Environmental injustice among Hispanics in Santa Clara, California: a human–environment heat vulnerability assessment

In the United States, there is a growing interest in understanding heat stress in lower-income and racially isolated neighborhoods. This study spatially identifies heat-vulnerable neighborhoods, evaluates the relationship between race/ethnicity and temperature exposure, and emphasizes differences among Hispanics by origin to capture environmental injustices in Santa Clara County (SCC), CA. The current methodology uses Landsat 8 via Google Earth Engine to measure the Land Surface Temperature (LST) and Normalized Difference Vegetation Index (NDVI) to assess the physical environment. The human environment is evaluated using the Modified Darden-Kamel Composite Socioeconomic Index to determine the spatial variability of socioeconomic status (SES) and the Index of Dissimilarity to determine the level of segregation between Hispanics and non-Hispanic Whites and among Hispanics/Latinos. The combination of these assessments comprises a comprehensive human–environment approach for health exposure evaluation by which to define environmental injustice. Results reveal socioeconomic inequalities and an uneven residential distribution between Hispanics and non-Hispanic Whites. Low NDVI and high LST values were found in Mexican neighborhoods, implying possible environmental racism. Almost half the Mexican population lives in highly segregated neighborhoods with low and very low SES, mainly located in East San Jose, where, historically, they have been ghettoized. Mexicans, in general, could be at a higher risk of heat stress and heat mortality during heat waves. Future work should examine additional variables (e.g., housing characteristics, crime, social cohesion, and collective behaviors) to comprehensively evaluate the at-risk Mexican population.


Introduction
The United Nations projects that by 2030, 60.4% of the world will be urbanized (UN Habitat, 2020). As cities expand, the urban population will increase energy consumption, greenhouse gas emissions, and environmental degradation. Exposure of cities to weather hazards, like heat waves, will increase and intensify with global warming. Although heat waves are a regional phenomenon, localized heatstress illnesses (e.g., cramps, heat exhaustion, and heat stroke, leading to organ system failure and death) are generally observed in urban areas (Grady, 2013). These adverse health outcomes Abstract In the United States, there is a growing interest in understanding heat stress in lower-income and racially isolated neighborhoods. This study spatially identifies heat-vulnerable neighborhoods, evaluates the relationship between race/ethnicity and temperature exposure, and emphasizes differences among Hispanics by origin to capture environmental injustices in Santa Clara County (SCC), CA. The current methodology uses Landsat 8 via Google Earth Engine to measure the Land Surface Temperature (LST) and Normalized Difference Vegetation Index (NDVI) to assess the physical environment. The human environment is evaluated using the Modified Darden-Kamel Composite Socioeconomic Index to determine the spatial variability of socioeconomic status (SES) and the Index of Dissimilarity to determine the level of segregation between Hispanics and non-Hispanic Whites and among Hispanics/Latinos. The combination of these assessments comprises a comprehensive human-environment approach for health exposure evaluation by which to define environmental injustice. Results reveal socioeconomic inequalities and an uneven residential distribution between Hispanics and non-Hispanic Whites. Low NDVI and high LST values were found in Mexican neighborhoods, implying primarily occur in areas with a high concentration of low-income individuals and a majority of racial/ ethnic minorities; and those living in poor housing conditions (e.g., inadequate housing materials) and poverty are more likely to suffer from the adverse effects of a changing climate (Rosenthal et al., 2014). Racial, ethnic minorities, and the poor, who have historically lived in segregated and densely populated neighborhoods, are thus expected to experience higher heat exposure, which will increase heat-related deaths (Jesdale et al., 2013); consequently, the problem is framed as environmental injustice (Mitchell & Chakraborty, 2015).
In the United States, there is substantial evidence that a high proportion of racial and ethnic minorities reside within low-income neighborhoods deprived of green areas and limited access to cooling resources (Gronlund, 2014). On average, non-Hispanic Whites tend to live in census tracts with lower temperatures than census tracts with a majority of people of color, thus reflecting heat exposure as an unevenly distributed environmental burden (Hsu et al., 2021). In Portland, Oregon, mean temperatures during a 2014 heat wave event were negatively correlated with the percentage of Non-Hispanic Whites (−0.1515°C for every 10% increase) but positively correlated with the percentage of African Americans and Hispanics (+0.3471°C for every 10% increase), at a US census block group level (Voelkel et al., 2018).
