Different approaches have been used to determine critical loads of air pollutants for the protection and recovery of ecosystems, including the application of steady-state and dynamic models and the development of empirical relationships or empirical critical loads. Critical load values derived from these approaches often have high uncertainty due to confounding factors, data limitations, and characteristics of the models themselves. Improving the characterization, quantification, and communication of uncertainty associated with critical load estimates, and reducing those uncertainties, have been established as important goals in future efforts to further develop and refine national-scale critical loads data. Machine learning models have shown great promise across a range of predictive tasks in the geosciences, substantially reducing inaccuracy in predictions by leveraging empirical relationships in input data. In this study, we investigate the potential application of machine learning techniques to determine critical loads of nitrogen and sulfur for forest ecosystems. Using the tree growth and survival database developed by Horn et al. (2018), we develop a machine learning regression model using the gradient boosted trees algorithm. The regression incorporates total nitrogen and sulfur deposition, as well as environmental factors such as mean annual temperature and precipitation. Using model interpretation techniques, including SHapley Additive exPlanations (SHAP) multi-variate partial plots, we interrogate the relationships between tree health and environmental factors including deposition, and we compare our relationships to those of Horn et al. We further examine the response surface of tree species health with changes in deposition to assess objective criteria for establishing a value of critical load.  Our results indicate opportunities for using machine learning to develop critical load values. Future work will expand the data layers considered - including soil acidity, soil texture, available water storage, elevation and topography - to examine how environmental factors affect tree species response to changes in atmospheric deposition. In addition to examining the structural responses to variations in atmospheric deposition, we plan to evaluate functional responses of forest ecosystems (e.g., carbon and nitrogen sequestration, forest biomass production) to changes in atmospheric deposition.