An Abstract: Using Spatial Data and Causal Trees to Analyze the Effectiveness of Environmental Aid Allocation

By conservative estimates, hundreds of millions of dollars of international aid is allocated to environmental protection each year. Despite the growing global importance of environmental conservation and many local case studies, the impact of this aid is poorly understood on a global level. This lack of information is due to the vast diversity of geographic regions and problems with comparing and understanding these different environments on a standardized global scale.

For my Upperclassmen Monroe Project this Summer I plan to use the SciClone High-Performance Computer in conjunction with spatial data from a wide variety of sources (including satellites, surveys, and more) to analyze the effectiveness in environmental aid allocation by international aid projects.  I will be applying a new machine learning technique, Causal Trees, which enables global-scope analyses that would otherwise be computationally infeasible. Similar to a clinical trial, this approach uses a matching algorithm to compare areas that received aid to “twins” where no aid was given.

The results of this study will help further the understanding of the effectiveness of current environmental aid in addition to providing information that can help improve future environmental aid allocation. Further, it will provide a helpful roadmap for other analysts seeking to use large sets of spatial data to analyze aid effectiveness.

If this study is successful, this research will help researchers to better identify the contexts in which aid positively changes environmental conditions.  Further, by leveraging the GeoQuery data platform at William and Mary, this project will provide a case study example of how to use data for impact evaluation for researchers across hundreds of institutions.

This research also has personal significance as a Computer Science and Data Science double major. As someone who aspires to pursue a career using machine learning and computer science to help aid in healthcare, environmental, and humanitarian issues, this study provides an exciting entry point into an applied case with a real audience. Further, I will be using data from the GeoQuery tool I help maintain and improve as a lab director in the William and Mary GeoQuery lab, which will help improve my understanding on the importance and limitations of the tool in real-world applications.