Can Digital Environmental Scans Adequately Measure the Environment?

Digital scans of the environment minimize financial burdens and time constraints associated with in-person environmental scans. Studies found digital scans to be an adequate replacement for physical scans in certain cases; however, an updated and comprehensive comparison between digital and in-person scans is needed (Ben-Joseph et al., 2013; Clarke et al., 2010; Edwards et al., 2013). Recent advancements in street-level imagery improve the capacity for remote analysis of environments. Yet, the efficacy of digital scans relative to in-person scans is not entirely known. It is necessary to determine whether digital platforms effectively capture the environment as digital scan studies employing Artificial Intelligence increase. Therefore, this study seeks to update discussion over the efficacy of digital scans through a comprehensive analysis of multiple street-level imagery platforms. Environmental correlates with pedestrian injury were digitally measured across ten locations in Halifax that were previously measured in-person (Schuurman et al., 2020). We also digitally assessed the environment as it relates to pedestrian injury based on a 2009 study in Vancouver across 32 sites (Schuurman et al., 2009). Last, a runnability index was created then measured across 30 sites both in-person and digitally in Vancouver. Digital assessments analysed imagery collected from Google Street View, Microsoft Streetside View, and Apple Map’s Look Around. Spearman rank-order correlation determined the level of agreement between in-person scans, serving as ground truth, and each of the digital scan platforms. Preliminary results indicate that Google Street View effectively measures some aspects of the environment; however, is less capable of harvesting fine-grained details. Google Street View shared high correlation with in-person scans for blue space, vegetation, green space, and general land use (ρ > 0.8). However, Google Street View poorly captured path quality, path width, and small path obstructions (ρ < 0.6). Between platforms, temporal accuracy and spatial coverage varied greatly. In general, Google Street View had the most coverage over the selected sites; however, Apple Map’s Look Around produced higher quality imagery. Findings indicate that much variation exists between imagery platforms. While in-person scans are the most effective tool for measuring the environment, digital scans are an adequate replacement if used with caution. Digital scans create opportunity for international comparative studies and can help access locations from the past. Significant investment into digital street-level imagery platforms is required to improve their capability and scope. One main study limitation was researcher bias. Since some of the measures assessed were qualitative (i.e., perceived attractiveness), scoring differed slightly between the two auditors. Inter-rater bias will be assessed in the future to determine how significantly ratings vary both digitally and in-person.