Zusammenfassung
Daily time-step normalized difference vegetation index (NDVI) time series from satellite-derived (NOAA/AVHRR, SPOT/ VEGETATION, TERRA/MODIS) and ground-based micrometeorological sensors were evaluated for a coniferous pine forest (Pinus sylvestris L.) located in Hyytiala, Finland. Micrometeorology-based broadband NDVI was calculated from observed upward and downward photosynthetically activity radiation (PAR) and global radiation measurements. The composite satellite-derived NDVI time series were smoothed with a best index slope extraction method (BISE) and adjusted Fourier transform (AFT) in order to downscale from the compositing period to daily scale.
The broadband and satellite-derived NDVIs were highly correlated during the main growth period (Julian days 90-270), but poorly correlated when the entire year was considered, i.e., large differences occurred during winter. High correlations were also found between the seasonal courses for broadband NDVI and daily air temperature. The analysis revealed that the onset of greenness in spring was consistently determined from broadband NDVI time series in different years, but that fluctuations in NDVI during the late season transition to winter dormancy prevented reliable prediction of the termination in physiological activity. Efforts to retrieve the same relationships during spring from satellite-derived NDVI failed.
After comparing the smoothed time series from different NDVI determinations, we examined the relationship between NDVI, gross primary production (GPP) and FAPAR. An obvious exponential relationship is found between broadband NDVI and GPP (R-2=0.72 for clear weather conditions; also detectable from the satellite sensors), while a linear relationship occurs between broadband NDVI and FAPAR (R-2=0.79). FAPAR in relation to satellite-derived NDVI is best described with a logistic curve under clear weather conditions, but the level of correspondence is low (R-2=0.53). Overall, broadband NDVI is a good index to describe physiological activity of the pine forest during certain periods, i.e. provides a means for obtaining other physiological parameters that are required by ecosystem models. However, during the late season, broadband NDVI estimated over the pine stand is influenced by more than vegetation physiological activity. Though satellite-derived NDVI is more difficult to link to GPP, it may still provide useful information under clear weather conditions. Satellite-derived NDVI remains our only choice for generalization in large-scale investigations. Thus, intensified examination of the influences of smoothing and downscaling of satellite-derived NDVI is inevitable. (C) 2004 Elsevier Inc. All rights reserved.
