This dataset contains monthly mean global atmospheric distributions of ozone mixing ratio and methane chemical loss rate (the largest sink for atmospheric methane is the hydroxyl radical, OH) from 105 model runs with three independent global chemistry climate models. The models include the Frontier Research System for Global Change version of the University of California, Irvine Chemical Transport Model (FRSGC/UCI CTM), the Goddard Institute for Space Studies Global Climate Model (GISS GCM) and the Community Atmosphere Model with Chemistry (CAM-Chem). All three models performed the same simulations for a one-year period (broadly representative of 2001 meteorology) under standardised conditions (40 TgN/yr surface NOx emissions, 5 TgN/yr lightning emissions, 500 TgC/yr biogenic isoprene emissions, 1760 ppb tropospheric methane). An ensemble of 105 simulations was performed that included a control run (run 0), a set of runs that were used to build Gaussian Process emulators (runs 1-80), and additional runs that were used to evaluate the emulators (runs 81-104). A spin-up of six months was performed for each run, and monthly mean model results were archived for the following 12 months. The meteorological conditions used were the same in each simulation.

The data supports the exploration of the sensitivity of tropospheric ozone and the chemical lifetime of methane in the troposphere (a proxy for the hydroxyl radical, OH) to eight variables: (1) NOx emissions from all surface sources (range: 30-50 TgN/yr), (2) NO emissions from lightning (range: 2-8 TgN/yr), (3) biogenic isoprene emissions (range: 200-800 TgC/yr), (4) dry deposition rates of all deposited species (range: +/- 80%), (5) wet deposition rates of all soluble species (range: +/- 80%), (6) atmospheric humidity as used in the chemistry scheme only (range: +/- 50%), (7) cloud optical depth (range: factor of 10), and (8) turbulent mixing in the planetary boundary layer (range: factor of 100). The design of the ensemble runs used a Latin Hypercube method to sample this eight-dimensional parameter space to achieve optimal coverage with only 80 simulations. A separate design was used to select an additional 24 simulations to evaluate the emulators built from the standard 80 runs. The specifications for each simulation (total annual emission rate for surface NOx, biogenic isoprene and lightning NO, and the scaling factors applied to native model dry and wet deposition rates, humidity, cloud optical depth, and boundary layer diffusion coefficient) are provided with the dataset.