An accurate detection of landslides was achieved herein with the application of a decision-tree model with explanatory variables calculated from satellite images and digital elevation models (DEMs) learned from two cases of landslides in Japan with different causes. The study areas were (1) Atsuma, Hokkaido, where the 2018 Hokkaido Eastern Iburi earthquake occurred, and (2) Asakura, Fukuoka, where the heavy rainfall in northern Kyushu in July 2017 caused damage.

In the decision-tree analysis, explanatory variables were calculated from optical satellite images and DEMs. Based on these data, three models were generated: the ‘Atsuma model’ using the Atsuma case as supervised data, the ‘Asakura model’ using the Asakura case as supervised data, and the ‘Atsuma+Asakura model’ using pixels extracted from the Atsuma and Asakura cases as supervised data. Each model was applied to test data that were randomly extracted from the target areas of Atsuma and Asakura, and the accuracy of the landslide detection was quantitatively evaluated with Cohen’s Kappa, F-measure, and overall accuracy.

The analysis results demonstrated that the Atsuma+Asakura model had the same level of detection accuracy as the Atsuma model, with Cohen’s Kappa of 0.701 and overall accuracy at 91.7% in the Atsuma case. In the Asakura case, the Atsuma+Asakura model’s detection accuracy had a Cohen’s Kappa of 0.559 and overall accuracy at 90.0%, which were at the same level as those of the Asakura model. These results indicate that the decision-tree model trained on multiple cases has high detection accuracy and generalization performance and is effective for detecting landslides caused by multiple factors. The effectiveness of the parameters calculated from the optical satellite images and DEMs for detecting landslides was also quantitatively confirmed by ROC curves.

Lidar observations at 532 nm were made in conjunction with Total Carbon Column Observing Network (TCCON) observations in Burgos, Philippines (18.52˚N, 120.65˚E) in Southeast Asia to investigate the effects of aerosols and thin cirrus clouds on column-averaged dry air mole fractions of greenhouse gases such as carbon dioxide (XCO₂) estimated from the Greenhouse gases Observing SATellite (GOSAT) and its successor GOSAT-2. Here, we report on stratospheric aerosol events observed from January 2017 to March 2020, and from January to March 2023. Non-spherical smoke particles originating from the Canadian forest fires in August 2017 were observed on 25 September 2017. The backscattering ratio (BSR) and the particle depolarization ratio (PDR) of these smoke particles were 1.20 and 0.15 at an altitude of 18.24 km. Furthermore, smoke particles were still detected around 20 km altitude on 2 April 2018, exhibiting a PDR of 0.07.

Three aerosol layers were observed on 2 October 2019, with respective BSR peaks of 1.40 (18.09 km), 1.57 (20.79 km), and 2.07 (25.14 km). The narrow aerosol layer at an altitude of 25 km was observed on 8 out of 12 nights from 11 September to 24 October 2019. These increases in stratospheric aerosols could be the result of the eruption of Mt. Raikoke (48.29˚N, 153.25˚E) in June 2019 and the eruption of Mt. Ulawun (5.05˚S, 151.33˚E) in August 2019.

During the whole lidar observation period, the maximum value of the integrated backscattering coefficient (IBC) of stratospheric aerosols from the tropopause to 33 km altitude was 4.45×10⁻⁴ sr⁻¹ on 29 September 2017, and the minimum IBC was 4.53×10⁻⁵ sr⁻¹ on 2 October 2017. Assuming a lidar ratio of 50 sr, we estimated the maximum value of the stratospheric aerosol optical depth (SAOD) to be 0.022 on 29 September 2017 and the minimum SAOD to be 0.0023 on 2 October 2017.

This study targets the surrounding areas of Okaya City, Nagano Prefecture (Okaya sites), and Kanzaki City, Saga Prefecture (Kanzaki sites), where heavy rainfalls caused landslides in August 2021, and uses optical satellite data from two types of optical satellites with different observation conditions and sensors. We extracted the sediment runoff range by calculating the difference value (ΔNDVI) and compared the accuracy, and investigated the difference in the ΔNDVI threshold for distinguishing new bare land from other areas.

The relationship between precision (x) and recall (y) can be approximated by the quadratic equation y＝ax²＋bx＋c, and there was a trade-off relationship between the two in that as the threshold rose, precision increased while recall decreased. Comparing the quadratic approximation curves at the Okaya and Kanzaki sites, the values at the Okaya sites were higher overall, suggesting that the value of the solar zenith angle in the satellite data may have had an effect. Furthermore, when comparing satellite data, the precision of the ΔNDVI threshold case with the maximum F-measure is 0.71 for Sentinel-2 and 0.72 for Planet, and the recall is 0.70 for Sentinel-2 and 0.75 for Planet. The average of the ΔNDVI threshold, which indicates the maximum F-measure, was 0.19 for Sentinel-2 and 0.22 for Planet, and differed depending on the satellite data used for analysis. The lower ΔNDVI threshold in Sentinel-2 was thought to be due to the lower recall due to coarser spatial resolution compared to Planet. These findings suggested that the ΔNDVI threshold that optimizes the extraction accuracy of the sediment runoff range differs depending on the satellite data used for analysis.