A pressing challenge in arsenic (As) adsorptive filtration is to decipher how the As atomic surface structure obtained in the laboratory can be used to accurately predict the field filtration cycle. The motivation of this study was therefore to integrate molecular level As adsorption mechanisms and capacities to predict effluent As from granular TiO2 columns in the field as well as its health impacts. Approximately 2,955 bed volumes of groundwater with an average of 542 μg/L As were filtered before the effluent As concentration exceeded 10 μg/L, corresponding to an adsorption capacity of 1.53 mg As/g TiO2. After regeneration, the TiO2 column could treat 2,563 bed volumes of groundwater, resulting in an As load of 1.36 mg/g TiO2. Column filtration and EXAFS results showed that among coexisting ions present in groundwater, only Ca(2+), Si(OH)4, and HCO3(-) would interfere with As adsorption. The compound effects of coexisting ions and molecular level structural information were incorporated in the PHREEQC program to satisfactorily predict the As breakthrough curves. The total urinary As concentration from four volunteers of local residences, ranging from 972 to 2,080 μg/L before groundwater treatment, decreased to the range 31.7-73.3 μg/L at the end of the experimental cycle (15-33 days).