Explainable Self-Organizing Artificial Intelligence Captures Landscape Changes Correlated with Human Impact Data

Abstract: Novel methods of analysis are needed to help advance our understanding of the intricate interplay between landscape changes, population dynamics, and sustainable development. Self organized machine learning has been highly successful in the analysis of visual data the human expert eye may not be able to see. Thus, subtle but significant changes in fine visual detail in images relating to trending alterations in natural or urban landscapes may remain undetected. In the course of time, such changes may be the cause or the consequence of measurable human impact. Capturing such change in imaging data as early as possible can make critical information readily available to citizens, professionals and policymakers. This promotes change awareness, and facilitates early decision making for action. Here, we use unsupervised AI that exploits principles of self-organized biological visual learning for the analysis of imaging time series. The quantization error in the output of a Self Organizing Map prototype is exploited as a computational metric of variability and change. Given the proven sensitivity of this neural network metric to the intensity and polarity of image pixel colour, it is shown to capture critical changes in urban landscapes. This is achieved here on imaging data for two regions of geographic interest in Las Vegas County, Nevada, USA. The SOM analysis is combined with the statistical analysis of demographic data revealing human impacts. These latter are significantly correlated with the structural change trends in the numerical data for the specific regions of interest. By correlating data relative to the impact of human activities with numerical data indicating structural evolution, human footprint related environmental changes can be predictably scaled.