The effect of forest canopy on remote sensing observations in the boreal region

Remote sensing techniques are often used for monitoring various processes in the boreal environment. Typical satellite sensor types for this purpose are Synthetic Aperture Radar (SAR), optical, and passive microwave sensors. Many of the observed targets on the ground are covered by forest canopy. Vegetation considerably influences the signal behavior, especially for the most commonly used microwave and optical wavelengths. It is therefore necessary to consider the effect of forest canopy on the observed signal in order to provide reliable estimations of geophysical phenomena on the ground. Various models describing the interaction of electromagnetic radiation with forest canopy have been developed, but many of these are overly complex with high ancillary data requirements. For retrieval purposes, simple models are preferred. This thesis aims at increasing the understanding of how vegetation, and particularly forest canopy, influence remote sensing observations in boreal environments. The focus is mainly on SAR instruments, but also passive microwave and optical sensors are investigated. The capability of a simple zeroth-order model in simulating the effect of vegetation on the remote sensing signal is first quantified by a spatial analysis of optical, SAR, and passive microwave remote sensing data. Then, the influence of vegetation in SAR remote sensing is further examined through two practical applications; mapping floods under various forest conditions, and detecting soil freezing/thawing in boreal forests. The results demonstrate that despite using a relatively simple model, the extinction of electromagnetic signals in forest canopy was well estimated. Due to both sufficient estimation accuracy and simplicity, the presented model can be considered applicable in near real-time monitoring applications. Floods were well detected in open areas due to specular reflection of the water surface and in dense forests due to double bouncing between flood surface and tree trunks. Yet, in low tree and sparse forest areas, the detection of floods was less successful. The forest backscattering model was capable of separating between the backscatter contributions originating from the ground surface and from the forest canopy, thus enabling the identification of frozen and thawed terrain in forests.