[S-TT39] Synthetic Aperture Radar and its application

[STT39-P14] Tropical peatlands drainage detection by using L-band of PALSAR-2

*Haemi Park¹ (1.Graduate School of Global Environmental Studies, Sophia University)

Keywords:Wetlands environment, Human activity, Deep learning

Peatlands are distributed in tropical rain forest and boreal forest areas. Although it has only 2 - 3% of whole terrestial area, the carbon storage in peatlands is known as huge amounts. Conservation of this peatland environment is extremely important in terms of the mitigation of global warming. In case of tropical peatlands, most of the areas are located in Indonesia and Malaysia. However, recently, it is reported that human activity is affecting to the loss of peat soils. Decompositions of peat soils are accelerated by drainage of peatlands. Therefore, to grasp the distribution of drainage canals on peatlands is necessary for adaptation to climate change and accomplishment of sustainable development goals. As the methodology, high resolution and microwave imagery are essential due to the meteorological characteristics of tropical area, for examples rain and clouds. For that reason, this study tried to develop a deep learning-based model for drainage canal detection by using PALSAR-2 HH polarization at 25m spatial resolution. Furthermore, in discussions part, a usefulness of microwave imagery for drainage canal detection was demonstrated through a comparison between both data distributions of LANDSAT-8 NDVI and PALSAR-2 HH. For the comparions, averages of NDVI from January to August 2022 and backscattering values of HH in 2022 were collected. As the result, averaged pixel values in drainage samples were 0.6 and -10.46dB for NDVI and HH, respectively. In case of non-drainage samples, it was 0.63 and -9.54dB, respectively. A methodology by using NDVI, which uses visible and near infrared, might have lower accuracy than microwave-based methodology because the difference of average values between drianage and non-drainage samples of the NDVI is smaller (0.03) than that of microwave (0.92). If the limitation of low data acquisition rate was considered, microwave imagery should be used for drainage canal detection model.