Acquiring, maintaining, disseminating, and utilizing quality data is key to adequate understanding and management of ecosystems. Modern remote sensing technology provides us an increasingly cost effective, unique opportunity for acquiring highly detailed information across every square meter of a landscape. The plethora of data available to scientists allows for use of high quality data on diverse topics, including the effects of fire through fire severity. However, the abundance of data may not be appropriate for certain modeling challenges and methods. We use aerial Light Detection and Ranging (LiDAR) acquired before local fires in conjunction with Remote Automated Weather Station data to derive fire severity. Observed fire severity was quantified as relative differenced Normalized Burn Ratio (RdNBR), a Landsat-derived parameter indicating basal area mortality. We utilized Pearson’s correlation coefficient, random forest importance metrics, and pairs plots to identify and isolate parameters of particular importance in predicting RdNBR. The severity metric was then modeled for prediction through principal component analysis (PCA) and random forest. Model performance was assessed using root mean square error (RMSE), root mean square prediction error (RMSPE), Bayesian information criterion (BIC), Akaike information criterion (AIC), bias, and adjusted R2. LiDAR variables representing the six to nine meter height class were generally found to be key, along with weather variables representing duff moisture content. Weather variables were overall found to be most important in modeling RdNBR. PCA-based models were found to perform best for prediction.