Measurement of surface visibility is important for the management of air quality, human health, and transportation. Currently, visibility measurements are only available through ground-based instrumentation, such as the Automated Surface Observing System (ASOS), and therefore lack spatial coverage. In analogy to the recent work of using satellite-based aerosol optical depth (AOD) to derive surface dry aerosol mass concentration at continental-to-global scale for cloud-free conditions, this study evaluates the potential of AOD retrieved from the MODerate Resolution Imaging Spectroradiometer (MODIS) for deriving surface visibility. For this purpose of evaluation the truncated (up to w16 km or 10 miles) and discrete (at the interval no less than 0.4 km or 1/4 mile) visibility data from hourly operational weather reports are not suitable, and the ASOS-measured one-minute raw surface extinction coefficient (bext) values have to be used. Consequently, a method for quality control on the bext data is first developed to eliminate frequent problems such as extraneous points, poor calibration, and bad formatting, after which reliable bext data are obtained to estimate the surface visibility. Subsequent analysis of the AOD and bext relationship on the East Coast of the United States reveals their average linear correlation coefficient (R) of 0.61 for all twelve (2000e2011) years of data at 32 ASOS stations, with the highest R value in summer and the lowest values in fall and winter. Incorporation of the Goddard Earth Observing System, Version 5 (GEOS-5) modeled vertical profile of aerosols into the derivation of visibility from AOD is evaluated for two methods, one scaling the modeled surface bext with the ratio of MODIS AOD to the modeled AOD, and another scaling the ratio of modeled AOD in the boundary layer to total columnar AOD with the MODIS AOD and assuming well-mixed aerosol extinction in the boundary layer. Analysis with three summers (2003e2004, 2006) of available GEOS-5 data and ASOS data reveals that the second method is superior, and generates a regression model that, after independent evaluation for summer 2005, is found to be statistically robust with R of 0.70 and a mean bias of 0.32 km in derived visibility. This study is among the first to demonstrate the potential of using satellite-based aerosol product over land to operationally derive surface visibility.