In environmental engineering field studies, data analysis plays an important role when presenting data into useful information that can be used by engineers and policy makers. However, traditional and currently used approaches have significant limitations due to the nature of the field data, such as high temporal variability, high spatial variability, and high heterogeneity. Such uncertainty may be better handled with more realistic statistical models than traditional statistical models with normal approximation. Additionally, a more robust incorporation of heterogeneity and variability may help to modify environmental fate models to achieve more accurate predictions. Therefore, this dissertation applied some advanced data analysis techniques to four case studies.First, reparameterization was applied to modify the Gaussian plume model to predict dispersion of air pollutant emission from a ground-level active-discharge releasing source. Cross-validation was applied for model selection. The results showed that predictive accuracy of the modified GPM was greatly improved compared with the original model. Second, dispersion of particulate matter was accessed, and a dispersion correction factor was developed to enhance the performance of the regulatory air dispersion model (AERMOD) for low-level sources. Cross-validation was used for model comparison. The results showed that predictive accuracy of the corrected model was greatly improved. Third, carbon amendments were applied to a historically contaminated field to investigate the feasibility for mitigating bioaccumulation. The effect of carbon amendments on bioaccumulation were evaluated. The results showed some evidence of the mitigation effect of compost, and in the meanwhile, the need of a robust statistical method was highlighted due to great spatial variability. Lastly, the Bayesian hierarchical model (BHM) was applied to the field measurement dataset to characterize pollutant concentrations and bioaccumulation. Cross-validation and information criteria were used to evaluate model performance between the BHM and traditional model. The results showed that the BHM was preferred for smaller predictive errors and ability to handle data with larger observational error. These case studies demonstrate the capability of advanced statistical methods for dealing with different environmental research problems. Such statistical methods will be useful for model modification with more specific situations, for data analysis with limited sample size and/or great variability and observational error, for environmental and ecological risk assessment, for evaluation of environmental mitigation strategies, for simulation of real-time pollutant distribution and forecasting with integration of monitoring and modelling approaches, and for minimization of sample size to meet with the accuracy requirement and lower the cost. In conclusion, advanced statistical methods are useful tools for environmental research.