Spatial predictions including spatially continuous data play a significant role in planning, risk assessment and decision making in the environmental management and conservation. They are, however, usually not readily available and often difficult and expensive to acquire, especially for vast, mountainous or deep marine regions. Therefore, spatial predictive modelling methods become essential tools for generating such data. Because of the rapid development in and application of remote sensing in both terrestrial and marine environments, increasingly more environmental variables become available for spatial predictive modelling. Hence in parallel to the advancement in data processing and computing capabilities, predictive models have been increasingly employed to generate spatial predictions in the environmental sciences. However, accuracy of predictive models is critical as it determines the quality of their predictions that form the scientific evidence for decision-making and policy. This presents an opportunity for scientists to develop and improve their predictive models, but also presents a challenge to select an optimal predictive model from a large number of available predictors. Therefore, scientists are encouraged to submit their findings in using various predictive modelling methods (e.g. geostatistics, machine learning methods, modern statistics, or novel modelling methods), especially on model selection, predictive accuracy assessment and uncertainty quantification, with an application case in the environmental sciences.