EDS Seminar: Mind the Gap
Abstract
Earth observation data provides insights into anthropogenic and natural land use with unprecedented detail, global coverage, and frequent resurveys. Moreover, deep learning models like AlphaEarth and SatClip enable us to leverage remote sensing data to study large-scale patterns in ecology, economics, and global change. Nonetheless, state of the art remote sensing data products fail to capture the fine-scale habitat characteristics that define species’ ecological niches and ultimately determine where species occur. As a result, fine-scale habitat can only be measured for small plots, and species distribution modeling at continental scales must rely on coarse ecological covariates derived from satellite imagery. In this presentation, I will first quantify the gap: poor performance of earth observation models for fine-grained habitat classification. I will then describe an opportunity to fill it: airborne lidar data captures detailed three-dimensional structure and is now available coast-to-coast in North America. Building deep learning models that can interpret the ecological structure in these point cloud data will unlock the ability to characterize fine-scale habitat structure across large areas.
Speaker Bio
Sam Lapp's research leverages emerging artificial intelligence, machine learning, and automated sensing technologies to understand natural ecosystems and inform the conservation of biodiversity. He is currently a postdoctoral research fellow at ESIIL, CU Boulder, where he is co-advised by Dr. Esther Rolf and Dr. Sara Beery. Sam's work emphasizes open science and inclusivity. He leads the development of Dipper, a desktop tool for passive acoustic monitoring analysis, and OpenSoundscape, an open-source Python package for bioacoustics analysis. He works closely with ecologists and conservation practitioners to apply these approaches to real-world data, enabling large-scale investigation of ecological processes and supporting conservation decision-making.
Please join us in SEEC S127 or on Zoom.