The project will apply to the modelling of natural phenomena techniques originating in the fields of pattern recognition/classification and image analysis, describing the prior knowledge as constraints on the set of possible macroscopic internal organisations of the field observed. This approach, which has already been used in other bio-engineering fields, is better suited to the use of the prior information for the on-line guidance of the platform, enabling the definition of extrapolation techniques to predict the overall field distribution from a set of localised measurements. However, in most cases these methodologies do not explicitly provide a measure of the spatial distribution of the uncertainty affecting the predicted distributions. The project will integrate into these shape description techniques an explicit measure of confidence in their outputs, which is the fundamental requirement for the definition of the model-based surveys.