On the relevance of accurate correction and validation procedures in the analysis of AVHRR-NDVI time series for long term monitoring

The effects of solar zenith angle (SZA) changes, short-wave calibration instability, and satellite switches on the interannual vegetation variability inferred from 1-km advanced very high resolution radiometer - normalized difference vegetation index (AVHRR-NDVI) were evaluated. The annual maximum value composite (MVC) of NDVI over middle and southern Italy was processed from afternoon viewings of NOAA 9, 11, and 14 for the period 1985 - 1999. MVCs were computed from imagery relative to the period of maximum photosynthetic activity. Raw data were compared with data corrected according to a procedure we recently developed and validated. Our results show that the SZA changes cause an error that increases as the NDVI value decreases. The maximum magnitude is 6.2% for forests and up to 41.8% for lakes. Calibration and intercalibration anomalies modify NOAA 11 NDVI values of 7 - 8% for all classes except for lakes and artificial surfaces, which show a variation of almost 11 - 12%; the NOAA 14 NDVI anomalies are set on about 11% with the exception of artificial surfaces ( similar to 17%) and sparsely vegetated areas ( similar to 9%). We also evaluated the improvement obtained from the use of the last update of calibration coefficients of NOAA 14. The positive impact of these coefficients is apparent for all the considered land covers, but a persistent overestimation of NOAA 14 calibration coefficients was still found. Altogether, the artificial anomalies can induce a relative error on NDVI variations up to 130%. This result strongly points out the relevance of performing an accurate correction of systematic errors and a careful validation before drawing conclusions from multitemporal AVHRR-NDVI analyses focused on the detection of long-term vegetation changes.