Satellite-based mapping of the growing season and bioclimatic zones in Fennoscandia

Aim To test whether satellite-derived NDVI values obtained during the growing season as delimited by the onset of phenological phases can be used to map bioclimatically a large region such as Fermoscandia. Location Fermoscandia north of about 58 degrees N and neighbouring parts of NW Russia. Methods Phenology data on birch from 15 research stations and the half-monthly GIMMS-NDVI data set with 8 x 8 km(2) resolution from the period 1982-2002 were used to characterize the growing season. To link surface phenology with NDVI data, new algorithms on a pixel-by-pixel basis that show high correlation with phenophases on birch were developed. Then, time-integrated values (TI NDVI) during the phenologically defined growing season were computed to produce a bioclimatological map of Fermoscandia, which was tested and correlated with growing degree days (GDD) obtained from 20 meteorological stations. The map was also compared vs. traditional bioclimatic maps, and analysed for error factors distorting NDVI values. Results The correlation between GDD and TI NDVI data during the phenologically defined growing season was very high. Therefore, the TI NDVI map could be presented as a bioclimatic map reflecting GDD. However, several major areas have interfering factors distorting NDVI values, such as the pixel heterogeneity caused by the altitudinal mosaic in western Norway, the mosaic of lakes in southeastern Finland, and the agriculture-dominated areas in southern Fermoscandia. Main conclusions TI NDVI data from the phenologically defined growing season during 1982-2002 in Fermoscandia can be processed as a bioclimatic map reflecting GDD, except for the areas distorting NDVI values by their strong ground-cover heterogeneity.