The use of highly reflective “cool” roofing materials can decrease demand for air conditioning, mitigate the urban heat island effect, and potentially slow global warming. However, initially high roof solar reflectance can be degraded by natural soiling and weathering processes. We evaluated solar reflectance losses after three years of natural exposure reported in two separate databases: the Rated Products Directory of the US Cool Roof Rating Council (CRRC) and information reported by manufacturers to the US Environmental Protection Agency (EPA)’s ENERGY STAR^® rating program. Many product ratings were culled because they were duplicative (within a database) or not measured. A second, site-resolved version of the CRRC dataset was created by transcribing from paper records the site-specific measurements of aged solar reflectance in Florida, Arizona and Ohio.

Products with high initial solar reflectance tended to lose reflectance, while those with very low initial solar reflectance tended to become more reflective as they aged. Within the site-resolved CRRC database, absolute solar reflectance losses for samples of medium-to-high initial solar reflectance were 2–3 times greater in Florida (hot and humid) than in Arizona (hot and dry); losses in Ohio (temperate but polluted) were intermediate. Disaggregating results by product type—factory-applied coating, field-applied coating, metal, modified bitumen, shingle, single-ply membrane and tile—revealed that absolute solar reflectance losses were largest for field-applied coating, modified bitumen and single-ply membrane products, and smallest for factory-applied coating and metal products.

The 2008 Title 24 provisional aged solar reflectance formula overpredicts the measured aged solar reflectance of 0–30% of each product type in the culled public CRRC database. The rate of overprediction was greatest for field-applied coating and single-ply membrane products and least for factory-applied coating, shingle, and metal products. New product-specific formulas of the form ρ_a′=0.20+β(ρ_i−0.20) can be used to estimate provisional aged solar reflectance ρ_a′ from initial solar reflectance ρ_i pending measurement of aged solar reflectance. The appropriate value of soiling resistance β varies by product type and is selected to attain some desired overprediction rate for the formula. The correlations for shingle products presented in this paper should not be used to predict aged solar reflectance or estimate provisional aged solar reflectance because the data set is too small and too limited in range of initial solar reflectance.