Quality Control of CTD Salinity Profiles Based on Information Diffusion Model

Abstract Traditional laboratory calibration methods for water samples obtained via shipborne CTD (conductivity, temperature and depth) observations have limitations in addressing data interpolation for underway observations and involve redundant tasks. Methods such as linear and cubic spline interpolation and optimal interpolation require large sample sizes and high-quality data, which makes precise interpolation difficult in scenarios with sparse samples or non-normally distributed data. This paper introduces information diffusion theory to overcome the limitations of traditional normal information diffusion models, such as the inadequate adaptability of empirical window widths (which are particularly unsuitable for non-normally distributed ocean data) and the tendency of optimal window widths to converge to local optima during multidimensional data interpolation. Leveraging the advantages of genetic algorithms in terms of global optimization and rapid error convergence, a genetically optimized information diffusion interpolation model is constructed. This model uses fuzzy mapping principles to treat reliable, sparse sample data as ‘information injection points’. These points are then expanded into fuzzy sets via diffusion function mapping. Probabilistic diffusion of sample information is achieved through the transformation of information and fuzzy relation matrices, along with fuzzy synthesis rules. Experiments involving the interpolation of salinity profiles, alongside comparative analyses against the standard information diffusion model and traditional interpolation methods, validate the feasibility and effectiveness of the proposed optimisation algorithm. This approach provides a novel technical solution to quality control issues in oceanographic surveys, such as missing CTD data and handling outliers. Not only is its core methodology applicable to salinity profile data processing, it also provides a reference for interpolating sparse data of other oceanographic parameters, such as temperature. This offers guidance for the objective processing and analysis of cruise-based oceanographic survey data.