From Agri4castWiki
Revision as of 10:07, 15 November 2013 by H Eerens (talk | contribs) (Compositing to daily images)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Main principles

The SEVIRI sensor on board of the geostationary platform MSG (Meteosat Second Generation) provides every 15 minutes a low resolution scan of the Eurafrican continent, the Middle East and the eastern tip of Brazil. The images have a sub nadir resolution of 3 km that deteriorates towards the poles, reducing to an average of 5 km over the European continent. All the raw data are collected by EUMETSAT (Darmstadt-Germany), pre-processed to a certain degree (calibration, cloud masking, addition of Lon/Lat planes,...) and transmitted in NRT to the network of SAFs (Satellite Application Facilities). For instance, the NWP SAF deals with numerical weather prediction, the HSAF with hydrological applications and the CM SAF with climatic monitoring. Task 2 however uses the data provided by the Land Surface Analysis group (LSA SAF), which is hosted at the IM in Lisbon-Portugal.

The EUMETSAT multi-mission ground segment consists of the following components:

  • a set of central facilities, located at EUMETSAT headquarters in Darmstadt, Germany
  • a primary and a back-up ground station for satellite control and data acquisition
  • a geographically distributed network of 7 so-called Satellite Application Facilities (SAFs).

At IM-Lissabon the LSA-SAF further derives a number of land-oriented products which are then freely distributed to the scientific world. They can be subdivided in three categories: agro-meteorological data (temperature, solar radiation, actual evapo¬transpiration, ...), vegetation products (fAPAR, fractional cover, LAI) and fire products (fire radiative power, ...). Some of these derived images still have high frequency (e.g. LST is generated at 15 minute intervals, radiation and ET every 30 minutes) while others are already “composited” to the daily time step (e.g. the three vegetation products). The data are treated separately for four separate regions, labelled as EURO (Europe), NAFR (northern Africa), SAFR (Africa below the equator) and SAME (South America). The derived images are still in the “raw satellite projection” but they can be remapped using the ancillary Lon/Lat-planes generated by EUMETSAT. LSA SAF distributes its results in the form of a continuous stream of small HDF5-files, which can be acquired either via “secure FTP” (sFTP) or via the EUMETcast broadcasting system. Each individual HDF5-file contains the information of one variable (e.g. solar radiation), time step (e.g. 48 files per day for ET but only one for LAI) and region (EURO, NAFR, SAFR, SAME).

Input data for MCYFS

For MCYFS, all agro-meteorological and vegetation products which are produced by the Land Surface Analysis (LSA) SAF hosted by the Portuguese Meteorological Institute in Lisbon are acquired. They are summarized in the table below. More information on the products can be found at the LSA-SAF website (see Links).

Characteristics of the MSG LSA SAF products used in MCYFS


Data from the regions EURO and NAFR are used to fill the entire European coverage. Some of the indicators are produced at a temporal resolution of 15 minutes, while others (such as LAI) are already delivered at the daily time step. For MCYFS, the indicators are composited to daily composites (S1) and 10-daily composites (S10), and remapped in the INSPIRE-LAEA projection with a fixed spatial resolution of 5km across Europe.


The general workflow for the processing chain is shown in the figure below.

The four MSG regions and the general pre-processing overview

The pre-processing of the MSG-SEVIRI data is different that for the other sensors in many ways. First, the data of the LSA SAF are distributed in 4 regions, of which 2 coincide with the ROI used in MCYFS, i.e. EURO and NAFR. Secondly, the information of the different variables (temperature, short/longwave radiation, ET, fAPAR, etc.) is sent by LSA SAF in separate files, which do not necessarily arrive at the same time for the different MSG-regions. For instance, at a given moment the radiation may be complete but the temperature lagging behind for EURO, while for NAFR the situation is exactly the opposite. Then, the data from the two regions EURO and NAFR need to be assembled into the ROI of MCYFS, remapped to INSPIRE-LAEA at 5km resolution and composited to daily and 10-daily values. Unlike for the other sensor, the processing of the data from MSG-SEVIRI is done entirely in a Windows environment (in the post-processing environment).

