Biodiversity mapping and modelling using remote sensing data: mapping and modeling patterns of breeding bird diversity across the United States

Biodiversity conservation is imperative in the face of increasing anthropic pressures and threats to forest ecosystems. Our ability to evaluate and monitor biodiversity is essential to ensure effective conservation. Forest structure is a key factor driving several processes in forest ecosystems. Stand structures affect microclimate, habitat quality and therefore biodiversity potential. Biodiversity indicators have been shown to be strongly correlated with the three-dimensional spatial pattern of vegetation (MacArthur and MacArthur, 1961). And the richness of wildlife has been related to canopy three-dimensional features (Carey et al., 1991). Establishing reliable models describing the link between biodiversity and forest structure would facilitate the implementation of sustainable management strategies and practices. Forest structure is generally described from field measurements. But only a limited number of plots can be inventoried, as this work is both costly and time-consuming. Remote sensing has the potential to provide quick and accurate measurements over large areas. The potential of LiDAR (Light Detection And Ranging) systems to measure forest structure and assess forest attributes is widely acknowledged (Næsset, 2004; Nelson et al., 1988). LiDAR are active systems providing precise distance measurements based on elapsed time between the emission of a laser pulse and the reception of the backscattered signal. The use of LiDAR in landscape ecology and biodiversity studies is a recent field of research. Metrics extracted from LiDAR data have been proposed for characterizing landscape pattern and structure (Mücke et al., 2010). The use of LiDAR data allows analysing relationships between biodiversity indicators and a broad range of structural metrics related to the 3D arrangement of vegetation. Indeed LiDAR data provides the opportunity to analyse the impact of forest structure surrounding field plots for which biodiversity indicators were measured. Some studies already explored the relationship between biodiversity indicators and forest structure metrics from LiDAR data (Lesak et al., 2011; Müller and Brandl, 2009; Müller et al., 2014; Zellweger et al., 2013). However, while the relationships between LiDAR metrics and faunal biodiversity have already been explored, floristic biodiversity has not yet been analysed. Furthermore, most studies did not integrate the ecological context in addition to 3D vegetation structure data, when the models explaining the biodiversity indicators were built. Ecological context here refers to abiotic variables, on which biodiversity indicators highly depend (Maestre et al., 2009). Complementing LiDAR metrics with abiotic variables improved model predictive power (Zellweger et al., 2014).\nThe aim of this study was to further evaluate the potential of LiDAR for floristic biodiversity monitoring. Floristic biodiversity was studied in terms of plant species abundance and richness of the different ecological groups. Bayesian statistical models, described by Zilliox and Gosselin (2013), were used to model the link between floristic biodiversity and both abiotic and biotic characteristics of the environment. In these models forest structure was initially assessed using traditional field measurements on circular plots with a 15 m radius (e.g. basal area, cover). Two specific objectives were identified for this study. Firstly, we evaluated the potential of LiDAR to replace forest structure indicators measured in the field and to improve the modelling of the link between floristic biodiversity and stand structure. Secondly, we took advantage of the capacity of LiDAR to assess forest structures at various scales, in order to improve our knowledge on the drivers of biodiversity and try to identify up to which distance the structure can influence local biodiversity. \nThe study site was a deciduous forest located in North-Eastern France (48.53° N, 5.37° E). Forest was studied under leaf-on conditions in a 60 km² area. The climate is semi-continental, and subject to an oceanic influence. The site was comprised of complex stands with multi-layered forests, dominated by European beech (Fagus sylvatica), Hornbeams (Carpinus betulus) and Sycamore maple (Acer pseudoplatanus). LiDAR data was collected from small-footprint airborne LiDAR with a high point density of 30 pt/m². 741 field plots located within a radius of 100 km around the study area and 49 field plots located within the study area were used to build the models. As the study site was too small to offer enough site type diversity, the first field dataset was used to model the impact of site type variation on biodiversity. Five abiotic variables were thus included in the model: mean annual temperature, solar radiation, topography, soil pH and soil water capacity. Temperature, solar radiation and topography were subsequently considered as constant over the study site. The second field dataset was used to include and test one by one diverse LiDAR metrics in the statistical model. Stand-level metrics were extracted from LiDAR data in order to describe vertical and horizontal distribution of forest vegetation. Metrics were extracted from circular plots within a 15 m radius as field plots, and also 50 m, 100 m and 200 m radius. Bayesian statistical models provide an estimate of the magnitude of the relationship between biodiversity indicators and ecological variables. We could evaluate the magnitude of the relationship between the floristic biodiversity indicators and the LiDAR metrics. Deviance Information Criterion (DIC) was used to compare models with each other. Several metrics were necessary to predict plant species abundance and richness models. Several LiDAR metrics measured at the plot level were found to have non-negligible relationships with floristic biodiversity. The study also highlights that forest structure in the neighbourhood of field plots can impact on biodiversity indicators measured at plot level.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.