Automatic model-based estimation of boreal forest stem volume from repeat pass C-band InSAR coherence

C-band repeat pass interferometric synthetic aperture radar coherence can provide high-accuracy estimates of boreal forest stem volume in spite of environmental dependence. Typically, the retrieval methods require a data set of in situ measurements for training the model linking coherence to stem volume. The drawback is the need for such data and the incapacity to take into account spatial variations of environmental conditions. Here, we demonstrate a model training method that does not require reference data. For the investigated case, the relative root mean square error of stem volume is 18% or as good as obtained using the traditional training method with in situ data. Index Terms--Coherence, European Remote Sensing (ERS), forestry, interferometry, stem volume, synthetic aperture radar.