Tidal Flat Extraction and Change Analysis Based on the RF-W Model: A Case Study of Jiaozhou Bay, East China

Abstract

Coastal tidal flats are important ecological resources. As the dividing line between marine and terrestrial ecosystems, tidal flats provide a large number of ecosystem services. However, with the excessive development of coastal areas, tidal flat resources have been drastically reduced, leading to the deterioration of coastal ecosystems. There is an urgent need to acquire accurate information on the changes in tidal flat resources. This research proposes a tidal flat extraction model (RF-W model) that combines the random forest (RF) method and waterline method, which aims to improve the accuracy of tidal flat extraction. This method can effectively eliminate the shortcomings of the RF method in determining the lower boundary of tidal flats and those of the waterline method in distinguishing river channels and tidal flats. The tidal flat extraction of Qingdao Jiaozhou Bay in 2020 is performed as an example of the model. The results show that the user’s and producer’s accuracies of the RF-W model were both the highest, indicating that the improved model can accurately extract tidal flat information. Then, we used the RF-W model to extract tidal flat information for Jiaozhou Bay in seven periods (1990, 1995, 2000, 2005, 2010, 2015, and 2020) and to study the spatiotemporal changes in the tidal flats and influencing factors from 1990 to 2020. The tidal flat area of Jiaozhou Bay showed an overall downward trend before 2015, and the area decreased by 21.9 km2, with a reduction in the rate of approximately 1.1%/year. After 2015, the tidal flat area rebounded slightly. The overall change in Jiaozhou Bay showed reclamation and expansion toward the sea. The reduction in the sand content of the rivers entering the sea, reclamation and cultivation, and land development were the main factors contributing to the reduction in the tidal flat area in Jiaozhou Bay. In addition, sea level rise due to climate warming is a long-term potential factor.