Sentinel4Nature: Estimating environmental gradients and properties using remote sensing

In 2014, the European Space Agency (ESA) launched the first Sentinel satellite as one of a series of complementary sensors that together form the Sentinel mission family. It is part of to-day’s most ambitious Earth Observation Program: Copernicus. At the same time, a new system for describing, mapping and analysing nature in Norway (NiN) has been developed (Halvorsen et al. 2015). One of the leading principles in NiN is to account for gradual transitions in nature and thus to focus on the underlying environmental gradients and properties (e.g. related to cli-mate, soil, etc.) that govern the occurrence of species and associated nature types.
The aim of the Sentinel4Nature project (financed by ESAs PRODEX funds) has been to develop and advance an approach to remote sensing that focuses on monitoring basic environmental gradients and properties and utilizes fusion of different data sources (different sensors as well as auxiliary data). In this context, the suitability of remote sensing to identify environmental gra-dients in the NiN classification system has been assessed.
Based on expert judgement and a literature review, it was estimated that satellite remote sensing can be a useful source of information for more than 50 % of the 61 environmental gradients in NiN. The majority of the most suitable gradients is related to land cover as well as presence of water or snow in the landscape.
From the most suitable gradients, 1) reduced growing season due to prolonged snow-lie and 2) tree canopy cover, were selected for case studies that were conducted in one to four study sites across Southern- and Central-Norway. For 1), the date of snow melt was estimated from a time series of Landsat8 and Sentinel-1 observations and for 2), the percentage of canopy cover per pixel was modelled using Sentinel-1 and Sentinel-2 data. In both cases fairly accurate models could be developed that improve the current possibilities to map or model vegetation structures or species occurrences. Fusion of imagery from different sensors (in particular radar and optical) significantly improved the model performance. Auxiliary data was less important than expected. However, data on terrain has been important during image enhancement and correction.
Although the presented methods already perform quite well, future adjustments and improve-ments in the processing chains and also parameter tuning have to be expected when used at larger extents and especially towards arctic environments of the Scandinavian peninsula.
KEY WORDS : Norway, Oslofjord, Lurøykalven, Hjerkinn, Sunndalen, environmental gradients, snow cover, tree canopy cover, remote sensing, Sentinel imagery, data fusion, modelling, NiN, PRODEX, ESA, NRS
NØKKELORD : Norge, Oslofjorden, Lurøykalven, Hjerkinn, Sunndalen, lokale komplekse miljøvariabler (LKM), snødekkebetinget vekstsesongreduksjon, tresjiktstetthet , fjernmåling, Sentinel, data fusion, modellering, NiN, PRODEX, ESA, NRS