Coastal aquaculture monitoring is vital for sustainable offshore aquaculture management. However, the dense distribution and various sizes of aquacultures make it challenging to accurately extract the boundaries of aquaculture ponds. In this study, we develop a novel combined framework that integrates UNet++ with a marker-controlled watershed segmentation strategy to facilitate aquaculture boundary extraction from fully polarimetric GaoFen-3 SAR imagery. First, four polarimetric decomposition algorithms were applied to extract 13 polarimetric scattering features. Together with the nine other polarisation and texture features, a total of 22 polarimetric features were then extracted, among which four were optimised according to the separability index. Subsequently, to reduce the “adhesion” phenomenon and separate adjacent and even adhering ponds into individual aquaculture units, two UNet++ subnetworks were utilised to construct the marker and foreground functions, the results of which were then used in the marker-controlled watershed algorithm to obtain refined aquaculture results. A multiclass segmentation strategy that divides the intermediate markers into three categories (aquaculture, background and dikes) was applied to the marker function. In addition, a boundary patch refinement postprocessing strategy was applied to the two subnetworks to extract and repair the complex/error-prone boundaries of the aquaculture ponds, followed by a morphological operation that was conducted for label augmentation. An experimental investigation performed to extract individual aquacultures in the Yancheng Coastal Wetlands indicated that the crucial features for aquacultures are Shannon entropy (SE), the intensity component of SE (SE_I) and the corresponding mean texture features (Mean_SE and Mean_SE_I). When the optimal features were introduced, our proposed method performed better than standard UNet++ in aquaculture extraction, achieving improvements of 1.8%, 3.2%, 21.7% and 12.1% in F1, IoU, MR and insF1, respectively. The experimental results indicate that the proposed method can handle the adhesion of both adjacent objects and unclear boundaries effectively and capture clear and refined aquaculture boundaries.

Net-pen aquaculture, one of the primary forms of aquaculture in shallow lakes, has generated substantial economic benefits but has also triggered eutrophication and ecological degradation. Therefore, the accurate and scientific dynamic monitoring of aquaculture areas is of the utmost importance. However, existing studies have predominantly employed annual monitoring, failing to fully exploit the unique advantage of synthetic aperture radar (SAR) imagery for seasonal, monthly or more frequent time scale monitoring. This study aimed to develop an approach for monitoring the monthly dynamics of net-pen aquaculture using Sentinel-1 time series images. We integrated threshold segmentation and temporal consistency checking to extract spatial data on the net-pen aquaculture from eight typical shallow lakes in Jiangsu Province, one of the eastern coastal provinces of China, from 2016 to 2021. This study yielded the following results: (1) The VV polarization image offers a particularly noticeable contrast between net-pen aquaculture structures and their surroundings and checking for temporal consistency can help alleviate the problem of distinguishing between changes in error caused by misclassification and actual changes that occurred. The overall accuracy of the net-pen aquaculture extraction results was over 85%. (2) This approach not only enables the analysis of changes in the aquaculture area and fence length but can also identify key time points. For instance, the net-pen aquaculture area in Gehu Lake experienced a rapid decline from May to November 2018, the decrease in while Eastern Taihu Lake occurred from September 2018 to June 2019. (3) The removal of net-pen aquaculture in lakes within the Jiangsu Province is primarily attributed to policy responses. In response to the ecological river and lake action plan implemented in Jiangsu Province, a decrease was observed in both the aquaculture areas and fence lengths across most lakes from 2016 to 2021. These results can provide a reference for lake management and ensuring the effective implementation of related policies.

Coastal aquaculture has become one of the main sources of animal protein and plays an important role in food and nutrition supplies and security around the world. Accurately mapping aquaculture areas is the basis for its sustainable management and use, and provides important support to policy development and implementation at regional, national, and global levels. Considering the concentrated and densely distributed characteristics of aquacultures, it is difficult to distinguish the dikes between aquacultures and identify small-scale aquacultures using medium-and low-resolution SAR images. GaoFen-3 (GF-3) is the first launched full-polarimetric C-band SAR satellite of China at metre-level resolution. This study aims to use a novel combination model to extract coastal aquacultures in the Yancheng coastal wetland, China on the basis of GF-3 Fully Polarimetric SAR Imagery. Polarimetric decomposition algorithms were applied to extract polarimetric scattering features and feature optimization was applied based on the separability index. To separate adjacent and even adhering ponds into individual aquaculture objects, we proposed a novel model that integrated two UNet++ subnetworks with the marker-controlled watershed (MCW) segmentation strategy to obtain more refined coastal aquaculture results. The accuracy assessment results demonstrated a considerable performance, with F1 greater than 95%, IoU greater than 90%, and insF1 higher than 85%. The experimental results indicate that the proposed algorithm achieved fairly high accuracy in aquaculture extraction and can effectively improve the boundary quality of individual aquacultures.

Insufficient three-dimensional (3D) geolocation accuracy of interferometric synthetic aperture radar (InSAR) scatterers on dikes often hampers the distinction between the dike crest and slopes. This limitation hinders effective monitoring and interpretation of deformation associated with different dike structures using the multi-temporal InSAR (MT-InSAR) technique. To address this, we propose a geolocation improvement method by integrating light detection and ranging (LiDAR) point cloud. We first construct and transform a 3D error ellipsoid of each InSAR scatterer by quantitatively estimating its geolocation uncertainty. Next, we design and employ a rotation matrix and projection model to extract LiDAR points located within each error ellipsoid. Then we use the mean position and height of the extracted LiDAR counterparts to improve the geolocation accuracy of the InSAR scatterers. The Houtribdijk, as our test site, is a 26.5 km long dike in the Netherlands, where we used both ascending and descending Sentinel-1 tracks acquired from 2018 to 2022. Results show that the original geocoded InSAR scatterers exhibit positional and height discrepancies between the two datasets, and their height variations fail to reflect the actual topographic features of individual slopes. After improvement, the heights of the InSAR scatterers agree well with LiDAR measurements, with root mean square errors reduced by up to 97%. Coefficients of determination with the AHN4-derived digital surface model increase from 0.22 and 0.24 to 0.82 and 0.78 for ascending and descending tracks, respectively, and reach 0.96 along the Houtribdijk. The percentage of boundary scatterers located within the dike extent also increases from 42% and 20% to 85% and 84% for the ascending and descending tracks, respectively. This test demonstrates that our method effectively improves InSAR geolocation and provides the spatial and geometric basis for deformation analysis of different dike structures.