Historical uneven urban development patterns, discriminatory realtor, lending, and zoning practices segregated racial and ethnic minorities in neighborhoods with poor housing conditions and limited green infrastructure, which have persisted and continue to expose them to higher ambient temperatures (Uejio et al., 2011). A study of 108 previously redlined neighborhoods showed temperatures of almost 7°C higher in redlined areas than in non-redlined ones (Hoffman et al., 2020). In Richmond, Virginia, previously redlined neighborhoods have the highest rates of heat-related emergency-room visits, as the number of visits increased by 2.5% for every one-degree increase in ambient temperature (Plumer & Popovich, 2020). In Phoenix, a higher number of heat distress calls has been associated with neighborhoods with a high proportion of Black, Hispanic, linguistically, and socially isolated residents, while in Philadelphia, neighborhoods with high heat mortality were more likely to have low housing values and a higher proportion of Black residents (Uejio et al., 2011).
There is also an association between race, socioeconomic status, and the ability of a population to cope with the heat as it determines their access to mitigating resources (e.g., air conditioning systems). A study performed at a city level indicated that, in four US cities, the prevalence of central air conditioning (A/C) was lower among Black households and likely correlated with socioeconomic characteristics (O'Neill, 2005). In the United States, neighborhoods with a high proportion of minority residents have been associated with low-income status due to institutionalized racism in the form of redlining and zoning ordinances (Gronlund, 2014). In Phoenix, neighborhoods with a high proportion of Hispanic residents, mainly of Mexican origin, were associated with low-income status (Jenerette et al., 2007), low educational attainment, and low use or availability of A/C systems (Harlan et al., 2013).
High levels of segregation result in the isolation of a minority group from amenities, opportunities, and resources that affect social and economic wellbeing (Massey & Denton, 1989) and, thus, their ability to cope with heat exposure. Highly segregated areas with higher proportions of non-white and lowincome residents tend to experience more significant cumulative environmental hazards (e.g., toxic release facilities, noise, heat) (Casey et al., 2017;Morello-Frosch et al., 2002). There is, however, a variation in heat-related outcomes by race. During a heat wave in California, Latinos reported increased cardiacrelated illnesses, African Americans had acute renal failure and electrolyte imbalance, and Asians had significantly elevated emergency visits for respiratory disorders (Green et al., 2010). Measures of residential segregation are necessary to understand the origins and persistence of environmental health disparities (Morello-Frosch & Lopez, 2006).
Residential segregation is the degree of separation between two or more groups within an urban environment (Massey & Denton, 1988). Residential segregation indices are an objective measure of racial and economic segregation that could serve as proxies for structural racism (Chambers et al., 2019). There are five key dimensions of residential segregation: evenness, exposure, concentration, centralization, and clustering (US Census Bureau, 2000). The Index of Dissimilarity is one of the most common measures of residential evenness in environmental justice studies but can only measure a region's segregation for two racial groups. Baxter (2010) used the Index of Dissimilarity (range, 0-1) to estimate the degree of ethnic residential segregation in Santa Clara, CA, by block groups in 1990-2000. Results showed higher segregation levels between Whites and Hispanics (0.54) than between Whites and Asians (0.45). Comparisons of 1990 and 2000 results showed an increase in residential segregation with fewer White residents in Asian and Hispanic neighborhoods, from 43 to 23% and from 32 to 17%, respectively. The number of White residents in White block groups remained stable (82%). The proportion of Hispanic residents in Asian neighborhoods decreased (16-13%), and the proportion of Asians in Hispanic areas increased (12-14%). As of 2010, Latinos in Santa Clara County showed the highest levels of segregation (Menendian & Gambhir, 2018).
Levels of segregation, however, vary among Hispanics by origin (Bean & Tienda, 1987). For example, Mexicans tend to live in isolated inner-city ghettos; consequently, this denies them equal access to schooling, jobs, and health care compared to other Hispanic groups (National Research Council, 2001). The association between segregation and health outcomes also varies among Hispanics by origin, nativity, and length of time in the United States (Do et al., 2017). The socioeconomic differences within Hispanic subgroups expose them to unique environmental hazards that may contribute to higher morbidity resulting in widening health disparities. Cubans, for example, have a college graduation rate three times higher than Mexicans, while Puerto Ricans have higher high school graduation rates than Mexicans (Williams et al., 2010).
