Who depends on swamp coolers in Tucson? Evaporative coolers in Tucson. Here we show who depends on them, where they fail, and what might happen next…
This is a story about swamp coolers. About who depends on them, where they fail, and what that means for the people inside those homes. They are cheap to run, great in dry heat, and the most common cooling technology in the city. But they fail when humidity rises. They pull unfiltered air indoors. And they concentrate in the neighborhoods that can least afford an alternative. We mapped them for the first time. This is what we found…
We start with the most basic questions. Where are the swamp coolers? Who lives in those neighborhoods?
Where the swamp coolers are
Click any neighborhood. Red means more swamp coolers. Blue means fewer.
The south and west sides of the city carry the highest prevalence. The foothills and east side carry the lowest.
This is not random. It tracks income, race, housing age, and renter status almost perfectly.
The spatial gradient is stark. Evaporative cooler prevalence ranges from near zero in the Catalina Foothills to over 60% in parts of the south side. The pattern is immediately recognizable to anyone who knows Tucson.
The gradients
Income
Lower-income neighborhoods have dramatically higher evaporative cooler prevalence. This is not a choice. It is a constraint. Swamp coolers cost a fraction of what refrigerated AC costs to install and run. They persist where the money to replace them does not.
Race
XXX
Built year
XXX
Spatial clusters
Our clustering analysis confirms that this pattern is not random. Red areas are clusters of high evaporative cooler prevalence surrounded by other high-prevalence neighborhoods. Blue areas are clusters of low prevalence.
The spatial autocorrelation is statistically significant. Swamp cooler dependence is a neighborhood-level phenomenon, not a household-level one.
The covenant legacy
Covenants shaped who could build wealth through homeownership and where investment flowed for decades. The housing stock in covenanted areas aged without the renovations that wealthier neighborhoods received. The residents who moved in after the covenants were lifted had less capital to upgrade cooling systems.
Neighborhoods inside historically covenanted areas have 7.8x higher odds of high evaporative cooler prevalence.
Average prevalence inside: 37.2%. Outside: 22.4%.
The evaporative cooler is not the cause of this inequality. It is its residue.
The hottest places
Now we add a layer. Not only do swamp-cooler-dependent neighborhoods have the weakest cooling technology, they also sit in the hottest parts of the city.
Double exposure
Evaporative cooler prevalence.
Summer maximum temperature (PRISM 4km).
Compare the two maps. The red areas on the left (high evaporative cooling) overlap substantially with the red areas on the right (highest temperatures). These neighborhoods have the worst cooling technology AND the most heat.
Less green does not mean more heat
Neighborhoods with less vegetation are hotter. This is the urban heat island effect mediated by tree canopy.
The neighborhoods without trees are the same ones stuck with swamp coolers. Less shade, more pavement, higher temperatures, and a cooling technology that was already inadequate.
Bivariate clusters
Red = the double-exposure hotspots. High evaporative cooling AND high temperature. These are the neighborhoods where the cooling gap is most dangerous.
When cooling fails
Swamp coolers work by evaporating water into dry air. When the monsoon arrives and humidity spikes, that process slows. The cooler keeps running but the house stays hot. This chapter quantifies how often that happens.
The failure heatmap
Each cell is the probability that a swamp cooler is failing at that hour of that day, averaged over 2018-2023. The monsoon signal lights up in July through September. Hover for temperature, humidity, and failure probability.
Compound exposure
Compound exposure = evaporative cooler prevalence multiplied by the number of failure days per summer.
A neighborhood where 40% of homes use swamp coolers and there are 25 failure days has a compound exposure of 10. That means 40% of households are effectively uncooled for 25 days each summer.
The burden falls on the south and west sides. The same neighborhoods. Every time.
Year to year
Some years are worse than others. The dashed line is the average.
Monsoon intensity varies. But the trend matters more than any single year. And the trend is the subject of the next chapter.
What the future holds
Climate projections from five CMIP6 models under two emissions scenarios. The question: which neighborhoods that are adequately cooled today will become inadequately cooled in 20 years?
Current vs. projected
Compound exposure today.
Projected under high emissions (SSP5-8.5, 2060s). Same color scale.
Both maps use the same color scale so you can compare directly. The red areas expand. Neighborhoods that are currently borderline become fully inadequate.
Threshold crossing
The dashed line is where we are today. The bars show where we are headed.
Under high emissions, the percentage of block groups exceeding the compound exposure threshold grows substantially by the 2060s.
These are not abstract numbers. Each percentage point represents real neighborhoods with real families whose cooling will become inadequate.
The trajectory
Projected cooler failure days per summer. Shaded bands = model uncertainty (10th-90th percentile across 5 CMIP6 models). Dashed line = observed baseline. The divergence between SSP2-4.5 (blue) and SSP5-8.5 (red) widens with time.
Who gets sick
The question underneath all of this. Does inadequate cooling actually make people sick? We linked heat-related EMS calls to cooling type, temperature, and demographics.
