve the quality of the nudging analysis including subsequent forecasts and of KENDA at a second step. Use of satellite data will be enhanced by exploiting information on clouds from SEVIRI infrared radiances and over land from AMSU-A microwave radiances, which are not assimilated hitherto.

Physical parameterisations of humidity and clouds will be considered as well as state dependent error correlations on the 2-3km scale of the model grid.

Motivation

The current use of satellite data within COSMO (and for most operational NWP applications) is restricted to the use of infrared and microwave sounding data over sea and for cloud free situations only. For infrared data the absorption and emissivity of clouds have to be simulated using observation operators, which is very demanding. Some developments for the assimilation of cloudy infrared radiances exist (e.g. at ECMWF), nevertheless further improvements for operational application are still needed.

Usually all channels that are affected by clouds are blacklisted using various cloud detection algorithms. Since 90% of the atmosphere is cloudy, most important information is lost in this way. Cloudy radiances will be used over sea and land. Microwave data is currently used over sea only. As microwaves are less sensitive to clouds, about 90% of the earth's surface can be observed. However surface emissivity has to be simulated, which is well known over sea, but critical over other surfaces such as land and ice/sea-ice.
Microwave observations over land are therefore in general blacklisted. As model domains of highly resolved limited area models are placed over land in most cases, only small fractions of the available data can be used.

The project will increase the amount of information derived from current operational satellite radiances and assimilate these information in order to improve the nudging analysis and finally KENDA for short range and very short-range forecasting. Meteorologically important situations are often abundant of clouds and a better representation of model clouds and humidity has an overall positive impact on forecast quality.

The centres of the COSMO domains, where the impact of lateral boundary values is lowest, are mostly located over land. The quality of the analysis over land, which has the potential to be improved with microwave data over land, has an essential impact on the forecast quality.

Proposed Actions:

The first part of the project deals with the use of cloud analyses derived from SEVIRI and hyperspectral infrared radiances (e.g. IASI, AIRS, CRIS) for data assimilation. The SEVIRI instrument on the geostationary METEOSAT satellites provides a rich source of information with high resolution in space (2-4 km) and time (15 min) over South and Central Europe.
Eight infrared channels contain information on temperature, humidity, and surface properties, respectively, and in combination products on cloud type, cloud fraction, top temperature, and height can be retrieved. An example is the cloud classification product of the NWC-SAF (Derrien and LeGleau 2005).

Hyperspectral infrared sounders on polar orbiting satellites provide low temporal and medium spatial resolution for limited area models, but with thousands of channels high spectral and therefore vertical resolution they are complementary to the SEVIRI imaging data.

Based on existing methods the cloud classification for SEVIRI will be improved using NWP data and also conventional data from surface synoptic reports, ceilometers, air planes, radiosondes, and possibly radars, as available. Additional validation and further accuracy for cloud top temperature and height can be achieved by collocations with data from hyperspectral infrared sounders, and this vertical information can be distributed in space in congruence to the cloud classification. Spatial situation dependent observation error correlations will be derived for IASI and AMSU data for this reason.

This cloud analysis is then conveyed into humidity to be used in the Nudging scheme and for KENDA, both over sea and also over land, although cloud detection is more demanding over land because of inaccurately known surface emissivities (similar to microwave frequencies).Relative humidity values in the COSMO-model are adjusted in order to generate model clouds matching the cloud analysis as far as possible.

Thereby vertical information from cloud top height and if available cloud base height is considered. The adaptation of relative humidity has to be parameterised depending on the integral analysis of cloud classification. Linear physical parameterisations of humidity and clouds (e.g. Lopez and Moreau 2004, Tompkins and Janiskova 2005) will be assessed and compared to the existing model-inherent physics in order to explore simple but effective parameterisations of relative humidity based on the cloud analysis.

In this way the specification of convective cells can be improved that are not yet precipitating and seen by the radars. Cloud information is also important in other weather situations such as low stratus conditions, and is generally an important weather parameter to forecast. Currently, the nudging is yet lacking any direct use of information on clouds.

The matter of the second part is the use of microwave data (AMSU-A, AMSU-B/MHS) over land. With microwave data important information in cloudy but not precipitating situations can be retrieved with low vertical resolution. The lower sensing channels carry most information on tropospheric temperature and humidity, however are significantly affected by the surface conditions and its emissivities.

Accurate surface emissivity values are required however at the exact instrument field of view in order to derive information on the atmosphere, which is the reason that microwave data is commonly assimilated over sea (sometimes over sea ice) only. The basic idea of the project is to derive the surface conditions over land from microwave window channels (AMSU 1,2,3, and 15) and use these emissivity values for the subsequent assimilation of the lower peaking channels.

These estimated surface emissivities could be further applied to provide satellite derived information on soil conditions at the surface. Other approaches using surface emissivity maps require extremely high resolution and frequent updates of surface conditions including soil moisture and vegetation, which are rarely available.

Tasks

Work has been separated into these distinct tasks:

References