Populations living in highly dense coastal cities with low acclimatization to heat will thus experience negative health outcomes (e.g., increased heat stress illnesses and premature mortality) (Grady, 2013) and strain energy resources as the energy demand for A/C increases (EPA, 2021). The spatial distribution of land surface temperatures and vegetation, income disparity, the cultural diversity of the population (multiracial, multinational, multilingual, multicultural), and the spatial segregation patterns increase the complexity of assessing heat vulnerability. The current effort applies existing and fully developed metrics of socioeconomic status and racial segregation to spatially identify heat-vulnerable neighborhoods and evaluate the relationship between race/ethnicity, the environment (e.g., temperature exposure), and socioeconomic status in Santa Clara County (SCC), CA, emphasizing differences among Hispanics/Latinos. Of the total population (1,911,226) in SCC, 32% (622,266) is Non-Hispanic White and 26% (498,253) is Hispanic/Latino, of which 85% (420,559) is Mexican, 6% (29,464) Central American, and 3% (15,104) South American (US Census Bureau, 2018).
California is among a few states where extreme heat days and nights have become more frequent (CalEPA, 2018;Taha, 2017), and SCC ( Fig. 1), located south of the San Francisco Bay in Northern California, has experienced, from 1950 to 2010, a surge in heat-wave events, which are also projected to intensify in the future (Gershunov & Guirguis, 2012). The mean daily apparent temperature in SCC during the warm season (May 1 to September 30), calculated from meteorological data acquired from 1999 to 2006, was 18.2°C with a range of 9.1-32.8°C (Basu & Malig, 2011). Within SCC, the city of San Jose had the highest increase in temperature values, with an average rise of 0.32°C per decade (Lebassi et al., 2009). The findings from this study will therefore inform other coastal as well as inland cities on the variability of temperature exposure among the Hispanic population, which, as of 2016, was the most significant ethnic or racial minority in the United States and, by 2060, they will represent 27.5% of the total population (Vespa et al., 2020).

Data and methods
The current methodology combines Landsat 8 satellite data and the US Census Bureau-American Community Survey (ACS) 2015-2019, Five-Year Estimates to determine the distribution of heatvulnerable populations in SCC. The first step assessed the physical environment via the calculation of the Land Surface Temperature (LST) and Normalized Difference Vegetation Index (NDVI). The second part evaluated the social environment with the use of the Modified Darden-Kamel Composite Socioeconomic Index (CSI) to determine the spatial variability of socioeconomic status (SES) (Darden et al., 2010).
The Index of Dissimilarity (D) (Massey & Denton, 1988) was calculated to determine the level of segregation within the Hispanic/Latino population. The combination of these assessments resulted in a comprehensive human-environment approach for health exposure evaluation by which to define environmental injustice.

Physical environment
Heat exposure was determined from LST and NDVI values, calculated from Landsat 8 imagery, accessed via the freely available (Google) Earth Engine website for 2015-2019, in accordance with the ACS (Five-Year Estimates) dataset. Landsat 8, with a time range from 1100 to 1115 Central Standard Time (CST), has a 30-m spatial resolution in the Visible, Near-Infrared (NIR), and Shortwave Infrared (SWIR) bands and 100-m resolution in the thermal infrared (TIR-1 and TIR-2) bands. The atmospherically corrected TIR and surface reflectance (SR) bands are resampled and made available at a 30-m resolution in Earth Engine.
The dataset also includes a Quality Assessment (QA) band, which contains per-pixel cloud, shadow, water, and snow information.
The procedure to retrieve LST and NDVI values first involved a per-pixel filter applied to Landsat 8 images, using its QA band to mask cloudy and cloud shadow pixels from the SR bands. The remaining pixels were then used to calculate LST distributions using the brightness temperature derived from the TIR-1 band. The NIR and Red bands were also utilized to perform an emissivity correction of LST via the calculation of NDVI values (Avdan & Jovanovska, 2016;Weng et al., 2004). The emissivitycorrected LST values were then converted from °K to °C and averaged for the 5-year study period. The averaged values reduce the effects of anomalously wet or dry years, as well as seasonal changes in vegetation and soil moisture, which could impact the calculated LST and NDVI values.
