Nitrate photolysis proceeds via two major channels at illumination wavelengths above 290 nm: NO3- + hv (+H+) --> NO2 + (OH)-O-. (1) and NO3- + hv --> NO2- + O(P-3) (2). A recent study determined the quantum yield of reaction 1 on ice by measuring NO2 production, but suggested their values might be lower bounds because of incomplete recoveries of NO2. We measured the quantum yield of pathway 1 using an alternate approach, i.e., by following the formation of (OH)-O-.. Our quantum yields for (OH)-O-. (Phi(OH)) at 263 K were independent of nitrate concentration and illumination wavelength (lambda > 300 nm), but were dependent upon pH. Values of Phi(OH) decreased from (3.6 +/- 0.6) x 10(-3) at pH 7.0 to (2.1 +/- 0.8) x 10(-3) at pH 2.0, where the listed pH values are those of the sample solution prior to freezing. Temperature dependence experiments (239-318 K; pH 5.0) showed that values of Phi(OH) in ice pellets and aqueous solutions were both well described by the same regression line, ln(Phi(OH)) = ln(Phi(1)) = -(2400 +/- 480)(1/T) + (3.6 +/- 0.8) (where errors represent +/-1sigma), suggesting that the photolysis of nitrate on ice occurs in a "quasi-liquid layer" rather than in the bulk ice. Our ice quantum yields between 268 and 240 K are 3-9 times higher, respectively, than Phi(1) values determined previously in ice. Applying our quantum yields to past field experiments indicates that nitrate photolysis can account for the flux of NOx from sunlit snow in the Antarctic and at Summit, Greenland, but that nitrate was only a minor source of the snowpack NOx measured during the Alert 2000 campaign in the Canadian Arctic. Additional calculations show that the photolysis of nitrate on cirrus clouds in the upper troposphere is a minor source of NOx that cannot account for the apparent underestimation of the ratio of NOx/HNO3 in current numerical models.