Environmental monitoring may be defined as a description of processes and activities performed to characterize and monitor the quality of the environment. Monitoring schemes may differ greatly in their spatial and temporal extent, but as an outcome of any environmental monitoring process, data are gathered exhibiting both a spatial and a temporal dimension. With this work, we aim to analyze the predictive accuracy when characterizing the spatio-temporal patterns of heavy metal deposition in mainland Portugal. The data set in use consists of measurements of heavy metal deposition in mosses, resulting from three nationwide surveys performed in 1992, 1996 and 2002. Firstly, we begin with an exploratory descriptive analysis and an exploratory spatial analysis of the data, using well known techniques of spatial interpolation. After, we propose to make a spatio-temporal prediction of heavy metal concentration for the most recent survey, allowing to incorporate georeferenced explanatory covariates of the process under observation, calling on an existing spatio-temporal prediction model. This model focuses on the spatial dimension by defining random fields for the mean, the scale and the residuals components, and incorporates the time dimension by means of strictly temporal random fields, which work as corrections for the temporal evolution of the process. Motivated by the fact that the data set in use in dense in the spatial dimension but sparse in the temporal one, a novel model-based approach is proposed for Gaussian data, corresponding to a saturated correlation model in the time dimension. The proposed model is derived in order to accommodate not exclusively geo-referenced covariates, but also covariates associated to the temporal behavior of the process. Regarding the results obtained in terms of predictive accuracy, a comparison of predictions from a purely spatial model with the ones from a spatiotemporal model showed that the latter improve the accuracy of predicted value. Moreover, if the comparison is restricted to the two spatio-temporal models, the new model proposal provides better results.