The National Land Cover Dataset (USGS, 2019) was used to extract NDVI and LST pixels that corresponded with values classified as developed (21-24), as shown in Fig. 1. This process filtered non-populated locations and allowed for the accurate calculation of mean LST and NDVI values per census tract. The extraction of these values reduces the elevation and vegetation effects, especially from the non-developed area of the hills. The ArcGIS Zonal Statistics tool was then used to calculate the mean LST and NDVI values at the census tract level to obtain the mean value per census tract.

Socioeconomic environment
Socioeconomic and race/ethnicity data used to calculate the Modified Darden-Kamel CSI and the Index of Dissimilarity, respectively, were acquired at the census tract level for SCC from the US Census Bureau, American Community Survey (ACS) 2015-2019 (US Census Bureau, 2020).
Modified Darden-Kamel CSI. Measures SES for the entire population in a study area and considers nine variables: (1) percentage of bachelor's degrees, (2) median household income, (3) percentage of managerial and professional status positions, (4) median value of dwelling, (5) median gross-rent of dwelling, (6) percent of homeownership, (7) incidence of low income (-), (8) unemployment rate (-), and (9) percent of households with vehicle (Darden & Rubalcava, 2018). The formula to calculate the Modified Darden-Kamel CSI is: where V ij is the jth SES variable for a given census tract i, V jDMA is the mean of the jth variable in the study area, and S ( V jDMA ) is the standard deviation of the jth variable in the study area. Census tracts with a higher socioeconomic position were assigned a higher score. These scores were then divided into five degrees (or classes) by use of the Dalenius and Hodges (1959) stratification method, which minimizes variation within each group. This process was conducted with the use of the stratification library and implemented in RStudio © (version 4.1.2).
Index of Dissimilarity. The Index of Dissimilarity was used to evaluate the levels of segregation between non-Hispanic Whites and the Hispanic/Latino population and within Hispanic subgroups, including Mexican, Central American, and South American. It calculates the proportion of Group A that would have to change their neighborhood (or census tract) to achieve an even distribution with Group B (Massey & Denton, 1988, p. 284). High D values reflect a high degree of residential segregation. The formula to calculate D is: where x i is the percentage of the total ethnic minority population in SCC (e.g., Hispanic/Latino) in census tract i, y i is the percentage of the total non-minority population in SCC (e.g., non-Hispanic White) living in the same census tract i, and k is the total number of tracts in the study area. The absolute differences between these percentages by census tract were then divided into five degrees (or classes) of segregation by use of the Dalenius and Hodges (1959) stratification method to minimize variations within each group, also done in RStudio © with the implementation of the stratification library. This process allowed us to visualize the spatial distribution of Hispanics in the study area. One-half of the sum of these absolute differences results in the level of segregation (D) for the study area.

Human-environment
Mean LST and NDVI values at the census tract level were combined with the Modified Darden-Kamel CSI and Index of Dissimilarity (D) results to evaluate the physical and social environmental risks and study human-environment interactions between heat exposure, socioeconomic status, and race/ethnicity. LST and NDVI values per SES first inform about differences in environmental exposure between the five classes. LST and NDVI values were then combined with the D values to evaluate the differences in environmental exposure by the degree of segregation for each considered race/ethnic group.
The Shapiro-Wilk test (Shapiro & Wilk, 1965) determined LST and NDVI values as non-normal; thus, Kruskal-Wallis tests (Kruskal & Wallis, 1952) were performed to determine how LST and NDVI values differ by SES and ethnicity, with 95% Confidence Intervals. The Univariate Moran's I Index (Anselin, 1995) measured global spatial autocorrelation of the LST and Modified Darden-Kamel CSI values, while the Bivariate Local Moran's I (Anselin et al., 2002) provided a local spatial autocorrelation measure between these two variables. All analyses used first-order Queen contiguity (Anselin & Rey, 2014) to determine neighboring census tracts.

Physical environment
Land Surface Temperature. Landsat 8-derived mean LST distributions (Fig. 2) Fig. 3) range from −0.45 to 0.89, where higher values indicate greenness or photosynthetic activity. Maximum values are found in southeast Los Gatos, Saratoga, and Los Altos Hills, which correspond with vegetated mountainous areas and golf courses, dispersed throughout the study area. Minimum values are primarily located in the Downtown and northern area of SCC, which correspond with industrial sites and airports. Below-average values (< 0.35) are found in the East San Jose areas and along major transit corridors (e.g., main boulevards and highways). Aboveaverage values (> 0.35) are found in the cities of Palo Alto, Los Altos, and the southwest San Jose Downtown area.

