Removal of inorganic arsenic from As-spiked Lake Washington water by sorption onto activated alumina, iron-oxide-coated sand (IOCS), and an ion exchange resin was studied in laboratory systems. The sorption results could be rationalized only by invoking unanticipated changes in arsenic speciation between As(III) and As(V) in the feed reservoir. The changes in As speciation were subsequently confirmed in independent experiments, and the factors controlling the changes were investigated. Oxidation and reduction reactions sometimes occurred in the same system sequentially over a period of days to weeks, and in other systems both oxidation and reduction seemed to be proceeding simultaneously.

Because As(III) has a much lower affinity for the adsorbents than does As(V), arsenic breakthrough was much more rapid when the influent arsenic was in the reduced form. In some cases, the combination of speciation changes in the influent and differential sorption of As(III) and As(V) generated breakthrough curves for total arsenic in which the breakthrough increased and then declined for a period before increasing again.

Reduction of As(V) to As(III) was facilitated by filtration of the sample and by incubating the sample in the light, and it was impeded by the addition of PO4 to solution. Neither filtration nor exposure to light had a discernible effect on the rate of As(III) oxidation. As(III) oxidation was facilitated by autoclaving the sample either before or after the As(III) was added.

Although arsenic speciation changes are reproducible and explainable when investigating the behaviour of a given water sample under well controlled laboratory conditions, they are not easily extrapolated to other laboratory systems or field conditions, because they result from chemical and microbially mediated reactions that are strongly interrelated and evolve with time.