The trend of vegetation phenology dynamics is a key factor in understanding vegetation growth and is strongly influenced by responses to long-term variation climate change. However, it still remains a challenge to understand how the trends changed over the long-term. Remote sensing-based vegetation indices (VIs) have been used to monitor phenological and seasonal changes in forest ecosystems. The normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI) are the most widely used. We detected the trend changes in forest phenology in South Korea from 2003-2019, including the start of the season (SOS), end of the season (EOS), and length of the season (LOS) periods in the forest ecosystem using MODIS vegetation indices. The forest area includes deciduous, mixed, and coniferous forest dominated by Quercus sp., Acer sp., Pinus sp. We used the 16-day composite MODIS/Terra and MODIS/Aqua vegetation indices. We generated an 8-day composite Surface Reflectance (MCD13A2) result utilizing terra and aqua data collected over a 16-day period, and the maximum value compositing (MVC) technique. After fitting the NDVI and EVI curves derived from Beck et al. (2006), the phenological thresholds including SOS, EOS, and LOS were extracted based on local extremes from the first derivative function. One MODIS pixel covers different forest types such as deciduous, coniferous, and mixed forest, and hence each pixel requires determining one representative forest type. We adjusted our 1:5000 digital forest-types map to a 1000:1000 scale.
Results/Conclusions
The study demonstrates the use of long-term satellite vegetation indices to detect phenology trends alongside specific forest types. For most forest types, the results indicated the on-average extension of LOS due to a slight advance of SOS and large delay of EOS over the 2003-2019 period. The magnitudes of pixels dominated by coniferous species in SOS was much higher than that by other species. The methodology used in this study could potentially reduce the uncertainties in carbon budgets by detecting the seasonality of forest ecosystems relative to forest type. The results achieved in the study demonstrate that the analysis of trend evolutions using long-term remote sensing data is essential to being able to reveal variations in phenology trends and their shifts.