Socioeconomic environment
Modified Darden-Kamel CSI. There are significant disparities between the populations living in Very High-(VH) and Very Low-(VL) socioeconomic status (SES) census tracts in Santa Clara County. Results (Table 1) show a positive linear trend for all variables except for poverty, which increases with decreasing SES, and unemployment, which shows minor variation. The Medium-(M) SES is, however, smaller than the overall mean value in SCC for median household Neighborhoods with VH-SES have, on average, a median income of $205,06, nearly three times that of neighborhoods with VL-SES, and both higher than the 2019 national median household income of $69,560 (Shrider et al., 2021). VL-SES households spend 28% of their median annual household income on rent ($1694/month), almost twice that of VH-SES households ($3,198/month). Households with a VH-SES are mostly homeowners (82%) with median house values of 1.8 million, further exacerbating the gap between the VH-SES and VL-SES groups. Although vehicle ownership is above 90% for all groups, VH-SES neighborhoods have higher vehicle ownership than VL-SES.
Similar values among groups are found for unemployment, with values that range between 4.1 and 4.3, with minimum ones for VH-SES that increase  as SES decreases, except for the M-SES class, with maximum unemployment of 4.4%. Despite similar unemployment levels, the percentage of families living in poverty in VL-SES is three times that of those with a VH-SES. The percentage of professional and managerial workers is similar to those with bachelor's degrees or higher education, with lower values for VL-SES areas, 28 and 31%, respectively, and increasing with SES to a maximum of 79% (bachelor's degree) and 77% (professional and managerial workers) for VH-SES areas.
Distributions of the Modified Darden-Kamel CSI results at the census tract level (Fig. 4) show neighborhoods with VH-SES in southwest SCC in Palo Alto, Los Altos, Los Altos Hills, Cupertino, Saratoga, Monte Sereno, and Los Gatos, while VL-SES ones are in the Downtown and East San Jose areas, as well as in the southeast cities of Morgan Hill and Gilroy. Stanford University was classified as having a VL-SES and considered an outlier. Areas with Low (L)-SES surround those neighborhoods with VL-SES.
Almost 25% of the census tracts were classified as areas with M-SES and 23% with H-SES. Of the total population, 12% live within neighborhoods with VH-SES, 24% with H-SES, 25% with M-SES, 22% with L-SES, and 17% with VL-SES.
The share of the Hispanic/Latino population and within Hispanic subgroups, including Mexican, Central American, and South American, by socioeconomic status in SCC (Table 2) shows an inverse relationship between SES for Hispanic/Latino, Mexican, and Central American groups. A similar proportion of Non-Hispanic Whites and South Americans is found for H-, M-, and L-SES. A higher proportion of South Americans is, however, found in the VL-SES group (13.4%) than of Non-Hispanic Whites (9.9%). The reverse occurs for VH-SES groups, with 9.1% of South Americans and 16.1% of Non-Hispanic Whites belonging to this group. Between subgroups, South Americans show small disparities between VH-and VL-SES (9.1-13.4%), followed by Central Americans While 16.1% of the Non-Hispanic White population lives within a VH-SES neighborhood, only 1.6% of the Mexican population lives in a VH-SES area, followed by Central Americans (3.6%) and South Americans (9.1%); these last two with values higher than that of the general Hispanic/ Latino population (2.3%). Mexicans are, however, overrepresented in L-and VL-SES neighborhoods, with 32.6-34.9% of their population in these groups, respectively. Central Americans have a lower share of their population in the L-and VL-SES (31.1-25.3%). As the SES decreases for all Hispanic/Latino groups, the percentage of the population per class increases, except for South Americans, who have the lowest percentage of its population within an L-and VL-SES neighborhood, a similar trend to that of the Non-Hispanic White population.