At the level of the pre-processing, only the following three steps are specific for MSG-SEVIRI:


The FTP-site of the LSA SAF is continuously scanned and all inputs are uploaded . The data of the different MSG-regions are immediately spread over two distinct folders. Then, the processing chain permanently looks for new data in these folders and performs the format conversion from HDF5 to ENVI. For the file format conversion from HDF5 to ENVI/GLIMPSE a specific tool was developed using the HDF5 library for reading the LSA SAF files. The tool is capable of interpreting the specific keywords in the header of the LSA SAF files and to generate the corresponding ENVI/GLIMPSE files (IMG+HDR). The output images are still in satellite projection.

Geometric corrections

The LSA SAF data are not yet geo-corrected and the images are still expressed in the "raw" Col/Rec system. All images are mutually congruent and accompanied by a high-quality IN-GRID, LL(Ci,Ri), which can be used for the remapping. This procedure is done separately for each variable after checking whether the results of the underlaying MSG region are ready. If so, these results are first reprojected to the desired OUT-system (INSPIRE 5km) and then combined/mosaicked to cover the MCYFS-ROI (see Compositing to daily images). This action is repeated independently for each variable.


Compositing to daily images

For the Land SAF products which are not delivered as daily values i.e. DSSF, ET and solar duration, daily values are calculated using the available images for a period covering the interval from 6:00 UTC for a particular date to 6:00 UTC the day after. For all products except solar duration the program MSGCompositer is used for the compositing. The tcl script creates a list of available files so that the required period is covered. The list can contain a file dated before 6:00 UTC of the date or after 6:00 UTC of the date + 1 if files for these dates and times are missing. If there are no files for a time equal to or smaller than 6:00 UTC of the date or equal to or larger than 6:00 UTC of date + 1, the program will use the values of the file with the nearest date and time and extrapolate these.

MSGCompositer reads in the available files and then uses linear interpolation to fill in possible missing values. Interpolation is limited to land pixels and to series in which the fraction of missing values and the size of the maximum gap are limited to the values maxMissingFraction (-, currently set to 0.25) and maxSizeGap (minutes, currently set to 240) which are specified in config.tcl. If there are too many missing values the daily value is set to the missing value flag. From the interpolated time series MSGCompositer then calculates the average value and if needed also the minimum and maximum value (e.g. for LST).

The sunshine duration is not delivered as such by the Land SAF but can be determined by counting the number of cloud free hours during day time. For this purpose the cloud mask is extracted from the quality information of the LST product and cloud free pixels are counted with the program


To determine daytime MSGCloudFreeHours reads in images with the latitude and longitude of the pixels to calculate sunrise and sunset in UTC. The calculation uses the following formulas for the fractional year (γ, radians), equation of time (eqtime, minutes), the solar declination angle (δ, radians) and the solar hour angle (ha, degrees):

gamma =\frac{2\pi }{365} \left(day\_ of\_ year-1\right)
eqtime=229.18(0.000075+0.001868\cos \gamma -0.032077\sin \gamma -0.014615\cos 2\gamma -0.040849\sin 2\gamma)
delta =-\arcsin \left(\sin \left(\frac{23.45\pi }{180} \right)\cos \left(2\pi \frac{day\_ of\_ year+10}{365} \right)\right)
ha=\pm \arccos \left(-\tan (\delta )\tan (latitude)\right)

For the positive value of ha the UTC time of sunrise in minutes is then:


The sunset time is obtained in the same way using the negative value of ha. Care has to be taken for situations where the product tan(δ)tan(latitude) is smaller than -1 or bigger than 1 which respectively correspond to situations where the sun doesn’t rise (no day time) and doesn’t set (no night time).

Compositing to 10-daily images

The procedure waits until all daily S1 are ready and then composites them to an S10-image.


The only post-processign step applied on the MSG-SEVIRI images is the extraction of databases with Regional Unmixed Means (RUM-values). This is done only on the 10-daily composites of : shortwave radiation (r), mean temperature (t), minimum temperature (n), maximum temperature (x) snow cover (q) and evapotranspiration (e).

Products overview

The table below provides an overview of all daily and 10-daily composite products that are generated for MCYFS from MSG-SEVIRI input data.

Daily and ten-daily composited data are generated in the form of images (I), quicklooks (Q) and databases of RUM-values (R).


More information on the image format, headers and rescaling can be found in the section Images. The column V refers to the filename suffix, and is explained in the section on Filename conventions.