Call rates and cooling
Heat-related call rate per 1000 days.
Evaporative cooler prevalence. Compare the patterns.
The relationship
Neighborhoods with more swamp coolers have higher rates of heat-related calls.
This is an ecological association, not proof of individual-level causation. But the gradient is clear. And it persists after controlling for income, demographics, and renter status in the regression models.
Failure days and health
The city sees more heat-related calls on days when swamp coolers are failing (left). And neighborhoods in the highest evaporative cooling quartile have the highest call rates (right).
The interaction
| variable | estimate | se | IRR | p |
|---|---|---|---|---|
| failure_day | -0.010 | 0.025 | 0.990 | 0.673 |
| z_evap | 0.260 | 0.009 | 1.297 | 0.000 |
| tmax | 0.001 | 0.001 | 1.001 | 0.445 |
| z_income | -0.090 | 0.010 | 0.914 | 0.000 |
| failure_day:z_evap | 0.011 | 0.021 | 1.011 | 0.603 |
The interaction term (failure_day x evap prevalence) tests whether the health impact of a cooler-failure day is disproportionately concentrated in neighborhoods with more swamp coolers. A significant positive interaction means the cooling gap translates directly into a health gap.
This is the key result of the health paper. The interaction term closes the loop from cooling type to climate conditions to health outcomes.
The health burden ahead
We know that neighborhoods with more swamp coolers have more heat-related calls today. We know that climate change will increase the number of days those coolers fail. The next question follows directly: how much worse will the health burden get, and who will carry it?
Projected health outcomes
Heat-related call rate today.
Projected under SSP5-8.5, 2060s. Same color scale.
Excess calls and growing disparities
Left: excess heat-related calls per summer under two scenarios. Right: the disparity ratio between the highest-minority and lowest-minority neighborhoods. If the ratio increases over time, warming amplifies existing racial health gaps.
Who bears the projected burden
| variable | high_burden_mean | low_burden_mean | wilcox_p |
|---|---|---|---|
| Evap. prevalence | 0.24 | 0.25 | 0.862 |
| Median income ($) | 63576.95 | 62021.06 | 0.953 |
| Minority (%) | 0.46 | 0.49 | 0.203 |
| Renter (%) | 0.29 | 0.30 | 0.929 |
| Year built | 1976.56 | 1979.64 | 0.282 |
| Current call rate | 0.00 | 0.00 | NA |
The neighborhoods projected to see the largest increase in heat-related calls are the same ones that already have the most swamp coolers, the lowest incomes, and the highest minority proportions. Climate change does not create new inequities here. It deepens the ones that already exist.
So what do we do about it?
The data tells us where the problem is, who it affects, and how it will get worse. The remaining question is what it would take to fix it.
Cooling access deserts
A cooling access desert is a neighborhood with both high evaporative cooler prevalence (top quartile) and low income (bottom quartile). These are the places where the need for better cooling is greatest and the capacity to pay for it is lowest.
| variable | desert_mean | city_mean | wilcox_p |
|---|---|---|---|
| Evap. prevalence | 0.59 | 0.24 | 0.00e+00 |
| Median income ($) | 30031.85 | 63158.25 | 0.00e+00 |
| Minority (%) | 0.72 | 0.47 | 0.00e+00 |
| Renter (%) | 0.34 | 0.30 | 4.28e-04 |
| Year built | 1965.11 | 1978.32 | 1.00e-07 |
Vulnerability and cost
Composite vulnerability index: evap dependence, income, minority status, renter proportion, and housing age.
| scope | units | cost_low_M | cost_mid_M | cost_high_M |
|---|---|---|---|---|
| City-wide | 70859 | 248.0 | 389.7 | 531.4 |
| Cooling deserts only | 13786 | 48.3 | 75.8 | 103.4 |
The cost of replacing swamp coolers city-wide is substantial. But the cost of targeting only the cooling deserts is a fraction of that, and it is where the health returns are highest.
Priority neighborhoods
Priority neighborhoods meet all three criteria: high evaporative cooler prevalence, low income, and high composite vulnerability. These are the block groups where cooling intervention should start.
| dimension | covenanted | non_covenanted | wilcox_p |
|---|---|---|---|
| Evap. prevalence | 0.372 | 0.224 | 0.00e+00 |
| Income | 46329.328 | 65091.516 | 1.19e-05 |
| Minority % | 0.564 | 0.462 | 3.82e-04 |
| Renter % | 0.372 | 0.290 | 3.00e-07 |
| Year built | 1960.951 | 1980.316 | 0.00e+00 |
| Vulnerability index | 0.569 | -0.065 | 0.00e+00 |
These are the neighborhoods where every dimension of disadvantage converges. High evap, low income, high minority, high renter, old housing. The covenant legacy persists across all of them.
The evaporative cooler is not the cause of this inequality.
It is its residue.
Data Diversity Lab | University of Arizona