Index of Dissimilarity. The Index of Dissimilarity (range 0-100) assigns higher values to neighborhoods (census tracts) with a high degree of residential segregation. Results (Fig. 5) show the spatial distribution of residential segregation of Hispanic/Latinos and by their origin: Mexican, Central American, and South American, compared to the Non-Hispanic White population. The highest levels of residential segregation are within the Central American community (Fig. 5c), with an Index of Dissimilarity (D) of 52.7, scattered around the study area but not found in the Los Altos and Los Altos Hills areas. Mexican neighborhoods (Fig. 5b) also have high residential segregation (D = 51.4) and are clustered in the East San Jose area and southeast SCC in Gilroy. South Americans (Fig. 5d), like Central Americans, are also scattered around the study area but showed the least amount of residential segregation (D = 42.0), even lower than the general Hispanic/Latino population (D = 47.4), as shown in Fig. 5a.

Human-environment
The LST and NDVI relation with SES shows that as SES decreases, mean LST values increase from 28.5 to 29.9°C (Fig. 6). The reverse occurs for NDVI values and SES, which decrease from 0.41 to 0.27 as SES decreases. Only VH-and H-SES areas exhibit LSTs lower than the SCC average (29.4°C). VH-and H-SES census tracts exhibit values above the SCC average (0.33). Kruskal-Wallis results show that LST and NDVI values are statistically significantly different for at least one of the groups (H = 53.44, df = 4, P = 0.00; H = 117.89, df = 4, P = 0.00, respectively). Pairwise comparisons show that LST values vary between groups except between M-and L-SES and L-and VL-SES.
Mean LST and NDVI results combined with stratified D values (Table 3) show a broader range of values for Mexicans than for Non-Hispanic Whites (29.3-30.2°C vs. 29.2-29.5°C). Areas with a VH concentration of the Hispanic/Latino population, particularly Mexicans, have a higher LST than areas with a VL concentration of the Hispanic/Latino population (29.8°C vs. 29.4°C). Minimum LST (28.5°C) and maximum NDVI (0.37) values were observed for areas with an H concentration of South Americans. Kruskal-Wallis results show that only LST values are different for at least one group (H = 14.22, df = 4, P = 0.007), while no statistically significant difference was found for NDVI values (H = 8.221, df = 4, P = 0.084). Pairwise comparisons showed no statistically significant differences in LST values between areas with a VH-Concentration of South Americans and VH-Concentration of Central Americans and between South Americans and Non-Hispanic Whites. Significant differences in LST values were found between Mexicans and South Americans and Mexicans and Non-Hispanic Whites.
The Univariate Moran's I Index results showed 0.610 and 0.581 values for the LST and Modified Darden-Kamel CSI measures, respectively, with P < 0.05; thus, values for both variables are clustered.  The spatial association between these two variables, explored via the Bivariate Local Moran's I, shows a value of −0.324 (P < 0.05). Results indicate a negative correlation between LST and SES (Fig. 7). High-high clusters of SES and LST values are located within neighborhoods with large low-rise and medium-rise buildings, with a medium concentration of Hispanic/Latinos, specifically South Americans.
Low-low clusters of SES and LSTs are mainly located east and southeast of SCC, where the proportion of the Hispanic/Latino population is insignificant (L to VL). Low-high clusters of low SES and high LSTs are in East San Jose, where most of the Hispanic/ Latino population resides, where historically, Mexicans have been ghettoized. Areas of high-low clusters of high SES and low LSTs are mainly located in the

Discussion
This study reveals socioeconomic neighborhood inequalities and an uneven spatial distribution of NDVI and LST values across Hispanic communities in SCC. Almost half the Mexican population lives in highly segregated neighborhoods with L-and VL-SES, mainly located in East San Jose, where, historically, they have been ghettoized. Neighborhoods with a VH-Concentration of Mexicans show higher LST and lower NDVI values than neighborhoods with a VH-Concentration of Whites. The spatial patterns of residential segregation and environmental and socioeconomic neighborhood inequality reflect high-temperature values and a low amount of green areas available in Mexican neighborhoods, implying possible environmental racism that could have resulted from historical racial/ ethnic processes (e.g., racism and discrimination).
