Fast Artificial Intelligence Detecting Climate Change Effects in Imaging Data - Archive ouverte HAL Accéder directement au contenu
Chapitre D'ouvrage Année : 2024

Fast Artificial Intelligence Detecting Climate Change Effects in Imaging Data

Résumé

Satellite images display visual data relating to natural or urban landscapes. In the course of time, they reveal structural alterations therein that may be a consequence of climate change. Capturing such visual data rapidly to make them available to citizens, professionals, and policymakers promotes awareness and assists decision making for action. We applied unsupervised Artificial Intelligence (AI) to the analysis of a time series of satellite images using Self-Organizing Maps (SOM). The Quantization Error (QE) in the map output is exploited as an indicator of change. Given the proven sensitivity of this neural network metric to intensity and polarity of image pixel contrast, and its proven selectivity to pixel colour, it is shown to capture critical changes in the water levels of Lake Mead across the years 1984–2008. The SOM-QE analysis is combined with statistical trend analysis to further highlight the magnitude and the direction of pixel contrast changes in specific regions of interest. We show that the structural change in the Lake Mead region is significantly correlated with demographic data for the same reference time period, translating the impact of human activities on the local environment. The results of this study show the usefulness of SOM-QE analysis as a parsimonious and reliable AI approach to the analysis of minimally pre-processed satellite images for rapid detection of environmental degradation as a consequence of human activities and climate change.
Fichier non déposé

Dates et versions

hal-04558203 , version 1 (24-04-2024)

Identifiants

Citer

Birgitta Dresp, John Wandeto. Fast Artificial Intelligence Detecting Climate Change Effects in Imaging Data. Lecture Notes in Networks and Systems, 956, Springer Nature Switzerland, pp.414-424, 2024, ⟨10.1007/978-3-031-56950-0_35⟩. ⟨hal-04558203⟩
0 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More