[Combining Multi-source Remote Sensing Data and Object-oriented Information Extraction for Arid Wetlands].

Area of Science:

• Remote Sensing
• Environmental Monitoring
• Geospatial Analysis

Context:

• Arid and semi-arid regions, particularly in Northwest China, rely heavily on wetlands for ecological stability and resource management.
• Effective wetland information extraction is crucial for monitoring ecological health, conserving biodiversity, and preventing desertification.
• The Ningxia Yinchuan metropolitan area along the Yellow River serves as a case study for wetland assessment in arid environments.

Purpose:

• To explore the effectiveness of red edge, radar, and topographic features in extracting wetland information in arid regions.
• To evaluate the RF-Pearson model for optimal feature selection combined with random forest and BP neural network algorithms.
• To accurately map and classify wetland types in the Ningxia Yinchuan metropolitan area.

Summary:

• Integrating Sentinel-1 radar, Sentinel-2 red edge bands, and topographic data significantly enhances wetland identification accuracy in arid zones.
• The RF-Pearson model combined with the random forest algorithm achieved the highest classification accuracy (89.79% overall, 0.8423 Kappa).
• River wetlands were identified as the dominant type, with natural wetlands covering a larger area than artificial wetlands in the study region.

Impact:

• Provides a robust methodology for wetland information extraction in data-scarce arid regions.
• Offers critical data for ecological monitoring, biodiversity conservation, and sustainable development planning in the Yellow River Basin.
• Highlights the importance of multi-source data fusion and advanced machine learning for environmental management.