A quantitative evaluation of the effects of the leaf area index (LAI) of the forest canopy on evapotranspiration and runoff from a forest is necessary for a deeper understanding of eco-hydrological processes and their mechanisms, with the aim to achieve integrated management of forests and water resources and regional sustainable development in semi-arid regions. In this study, the lumped eco-hydrological BROOK90 model was used to simulate the responses of hydrological processes to varying canopy LAI of a Larix principis-rupprechtii plantation in the small semi-arid watershed of Diediegou, in the northern region of the Liupan Mountains in Ningxia, under various conditions of annual precipitation (wet, normal, and dry years). Annual evapotranspiration, annual canopy interception, and annual forest transpiration were all significantly positively correlated with the canopy LAI of the plantation (R²>0.99, P<0.01), while both annual soil evaporation and annual runoff were significantly negatively correlated with canopy LAI (R²>0.99, P<0.01). All hydrological processes that we studied varied with canopy LAI as exponential functions regardless of the annual precipitation in the year of study. The response of each variable to canopy LAI was associated with a threshold canopy LAI. When the canopy LAI was below this threshold, the change in each variable with respect to LAI was rapid; however, when the canopy LAI was above the threshold, the change in each variable was measurably slow, and gradually stabilized. Additionally, the thresholds for canopy LAI differed with differences in annual precipitation. In general, the thresholds of canopy LAI for the variables studied were higher in wet years than in dry years, especially for canopy interception and soil evaporation. For example, the threshold of canopy LAI for annual evapotranspiration was 1.6, 1.5, and 1.3 in wet, normal, and dry years, respectively; the threshold of canopy LAI for canopy interception in dry years (4.3) was significantly lower than that in wet (6.2) and normal years (6.4), and the threshold of canopy LAI for soil evaporation was distinctly higher in wet years (4.3) than in dry (3.1) and normal years (3.0). Moreover, the rates of change of each variable with increasing LAI were higher in wet years than in normal and dry years, indicating that change in each variable was more dependent on canopy LAI in wetter years, but more limited by insufficient precipitation and soil water in drier years. Meanwhile, our results indicated that the effects of the decrease in evapotranspiration and increase in water yield in a given forest stand caused by a reduction in canopy LAI (such as by thinning of the leaf canopy) are less than expected, because the decrease in the rate of annual evapotranspiration was smaller than the rate of decrease of canopy LAI. For example, when canopy LAI reduced from 4.2 to 2.0 (a relative reduction of 52.4%) in a normal year with an annual precipitation of 418.5 mm, the annual evapotranspiration decreased from 357.2 mm to 333.9 mm (an absolute reduction of 23.3 mm and a relative reduction of 6.5%). The results of this study can provide guidance to scientific forest management in semi-arid regions.