Deforestation Monitoring with SAR and Optical Fusion in the Amazon

Tropical forest ecosystems represent critical global reservoirs of terrestrial carbon and biodiversity, yet their structural integrity remains severely compromised by accelerating deforestation and selective biomass degradation. In the Amazon Basin, timely detection of canopy disturbance is indispensable for early warning monitoring and conservation governance. However, the operational efficacy of conventional optical satellite observation, such as the Landsat and Sentinel-2 constellations, is persistently hindered by dense atmospheric contamination and cloud cover during critical disturbance windows [1], [4]. Consequently, relying solely on passive optical sensors creates significant observational latency that obstructs immediate environmental intervention [7].

Active microwave remote sensing using Synthetic Aperture Radar provides an essential alternative to optical imagery due to its all-weather, day-and-night imaging capability and sensitivity to vegetative moisture and structural geometry [1], [5]. Multiple radar wavelengths, including C-band and L-band systems, offer complementary diagnostic perspectives: while shorter wavelengths capture upper canopy changes, longer wavelengths penetrate deeper into dense vegetation layers to expose biomass removal and ground disruption [5]. Nevertheless, SAR signals remain susceptible to moisture fluctuations and seasonal dielectric variations, necessitating robust filtering and stabilization mechanisms to maintain detection reliability across large tropical domains [7].

The integration of optical and SAR observations creates a multi-source analytical framework capable of mitigating the individual shortcomings of single-sensor systems. Joint processing protocols combine the rich spectral and biochemical sensitivity of optical bands with the cloud-penetrating structural metrics of microwave backscatter [2], [4]. Methodological approaches spanning feature-level fusion, harmonic time series stabilization, and adaptive thresholding significantly improve change detection accuracy and minimize false positives along dynamic deforestation frontiers [3], [7]. Establishing a systematic fusion architecture is therefore vital for continuous, high-confidence tropical forest monitoring.

This dissertation develops a comprehensive methodological framework for automated deforestation monitoring through the synergistic fusion of multi-sensor SAR and optical time series across the Brazilian Amazon. By systematically evaluating sensor interoperability, stabilization algorithms, and spatial decision frameworks, the investigation provides reproducible protocols for near real-time disturbance tracking [7]. The resulting findings bridge theoretical scattering physics and operational Earth observation, supporting scalable environmental protection architectures across tropical forest biomes [1].