Detecting Deforestation Using Robust Online Bayesian Monitoring

Near real time change detection is important for a variety of Earth monitoring applications and remains a high priority for remote sensing science. Data sparsity, subtle changes, seasonal trends, and the presence of outliers make detecting actual landscape changes challenging. Adams and MacKay introduced Bayesian Online Changepoint Detection (BOCPD), a computationally efficient, exact Bayesian method for change detection. Incorporation of prior information allows for relaxed dependence on dense data and an extensive stable period, making this method applicable to short time series and multiple changepoint detection. In this paper, we conduct BOCPD with a multivariate linear regression framework that supports seasonal trends and outlier robustness. We adapt BOCPD to introduce Robust Online Bayesian Monitoring (roboBayes), which includes a mechanism to make BOCPD robust against occasional outliers, such as corrupted images, without compromising the computational efficiency of an exact posterior change distribution nor the detection latency. Without this robustness feature, the change detection algorithm cannot be employed autonomously without an inflated false positive rate. We show via simulations that the method effectively detects change in the presence of outliers. The method is applied to monitor deforestation in Myanmar where we show superior performance compared to current online changepoint detection methods.