TRACKING URBAN HEAT

The project aims to explore the relations between human (demographic) vulnerability and the urban heat island (UHI) effect in Dallas, Texas. Anthropogenic heat flux and increased thermal admittance and storage of land surfaces as a result of urbanization have intensified the UHI effect in the Dallas - Fort Worth metropolitan area between 2001 and 2011 (Winguth & Kelp, 2013). In Jul 2011, a heat wave with a dual-peak maximum heat island was recorded in the city (Winguth & Kelp, 2013). The study focuses on the City of Dallas, instead of the Dallas - Fort Worth metropolitan area, because the city itself is more urbanized (excluding the Lake Ray Hubbard area) than its surroundings, based on the  2020 Global Land Use / Land Cover dataset  (Esri, FAO, NOAA, 2020). Its higher development density and a lower percent of green cover in the city contribute to a more pronounced UHI effect. By comparing recent patterns of human (demographic) vulnerability and UHI distribution in Dallas, the study aims to (1) identify potential correlations between heat and community demographics and (2) locate neighborhoods with high urban heat vulnerability in Dallas as a function of physical exposure and sensitivity.

*The hottest months were defined as months with a mean temperature equal to or above 86 °F (30 °C). ** The period between 2015 and 2019 was chosen to match the same study period of the 5-year ACS from which the demographic variables were extracted.

Since human vulnerability to urban heat is closely associated with high air temperatures in urbanized areas, we mapped land surface temperatures during the hottest months between 2015 and 2019. We decided to aggregate daily ST images into monthly intervals due to the high proportion of Landsat 8 Provisional Surface Temperature (ST) images with a high percentage of cloud cover which yielded a significant amount of incomplete land pixels. To address this issue, we derived mosaic images by calculating the mean values of all ST images with cloud cover of 10% or less during the hottest months. To determine the hottest months between 2015 and 2019, we investigated the past monthly mean temperature data of the Dallas-Fort Worth Area, which is the most detailed weather forecast area which includes Dallas provided by the Applied Climate Information System (ACIS) of the National Oceanic and Atmospheric Administration (NOAA).

The six demographic variables used in this study are indicators used nationwide to assess urban heat vulnerability (Reid et al., 2009, Kathryn et al., 2020, Chuang & Gober, 2015). Via U.S. Census API, we sourced these variables from the 5-year American Community Survey (ACS) (2015-2019). To study differences across census tracts, we converted these variables to percent values to remove biases incurred by the variability of total / focal populations across the tracts. We also tested these variables in terms of their correlation. Most of the variables are not highly correlated. Despite the fact that the percent population with less than high school education and the percent of households under poverty level are correlated, they reflect different demographic characteristics that are not always in a direct causal relationship.

Using CenPy, we downloaded data of each variable from the Census API's 5-year American Community Survey 2015-2019. To allow comparison with the land surface temperature distribution across the same census tracts (tracts that intersects with or are contained within the City of Dallas boundaries), we used an inner join in Pandas based on the GEOID of the heat and demographic tables. The resulting descriptive table of the six demographic variables is included below:

• Aggregation by census tracts generalized the variations in terms of demographic characteristics and local environmental conditions within a single tract
• The 10-m resolution of the 2020 Sentinel-2 Land Use / Land Cover data is too coarse to capture small elements that influence local land surface temperature distribution
• Further experiementation of the multivariate linear regression models should be conducted to identify any outliers that also influence urban heat