A study on the classification of coastal wetland vegetation based on the Suaeda salsa index and its phenological characteristics

The vegetation of coastal wetlands exerts a significant influence on the functioning of the ecosystem. The study of effective vegetation indices and phenological metrics to enhance the separability of the classification characteristics between different vegetation types is a fundamental prerequisite for the expeditious and accurate classification of coastal wetland vegetation and the acquisition of the vegetation structure, which is of paramount importance for the conservation of coastal wetlands. The conventional approach to vegetation classification, based on phenological characteristics, relies primarily on the Normalized Difference Vegetation Index (NDVI) to assess vegetation growth. However, this method lacks the capacity to discern changes in vegetation colour. Therefore, in this study, based on the analysis of the spectral separability of Suaeda salsa (S. salsa) and native vegetation, two new Red Suaeda salsa Indices (RSSI and RSSI(1)) were constructed by selecting the red, green and near-infrared bands of the Sentinel 2 multispectral band, and then, based on the sample points, we constructed the RSSI time series and fitted using Fourier function fitting, and extracted (a) Difference Of RSSI (DOR), (b) Sum Of RSSI (SOR), (c) Ratio Of Green-up RSSI (ROGR), and (d) Ratio Of Senescence RSSI (ROSR) from the phenological fitting curve. The four phenological metrics were combined with spectral and textural features to classify the vegetation using the random forest (RF) algorithm. In order to demonstrate the efficacy of the constructed vegetation indices, this study employed both the NDVI and the existing Suaeda salsa Vegetation Index (SSVI) to calculate the phenological metrics for classification purposes. The results showed that: (1) The classification results of the phenological metrics using any of the S. salsa indices were significantly better than those of the NDVI, with RSSI being the best, with the accuracy of S. salsa being improved by 10 % for the producers and 30 % for the users, and the overall accuracy being improved by 13 %, the Kappa coefficient increased by 0.19. (2) The results of RSSI(1) and SSVI were more consistent with each other, with the overall accuracy increased by 10 % and the Kappa coefficient increased by 0.14. (3) When the combination of phenological metrics obtained from four vegetation indices was used for classification, the results were slightly better than those of a single vegetation index, with the overall accuracy increasing by 2 % and the Kappa coefficient increasing by 0.03. The study showed that RSSI time series phenological characteristics possess greater potential for the classification of coastal wetland vegetation.