Results of this study thus suggest that the ethnic composition of a neighborhood is strongly linked to LST values. Mexicans, in general, could be at a higher risk of heat stress and heat mortality during heat waves. Although the study area is racially diverse, there is an intrinsic ethnic division of labor (Mehrens, 2015), as shown in Table 1. Since the "Factory Valley," when the region dominated the manufacture of semiconductors, thus earning the name Silicon Valley (Cheyre et al., 2015), around 70-80% of people working in manufacturing jobs were immigrants, women, and people of color, who continued being the backbone of Santa Clara County's economy long after the Gold Rush (Pellow & Park, 2002). For example, Siegel (1995) states that, during the 1970s, most managers were White (88%), and only 4 and 5% were Latino and Asian, respectively. By the 1980s, the number of Asians in managerial positions increased from 5 to 10%, but the percentage of Latinos remained. From the 1970s to the 1980s, however, the percentage of White laborers decreased from 41 to 19%, while the number of Latinos remained almost the same (34 to 36%) (Siegel, 1995). In the current tech industry, the ethnic division of labor remains in the top three most profitable companies-Google, Facebook, and Apple-where only 2-7% of the workforce is Hispanic/Latino (Mehrens, 2015). Hispanic/ Latinos make up 69% of janitorial workers. As shown in this study, only 30% of the population in VL-SES tracts hold a professional or managerial occupation, of which nearly 30% are Hispanic/Latino.
There is a massive disparity between the population living in VH-and VL-SES and high living costs. The median rent in SCC was $2,363 a month, yet the minimum wage in SCC is $15.65/hour. Skyrocketing house prices have forced many people into homelessness, and the area hosts the largest homeless camp in the continental US (Johnson, 2013): Of the 4,350 homeless people, 3,219 were unsheltered. Most homeless individuals are male (70%), and 46% self-identified as Latino; among homeless families with children, 78% were singlemother families, of which 56% were Hispanic/Latino. Approximately 47% of unaccompanied children also identified as Hispanic or Latino (Applied Survey Research, 2017). The Hispanic/Latino population, most living with a median household income of $73,000, although with low levels of unemployment (4.3%), live in poverty and are at risk of eviction and homelessness.
Homeless individuals showed higher rates of preexisting psychiatric illnesses and comorbid conditions (e.g., cancer, cerebrovascular, infectious diseases), which increases their heat mortality risk (Ramin & Svoboda, 2009). Their inability to access drinking water and lack of healthcare make them susceptible to high temperatures and further exacerbate their illnesses (Nicolay et al., 2016). Homelessness and age are critical variables that drive heat-related mortality (Putnam et al., 2018). Homeless populations do not perceive themselves as vulnerable, avoid water distribution routes, and are frequently asked to leave cooled public areas (Benmarhnia et al., 2018).
Inequalities exist within the Hispanic/Latino population.
Although the Hispanic/Latino population is, on average, highly segregated and impoverished, Mexicans represent the highest share of VL-SES. Residential segregation in Silicon Valley dates to the late 1800s, during the gold rush. Southwest San Jose became a valuable player with the discovery of a quicksilver mine in the Almaden Valley. Mexicans and Native Americans, considered second and third-class citizens, worked the mines in dangerous conditions. Over time, the area around the Almaden mine would become known as the segregated neighborhood of Mexican workers. According to this study, a high concentration of the Hispanic/Latino population, primarily Central Americans, is located around the Almaden mine and less in the East San Jose area, but they exhibit M-to VH-SES but are the most segregated Hispanic subgroup.
By the 1940s, most Latinos lived in the segregated muddy fields on the East Side of San Jose. Nearly 30% of the Mexican population, however, lives in VL-SES and concentrates in East San Jose, also known as "Sal Si Puedes" (get out if you can) (Heppler, 2018), but most Central Americans concentrate southwest SCC towards the Almaden neighborhood. South Americans are the least segregated group but do not concentrate in the same areas where Mexicans and Central Americans commonly settle. The percentage of South Americans in VH-SES (9.1%) is four times higher than that of Mexicans in VH-SES (1.6%). According to results from the Modified Darden-Kamel CSI, most Mexicans (67%) exhibit VL-and L-SES.

Heat disparities within Hispanics
Although areas with a VH concentration of Hispanics have higher LST values than Non-Hispanic Whites, Mexicans have the highest LST values (1.2°C higher than non-Hispanic Whites), thus reflecting being overburdened by heat and, as previously discussed, by poverty. The most considerable differences in LST values (1.4°C) are shown between the VH-Mexican and H-South American groups. Results also showed the same statistically significant temperature (1.4°C) difference between VH-and VL-SES. The VL-SES group is also 1°C above the mean SCC value.
