The impact of long-term exposure and persistence of pesticides in the environment on human biology is not completely understood. With the proliferation of pesticide application technologies, there have been documented associations between exposure to every major functional group of pesticide and adverse health effects in humans such as cancer and neurological disease. Of observed pesticide-induced cancer health-risks, hepatocellular carcinoma (HCC) has shown some of the most significant associations. Epidemiological study is complex, especially when examining pesticide health risks. It is difficult to understand the significance behind interactions between the large list of pesticide compounds, external environmental factors, and biological variables. Therefore, complexity is a driving factor of uncertainty in pesticide epidemiology research. Using cancer data from New York State Department of Health (NYSDOH) aggregated to census block groups and NYS pesticide data from Cornell University aggregated to zip-codes, this study developed a Geographic Information System (GIS) based statistical model to investigate the possibility of an association between pesticide applications and higher indices of HCC sites in NYS. Model development progressed from simple linear regressions (such as Generalized Linear Regression (GLR)) to analysis using Local Bivariate Relationships (LBR), Geographically Weighted Regression (GWR), and a final model utilizing random forest-based classification and regression. Modeling was performed over all of NYS, including localized Areas of Interest (AOIs) around Rochester, Syracuse, and Buffalo. Additional models were performed on clusters generated using Multivariate Cluster Analysis (MCA). Models based on LBR indicated clusters of statistically significant relationships, including importance of pesticide exposure in explaining variance in HCC indices between zip-codes in random forest models. These results are evidence of possible association, though it must be noted that further study is needed to establish significant correlation or causality. The methods developed in this study serve as a framework and showcase of geospatial statistical analysis in environmental epidemiology.

Library of Congress Subject Headings

Liver--Cancer--New York (State)--Monroe County; Pesticides--Health aspects--New York (State)--Monroe County

Publication Date


Document Type


Student Type


Degree Name

Environmental Science (MS)


Karl Korfmacher

Advisor/Committee Member

Todd Pagano

Advisor/Committee Member

Leslie Kate Wright


RIT – Main Campus

Plan Codes