[A-CG36] Terrestrial monitoring using new-generation geostationary satellites

[ACG36-P06] Investigation of the smoothed peak NDVI trend over Tsambagarav national park based on Landsat time series data

*Kenta Taniguchi¹, Tomomi Sugiyama¹, Chikara Uchida¹, Shigeki Sano¹, Kenta Obata², Hiroki Yoshioka² (1.Nippon Koei Urban Space Co.,Ltd., 2.Aichi Prefectural University)

Keywords:NDVI, reflectance, temperature, Landsat, Tsambagarav National Park

Tsambagarav National Park, located in Mongolian Altai, includes the glacier area of Tsambagarav mountain. This mountain area is protected for the conservation of its native species and is therefore been an area to be monitored. In recent years, however, a rapid change in the regional climate has become a serious threat for the long-term preservation of this enriched environment. In this study, a series of long-term Landsat Earth observation missions was used to monitor the area, mainly to investigate seasonal/decadal trends over the protected region. We employed the normalized difference vegetation index (NDVI) as an indicator of surface conversion in the glacial area. Furthermore, among various NDVI metrics, we focused on the maximum values of NDVI (Max NDVI) because the Max NDVI occurs in the summer season due to the low snow/ice coverage. The major difficulty of processing data for this purpose is to reduce the noise caused by discrepancies in sensor specifications, observation conditions, and inter-seasonal biases caused by global warming. In order to reduce noise, we used a temporal averaging filter to capture seasonal trends of NDVI which tend to be hidden otherwise. The objective of this study is to introduce a simple method to transform Landsat time series data into a de-noised, smoothed curve that makes the summer peak in the NDVI trend clearer. This study used the Landsat Collection-2 Level-2 Science Product (L2SP), which enables us to see the surface reflectance and temperature (Ts) of our area of interest after atmospheric correction using a standardized pixel grid. Our method consists of three steps. First, a set of L2SP data was downloaded by limiting observation to only the three months of summer. In addition, a manual filtering process was performed to discard data of extreme conditions, such as nearly 100 % snow coverage and data influenced by malfunctions in the L7 ETM-plus scan mirror corrector. Second, four types of images (red, NIR, NDVI, and Ts) were converted to their mean values by using a temporal averaging filter with a certain time range. Third, a series of averaged images are processed to obtain linear regression coefficients (slope and offset) as a function of the year at each pixel for four individual images. Finally, we considered the slope as an indicator of decreasing and increasing trends. A set of numerical experiments was carried out using 40 scenes from June, July, and August. Temporally averaged images of red, NIR, NDVI, and Ts over a period of 25 years were produced by setting the window size to 11 years. After filtering L2SP, four types of slope images were generalized, namely, the peak NDVI, red, NIR, and Ts. The results show an overall positive trend for Ts and NDVI, while we observed a negative trend for red and NIR band images. These results imply that climate change (warming), greening, and darkening trends have occurred in Tsambagarav National Park. Furthermore, comparisons of the smoothed peak NDVI trend and false color composite by red and NIR images indicate two different causes of NDVI trends: (1) greening in grass and valley areas and (2) darkening in the glacial area in Tsambagarav National Park. These findings suggest that the combinational use of the smoothed NDVI trend with red-NIR-based false color images and Ts is useful for obtaining a better understanding of the greening trend in Tsambagarav National Park.