[S-TT57] [EJ] Synthetic Aperture Radar

[STT57-06] Development of early deforestation detection algorithm (advanced) with PALSAR-2/ScanSAR for JICA-JAXA program (JJ-FAST) 3 – Time series analysis in South America –

*Manabu Watanabe¹, Christian Koyama¹, Masato Hayashi², Hiroshi Miyoshi², Masanobu Shimada^1,2 (1.School of Science and Engineering, Tokyo denki university, 2.JAXA)

Keywords:Forest monitoring, ALOS-2, Polarimetry

Time series PALSAR-2/ScanSAR data were used for detecting early-stage deforestation. The data were taken 9 times/year, following ALOS-2 systematic observation strategy [1], and it covers global areas, including major tropical forest in the world. By using this data, JICA and JAXA launched a service, “JICA-JAXA Forest Early Warning System in the Tropics (JJ-FAST)” in November 2016 [1]. The system is on the web, and is freely accessible from a smartphone or other devices. Decrease of Gamma⁰_HV are observed after deforestation, and current algorithm uses HV polarization data and two data taken in different timing. HH polarization, and time series data will be used for the future operation. Time series data obtained in South America with HH and HV polarizations were used, and compared it to the Landsat data to clarify the ability to detect the early-stage deforestation, where fallen trees were left on the ground. The Sigma⁰_HH value increased by 1.2 dB in areas undergoing the early stages of deforestation. The detection timing is almost same as that using the optical sensor. On the other hand, the Sigma⁰_HV value decreased by 3.2 dB for late-stage deforestation areas, where fallen trees were removed. The detection timing is about a few month after the detection of deforestation by Sigma⁰_HH, or optical sensor. Many errors of the deforestation detection were observed at wet forest areas. Temporal variations of Gamma⁰ were observed for the area, which induces the deforestation detection errors. The variations of Gamma⁰ shows some correlation with precipitation for both HH and HV polarization [3], and flooding, variation of moisture for soil and trees may be the possible cause for the Gamma⁰ variation.

[1] ALOS-2 systematic observation strategy,
http://www.eorc.jaxa.jp/ALOS/en/top/obs_top.htm Accessed February 16, 2017
[2] JJ-FAST, http://www.eorc.jaxa.jp/jjfast/jj_index.html, February 16, 2017
[3] Manabu Watanabe, et al., Multi-temporal Fluctuations in L-band Backscatter from a Japanese Forest, IEEE Trans. Geosci. Remote Sensing, 53(11), 5799-5813, 2015