The physical and kinetic speciation of Cu and Zn in three impacted marine estuaries was examined. Contrasts in sources of metal-binding ligands, solution chemistry, and hydrologic forcing between and withinthethree study systems (Cape Fear River Estuary, North Carolina; Norfolk-Hampton Roads-Elizabeth River, Virginia; San Diego Bay, California) were exploited to enhance our understanding of Cu and Zn speciation. Trace metal-optimized tangential-flow ultrafiltration at 1 kDa nominal molecular weight limit (NMWL) was used to fractionate <0.4 microm species into colloidal and "dissolved" pools. Colloidal species of dissolved organic matter (DOM) and copper were significant and often the dominant pools in each of the three study systems. Characteristic colloidal fractions of both DOM and Cu ranged from near 70% of <0.4 microm concentrations in Cape Fear to 50% in San Diego Bay. Colloidal Cu and DOM were strongly coupled, and variability in observed <0.4 microm Cu concentrations was closely related to the concentrations of colloidal-associated metal. Colloidal fractions were much smaller for Zn than that of Cu; ranging from 10-30% in Cape Fear to less than 5% in San Diego Bay, and no relationship to DOM was observed. Kinetic separations on Chelex resin revealed the presence of large nonlabile pools of Cu in each of the study systems, with the highest fractions (70-100%) in Cape Fear and Norfolk and lowest (30-50%) in San Diego Bay. A close relationship was observed between colloidal and nonlabile Cu species, implying slow reactivity of colloidal-bound Cu. The fraction of filterable Zn labile to Chelex averaged 97%, 85%, and 60% in San Diego, Norfolk, and Cape Fear, respectively. Anthropogenic Zn appeared almost exclusively in the <1 kDa fraction, while anthropogenic Cu was distributed between dissolved and colloidal pools. Copper particle-partition coefficients (Kd) followed the trend: San Diego >> Norfolk > Cape Fear and were inversely correlated with DOC concentrations. Colloid-based partition coefficients were significantly greater, in many cases an order of magnitude greater, than particle-based partition coefficients. The partitioning data suggest the presence of metal-enriched bacterial-derived exudates and/or discrete metal phases in colloidal-sized particles in impacted regions of these estuaries. The strong relationships observed between Cu and DOC indicate that Cu partitioning behavior over a range of estuarine environments may be modeled effectively with a limited set of coefficients. Our measurements of metal lability and size distribution imply that the fraction of <0.4 microm Zn that is likely to be bioavailable is greater than that for Cu, especially in impacted regions of the study systems.