Large-Scale Automated Sustainability Assessment of Infrastructure Projects Using Machine Learning Algorithms with Multisource Remote Sensing Data

Abstract

Considering the magnitude, lifespan, and environmental impacts of physical infrastructures, integration of sustainability with development policies has proved to be indispensable; accordingly, several rating systems were nationally developed to enhance implementing sustainability into physical infrastructures. Lack of automation and strategic outlook of the conventional approach to infrastructures’ sustainability assessment, exacerbated by the lengthy and costly processes involved, highlights the necessity of adopting comprehensive and innovative measures. This paper principally aims at extending the scope of sustainability rating systems such as Envision by proposing a framework for large-scale and automated assessment of infrastructures. Based on the proposed framework, a single model was developed incorporating remote sensing and GIS techniques alongside the support vector machine (SVM) algorithm into the Envision rating system. The proposed model adds a certain degree of automation in assessment process regarding the criterion N.W.1.2 of Envision rating system (i.e., provide wetland and surface water buffers) as a starting point toward entire automation of the Envision system. Given the quantitative scale of the criterion N.W.1.2, our model automatically extracts (1) wetlands, (2) waterbodies, and (3) roadways through Optical Satellite_Sentinel-2A, Synthetic Aperture Radar (SAR) Satellite_ALOS-1 imagery, and shapefile from Florida Department of Transportation (FDOT). The image-based model then examines whether certain applicable specifications of Envision scoring system are met. The level of achievement is determined, and the final score in the criteria N.W.1.2 is calculated afterward. The results indicate that more than half of the existing road segments in the study area failed to obtain the minimum required score, regulated by Envision. This emphasizes the criticality of considering sustainability indicators in future infrastructure planning. In addition, the validated results confirm the feasibility of automation of other indicators of the Envision system that will help authorities see the bigger picture and make more sustainable decisions for future practices and policies.