A roof with high solar reflectance and high thermal emittance (e.g., a white roof) stays cool in the sun, reducing cooling power demand in a conditioned building and increasing summertime comfort in an unconditioned building. The high initial solar reflectance of a white membrane roof (circa 0.8) can be lowered by deposition of soot, dust, and/or biomass (e.g., fungi or algae) to about 0.6; degraded solar reflectances range from 0.3 to 0.8, depending on exposure. We investigate the effects of soiling and cleaning on the solar spectral reflectances and solar absorptances of 15 initially white or light-gray polyvinyl chloride membrane samples taken from roofs across the United States. Black carbon and organic carbon were the two identifiable strongly absorbing contaminants on the membranes. Wiping was effective at removing black carbon, and less so at removing organic carbon. Rinsing and/or washing removed nearly all of the remaining soil layer, with the exception of (a) thin layers of organic carbon and (b) isolated dark spots of biomass. Bleach was required to clear these last two features. At the most soiled location on each membrane, the ratio of solar reflectance to unsoiled solar reflectance (a measure of cleanliness) ranged from 0.41 to 0.89 for the soiled samples; 0.53 to 0.95 for the wiped samples; 0.74 to 0.98 for the rinsed samples; 0.79 to 1.00 for the washed samples; and 0.94 to 1.02 for the bleached samples. However, the influences of membrane soiling and cleaning on roof heat gain are better gauged by fractional variations in solar absorptance. Solar absorptance ratios (indicating solar heat gain relative to that of an unsoiled membrane) ranged from 1.4 to 3.5 for the soiled samples; 1.1 to 3.1 for the wiped samples; 1.0 to 2.0 for the rinsed samples; 1.0 to 1.9 for the washed samples; and 0.9 to 1.3 for the bleached samples.