<p><strong>Abstract.</strong> We assess the major factors contributing to local biases in the hydroxyl radical (OH) as simulated by a global chemistry-climate model, using a single-column photochemical model (SCM) analysis. The SCM has been constructed to represent atmospheric chemistry at Lauder, New Zealand, which is representative of the background atmosphere of the Southern Hemisphere (SH) mid-latitudes. We use long-term observations of variables essential to tropospheric OH chemistry, i.e. ozone (O<sub>3</sub>), water vapour (H<sub>2</sub>O), methane (CH<sub>4</sub>), carbon monoxide (CO), and temperature, and assess how using these measurements affect OH calculated in the SCM, relative to a reference simulation only using modelled fields. The analysis spans 1994 to 2010. Results show that OH responds approximately linearly to correcting biases in O<sub>3</sub>, H<sub>2</sub>O, CO, CH<sub>4</sub>, and temperature. The biggest impact on OH is due to correcting H<sub>2</sub>O, using radiosonde observations. This is followed by correcting O<sub>3</sub>. Its impact is decomposed into a kinetics effect and a photolysis effect; both are of similar magnitude. The OH sensitivity to correcting CH<sub>4</sub> and CO biases is inversely related to the relative changes applied to these two species. The work demonstrates the feasibility of quantitatively assessing OH sensitivity to biases in longer-lives species, which can help to explain differences in simulated OH between global chemistry models and relative to observations. In addition to clear-sky simulations, we have performed idealised sensitivity simulations to assess the impact of clouds (ice and liquid) on OH. The results indicate that the impacts on the ozone photolysis rate and OH are substantial, with a general decrease of OH below the clouds relative to the clear-sky situation, and an increase above. The effects of liquid and ice clouds are less-than-additive. Using the SCM simulation we calculate recent OH trends at Lauder. For the period of 1994 to 2010, all trends are insignificant, in agreement with previous studies.</p>