Spatio-Temporal Classification Framework for Mapping Woody Vegetation from Multi-Temporal Sentinel-2 Imagery

Abstract

The inventory of woody vegetation is of great importance for good forest management. Advancements of remote sensing techniques have provided excellent tools for such purposes, reducing the required amount of time and labor, yet with high accuracy and the information richness. Sentinel-2 is one of the relatively new satellite missions, whose 13 spectral bands and short revisit time proved to be very useful when it comes to forest monitoring. In this study, the novel spatio-temporal classification framework for mapping woody vegetation from Sentinel-2 multitemporal data has been proposed. The used framework is based on the probability random forest classification, where temporal information is explicitly defined in the model. Because of this, several predictions are made for each pixel of the study area, which allow for specific spatio-temporal aggregation to be performed. The proposed methodology has been successfully applied for mapping eight potential forest and shrubby vegetation types over the study area of Serbia. Several spatio-temporal aggregation approaches have been tested, divided into two main groups: pixel-based and neighborhood-based. The validation metrics show that determining the most common vegetation type classes in the neighborhood of 5 × 5 pixels provides the best results. The overall accuracy and kappa coefficient obtained from five-fold cross validation of the results are 82.97% and 0.75, respectively. The corresponding producer’s accuracies range from 36.74% to 97.99% and user’s accuracies range from 46.31% to 98.43%. The proposed methodology proved to be applicable for mapping woody vegetation in Serbia and shows a potential to be implemented in other areas as well. Further testing is necessary to confirm such assumptions.

Comparing PlanetScope and Sentinel-2 Imagery for Mapping Mountain Pines in the Sarntal Alps, Italy

“…Due to the low requirements of nutrients and heat [4][5][6], mountain pines primarily colonize the treeline ecotone [2] but can also occur in more complex terrain (e.g., talus or mudflow slopes) [5] or pioneer subalpine open areas (e.g., former clearings or pastures) [7]. The mountain pine is widespread and native to the subalpine zone of the eastern and southern European Alps, the Carpathians and the Dinaric Alps [1,[8][9][10], while small-scale populations also exist in the Apennine mountains, Jura mountains and Vosges [3,7,11].…”

“…Based on this, up-to-date and accurate maps of mountain pines are needed to monitor and understand climatically and anthropogenically induced changes. However, such maps are rare and traditionally require time-and labor-intensive field surveys in alpine terrains [5], which strongly limit their spatial coverage and repeatability [10,20]. Using optical remote sensing data, alpine vegetation species can be mapped more comprehensively and cost-effectively.…”

“…Mountain pines tend to grow in dense patches ranging from several square meters up to multiple hectares [1,5]. However, the physiognomic resemblance of mountain pines and other subalpine conifers [3,21] results in a high spectral similarity, which causes ambiguities in multispectral Earth observation data [10,[22][23][24], especially in areas with a varying position of the treeline and strong intertwining between mountain pines and forested areas. Depending on the study area and patch size, increasing the spatial resolution may have a positive effect on the classification accuracy.…”

“…In general, studies focusing on the mapping of subalpine vegetation have shown benefits from a high spatial sensor resolution, ranging from submeter aerial imagery [22,25] to, most commonly used, spaceborne data, e.g., Sentinel-2 (10-20 m) [8,10,20,23,24,26], SPOT 5 (10 m) [27], RapidEye (5 m) [28] and GeoEye-1 (2 m) [29]. Besides sensor choice, the usage of additional features results in an increased accuracy of alpine vegetation classes.…”

“…The combination of multi-and hyperspectral datasets with features derived from airborne light detection and ranging (LiDAR) information (e.g., digital terrain model (DTM) and canopy height model (CHM)) improved the mapping of treeline vegetation [26][27][28][29]. Additionally, image-derived features increased class separability and classification accuracy, such as textural properties based on a gray level co-occurrence matrix (GLCM) [22,24,25,28,30] or phenological information derived from spectral indices [6,12,20,23,24] or multi-temporal imagery [8,10,20,23,24,28]. In studies classifying alpine vegetation, the usage of additional features resulted in an increased accuracy on the classification of mountain pines [10,[23][24][25]28].…”