An information theory-based approach to assessing spatial patterns in complex systems.

Heriberto Cabezas,Wen-Ching Chuang,Tarsha Eason,Shana Sundstrom

doi:10.3390/e21020182

Abstract

Given the intensity and frequency of environmental change, the linked and cross-scale nature of social-ecological systems, and the proliferation of big data, methods that can help synthesize complex system behavior over a geographical area are of great value. Fisher information evaluates order in data and has been established as a robust and effective tool for capturing changes in system dynamics, including the detection of regimes and regime shifts. The methods developed to compute Fisher information can accommodate multivariate data of various types and requires no a priori decisions about system drivers, making it a unique and powerful tool. However, the approach has primarily been used to evaluate temporal patterns. In its sole application to spatial data, Fisher information successfully detected regimes in terrestrial and aquatic systems over transects. Although the selection of adjacently positioned sampling stations provided a natural means of ordering the data, such an approach limits the types of questions that can be answered in a spatial context. Here, we expand the approach to develop a method for more fully capturing spatial dynamics. The results reflect changes in the index that correspond with geographical patterns and demonstrate the utility of the method in uncovering hidden spatial trends in complex systems.

Highlights

Today’s digital landscape presents a world full of information with more access to geotagged datasets
Transitions may be defined as declines in FI between two stable regimes [58], we identify them as periods characterized by a relatively high standard deviation and coefficient of variation in FI (↑σFI, ↑coefficient of variation of FI (cvFI); [67])
With the rise in availability of large-scale geospatial datasets coupled with the complexity of Concluding challenges in aand more connectedRemarks global society, there is a need for methods that afford the ability to examine patterns and trends in multiple variables without requiring the use of modelling, restrictive

Summary

Introduction

Today’s digital landscape presents a world full of information with more access to geotagged datasets. In this era, analysts are less likely to struggle with a lack of available data. Instead, they are taxed with overwhelming amounts of information and charged to make good use of the data. Geospatial assessment is a well-developed and growing field. Spatial analyses typically involve the visual assessment of mapped parameters, use of zonal/spatial statistics (e.g., Moran’s I), or Entropy 2019, 21, 182; doi:10.3390/e21020182 www.mdpi.com/journal/entropy

Objectives

Methods

Results

Conclusion