As one of the main drivers of biodiversity loss, deforestation is a major issue in Myanmar and has been increasing since the democratization of the country in the 1990s. Efficient enforcement of forest regulations is often unreliable due to the temporal latency of available forest loss data. Recent, near real-time (NRT) monitoring methods have reduced this latency, but the most consistent methods can only identify daily deforestation at least 6 ha in size. Illegal logging in Myanmar, and elsewhere, often takes the form of smaller scale selective removals. For remote sensing to be relevant for policy enforcement, NRT monitoring methods must be refined to detect deforestation sooner, and at finer spatial scales. The overarching goal of this project is to make progress toward this reduced-latency NRT monitoring by combining recent developments in data availability, high-performance computing, and advanced statistical methods. We therefore propose to develop a continuously validated monitoring system that assimilates multi-source remotely sensed imagery to provide daily updated deforestation probabilities for two protected areas in Myanmar. This effort is organized into two main objectives: 1) Use a Bayesian ensemble approach and multi-source imagery to reduce the latency and improve the spatial resolution of NRT deforestation monitoring; and 2) Create a continuously ground-validated application system using the probability maps. Specifically for Objective 1, daily deforestation probability maps will be calculated for the study sites within Myanmar. Training and validation data will be obtained from in-country partners via collaborators at the Smithsonian Conservation Biology Institute. Posterior probabilities per pixel will be determined by computing the likelihood of disturbance of all available multi-source imagery combined with a prior disturbance probability based on pixel-specific covariates. Then for Objective 2, the daily maps will be available via a Google Earth Engine application. The application will incorporate user (forest manager) interactive feedback by refining the training data, which will fix the current data creator / data user paradigm by closing the loop between the two actors.