Community Resilience to Climate Change: Theory, Research and Practice

150 study) can provide a detailed representation of heat exposure at a smaller scale [30,31,32,33]. From an environmental justice point of view, the existing research emphasizes point source pollution [34,35,36], though it has become clear that climate change affects communities differentially and creates novel impacts never before witnessed in traditional environmental research. As such, we argue that climate change is catalyst for injustice. While some of the effects can be easily observed at the national level, particularly in the world’s poorest countries [37,38], there is relatively little understanding of the impact at a more granular scale. Needed are approaches—methodological, conceptual, and pragmatic—that help us to identify those communities disproportionately affected by UHI, and strategies that can help to reduce vulnerabilities. This study provides new evidence of disproportionate exposure to climate change at a local level, as well as access to refuge, an understudied facet of adaptive capacity. Previous studies indicate a relationship between socio-demographic factors and heat-related morbidity/mortality [39,40]. Income is quite predictive of vulnerability (inverse relationship) [41,42,43], though it is possible that other indicators also play a role. If so, urban heat exposure may be framed as an environmental justice or ‘climate justice’ [44] issue, disproportionately affecting marginalized socio-demographic groups with limited adaptive capacity. This study aims to identify such populations in Portland, Oregon by assessing (1) disproportionate heat exposure among socio-demographic groups; and (2) disproportionate access to refuge (either public refuge facilities or residential central air conditioning), resulting in heightened or lowered adaptive capacity. An in-depth statistical and spatial analysis will reveal significant, inequitable relationships, validating the application of an environmental justice lens in addressing urban heat resilience. 2. MATERIALS AND METHODS 2.1. Study Area Our assessment occurs in a Pacific Northwest city of the United States. The City of Portland, Oregon is located at approximately 45.5° North, 122.6° West, at the confluence of the Willamette and Columbia Rivers. The city covers approximately 345 square kilometers, with of a population of nearly 640,000 as of 2016 [45]. The City of Maywood Park, Oregon is located within northeast Portland and, though an enclave of the City of Portland, has been excluded from the study. Due to the fact that summer temperatures have an average monthly highs of 22.7 °C, 26.4 °C, and 26.7 °C for June, July, and August, respectively [46], Portland offers several advantages to conducting an assessment of disproportionate effects of urban heat. First, historical high temperatures have rarely exceeded 35 °C, which reduces public concern for heat related illnesses. Second, as of 2013, fewer than 35% of households had air conditioning [47], and communities may not have immediately available private residences to consider refuge from heat waves. Finally, Portland Climate Action Plan (2015) explicitly addresses the importance of reducing disproportionate exposure to urban heat waves, yet few actions have materialized. 2.2. Data Three main data types were used in this study, all at the U.S. census block group level: socio-demographic indicators, distribution of ambient temperatures in the study location, and refuge availability. Socio-demographic data used include: income (percent of the population below 50% of the poverty line); race (percent of the population who do not self-identify as white); education (percent of the population over 25 years old without a high school degree or equivalent); age (percent of the population over 65 years old that lives alone); and English speaking ability (percent of the population that claims to have poor or no English skills). Obtained from the U.S. Census Bureau’s American Community Survey 5-year estimates, 2009–2013 [47], these data reflect categories into which individuals have self-identified. Another, non-census piece of data used to highlight socio-demographic status is presence of affordable housing, obtained from Oregon Metro’s Regional Land Information System [47]. In order to differentiate “low” and “high” categories used in the analysis, a model-based clustering algorithm was used to split each variable [48]. Following an established protocol [13], we collected approximately 60,000 temperature readings during one day of an extreme heat event on 25 August 2014, in Portland, Oregon, when the average temperature during the hottest hour of the day was in the 90th percentile of 30-year historic daily temperatures for the study region. We sampled temperatures for one hour at 3 times during the day (6 a.m., 3 p.m., and 7 p.m.) using vehicle traverses (cars with a mounted temperature sensor and global positioning system (GPS)) in six predetermined sections of the city. The temperature sensor consisted of a type T-fine (30 gauge) thermocouple in a plastic shade tube (12 cm in length and 2.5 cm in diameter) mounted on the passenger-side window approximately 25 cm above the roof of each of 5 vehicles deployed. Each temperature sensor was connected to a data-logging device with an estimated system accuracy of ±0.5 °C and a 90% response time of less than 60 s in 1 m/s airflow. A GPS unit on each vehicle paired temperature measurement and location. Based on the results from the temperature collection and subsequent modeling, we created three separate heat surfaces, which are continuous descriptions of temperature variation across the study region, corresponding to the three time periods. The resulting maps consisted of a 32-bit floating point 1-meter raster format and contain 449,359,188 pixels for each of the three time periods [13,49]. These three urban heat models were created using random forest machine learning on temperature data collected using vehicle-based traverse measurements. Multiple land uses are included in the model (e.g., tree cover, building volume), and the temperatures derived are representative of the underlying urban form. Earlier research suggests that evening temperatures can have the greatest impact on