An improved dynamic threshold method for determining the start of the vegetation greening season in remote sensing monitoring: The case of Inner Mongolia

• Published in: Ecological informatics Vol. 78; p. 102378
• Main Authors: Li, Haoxin, Li, Frank Yonghong, Guo, Jingpeng, Gao, Xiaotian
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-12-2023
• Subjects: Accuracy; Dynamic Threshold method; Enhanced vegetation index; Maximum curvature; Vegetation phenology; Enhanced vegetation index; Maximum curvature; Vegetation phenology; Accuracy; Dynamic Threshold method
• ISSN: 1574-9541
• DOI: 10.1016/j.ecoinf.2023.102378

The start of the greening season (SOS) is the beginning of vegetation phenology, and its time of occurrence affects the carbon cycle and organic matter production of the ecosystem. Accurate retrieval of vegetation phenology in remote sensing monitoring is important for sustainable management of vegetation resources. The dynamic threshold method (DTM) is one of the mainstream methods to retrieve vegetation phenology. However, it suffers the following two shortcomings. First, a subjective threshold percentage of the phenology nodes is used based on the user's empirical thresholds, and the determined threshold is not universal. Second, a common threshold is used for identifying the phenology globally with little consideration of the physiological characteristics of different vegetation types, thus blurring the phenology variability among vegetation types. Here, we developed an improved dynamic threshold method (IDTM) for the retrieval of the start of the greening season and tested it using temperate grassland and forest vegetation in Inner Mongolia. The algorithm in the IDTM uses the curvature of the EVI2 time series curves of different vegetation types to determine the threshold for the SOS, thus eliminating the subjectivity of the original DTM. The predicted SOS of the IDTM is in better agreement with the ground observations than the MCD12 Q2 vegetation phenology product or the DTM with different empirical thresholds. In terms of spatial correlations, the SOS of the IDTM was significantly correlated with that of the MCD12 Q2 phenology product (r = 0.72, p < 0.05). Our study indicates the efficacy of the IDTM in predicting the SOS of vegetation and that the average SOS of vegetation in Inner Mongolia is in early May (on Day 126) but has a large variation across vegetation zones.
[Display omitted]
•An improved dynamic threshold method is proposed to identify the start of season.•It uses the curvature of the EVI2 time series data to determine the SOS threshold.•The process of determining the SOS threshold is objective & vegetation-type specific.•It is validated against the ground-observed vegetation phenology in Inner Mongolia.•It outperforms the MCD12 vegetation phenology product and dynamic threshold method.