The application of hydroacoustic measurements for preparation of spatial distribution maps of benthic communities is reported. For the present study common mussels (Mytilus edulis), neptune grass (Posidonia oceanica) and Cymodocea nodosa, serving as canonical species of many European marine ecosystems, were selected. These species are supposed to be good indicators of marine ecosystem health. The hydroacoustic measurements comprise preprocessed echo sounder recordings and side-scan sonar data forming a large and unique collection of datasets based on 4 field campaigns in Øresund and the Meditteranean. combination of geostatistical methods for spatial interpolation of the echo sounder observations and a set of classification rules, based on discriminant analysis of the feature space of=20 the observations, is found to yield reliable distribution maps when compared to groundtruth data. The data-driven methodology developed is shown to be adaptive to instationarities in the echo sounder observations and is recommended as a substantial improvement of existing methods of sea floor mapping based on echo sounder data. Elaborations of the developed methodology are studied, comprising the use of geostatistical simulation, Markov random fields and Boolean models. Geostatistical simulation provides a means of assessing the variability of random field functionals such as the estimated distribution area of a benthic species. The Markov random field allows the spatial distribution of the benthic communities to be modelled as a less smooth or regular phenomena than assumed when using geostatistical models. The use of Markov random fields in a Markov chain Monte Carlo simulation framework enables an alternative means of assessing variability of image functionals that is based on a sound theoretical basis. The estimates of variability obtained for estimated distribution areas with the two approaches compare satisfactorily. The Boolean models are suggested as a point of departure for embedding models of spatial patterns on the minor scales of observations to be used in up-scaling approaches to enhance the quality of the distribution maps and to be combined with biogeochemical models describing spatiotemporal population dynamics. Finally, the use of side-scan sonar data is illustrated in a data fusion exercise combining side-scan sonar data with the results based on echo sounder measurements. The feasible use of side-scan sonar for mapping of benthic communities remains an open task to be studied in the future. The data processing methodology developed is a contribution to the emerging field of hydroacoustic marine biology. The method of penalised maximum pseudo-likelihood for estimation of the Ising model under a huge amount of missing pixel data is a contribution to statistical image analysis. Furthermore, the estimation method developed for non-stationary Boolean models that combines scale-space kernel smoothing with the so-called method-of-moments applied to stationary Boolean models is a contribution to stochastic geometry.