The onset of the phenological phases, such as budburst, leaf development, flowering, fruiting, and leaf senescence, is evoked by the genetically determined internal periodicity of vegetation and significantly affected by climate conditions. Therefore, the timing of the onset of phenological phases is considered as a good indicator of climate change impacts. The phenological phases of ecosystems can be observed from satellites using the change in spectral radiance that is mainly driven by leaf optical properties, leaf arrangement and total leaf area in canopy.

In this study we used Normalized Difference Vegetation Index (NDVI) as an indicator of beech forests (Fagus sylvatica L.) seasonal dynamics in five test sites in Slovakia. During one growing season we analysed the phenological phases at each test site using three different approaches: i) in situ phenological observations, ii) digital hemispherical images taken to characterize the changes in the leaf area index (LAI) and in the plant area index (PAI), iii) NDVI calculated from space-borne MODIS sensor data. The estimates of LAI and PAI directly depend on the precision of canopy gap fraction (transmittance) estimates derived from the hemispherical images. Hence, we tested pixel classification based on the subjective decision of an operator and a recently proposed linear conversion of camera raw data (LinearRatio) which has been shown to produce comparable results to commonly used plant canopy analyzers.

The results showed that NDVI values reacted sensitively to the changes of vegetative phenological phases. The most rapid increase of NDVI was recorded during the leaf unfolding phenophase. After reaching its maximum, the NDVI values in all test sites started to decrease slowly during the summer. This was followed by the rapid decrease during leaf senescence in the autumn. The main changes in NDVI were well explained by the changes in LAI; however, the impact of LAI estimation method was significant. The canopy transmittance calculated from subjectively classified hemispherical images started to increase already in May, whereas the LinearRatio-based gap fraction continued to decrease until the end of July, which was in concordance with the observation performed in situ. The LAI estimates using canopy gap fraction from LinearRatio procedure did not indicate the saturating relationship with NDVI at high LAI values (LAI > 3) as reported by many authors. According to our results, MODIS NDVI can be used to observe phenological phases in mature beech forests. For the calculation of transmittance, which is required to estimate leaf area index from hemispherical images, we recommend using LinearRatio based methods.

Key words: beech, NDVI, LAI, PAI, phenological phases