Doctoral Dissertation Research: Overcoming Scale Disparities Through Sub-Pixel Remote Sensing Classifications and Spatial Pattern Metrics

Suny At Buffalo, Amherst NY

Investigators

Abstract

This DDRI project investigates a new data aggregation technique that uses a combination of sub-pixel remote sensing data and landscape metrics to retain a greater amount of spatial information compared to traditional aggregation techniques. Geography has a long tradition of grouping spatial data into areal units (e.g., census tracts, pixels, etc.) that form the basis for observation and measurement. Data aggregation is necessary to map real-world phenomena and perform spatial analysis, but the process introduces statistical biases that adversely affect results and create obstacles for integrating datasets from different sources and across disciplines. These biases are familiar to geographers as the modifiable areal unit problem (MAUP) and are known to ecologists as the root of ecological fallacies. The data aggregation technique developed in this research will capture the increased land cover information present in sub-pixel remote sensing classifications and retain this information as smaller spatial units are combined into larger units, thereby reducing statistical biases. The study will then test the results of the data aggregation technique by predicting land cover patterns at very fine spatial resolutions. The accuracy of these predictions will depend on how much statistical bias was removed through the data aggregation technique. Lastly, the research will demonstrate a practical ecological application for these findings by using them to perform super-resolution mapping. Super-resolution mapping is a subfield of remote sensing concerned with mapping data at a finer resolution than the original sensor collection. The broader impacts of this research include advancing fundamental spatial science by redefining data aggregation while also providing a theoretical understanding of the causation of MAUP and ecological fallacies when using traditional remote sensing classifications. The research will demonstrate how advanced remote sensing techniques can help mitigate those biases. The project will promote interdisciplinary collaborations by ultimately allowing studies to integrate global spatial data sources from remote sensing with ground-based ecological spatial data more accurately. Eventually these advances can be used to study individual behaviors and impacts to the land using remote sensing and ground-based survey data. The long-term impacts reach beyond the theoretical aspects of geographical and spatial sciences and will impact applied scientific research including agent-based modeling for human impact analysis, spatial epidemiology, and population dynamics. Community outreach will include developing a lesson plan for a special education Earth Science class to teach students applications of geospatial technologies and highlight opportunities for future career paths. As a Doctoral Dissertation Research Improvement award, this project will provide support to enable a promising student to establish an independent research career.