The evaluation of the ecotoxicological risk of nickel (Ni) in surface water is hampered by a lack of speciation data. Six surface waters were sampled and speciation of Ni(II) was measured by the Donnan membrane technique (DMT) combined with radiochemical determination of 63Ni. The free Ni2+ ion fraction in the dissolved (<0.45 microm) phase was determined at background Ni concentration ((4-8) x 10(-8) M) and at concentrations in the range of toxicity thresholds for the Ni sensitive species Cerodaphnia dubia (5 x 10(-8) to 2 x 10(-6) M). The free ion fraction ranged from 4 to 45% at background Ni and increased with increasing Ni concentration and water hardness and with decreasing pH. The equilibration time after addition of Ni2+ (3h-7d) did not significantly change the measured free ion fraction. Predictions of the Humic-Ion Binding Model WHAM (Windermere Humic Aqueous Model) VI overestimated the observed free Ni2+ fraction (median>two-fold), even when assuming that all dissolved organic matter (DOM) was present as fulvic acid (FA). The impact of several model parameters affecting the prediction of Ni speciation were evaluated, including the solubility product of Fe(OH)3, which affects the Fe competition for complexation by DOM. The best fit (R2=0.88) was obtained by increasing only the distribution term DeltaLK2, which modifies the binding strength of multi-dentate sites, to accommodate the observed dependence of free ion fraction on Ni concentration.