The McMurdo Dry Valleys in the Ross Sea Region of Antarctica represent a globally unique desert ecosystem where water availability is likely to change under global warming scenarios, thus influencing the distribution and abundance of biota. Using the Random Forest machine learning model, we focused on the spatial distribution of macroscopic terrestrial biota (moss, cyanobacteria, lichen, springtails and mites) in the Dry Valleys. A wetness index, explicitly driven by satellite-derived glacier surface temperatures and meltwater routing, was used along with biological survey data collected over six field seasons (2009–2014) as part of the New Zealand Terrestrial Biocomplexity Survey (n = 886 sites). Our analyses use the full extent of survey data available and include the larger Taylor, Wright and Victoria valleys, as well as data from the previously studied Miers, Marshall and Garwood valleys. The overall model accuracies were mixed (kappa statistic: 0.34% and 17.3% variance explained). However, the resulting predictive maps derived from the model and the influence of the different explanatory variables align with field observations and theoretical expectations. The models show that distance from coast was an important driver for the biota, as well as elevation and temperature. The predictive maps provide an initial model of the distribution of biota in the Dry Valleys and can guide future sampling as well as inform conservation and management strategies. Our research highlights the importance of biological survey data for use in spatial predictive modelling as well as the need to obtain representative samples from a wide range of different habitats (e.g. wet vs dry).