Recognition and Characterization of Forest Plant Communities through Remote-Sensing NDVI Time Series

Abstract

Phytosociology is a reference method to classify vegetation that relies on field data. Its classification in hierarchical vegetation units, from plant associations to class level, hierarchically reflects the floristic similarity between different sites on different spatial scales. The development of remotely sensed multispectral platforms as satellites enormously contributes to the detection and mapping of vegetation on all scales. However, the integration between phytosociology and remotely sensed data is rather difficult and little practiced despite being a goal for the modern science of vegetation. In this study, we demonstrate how normalized difference vegetation index (NDVI) time series with functional principal component analysis (FPCA) could support the analyses of phytosociologists. The approach supports the recognition and characterization of forest plant communities identified on the ground by the phytosociological approach by using NDVI time series that encode phenological behaviors. The methodology was evaluated in two study areas of central Italy, and it could characterize and discriminate six different forest plant associations that have similar dominant tree species but distinct specific composition: three dominated by black hornbeam (Ostrya carpinifolia) and three dominated by holm oak (Quercus ilex). The methodology was also able to optimize the ground data collection of unexplored areas (from a phytosociological point of view) by using a phenoclustering approach. The obtained results confirmed that by using remote sensing, it is possible to separate and distinguish plant communities in an objective/instrumental way, thus overcoming the subjectivity intrinsic to the phytosociological method. In particular, FPCA functional components (NDVI seasonalities) were significantly correlated with vegetation abundance data variation (Mantel r = 0.76, p < 0.001).

Earth Observation and Biodiversity Big Data for Forest Habitat Types Classification and Mapping

“…It combined nationwide biodiversity database, environmental information, and satellite EO data (Sentinel-2 MSI at 20 meters spatial resolution) to train a supervised machine learning model for forest habitat classification. Previous studies on forest habitat mapping in Italy used satellite data for classification at coarse spatial and thematic resolutions, often investigating small areas (regional or sub regional scale areas) [37,90,103] or focusing on single tree species [96]. 3.…”

“…The phenological metrics have only been used as predictors for T1 classification (i.e., deciduous broadleaved), since plant phenology of evergreen plants has lower temporal fluctuation in terms of LAI values, which makes their estimation uncertain. Other studies attempted to use EO time series to describe plant phenology in vegetation mapping, without an in-depth exploitation through the phenological metrics estimates [90]. The median value of each temporal variable calculated for the years 2016-2019 has been selected as temporal predictor used for the training, validation and testing phases of the classification algorithm.…”

Practice Must Be Backed up by Theory! A Special Issue on Plant Community Ecology

“…Nevertheless, we must recognize that since new methods are continuously proposed in modern plant community ecology studies, we can benefit from the integration of new instruments with long consolidated, traditional approaches. The results of Pesaresi et al [5] confirm that by using phenological behaviors described by NDVI time series, it is possible to separate and distinguish plant communities in an objective/instrumental way, thus overcoming the subjectivity intrinsic to the phytosociological method. In the same way, databases that allow classical floristic data to converge in big data sets are useful to understand and describe the big picture of the ecoregions [6].…”

Earth Observation for Phenological Metrics (EO4PM): Temporal Discriminant to Characterize Forest Ecosystems

“…Forest plant communities, together with their own typical floristic composition, show exclusive phenological dynamics recognizable by vegetation indices time series [42]. Both vegetation indices time series [42,61] and land surface phenological metrics [61,92] has been successfully used to classify plant communities and natural habitats, discriminating vegetation patterns dominated by species with similar phenology features.…”

“…Phenology can be investigated by exploiting the optical response of vegetation through the seasonal variations of spectral and biophysical indices [35][36][37] by analyzing the vegetation index time series. Vegetation indices are the basic information for phenological researches, e.g., by analyzing the vegetation index time series to derive information [38][39][40][41][42], by assessing temporal statistics [43,44], or by estimating phenological metrics (as the onset of greenup, the length of the growing season, and the offset of the season) [45][46][47].…”