Title:

Accuracy of Remote Sensing Estimates of Multiple Forest Biodiversity Metrics

Forests support over two-thirds of the Earth’s terrestrial biodiversity. Forest biodiversity is typically assessed by quantifying the number (richness) and distribution (evenness) of dominant canopy species present, but biodiversity can also be characterized by the variation in species function (functional diversity), genetic makeup (phylogenetic diversity), and physical structure (structural diversity). Obtaining accurate estimates of these biodiversity measures through common field methods has proven difficult and costly, with limited spatial coverage. Satellite and airborne sensors offer an opportunity to broaden the scale of biodiversity estimates. In this study, we test how well multiple measures of biodiversity can be estimated remotely with high-resolution areal imagery from Bartlett Experimental Forest in Bartlett, New Hampshire. Remote sensing estimates of biodiversity included the mean and standard deviation of reflectance, diversity of spectral classes from unsupervised k-means classifications, vegetation indices from 5m hyperspectral imagery, and topographic indices from Light Detecting and Ranging (LiDAR). These estimates were validated against biodiversity metrics calculated from ground-based inventory data and airborne LiDAR for estimates of structural diversity. Our results suggest that there is a potential to estimate functional diversity using imagery reflectance (50% to 60% accuracy), while other forms of diversity were more difficult to estimate (accuracies less than 20%). To find ways to remotely estimate the other biodiversity metrics, the next steps would involve adding a temporal component where we test if change in season would improve estimates of different biodiversity metrics.

Name

Comment*

Comments are viewable only by submitter

Captcha*