Quantifying the Long-Term Expansion and Dieback of <italic>Spartina Alterniflora</italic> Using Google Earth Engine and Object-Based Hierarchical Random Forest Classification

Abstract

The overwhelming spread of Spartina alterniflora (S. alterniflora) (smooth cordgrass) over recent decades has put many native plant communities and coastal environments at risk. Therefore, long-term monitoring of S. alterniflora dynamics is necessary to better understand and manage the invasion of the species. However, it is difficult to map Spartina saltmarshes in China on an annual or multiyear epoch basis. To address this issue, we developed a classification approach integrating Google Earth Engine (GEE) and object-based hierarchical random forest (RF) classification, and we applied this approach to quantify the expansion and dieback of S. alterniflora at Dafeng Milu National Nature Reserve, Jiangsu, China during 1993–2020. Results showed that the area of S. alterniflora expanded from 24.48 ha in 1993 to 1564.96 ha in 2010. However, after ecological hydrological engineering and an increase in Elaphures davidianus (Père David's deer) numbers in 2011, the S. alterniflora area decreased significantly to 944.28 ha in 2020. During 2011–2020, the S. alterniflora area decreased substantially at a rate of 67 ha per year and by 86% in one area studied in Dafeng Milu National Nature Reserve. In 2020, the 944.28 ha of S. alterniflora in the reserve was mainly distributed in mudflats by the sea. Overall, these results show that it is feasible to identify S. alterniflora using the GEE platform and object-based hierarchical RF classification; moreover, this approach could improve understanding and management of this invasion species.

Keywords

• Dafeng milu national nature reserve
• Google Earth Engine
• landsat
• object-based image analysis
• random forest (RF)
• spartina alterniflora