LiDAR metrics and Sentinel-2 derived NDVI for Conservation Status Mapping

Published: 1 October 2021| Version 1 | DOI: 10.17632/gpy8kydxx7.1
Contributors:
Maria Teresa Lamelas,
,

Description

This study aims to assess the ability of single photon LiDAR (14 points m2) and Sentinel-2 data to classify the conservation status of oak forests in Natura 2000 habitats directive sites (codes: 9160, 9230 and 9240). The study area encompasses four Special Areas for Conservation in Navarra province (Spain): “Montes de Valdorba”, “Sierra de Lokiz”, “Belate” and “Robledales de Ultzama”. To conduct the analysis a random sample of pixels was selected using a stratified mapping based on the conservation status of the study zones performed by GAN-NIK (Gestión Ambiental de Navarra S.A.), terrain slope, height and cover of forest canopy.

Files

Steps to reproduce

The classification and mapping of CS in oak groves habitats (codes 9160, 9230 and 9240) was derived using a four step methodological approach (Figure 2). Firstly, LiDAR and Sentinel-2 data was collected and processed in order to generate the biodiversity related LiDAR metrics and radiometric indices. Secondly, a representative random sample to cover the variability of the study sites was created in order to compare, in a third step, between conservation status using the Kruskal-Wallis and Dunn statistical tests. Ultimately, a selection of the most suitable metrics with an ecological significance was used to classify and map the CS for the three mentioned Natura 2000 areas. The ALS data was acquired by the Spanish National Plan for Aerial Orthophotography (PNOA) in 2017 using a Single Photon LiDAR (SPL100). The point cloud return heights were normalized by subtracting the elevation data from the 2 m grid resolution digital elevation model (DEM) provided by IDENA, using FUSION LDV 4.0 open source software (McGaughey, 2018). A full suite of statistical metrics, commonly used within forestry, related to vertical and horizontal structure metrics were generated using FUSION LDV 4.0 at a spatial resolution of 20 meters. A threshold value of 0,2 m height was applied, considering the RMSE in z values of ALS-PNOA data according to Domingo et al. (2020), to remove ground and understory returns. Stratified density metrics were computed every 0.5 meters height. Based on this information, the biodiversity LHDI (LiDAR Height Diversity Index) based on the Shannon-Weiner diversity index (H’) and LHEI (LiDAR Height Evenness Index) based on Pielou’s index were generated (Listopad et al., 2015). The Normalized Differential Vegetation Index (NDVI) has been derived from Sentinel-2A image captured in 26th September 2019. Chlorophyll absorption in the red band and the relatively high reflectance of vegetation in the near infrared band (NIR) are used to calculate the NDVI according to equation 3. For this purpose, atmospherically corrected bands 4 and 8 were used (level 2A). The variability of a representative sample of the forest was determined with four variables: (1) the canopy cover (using the FCC metric), (2) the terrain slope (3) the canopy height (using the P95 metric) and (4) the CS of the habitats, obtained through a stratified random sampling (Næsset and Okland, 2002).

Categories

Database

Licence