Measuring the Built Walking Environment through Image Calibration Using a Hybrid Audit Approach

Abstract

The built environment (BE) influences active travel, and therefore it is studied by researchers in various fields, including transportation and health sciences. Thus a reliable and quantitative estimate of the built environment is desirable. Existing BE audit tools capture such data qualitatively and may not apply to urban contexts in which a large majority of the street does not contain any sidewalk or footpath. This study developed a BE audit measurement process with the hybrid approach of conducting surveys physically as well as virtually. Ten physical auditors captured the built walking environment in 65 links of Varanasi, India using video cameras. Subsequently, these videos were processed by three independent desktop observers using open-source image calibration software. Results from the extracted BE features indicate that the audit process was able to capture a wide variety of pedestrian features and the variation in the features across various land-use contexts. Extensive interrater and intrarater reliability testing of the extracted data showed an average value of both interclass correlation (ICC) and Fleiss’ kappa score over 0.8, indicating a substantial degree of agreement among the raters of the audit process. This confirmed that the audit process was highly reliable. It also was affirmed that the audit process can be administered by nontechnical surveyors, provided that proper training is given beforehand.