This paper presents results from 34 years of the Glendhu Experimental Catchment Study, established in 1979 by the former New Zealand Forest Service in upland east Otago in New Zealand's South Island to determine the hydrological consequences of converting indigenous tussock grassland to plantation forestry. A traditional paired catchment approach was adopted; after a 2.5‐year pretreatment period, one catchment (GH2, 310 ha) was planted over two thirds of its area in Pinus radiata, and an adjacent catchment (GH1, 216 ha) was left in tussock as a control. The average annual reduction in water yield from the planted catchment between canopy closure in 1991 and 2013, compared with that in tussock, was 273 mm (33%). Annual water yields from the planted catchment continued to decline relative to the tussock catchment until 2010. Since then, the difference in annual water yields between the two catchments has narrowed. Ripping before planting caused some redistribution of the total streamflow from stormflow to baseflow. Following canopy closure, afforestation has reduced the low flow (Q95) by an average of 26% compared with the tussock catchment. Average peak flows for small events (2–5 L/s/ha) were reduced by 78%, but only by 37% for larger, less frequent storms (>15 L/s/ha), suggesting that peak flows during high magnitude storms are less dependent on the prevailing land cover.