Student Research - Mission Earth

This study investigates how the Urban Heat Island Effect (UHIE) could be predicted in San Diegan professional and collegiate athletic fields using remote sensing satellite imagery to better improve athletic safety and health, while also contributing to the broader understanding of which UHIE parameters are most significant in its warming effects. Parameters were averaged at 25 fields for each August between 2016-2025 to create a machine learning model capable of predicting land surface temperature. Each field was assigned a 300 meter surrounding buffer to compare variations between the field and its surroundings. Data was analyzed on ArcGIS from USGS Sentinel 2 Level 2A and Landsat 8/9 images. Six predictive values were extracted: within the entire area of a site, NDVI (Normalized Difference Vegetation Index) and albedo were extracted; within the isolated field and its isolated buffer, NDMI (Normalized Difference Moisture Index) and NDBI (Normalized Difference Build-Up Index) were extracted. Land surface temperature (LST) at the isolated field and within the entire area of a site was extracted as a response variable. The random forest model was trained from the 250 extracted data points, using an 80/20 train to test ratio. The model was most accurate in predicting LST at real grass locations and concrete tennis court locations, while least accurate in predicting LST at synthetic turf fields. Despite the limitations of our data collection, our model achieved a robust R squared score of 0.56, highlighting that macro-level satellite remote sensing is a highly viable and accessible tool for screening regional microclimate risks.