Open AccessModeling air temperature through a combination of remote sensing and GIS data
Author(s) -
Cristóbal J.,
Ninyerola M.,
Pons X.
Publication year - 2008
Publication title -
journal of geophysical research: atmospheres
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.67
H-Index - 298
eISSN - 2156-2202
pISSN - 0148-0227
DOI - 10.1029/2007jd009318
Subject(s) - environmental science , mean squared error , remote sensing , latitude , altitude (triangle) , albedo (alchemy) , evapotranspiration , normalized difference vegetation index , multivariate interpolation , air temperature , regression analysis , mean radiant temperature , meteorology , geography , climate change , geology , statistics , mathematics , art , ecology , oceanography , geometry , geodesy , performance art , bilinear interpolation , biology , art history
Air temperature is involved in many environmental processes such as actual and potential evapotranspiration, net radiation and species distribution. Ground meteorological stations provide important local data of air temperature, but a continuous surface for large and heterogeneous areas is also needed. In this paper we present a hybrid methodology between Remote Sensing and Geographical Information Systems to retrieve daily instantaneous, mean, maximum and minimum air temperatures (2002–2004) as well as monthly and annual mean, maximum and minimum air temperatures (2000–2005) on a regional scale (Catalonia, northeast of the Iberian Peninsula) by means of multiple regression analysis and spatial interpolation techniques. To perform multiple regression analysis we have used geographical and multiresolution remotely sensed variables as predictors. The geographical variables we have included are altitude, latitude, continentality and solar radiation. As remote sensing predictors, we have selected those variables that are most closely related with air temperature such as albedo, land surface temperature (LST) and NDVI obtained from Landsat‐5 (TM), Landsat‐7 (ETM+), NOAA (AVHRR) and TERRA (MODIS) satellites. The best air temperature models are obtained when remote sensing variables are combined with geographical variables: averaged R 2 = 0.60 and averaged root mean square error (RMSE) = 1.75°C for daily temperatures, and averaged R 2 = 0.86 and averaged RMSE = 1.00°C for monthly and annual temperatures. The results also show that combined models appear in a higher frequency than only geographical or only remote sensing models (87%, 11% and 2% respectively) and that LST and NDVI are the most powerful remote sensing predictors in air temperature modeling.