Abstract
Synthetic aperture radar (SAR) enables the mapping of flooding over large areas, regardless of cloud and weather conditions. Simplified classification approaches based on threshold levels of backscattering intensity are typical in operational flood monitoring. Backscattering intensity from floods over open non-forested areas is typically lower, whereas backscatter from forest floods is higher compared to non-flooded areas. However, distinction of flooded areas from non-flooded surface in semi-forested areas with low canopy closure (CC) or low tree height (TH) is expected to be difficult due to confounding effects of the different scattering mechanisms. The aim of this study was to investigate X-band SAR backscattering in flooded boreal forests with varying TH and CC, and to quantify in which cases floods would be difficult to detect with typical threshold-based classification methods. To further understand the SAR signal behavior in flooded forests, the total backscatter was modeled using the HUT (Helsinki University of Technology) semi-empirical forest backscattering model. HH-polarized Cosmo Sky-Med acquisitions from four different locations in Finland were analyzed against airborne LiDAR based forest data, ground observations and a high resolution digital elevation model. Floods were well detected in open areas and dense forests. However, as hypothesized, when TH was higher than zero but lower than 4–5 m, or when CC was higher than zero but lower than 15–20%, the detection was less successful. TH was found to have slightly more influence on the capability of X-band SAR to detect floods than CC. In our four test areas, namely Kittilä, Kolari, Pudasjärvi and Evo, 85.3%, 89.1%, 82.7% and 73.9% of the total floods were detected, respectively, by a simple threshold value method. The model was able to successfully estimate the different backscattering components of the flooded forests in Kittilä and Pudasjärvi, where the number of observations from all forest conditions was sufficient.
Original language | English |
---|---|
Journal | Remote Sensing of Environment |
Volume | 186 |
Pages (from-to) | 47-63 |
Number of pages | 17 |
ISSN | 0034-4257 |
DOIs | |
Publication status | Published - 1 Dec 2016 |
MoE publication type | A1 Journal article-refereed |
Fields of Science
- 1172 Environmental sciences