Data Complexity and Spatial Scaling: Prediction Accuracy and Implications for Emerging Landscape Paradigms

点击翻译按钮获取中文摘要

This research will contribute new knowledge regarding the merger and aggregation of diverse geographic datasets and how their data quality loss can be minimized during these processes. As datasets are merged across varying spatial scales, statistical biases, commonly known as the Modifiable Areal Unit Problem (MAUP) occur and are impacted by the composition and configuration of the datasets. The investigator will study how these biases can be overcome in order to allow datasets to be accurately scaled to different resolutions. By developing methods that use measures of compositional and configurational heterogeneity, the researcher will provide new insights regarding how to assess the potential of a dataset to be accurately scaled to coarser and finer resolutions in order to foster integration. The investigator will raise awareness of accuracy issues surrounding appropriate and correct use of ecological, social, and geographical datasets for integrated analysis, and she will offer solutions to promote better interdisciplinary integration as well as methods for exposing young scientists to these issues through learning modules.This project will advance the spatial sciences by drawing on parallel theories in geography and landscape ecology. The research will seek to answer three core questions: (1) How do data complexity and spatial heterogeneity impact statistical biases associated with MAUP? (2) Can standards for data reduction/complexity improve prediction? (3) Do various landscape paradigms respond differently to changing heterogeneity? The investigator will use multiple methods, including innovations in surface paradigms in the field of landscape ecology, advances in remote sensing spectral unmixing, and scale-dependency of spatial pattern metrics to answer these questions. The research findings will result in identification of different forms of data reduction and complexity aggregation methods across multiple resolutions; development of a rapid assessment method for predicting downscaling accuracy; and establishment of a statistical basis for using the emerging continuous surface paradigm of landscape analyses. These outcomes have the potential to transform how scientists across environmental, social, and economic disciplines approach aggregation and scaling of various types of geospatial data.