Epidemiological studies have found that a 1°C increase in maximum temperature can increase ambulance response calls due to heat-related illnesses by 29% (Bassil et al., 2011), heat-related emergencyroom visits by 2.5% (Plumer & Popovich, 2020), and heat-related mortality from 1 to 3% (Hajat & Kosatky, 2010). Based on findings from our study, low-income Mexicans would have the highest rates of heat-related outcomes. These adverse effects are most likely to be cardiac-related illnesses, as it has been the primary reported outcome among Latinos (Green et al., 2010). It is necessary, however, to disaggregate these health effects by country of origin, as there may be differences among Hispanics due to socioeconomic disparities as well as population age-structure by origin within this ethnic group.
The literature is unclear if Hispanics are at higher risk of hospital admissions. Some studies have found Hispanics at higher heat risk, while others suggest a lower heat health effect (Gronlund, 2014), which may be related to the differences in socioeconomic status between Hispanics. In California, when compared to Whites, Blacks showed elevated heat-related risks, but this was not the case for Hispanics (Basu & Ostro, 2008). Similarly, Klinenberg (2002) also found that suburbs in Chicago composed mainly of African American residents exhibited a higher mortality rate than those primarily consisting of Latino residents. The low heat-related risk among the Hispanic population may be explained in part by the Hispanic Health Paradox (HPP). HPP indicates that although Hispanic immigrant populations in the US experience high economic deprivation, they tend to have better health outcomes than their native-born counterparts and other racial/ethnic subgroups, including the white-Anglo majority (Kim et al., 2014).

Conclusions
From the gold rush to the Apple janitorial strike, Silicon Valley was built by discriminatory practices and policies (Pitti, 2004). The capitalist wage system and the dominance of Euro-American culture helped marginalize non-white groups in California (Almaguer, 2009). Income inequalities emerged in the region because of economic and political policies that resulted in specific residential patterns. Fiveyear Landsat 8 satellite data and the Census Bureau-American Community Survey (ACS) 2015-2019 were used to assess the physical and social environment via the calculation of LST, NDVI, Modified Darden-Kamel CSI, and Index of Dissimilarity. This study used the Index of Dissimilarity to estimate the degree of Latino/Hispanic residential segregation and the Darden-Kamel CSI to calculate SES by census tract. The combination of these indicators allowed for the identification of vulnerable communities. Results showed differences in residential segregation, socioeconomic status, and high-temperature exposure (LST and NDVI) among Hispanics by origin. Hispanic/Latinos, primarily Mexicans, live in highly segregated areas and are exposed to high LST and low NDVI values. The Univariate Moran's I Index and Bivariate Local Moran's I results show a statistically significant spatial correlation and thus allowed for the identification of heat vulnerable populations, who in SCC are more likely to be Mexicans living in the historic neighborhood of Sal Si Puedes, where institutionalized racist and discriminatory processes have historically segregated them.
Relatively small differences in LSTs by SES and ethnic group may result from the aggregation of mean annual values that could homogenize extreme cold and heat days. Although non-developed areas were filtered to produce a more accurate temperature measurement of populated areas, this study does not consider urban morphology (e.g., land use and building form), which may vary by class and ethnicity, as studies have found higher population densities in Hispanic neighborhoods in the US. Future studies should consider averaging LST values for warm months only (e.g., May-September) and describing the built environment (e.g., population density, building materials), which also may vary by SES. Additional variables such as homelessness, housing characteristics, crime, social cohesion or isolation, and collective behaviors, and their agestructure differences within the Hispanic population, need to be included in more comprehensive heatrelated studies to provide a more robust evaluation of the at-risk population in urban areas, where heat stress is expected to intensify.
Proposed solutions to reduce heat exposure include urban greening projects, shading sidewalks, parking lots, and other paved surfaces, and using trees or additional structures (e.g., sailcloths and solar panels) (Rodriquez & Chapman, 2013). Urban greening is not a simple affair: it involves major infrastructural, management, maintenance, and watering complexities that impose costs and challenges for low-income communities. Recommendations thus also include promulgating local codes to accelerate the adoption of cooling strategies. This regulatory approach could be complemented with the implementation of heatemergency preparedness plans at the community level and the placement of cooling centers. Local communities must have input in the development of these mitigation and adaptation strategies (Wilhelmi & Hayden, 2010). A multi-scale multi-stakeholder approach would be the best path to ensure hazard reduction, especially in low-income and racial/ethnic minorities.