Map Viewer
Data & Products
Missions & Sensors
About WDC
User Account

WDC Home

Local Weather

Daily total water vapour column over Europe and Africa from MSG/SEVIRI

Total water vapour column from SEVIRI at WDC-RSAT

Description and Product example:
A retrieval scheme for total water vapor column (TWC) was developed which is applicable for the MSG (Meteosat Second Generation) instrument SEVIRI and suitable for operational near-real-time use on a daily basis. TWC is derived over cloud-free land surfaces from thermal infrared split window channels at 10.8 and 12 µm in a pixel wise resolution once daily.

Daily Total Water Vapour Column
for Europe
Daily Total Water Vapour Column
for MSG field of view

Fig. 1: Example image of daily TWC derived from MSG-SEVIRI on 22 July 2004 for Europe and the MSG field of view covering Europe, Africa, the Arabian Peninsula and parts of South America.



A physically based algorithm for the retrieval of total water vapor column (TWC) over cloud-free land surfaces proposed by Kleespies and McMillan (1990) was evaluated and extended for use in atmospheric correction and surface irradiance calculation schemes. Thermal infrared split window channels at 10.8 and 12.0 µm of the MSG - SEVIRI (Meteosat Second Generation - Spinning Enhanced Visible and Infrared Imager) instrument are used.
The proposed algorithm takes advantage of the improved measurement capabilities of the MSG-SEVIRI instrument with its 15 minute temporal resolution and its radiometric accuracy of 0.25 K and 0.37 K in the 10.8 and 12.0 µm channels. The temporal resolution allows exploitation of the daily land surface temperature variation. There is no further need for explicit auxiliary information on air and land surface temperatures, which is difficult to obtain on an operational basis.
Updated coefficients for the split window parameterization were derived based on simulations of 'top-of-atmosphere' SEVIRI brightness temperatures for the globally representative Thermodynamic Initial Guess Retrieval (TIGR3) set of radiosonde profiles. It turns out that the linear dependency on the transmission ratio in both split window channels as originally proposed by Kleespies & McMillan (1990) has to be extended towards a non-linear approach in order to make it applicable to the full range of global atmospheric conditions.


The new TWC algorithm was tested with MSG-SEVIRI data for European and African regions for the period March - August 2004 and compared with radiosonde data. The results show that the algorithm is capable of producing TWC values with a mean bias of -0.2 mm and a RMSE of 6.8 mm (Fig. 2). From the total amount of 2583 coincidences for all viewing zenith angles both for winter and summer conditions, 82 % were within a 5 mm and 94 % were within a 10 mm difference interval between MSG-based and radiosonde-based TWC.

This RMSE value was very close to the accuracy range of 4 to 7 mm inherent in the radiosonde data set itself, which was assessed with a semi-variogram analysis. Therefore, we assumed that the observed scatter in satellite based TWC values versus radiosonde based TWC values is due to comparing two data sets with nearly comparable accuracy. Further analysis revealed neither a significant seasonal dependence in relative TWC errors nor a latitudinal structure within the comparison data set. Also, no significant dependence on air temperature profile characteristics was found.

Fig. 2:Two-dimensional histogram for MSG SEVIRI-based TWC versus European and African radiosondes obtained from the British Atmospheric Data Center (BADC) for the period March to August 2004. 94 % of the differences can be found within a 10 mm range and 82% of the differences are within a 5 mm range. While there is a remarkable scatter with a low correlation coefficient of 0.59, there is also a remarkable maximum of coincidences found around the identity line.

A second comparison with European GPS TWC measurements reveals the best agreement with the 11 UTC measurement with a bias of -3.0 mm and a RMSE of 6.0 mm (Fig. 3). Two-hourly GPS measurements allow also for comparison of the cloud-free GPS-based daily mean value with the MSG-based TWC resulting in a bias of -2.6 mm and a RMSE of 6.8 mm. Comparison with GPS measurements shows also that large differences up to 20 mm canoccur in higher latitudes above 50 N in Europe. It is assumed that these cases can be attributed to failures in the cloud clearing procedure in cases with both large sun and satellite zenith angles.

Fig 3: Two-dimensional histogram for MSG SEVIRI based TWC versus 11 UTC GPS measurements based on 287 coincidences for the period from March to August 2004. TWC differences below 5 mm are found in 90% and below 10 mm in 97% of the cases.

Users and relevance:

A retrieval method for daily TWC measurements was elaborated which is accurate enough to be used for surface irradiance calculations in atmospheric correction and solar energy applications. The algorithm can be used in near-real-time operations without needing access to a numerical weather model output as a priori information. The restrictions to cloud-free regions are acceptable for these applications, as in case of clouds the cloud itself is the dominating element influencing surface irradiance. Also, the restriction to land surfaces is acceptable as solar energy power plants are currently not foreseen to be installed offshore.

Data access:

Currently, daily data and quicklooks from 2004 and 2005 are available here. A major reprocessing is ongoing and data can be made available in future.


For questions and more information please contact the SEVIRI team at DLR


Kleespies, J.T., McMillin L.M. (1990). Retrieval of precipitable water from observations in the Split Window over varying surface temperatures. Journal of Applied Meteorology, 29, 851-862

Schroedter-Homscheidt, M., A. Drews, S. Heise (2007) Total water vapor column retrieval from MSG-SEVIRI split window measurements exploiting the daily cycle of land surface temperatures, Remote Sensing of Environment, 249–258, doi: 10.1016/j.rse.2007.05.006 [pdf]