Energy and the environment are inextricably linked. Consumption of fossil fuels has made air pollution the largest environmental risk to human health, while renewable energy generation extracts natural resources to power the economy and mitigate global climate change. My dissertation, consisting of four studies, focuses on developing and applying data-driven methods to address the challenges at the nexus of energy and the environment, which capture the complex relationships in data with relatively inexpensive computational cost and ease of implementation compared with the traditional physically based approaches.The first three studies center around resolving the source-to-receptors relationship for air pollution problems in the current energy system, followed by the last one characterizing system resiliency for the future energy system dominated by renewable resources. In the first study, I developed machine learning-based models to accurately predict unit-level NOx, SO2 and CO2 emission rates from fossil-fueled electric generation units, providing not only a critical link between power systems and air quality modeling, but also the capability to electronically audit power plant data, identify potential regulatory compliance issues and reduce emissions. In the second study, I applied SHAP (SHapley Additive exPlanation) as a unifying framework to interpret different types of data-driven air quality models by quantifying contributions of predictor variables to pollutant concentrations. With enhanced interpretatbility, this unifying framework has the potential to enable quantification of emission source contributions using data-driven methods. In the third study, I created a data-driven screening tool to process data from distributed air quality monitoring networks and generate insights indicative of hyperlocal emission sources. By identifying the local anomalous concentrations and the associated local drivers, I demonstrated the capability of this screening tool in identifying both areas and emission sources of interests within the network, which can lead to actionable strategies and ultimately cleaner air. In the fourth study, I proposed a new concept named resources drought to characterize the least favorable conditions for wind and solar generation. Analyzing 39 years worth of climatology data, I elucidated the resources drought conditions for the continental U.S. and created a toolkit to inform energy system resiliency planning under high-renewable penetration scenarios.