ARTICLE | doi:10.20944/preprints202105.0199.v1
Subject: Keywords: urban structure, remote sensing, temporal change, NYC
Online: 10 May 2021 (14:26:15 CEST)
Surface temperature influences human health directly and alters the biodiversity and productivity of the environment. While previous research has identified that the composition of urban landscapes influences the physical properties of the environment such as surface temperature, a generalizable and flexible framework is needed that can be used to compare cities across time and space. This study employs the Structure of Urban Landscapes (STURLA) classification combined with remote sensing of New York City’s (NYC) surface temperature. These are then linked using machine learning and statistical modeling to identify how greenspace and the built environment influence urban surface temperature. It was observed that areas with urban units composed of largely the built environment hosted the hottest temperatures while those with vegetation and water were coolest. Likewise, this is reinforced by borough-level spatial differences in both urban structure and heat. Comparison of these relationships over the period between2008 and 2017 identified changes in surface temperature that are likely due to the changes in prevalence in water, lowrise buildings, and pavement across the city. This research reinforces how human alteration of the environment changes ecosystem function and offers units of analysis that can be used for research and urban planning.
ARTICLE | doi:10.20944/preprints202104.0588.v2
Subject: Earth Sciences, Atmospheric Science Keywords: Air Pollution; STURLA; Urban Structure; Mobile Monitoring; Spatial Prediction
Online: 5 May 2021 (12:41:17 CEST)
Understanding the relationships between land cover/urban structure patterns and air pollutants is key to sustainable urban planning and development. In this study, we employ a mobile monitoring method to collect PM2.5 and BC data in the city of Philadelphia, PA during the summer of 2019 and apply the Structure of Urban Landscapes (STURLA) methodology to examine relationships between urban structure and atmospheric pollution. We find that, while PM2.5 and BC vary by STURLA class, many of the differences in pollutant concentrations between classes are not significant. However, we also find that the proportions in which STURLA components are present throughout the urban landscape can be used to predict urban air pollution. Among frequently sampled STURLA classes, gpl hosted the highest PM2.5 concentrations on average (16.60 ± 4.29 µg/m3), while tgbwp hosted the highest BC concentrations (2.31 ± 1.94 µg/m3). Furthermore, STURLA combined with machine learning modeling was able to correlate PM2.5 (R2= 0.68, RMSE 2.82 µg/m3) and BC (R2 = 0.64, RMSE 0.75 µg/m3) concentrations with the urban landscape and spatially interpolate concentrations where sampling did not take place. These results demonstrate the efficacy of the STURLA methodology in modeling relationships between air pollution and land cover/urban structure patterns.