Search

• GF-1 and GF-6 remote sensing data have potential for wetland change detection. • Spatial-temporal attention method can identify information changes in cultivated land to wetland areas. • Sanjiang National Nature Reserve has large wetland restoration areas located near water sources. • The results of the study have important implications for the restoration and sustainable development of wetlands. Wetlands are important ecosystems, and detecting changes in wetlands plays a very important role in the management of wetland resources. Therefore, we selected Sanjiang National Nature Reserve as the study area and proposed a method in which bi-temporal GaoFen (GF) remote sensing data was combined with a spatial–temporal attention neural network (STANet) model, coupled with the selection of spectral variables. We used the bi-temporal images from GF-1 and GF-6 and compared three band selection methods, Relief F, principal component analysis, and RGB combination, and two STANet-based change detection methods (BAM: basic spatial–temporal attention module and PAM: pyramid spatial–temporal attention module) to improve the change detection of the wetland restoration. The results show that the accuracy of wetland change detection is related to different combinations of spectral variable selection and STANet models. The combination of Relief F feature selection and the BAM model led to the highest accuracy of the wetland change detection, 71.72 % for F1-Score and 76.92 % for MIou. The vector results of the change detection statistics show that the area of the Sanjiang National Nature Reserve cultivated lands converted to wetlands is 46.05 km2. This study implies that the STANet method has the capacity of detecting complex wetland restoration. Red, Bule, NIR and Red Edge1 bands of GaoFen data can provide effective information in the detection of changes in wetlands. The results of the study can be used as a reference for rational planning of wetland restoration, seeking a balance between cultivated lands and wetlands, which is important for the sustainable development of wetland. [ABSTRACT FROM AUTHOR]

• Dongting Lake has experienced land covers changes due to extreme hydrologic events. • We use qualitative and quantitative methods for change detection in Dongting Lake. • A new change detection method Bi-SRUNet++ is proposed for change detection. • This method has the great potential to improve the accuracy of wetland change detection. • The obtained information is important for wetland management and sustainable development. The utility of semantic change detection in myriad change scenarios has garnered considerable attention in contemporary research; however, its applicability in monitoring alterations in wetland ecosystems remains incompletely elucidated. To surmount the constraints associated with binary change detection methodologies—chiefly their insufficiency in the extraction of bi-temporal attributes—we introduced the Bi-Temporal Semantic Reasoning UNet++ (Bi-SRUNet++) algorithm. This algorithm leverages the architectural strengths of UNet++ as its foundational network to precisely delineate features pertinent to multi-class change detection. As a preliminary step, the study focused on the Dongting Lake wetland in China and conducted an analysis of feature trends predicated upon the monthly Normalized Difference Water Index (NDWI) and Normalized Difference Vegetation Index (NDVI), as derived from Landsat 8 data spanning 2021–2022. Subsequently, the optimal temporal phases for change detection were ascertained through differential analyses between NDWI and NDVI metrics. Implementing the Bi-SRUNet++ algorithm on a pair of Sentinel-2 images, captured during the optimal phases, yielded augmented change information. Comparative evaluations reveal that the Bi-SRUNet++ algorithm, conceptualized on the framework of Bi-Temporal Semantic Reasoning Network (Bi-SRNet), surpasses the performance indices of its counterpart Semantic Segmentation and Change Detection Late Fusion (SSCD-l). Furthermore, the incorporation of the UNet++ backbone network amplifies the algorithm's capacity for semantic feature extraction, thereby enhancing the efficacy of Bi-SRUNet++ in wetland change detection. The analysis divulges that the total altered area of Dongting Lake during the 2021–2022 period amounts to 1187.97 km2, comprising a water loss of 1186.16 km2, a 715.34 km2 transformation into vegetation, and a conversion of 469.96 km2 into mudflats. The codes and partial dataset in this paper are available at: https://github.com/vivianmiumiu/Bi-SRUNetplusplus-for-SCD. [ABSTRACT FROM AUTHOR]