Speaker:Dr. Ian McGregor, Cary Institute
The proliferation of satellite imagery has facilitated the analysis of deforestation trends using multi-source time series approaches. This combination of long-term data sources can be crucial for near real-time (NRT) deforestation monitoring in tropical regions with extensive cloud cover. Yet despite recent progress, methods in multi-source, NRT monitoring have experienced issues with trade-offs between accuracy and latency, the lack of external spatial data, and general inability to incorporate users into the workflow.
This talk will discuss McGregor's doctoral research which focused on three main topics: 1) a novel, NRT detection algorithm that was created to explicitly assess multiple spectra of trade-offs; 2) landscape processes that were incorporated into the algorithm using a Bayesian approach; and 3) an expansion of the methodology used to develop a continuously-improving model by integrating input from in-situ users.