Analysis of a long-term Time Series of 1 km NDVI Data derived from AVHRR over Europe and North Africa - Impact of Sensor Correction

Abstract

Since the last three decades, numerous scientists have analysed the Normalized Difference Vegetation Index (NDVI) in order to identify long-term trends. For the understanding of these developments, especially in the light of global climate change, data of the Advanced Very High Resolution Radiometer (AVHRR) proofed to be a unique data source for time series analysis. However, the large number of different NOAA satellites carrying AVHRR instruments demands for harmonisation approaches to quantify developments on the earth’s surface more precisely and to make data from different sensors comparable. The present thesis evaluated the impact of sensor correction as well as general spatio-temporal trends in the period from 1989 to 2013 over Europe and North Africa. Therefore, a framework for the preprocessing from L0 to L1 data and the averaging of the data into monthly NDVI composites was built. The results of the temporal analysis revealed an increasing NDVI trend for all analysed regions. From 1989 to 2013 the NDVI rose by 57, 69 and 224 % for the northern, the mediterranean and the deserts of the study area. Significant differences could be found compared to the sub-period 1989-2000, were most parts of the investigated area exhibited decreasing NDVI values. Concerning, the spatial pattern of NDVI, no significant change in the distribution of the vegetation index could be identified. The evaluation of the impact of sensor correction showed that the relative impact of sensor harmonisation differed throughout the study area. The strongest impact was measured in arid desert regions, where the corrected NDVI value differed up to 100 % . On the contrary, vegetaded parts of the study area were less influenced by the sensor correction only between 1 to 15 %. Main differences appeared due to the differences of the spectral response functions (SRF) that appeared due to differences in varying AVHRR instruments and due to BRDF effects, which were not quantified throughout this work.