Monitoring phenological key stages and cycle duration of temperate deciduous forest ecosystems with NOAA/AVHRR data

In this study we attempted to monitor two main key stages in the phenological cycle of deciduous forests-budburst and senescence-using the normalized difference vegetation index (NDVI) derived front NOAA/AVHRR. These stages induce rapid (time scale of ct month), large (>0.3) and nearly linear NDVI variations. The method we developed consists of a fit of NDVI predicted by line segment to AVHRR-NDVI time series. It made it possible to derive the budburst and senescence timing, and then the phenological cycle duration. A relationship found in the literature between leaf area index (LAI) and NDVI showed that LAI was about 1 for the satellite-derived budburst and about 1.5 for the satellite-derived senescence. We tested the method on three nearly monospecific (Quercus petraea and Fagus sylvatica L.) forests located in France using a 6-year NOAA/AVHRR archive during the 1989-1994 period. The satellite-derived phenology revealed differences in relational to the composition and the climatic features of the study areas: 1) the phenological cycle duration of oak was longer (34 days) for southern than for northern forests, 2) in the North of France, beech trees were buckling earlier (5.6 days) than oak trees, 3) the interannual variability of budburst was significantly lower for beech than for oak trees. The comparison with ground phenological observations found in the literature showed the spatio-temporal coherence of the satellite-derived phenology A good correlation was also found between the satellite-derived budburst and the budburst timing predicted from air temperatures using the thermal time model. These tests provided a first validation of the method we developed to monitor the phenological cycle of deciduous forests with NOAA/AVHHR data. The limitations of the method and the perspectives for modeling temperature deciduous ecosystems are finally discussed. (C) Elsevier Science Inc., 1998.