This study examines how parks and artificial lakes help cool down neighborhoods in hot urban areas, focusing on Dobson Ranch in Phoenix, Arizona. In June 2024, we collected over 274,000 air temperature readings near the ground (about 2 meters high) using a car equipped with sensors. These data were used to train a machine learning model that accurately predicts air temperature across the neighborhood.

We found that on June 16, 2024, the park was about 1°C cooler than the surrounding area during both midday (11:00 AM) and midnight (12:00 AM). The nearby artificial lake provided even stronger daytime cooling (around 2.4°C), but slightly warmed the area at night (by 0.3°C). The park's cooling effects also extended to nearby areas—up to 1.34°C cooler downwind during the day and 0.5°C at night within 50 meters.

Our analysis also revealed that using only satellite data to estimate heat exposure can be misleading. We found a weak relationship between satellite-based surface temperature and actual air temperature on the ground. We also highlight that car-based measurements may overestimate nighttime temperatures and underestimate daytime heat if timing is not carefully considered, which is important for accurate urban heat assessments.