Semiautomatic detection of coastal mangroves with Landsat Level-2

Objective: A model for rapid detection of coastal mangrove cover was devised. The idea is that it can be applied by users with basic knowledge of remote sensing and GIS. Design/methodology/approach: The model is based on calculating the first 3 principal components (PC) from bands corresponding to the visible, near infrared, and shortwave infrared regions in Landsat Level-2 images. The model was tested for three RAMSAR sites located in different hydroclimatic conditions in Mexico: Laguna Guasima on the upper Gulf of California coast, Puerto Arista on the Pacific Ocean coast, and Laguna Madre on the Gulf of Mexico. Results: It was found that the first PC in the three RAMSAR sites explains 80 to 90% of the variation and corresponds mainly to areas that include crop fields or urban infrastructure. The second PC, with cumulative variance of 8 to 14%, corresponds mainly to mangrove cover, and the PC with the lowest percentage of cumulative variance (< 5.0%) is invariably open water. Limitations on study/implications: The advantage of using Landsat Collection Level 2 is that there is an archive managed by the USGS of imagery from virtually all over the world that is over 50 years old. This model is not able to identify mangrove species. Findings/conclusions: The advantages of this proposed model are: 1) it uses Collection 2 Level-2 images, which have radiometric and atmospheric corrections; 2) since it is carried out in ArcGIS Model Builder, it can be automated, making it intuitive and enabling the results to be exported to a Python script; and 3) the model can be replicated accurately with the QGIS model builder tool.