Characterization of Lake Erie Coastal Terrestrial-Aquatic Interface (TAI) with a Machine Learning Functional Zonation Approach and Spatial Data Synthesis
Abstract
Coastal terrestrial-aquatic interfaces have a significant role in the global biogeochemical cycle. However, these systems present strong variations in geomorphology, which in turn influences topography and ultimately vegetation communities and biogeochemical gradients. To gain a predictive understanding of the functioning of such ecosystems, it is critical to characterize the various processes and their interconnectivity. In this study, we aim to transform the spatiotemporal characterization and monitoring of coastal ecosystems and subsurface environments, taking advantage of recent advances in sensing technologies and machine learning (ML). We propose a novel ML-based coastal functional zonation approach to tractably harmonize diverse spatial data layers to capture the self-organization and co-variability of the above/below-ground terrestrial systems from bedrock to canopy including geology, soil, and plant characteristics/dynamics. Critical data layers include publically available layers of topography (SRTM), soil (USDA), wetland land-cover maps, and long-term time-series of satellite images. We analyze those properties as a function of along-shore distance and transverse distance to investigate the heterogeneity of coastal ecosystem structures. We demonstrate this approach based on the datasets along Lake Erie. Results of the co-variability analysis show that topographic metrics (elevation, slope, topographic position index) and soil texture influence the spatial distribution and connectivity of wetland land-covers (emergent, shrubs, and forest) with emergent in sandy areas with lower slope. The time series analysis of 20 years of enhanced vegetation index shows that emergent wetlands are more sensitive to yearly variations in precipitation. The clustering analysis allows us to identify regions with determined environmental characters. Dominant species and variations in topographic properties (slope and elevation) are the main drivers of these clusters.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.B34A..04E