The hydroxyl radical (OH) is the main oxidant in the troposphere and controls the lifetime of many atmospheric pollutants, including methane. Global annual-mean tropospheric OH concentrations ([OHoline ]) have been inferred since the late 1970s using the methyl chloroform (MCF) proxy. However, concentrations of MCF are now approaching the detection limit, and a replacement proxy is urgently needed. Previous inversions of GOSAT (Greenhouse Gases Observing Satellite) satellite measurements of methane in the shortwave infrared (SWIR) have shown success in quantifying [OHoline ] independently of methane emissions, and observing system simulations have suggested that satellite measurements in the thermal infrared (TIR) may provide additional constraints on OH. Here we combine SWIR and TIR satellite observations from the GOSAT and AIRS instruments, respectively, in a 3-year (2013–2015) analytical Bayesian inversion optimizing both methane emissions and OH concentrations. We examine how much information can be obtained about the interannual, seasonal, and latitudinal features of the OH distribution. We use information from MCF data and the ACCMIP ensemble of global atmospheric chemistry models to construct a full prior error covariance matrix for OH concentrations for use in the inversion. This is essential to avoid an overfitting of the observations. Our results show that GOSAT alone is sufficient to quantify [OHoline ] and its interannual variability independently of methane emissions and that AIRS adds little information. The ability to constrain the latitudinal variability of OH is limited by strong error correlations. There is no information on OH at midlatitudes, but there is some information on the NH/SH interhemispheric ratio, showing this ratio to be lower than currently simulated in models. There is also some information on the seasonal variation in OH concentrations, although it mainly confirms the variation simulated by the models.
