BIOSCANN - A national scale predictive model for the detection of high-conservation value sites from remote sensing data

As we approach the short- to medium-term implementation deadlines of major international biodiversity agreements, such as the United Nation's Kunming-Montreal Global Biodiversity Framework and the European biodiversity strategy, there is a pressing need for high-quality data products to guide large-scale conservation prioritization decisions. In this study, we implement a deep learning segmentation approach for detecting high-conservation-value forests using a nationwide inventory dataset and a set of remote sensing data products. Using this model we produce a national data product showing remaining high-conservation value forests in Sweden, with a predictive accuracy of 91% and a high-detail spatial resolution of 10-meter pixel size. Sweden serves as a good test case for the developed approach, as it is home to a large portion of Europe's remaining old-growth forests and is also characterized by robust biodiversity and environmental data availability. Our approach allowed us to identify over 50,000 km² of potential high conservation value forest (HCVF) at high confidence, which has the potential to considerably improve efficiency in manual inventory efforts. With its high accuracy and spatial resolution, our data product offers substantial utility for decision-makers at different administrative scales, and directly addresses the goals set by large international biodiversity conservation plans. The implemented approach demonstrates the utility of mapping structural ecosystem metrics to identify sites of particular conservation interest. This is partly enabled through the availability of high-resolution airborne LIDAR data, which are available as national data products in several countries. With improving availability of new global remote sensing data products, the prospect of mapping ecosystem structural intactness becomes increasingly feasible on a global scale. While this project focuses on detecting high conservation value forests in Sweden, the presented model serves as a proof-of-concept implementation that can be adapted and applied for modeling other regions and habitat types.
An interactive version of the data product can be found here: https://gee-hcvf-andermann.projects.earthengine.app/view/hcvf-viewer.