This research is aimed at identifying and evaluating experimental factors which influence variations in results from sludge sampling and analysis, and investigating the feasibility of an in-situ solution for sludge characterisation. Due to the challenges present in removing sludge from wet nuclear storage facilities for laboratory-based analysis, it is difficult to generate characterization maps with high confidence as typically only very limited measurements can be made. This research presents a method of estimating the percentage confidence in sludge characterization results, specifically focusing on the Particle Size Distribution (PSD) as a representation of other sludge properties relevant to UKâs nuclear decommissioning programme. Also introduced is a novel algorithm, referred to as Recursive Relative Accuracy (RRA), which is shown to provide an indication of the benefits of taking more samples. Access to real nuclear site data is restricted; hence the chapter adopts the use of a real-life non-radioactive corroded magnesium sludge simulant tank. Sludge samples were collected from computer modelled and real-life simulant sludge beds under varying experimental conditions such as: the number of sampled locations, sampling strategy, the penetration depth and the selective bias of the sampling device used. These samples were in some cases, analysed for their PSD under varying experimental conditions such as: type of instrument model used, the concentration of samples and the dispersion medium used. Using PSD data measured at sampled positions and inferring data at non-sampled positions, three-dimensional sludge characterisation maps were obtained and analysed. It was observed that the factor âdepth of penetration of the sampling deviceâ contributes about 48 % to result variability, while the instrument model used and the sample concentration chosen each contribute about 40 % and 7 %, respectively, to variations in the PSD laboratory analysis results. The feasibility study conducted on ultrasonic spectroscopy failed to confirm the existence of an analytical model for interpreting PSD data in-situ. The average error margin in the inference of PSD mean values was over 600 Âµm.