RADD Forest Disturbance Alerts

RADD (Radar for Detecting Deforestation) Forest Disturbance Alerts is a near-real-time tropical forest disturbance product based on Sentinel-1 C-band SAR backscatter, developed and maintained by Wageningen University and Research (WUR) and distributed through WRI Global Forest Watch^[1][2]. The algorithm detects forest canopy loss independent of cloud cover or illumination conditions, giving it operational advantage over optical-based alert systems in persistently cloudy humid tropical environments.

The product is distributed at 10 m pixel spacing, matching the Sentinel-1 GRD IW pixel grid^[1]; the native SAR ground resolution is approximately 20x22 m^[3], so individual alert pixels represent sub-resolution detections confirmed through a multi-temporal Bayesian framework. Each pixel receives a pixel-specific Gaussian Mixture Model fit to the pre-disturbance backscatter distribution, and a disturbance flag is confirmed only when the posterior probability exceeds a set threshold, reducing speckle-driven false positives^[1]. The minimum detectable unit is 0.2 ha (approximately 20 pixels).

The product launched in 2021 covering six Congo Basin countries^[2] and had expanded to approximately 50 countries across the humid tropics by early 2023^[3]. Coverage is restricted to humid tropical forest biomes; the algorithm is not designed for boreal or temperate forest systems. Accuracy and revisit performance figures published in Reiche et al. (2021) reflect the Sentinel-1A+B constellation era; Sentinel-1B was retired in August 2022 and Sentinel-1C was added in late 2024^[4]. Post-fleet-change revisit cadence and detection accuracy have not been formally re-validated in the peer-reviewed literature accessed.

RADD alerts are integrated into Global Forest Watch alongside GLAD-L and GLAD-S2 optical alerts. Reiche et al. (2024) demonstrated that combining SAR and optical alert streams improves both detection timeliness and confidence relative to any single system^[4]. The product is available under a CC BY 4.0 licence^[2][5]. Primary access is through the Global Forest Watch map interface and the GEE Community Catalog dataset^[3].