From pixels to rafts: assessing pan-sharpening techniques for oyster aquaculture detection with Landsat data

Abstract Oyster aquaculture is a critical component of coastal economies, yet its floating raft structures are typically smaller than the spatial resolution of medium-resolution satellite imagery, posing a persistent engineering challenge for large-scale monitoring. This study investigates whether pan-sharpening can reliably enhance the detectability of oyster rafts in Landsat imagery under radiometrically homogeneous, water-dominated coastal conditions, using the Maowei Sea—China’s largest natural oyster breeding ground—as a representative study area. Four commonly used pan-sharpening methods—Colour Normalization (CN), Gram–Schmidt (GS), Nearest Neighbour Diffusion (NND), and Principal Component Analysis (PC)—were systematically evaluated over dense oyster farming areas using both spectral and spatial quality metrics. Quantitative results show that GS consistently achieves the lowest spectral distortion (SAM = 0.049; ERGAS = 0.088) and the highest spectral consistency (mean CC > 0.95 across visible and near-infrared bands), while maintaining spatial continuity of raft features. NND preserves high-frequency spatial information but exhibits less stable spectral behaviour across bands. In contrast, CN and PC introduce substantial spectral degradation and geometric inconsistencies, with SAM values reaching 0.834 and negative inter-band correlations in some cases. Field measurements indicate that most oyster rafts are 8–15 m wide, significantly smaller than a Landsat multispectral pixel (30 m). After GS-based pan-sharpening, these sub-pixel structures become discernible as coherent half- to full-pixel features, markedly improving their spatial continuity and visual interpretability. The results demonstrate that the effectiveness of pan-sharpening in coastal aquaculture monitoring is governed by algorithm–scene compatibility rather than spatial enhancement alone. By identifying GS as a stable engineering solution under strict radiometric and geometric constraints, this study extends the practical utility of long-term, freely available Landsat archives for fine-scale aquaculture monitoring and coastal resource management.