University of Twente Student Theses


Characterising relationships between weather and land surface variables using Meteosat Second Generation products

Cassar, Daphne Anne (2009) Characterising relationships between weather and land surface variables using Meteosat Second Generation products.

[img] PDF
Abstract:Geostationary satellite data offers new ways to look at land use and land cover from space. The SEVIRI sensor on board the geostationary Meteosat Second Generation (MSG) satellite is an attractive option for such studies requiring frequent observations, with its ability to capture data every 15 minutes in 12 spectral bands. The overall aim of this work was to explore some new MSG products in observing temporal relationships between precipitation, vegetation and land surface temperature in the south of Portugal, a region known to be at risk of desertification. Many studies considering the behaviour of these variables have been conducted in Africa. However, such relationships are location dependent, promoting the need for further studies in other regions, to better understand the temporal relationships between vegetation, surface temperature and local climate variability like precipitation. The study utilised four Meteosat products that are currently freely available to the scientific community: MSG level 1.5 data processed to NDVI, fractional vegetation cover (FVC), multi sensor precipitation estimate (MPE) and land surface temperature (LST). All data series were collected and processed for the 1 January 2007 to 15 October 2008 period. All of these products except FVC were validated by regression analysis, for MPE and LST using ground observation data and NDVI using the SPOT S10 satellite product. All regressions were found to be significant (at α = 0.05), though the MPE product showed considerable problems in estimating rainfall for high levels of precipitation occurring in the region. Data was composited into ten day intervals using a maximum function for FVC, NDVI and LST and a sum function for MPE. Cross correlation analyses were then used to establish the strongest correlations of paired variables at various time lags. Overall there were many significant correlations (p<0.05), like strong positive correlations of MPE with NDVI and FVC at lags of around 40 days. LST showed both positive and negative correlations with NDVI, which was as expected but the LST and FVC pair did not correlate in the same manner. The real meaning of this is in practical interpretation of the results, which was attempted by considering the rainy and growing seasons for the region, and keeping in mind that there are other factors influencing the behaviour of the analysis variables, like irrigation and evaporation. It was concluded that the products were useful for this type of study, but it would benefit from longer time series when these data become available.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page