Mapping the spatial-temporal variability of tropical forests by ALOS-2 L-band SAR big data analysis

Abstract Insufficient knowledge about spatial-temporal forest heterogeneities in the tropics is a major impediment to better estimation of carbon storage and prevention of forest loss by remote sensing. Seasonality induced fluctuations can severely impede the reliability of synthetic aperture radar (SAR) applications like forest classification, deforestation detection and biomass estimation. In the past, the seasonality effects on SAR data acquired over tropical forests have been largely ignored due to the lack of consistent time-series data for the entire region. To overcome this knowledge gap, this paper presents the first comprehensive statistical assessment of L-band SAR backscatter variability in tropical forests by means of a homogeneous big data analysis on multitemporal ALOS-2/PALSAR-2 dual-polarization ScanSAR data. Virtually the entire tropical forest region was monitored from 2016 to 2018 with 42 days intervals. Approximately 250 TB of ScanSAR data were computed on a 0.05 decimal degrees grid to produce forest maps which provide more detailed and reliable information about the spatial and temporal behavior of tropical forests. The first, easy to understand global forest variability maps provide unseen insights into forest structures for the entire tropical belt from local to continental scale. While the differences between the evergreen rainforest and the (semi-) deciduous dry forest is obviously large in the spatial domain, in the temporal domain seasonal changes can easily cause variations in the same order of magnitude within the same forest class. Depending on the area, natural environmental factors like flooding or leaf shedding can result in seasonal gamma naught ranges exceeding 3 dB in both HH and HV polarization. The findings of this study should have important implications for all future remote sensing research efforts on tropical forests. With drastically increasing availability of more frequent high-resolution L-band observations in the coming decade, these characteristic forest variabilities must be considered in the development of active and passive microwave remote sensing algorithms. The better understanding of the underlying dynamics and processes is expected to improve forest remote sensing applications for the upcoming L-band SAR satellite generation including ALOS-4, NISAR, and Tandem-L.