By combining satellite and ground-based measurements, scientists have created new maps that identify which neighborhoods experience the most extreme heat on the hottest days of summer.
In the summers of 2017 and 2018, citizen scientists in Richmond, VA, the District of Columbia, and Baltimore, MD, gathered temperature data all over their cities on days when temperatures reached at least 95°F. The results, as outlined in a NOAA article from 2018, show that air temperatures in some areas of the city can be up to 17°F hotter than other areas during the same time of day.
These hotter areas are known as “urban heat islands,” referring to the fact that cities’ unshaded roads and buildings absorb more heat during the day and radiate it back into the surrounding air as heat, significantly increasing the local air temperature relative to other, more vegetated or shaded areas. On days when local temperatures climb above 95°F, the additional heat emitted by paved and concrete structures can produce dangerously hot temperatures in some neighborhoods.
This work was funded in part by NOAA’s Climate Program Office and NOAA’s Office of Education. The results from the three campaigns are outlined in a new paper in the journal Climate, titled Integrating Satellite and Ground Measurements for Predicting Locations of Extreme Urban Heat. Vivek Shandas, an Urban Studies Professor at Portland State University, is the paper’s lead author.
Shandas and his team devised a new, engaging method for measuring temperatures. Instead of relying solely on technical analysis of satellite imagery, which has been long-applied to characterize urban heat, campaign volunteers drive around designated parts of a city in their car or bike with temperature sensors attached, resulting in over one-hundred-thousand data points. Then, they combine these ground-based data with satellite images of land cover, and a unique machine learning algorithm to construct detailed maps of the urban heat island effect for each city.
Washington, DC, heat maps at 6 am (left) 3 pm (middle) and 7 pm (right). Temperatures are represented in fahrenheit. Overall, the city is hottest in the afternoon, and coldest in the morning.
These maps provide municipal planners, public health practitioners, urban foresters, and community members an opportunity to identify ways to safe-guard their residents and structures from exposure to extreme heat. According to Jeremy Hoffman, a climate and earth science specialist at the Science Museum of Virginia, “We would like to protect the people who are disproportionately affected by heat waves in these urban areas,” he said. “From a public health standpoint, there are interventions that can be done to keep people safe now and into the future, as we’ve seen our heat waves become longer and more intense.”
What’s different about these new maps?
Shandas notes that his new urban heat island maps are different from satellite-derived temperature maps in several key ways. “First, we measure air temperature at about 6 feet above the ground, which is the temperature humans actually experience,” he explains. “Satellites, on the other hand, measure ‘skin temperature’ of whatever they’re observing, which can be road, ground, tree tops or rooftops. The surface skin temperature of black asphalt on a very hot day in summer can be as much as 200°F — you could fry an egg on that surface — but we don’t really experience such extreme temperatures because the air cools off rapidly as you go several feet above that surface.”
Also, finding high-resolution satellite sensors (like Landsat or ASTER) that are cloud-free on a given city during the summer pose additional challenges. Most satellites circle the earth every week or two, and at varying times of the day, sometimes well before the hottest time of day when the contrasts in temperature across the cityscape are most extreme. But Shandas designed his mapping campaign to be flexible, using local citizen science volunteers, so that he and his team can wait for the hottest day of summer, and spring into action with as little as a week or less notice. Additionally, they collect measurements all over the city at three times of day — 6:00 am, 3:00 pm, and 7:00 pm. Shandas explained that those three specific times were chosen because, “The morning is the point in which all the heat absorbed in the materials have had an opportunity to be released, which indicates the coolest moment in the city. Around 3:00pm in North America is typically the hottest period of the day and, because of this peak, we can capture how fast places are heating up. As the heat dissipates in the evening (7 pm), we pick up on rates of cooling and see what materials cool faster than others.”
Baltimore, MD heat maps at 6 am (left) 3 pm (middle) and 7 pm (right). Temperatures are represented in Fahrenheit. More areas around Baltimore heat up throughout the day and drive higher temperatures.
The hotter the day, the more precise the spatial patterns of temperature differences across the city. If the cities were traversed on a cooler day, the spatial patterns would be the same but the team wouldn’t be able to measure the maximum potential differences in temperature. For example, a city on a 110-degree day might have temperatures at 115 degrees in some places, and 95 degrees in others, thus creating a 20-degree range. But on a 90-degree day, some places could have temperatures of 93 degrees and 83 degrees in others, producing only a 10-degree range. The spatial pattern of the hottest and coolest places remains the same though the temperature differences vary.
Some previous studies of urban temperature have relied on interpolation, or where observations were used to create a representation of the real-world temperature signal by fitting a line between the points observed. Interpolation takes two data points and averages them to find a data point in between, which works for a wide open area, but in a city where there are parks, skyscrapers, parking lots, and more in between, interpolation doesn’t accurately reflect the temperature. Moreover, surface skin temperatures can at times be upwards of 200-300 degrees, so the satellite measurements taken are not the most accurate indication of the air temperatures humans actually experience several feet above the surface.
By using a model-based approach and measuring ambient air temperature, they can better describe temperatures felt throughout the day. A model-based approach takes specific land cover types — such as buildings, trees, and vegetation — into account and allows Shandas and his team to predict the temperature over the entire study area. To test the accuracy of their model, they use about 70 percent of the observed data (a total of about 85,000 temperature readings) and a machine learning algorithm to predict the remaining 30 percent. While these ‘hold-out’ methods of testing a model are common, their results are often upwards of 98% accuracy.
This campaign has been performed in nine cities so far: Albuquerque (NM), Portland and Eugene (OR), Tacoma (WA), Richmond (VA), Baltimore (MD), the District of Columbia, Hermosillo (Mexico), and Doha (Qatar). These cities have helped Shandas and his team refine their approach, which has now become almost fully automated. During the summer of 2019, for example, Shandas and his team are expecting to increase the number of campaigns. “Previously, our whole team needed to travel to each area, often relying on weather forecasts and availability of local volunteers, but now we have changed all that to improve the effectiveness of these field campaigns,” Shandas said. “We have created a software system that serves as a guide for organizing volunteers, executing campaigns, and uploading data. We also provide webinar support to organizers and trainings through video-conferencing.” These changes will mean that more cities can engage their communities, and generate high-resolution descriptions of urban heat each summer. A long-term goal is to reduce the cost of the campaign while scaling up the number of cities mapped in future years. “We have the capacity to support upwards of 20 cities during the summer of 2019,” Shandas added.
These campaigns offer a relatively simple, cost-effective way for cities, non-profits, community organizers, and citizens to map their exposure to extreme heat and get involved in climate action. These maps can also inspire coordinated strategies — small and large — to build resilience to climate change, such as updating cityscape designs for more green spaces and planting more trees, installing splash parks and cooling centers, and other initiatives to protect people and property from the damaging effects of extreme heat.
Full-resolutions maps from the team’s previous urban heat island mapping campaigns are available here: https://capastrategies.com/capa-heat-watch/
Read the full paper here: Integrating Satellite and Ground Measurements for Predicting Locations of Extreme Urban Heat