The main factor determining evapotranspiration is the net amount of radiation, which is modulated by the presence or absence of clouds.
Meteostat’s visible and infrared spectral ranges can only provide information about the earth surface where there is no cloud cover (otherwise, the measured radiances are those reflected from the top of the clouds, not the earth’s surface). Therefore, the first step is detecting clouds. An algorithm is used to separate the pixels of the images into clear sky and clouds. The cloudy pixels are further classified as representing low, intermediate or high amounts of cloud cover.

Where there is no cloud cover, the raw reflectance data of the earth’s surface must be corrected for the influence of the atmosphere using radiative transfer models, with both radiosonde and synoptic data furnishing supporting meteorological information. For cloudy pixels, an indirect method has been developed which exploits the results of the cloud classification. Net radiation is assessed for both clear and cloudy pixels.

Measurements from a reference micro-meteorological station are first assessed to determine the relation between the net surface radiation and the evapotranspiration. It is assumed that this relation, which is linked to the available soil humidity, is representative for the entire Belgian territory. Finally, the evapotranspiration is then computed for each pixel.

Because evapotranspiration depends on the annual and daily cycles of the solar energy and is modulated by the presence of clouds, it is important to take all available high-frequency meteorological data into account.

The data obtained by this method can be used in several ways:
- to analyse the temporal variability of evapotranspiration (how does it change with time);
- to investigate the spatial distribution of evapotranspiration (how much in different places);
- to validate the algorithms used in meteorological and climate models;
- to study the water budget at different spatial scales.

Analysis of the hourly evapotranspiration rates shows that successive images may be almost uniform or, on the contrary, differ widely depending on the position and evolution of cloudy areas. This spatial information could never be obtained by a simple spatial interpolation of evapotranspiration values computed in a few meteorological stations, but can only be achieved using geostationary satellite data. Temporal mean results of the evapotranspiration over the daylight period or over entire months may then be calculated by integrating the hourly evapotranspiration values.

Daily total of ETR (mm d-1)
Figures 1 to 5 show the daily total evapotranspiration values for four successive days in August and September 1997. Cloud cover was high on August 30th (2) and thus the evapotranspiration values remain very low all over the country. By contrast, the next day was sunnier and higher values (around 2.5 to 3.5 mm d-1) were attained. Only the eastern part of the country had good weather on September 1st (4), which explains the greater evapotranspiration values for this region. The following day (5), evapotranspiration was more intense in the western part, with the highest values obtained at the seacoast.

Since July 1995, this fast method of evapotranspiration assessment over Belgium has been performed daily by means of an automatic computer procedure using data from the previous day.
The data thus obtained can be used in a number of fields:

On a local scale, evapotranspiration data can be used to calculate the irrigation needs for crops in dry periods. Matching irrigation and rainfall amounts to crop evapotranspiration can be compared to transactions performed on a bank account. The soil is the "bank" for holding water. Rainfall and irrigation are deposits into the account, while evapotranspiration represents withdrawals from the account. This approach has even been called "checkbook" irrigation scheduling.

On a larger scale, evapotranspiration integrating the effects of several meteorological variables, is a complex key variable of the hydrological and meteorological models (i.e. important for weather forecasting and climate studies). Furthermore, possible climate changes are likely to result in evapotranspiration changes and feedback between the atmosphere, the soil and the vegetation. Such changes may very well forebode major shifts in the distribution and abundance of vegetation communities.