Optics and microwave detection of forest restoration after fires

Abstract

Рассмотрена возможность обнаружения лесного подроста с помощью поляриметрических разложений и анализа временны´х рядов радиолокационных изображений ALOS-2 PALSAR-2 и Sentinel-1. Представленные результаты показывают принципиальную возможность уверенно определять рост лесных насаждений. The problem of large-scale assessment of forest restoration after artificial deforestation and wildfires is relevant in connection with climate change and the corresponding desire of the world community for further low-carbon development. One of the promising methods aimed at solving this problem is remote sensing of the Earth using space-based synthetic aperture radars (SAR). The paper proposes a comprehensive approach to assessing the dynamics of forest plantings growth using time series analysis of ALOS-2 PALSAR-2 and Sentinel-1 radars data, as well as Sentinel-2A/B optical sensors data. It is shown that with the help of model-based decomposition (Freeman-Durden decomposition) of fully polarimetric data in L-band (ALOS-2 PALSAR-2), the increasing of volume scattering component and the corresponding decrease in the surface component can confidently identify the growth of young forest. However, the data of ALOS-2 PALSAR-2 with dual polarization are not able to separate forest undergrowth from other types of vegetation over a seven-year observation interval. This is also true for the C-band. Thus, the polarimetric Cloud- Pottier decomposition of Sentinel-1 data allowing only separation for the areas with vegetation from the treeless ones. Time series analysis of radiometrically corrected radar backscattering at vertical co-polarization in this band, imaged in the winter period of time, allows reliable determining of the dynamics for the growth of forest plantations. The use of freely available Sentinel-2A/B multispectral sensors data makes it possible to further divide the identified undergrowth by species composition and exclude classification errors of radar data in treeless areas, which show an increasing of volume backscattering component on model-based polarimetric decompositions.