Characterization of intra-urban spatial variation in observed summer ambient temperature from the New York City Community Air Survey

Determining the relationship between within-city spatial variation of ambient temperatures and land-use characteristics would be useful for developing mitigation strategies for urban heat island effects in the warming climate, but relatively few published studies collected high-density ambient temperature data in large cities to date. We applied land-use regression (LUR) modeling to the ambient temperature data collected during 2009-2016 as part of the New York City Community Air Survey (NYCCAS), a high density air pollution monitoring network with up to 150 locations within New York City (NYC). The 3-5 AM average summer temperature, seasonally adjusted, was selected to avoid the influence of variation in shading across monitoring units. Environmental indicators, including normalized difference vegetation index (NDVI), high resolution land-cover, imperviousness and building density, were characterized at multiple buffer distances around each monitoring site, screened for explanatory value, and combined in a stepwise multivariable regression model using the first 7 years of data. We validated the model against 2016 data. Summer average temperature over the 8 years ranged from 17.1 to 25.8 degrees C and the spatial variation was well-explained by a nonlinear term of km(2) tree, shrub, and grass cover at 200 m (negative association), a measure of live green vegetation (NDVI) within 1000 m (negative), interior built space within 600 m (positive), the summer average daily minimum temperature at the National Weather Service station at LaGuardia airport, and a nonlinear term based on the XY coordinates to model unexplained spatial autocorrelation (R-2 = 0.82). This model predicted temperature very well at NYCCAS sites for summer 2016 (R-2 = 0.87). We observed that there was a range of baseline vegetative cover where increasing cover was associated with the greatest decreases in temperature, suggesting specific urban micro-environments where increased tree canopy may be the most effective in reducing ambient temperature. Within-city variation of nighttime temperatures is well-explained by features of the built and natural environment. Interventions that target this combination of factors, such as converting paved vacant lots and derelict buildings into vegetated space or replacing city infrastructure that uses impermeable surfaces with vegetated ones wherever possible, may help to decrease ambient temperature in NYC. Our findings suggest that evaluations of neighborhood greening may not show temperature benefits until a certain level of live vegetative cover is established.