<p>The phyllosphere bacteria play a crucial role in global greenhouse gas emissions and sequestration, but the spatial interactions between phyllosphere bacterial diversity, host leaf traits and environmental variation remain poorly understood. This gap is mainly due to methodological limitations in linking the spatial pattern of bacterial diversity to leaf traits. Here, we present machine learning models based on visible and near-infrared (VIS-NIR) leaf hyperspectral proximal sensing that are used to independently predict the phyllosphere bacterial alpha-diversity indices and leaf traits under different abiotic conditions for both sides of the leaf. We demonstrate that the models can effectively represent leaf traits and bacterial alpha-diversity indices for different abiotic environmental conditions. The cross-correlation of the spatial patterns as a result of the spatial application of the independent models reveals fine-scale associations between leaf traits and bacterial colonization patterns. Our findings highlight the great potential of hyperspectral proximal sensing for understanding the relationship between leaf bacterial richness and leaf resources within the leaf microecosystem. Ultimately, this will enhance our capacity to quantify the contribution of the leaf bacteria to the greenhouse gas balance of the forest canopy in a changing climate.</p>