29 December 2025 in IEEE Transactions on Geoscience and Remote Sensing
With the increasing frequency of port trade activities, oil spill incidents pose severe threats to marine ecosystems. The complex port environments, characterized by diverse land cover types and intricate land-water interfaces, present significant challenges for oil spill detection (OSD). To address this, we propose a parameter-efficient fine-tuning (PEFT) approach based on the segment anything model (SAM) for port OSD tasks utilizing unmanned aerial vehicle (UAV) remote sensing imagery. The proposed DADS-SAM freezes most parameters of SAM and performs PEFT only on adapters. Dynamic adaptive low-rank adaptation (DALoRA) enhances the model’s adaptability to port OSD tasks by incorporating dynamic adaptive low-rank matrices into the self-attention modules of the SAM encoder, while dual scale convolutional adapter (DSCA) improves the model’s capability to capture local features and spatial dependencies in oil spill regions. Since the original SAM model lacks class-specific segmentation capabilities, we further design a multi-class segmentation head (MCSH) for oil spill segmentation. This head obtains the output masks from the SAM decoder, adjusts their dimensions, and transforms them through two linear layers to generate class-specific segmentation results. To the best of our knowledge, this study is the first to apply a PEFT approach to port OSD tasks. The proposed method was evaluated on both publicly available port oil-spill datasets and marine oil-spill datasets acquired via synthetic-aperture radar (SAR). Quantitative comparisons demonstrate that DADS-SAM outperforms conventional approaches and achieves state-of-the-art performance across all evaluated metrics. Models and data are publicly available at https://github.com/PetrichorAsh/DADS-SAM.
9 January 2026 in Expert Systems with Applications
18 February 2026 in SSRN Electronic Journal
Sea ice is a key component of the climate system, and reliable monitoring of sea ice thickness is essential for understanding coastal ice evolution. Existing global navigation satellite system-reflectometry (GNSS-R) inversion approaches often suffer from reduced accuracy and instability under low-SNR conditions, weak interferometric structure, and heterogeneous ice states. To address these challenges, we develop a spectral-estimability–driven dual-path adaptive inversion algorithm for ground-based GNSS-R sea ice thickness retrieval. A quantitative spectral-estimability criterion evaluates the dominant interferometric peak using its signal-to-noise ratio, spectral sharpness, and physical admissibility, enabling explicit discrimination between estimable and non-estimable windows. The algorithm then adaptively activates either a multi-channel, physics-constrained ESPRIT branch for super-resolution frequency extraction under high-quality spectra, or a physics-parameterized complex-domain nonlinear least-squares (NLS) branch for joint amplitude–phase fitting when the spectrum is diffuse or low-SNR. Multi-satellite L1/B1 estimates are fused via variance-inverse weighting and robust temporal statistics to reduce single-frequency bias and suppress outliers. Ground-based experiments at multiple coastal sites achieve centimeter-level accuracy, with RMSEs of 1.5–1.7 cm for thick ice and 2.4–2.6 cm for thin ice. Dual-frequency fusion reduces RMSE by 30% relative to single-frequency retrievals and markedly diminishes systematic bias, demonstrating strong robustness under challenging signal and environmental conditions.