Accurately partitioning evapotranspiration (ET) into evaporation (E) and transpiration (T) remains challenging, but is essential for understanding ecohydrological processes and sustainable management of water resources. This study aims to partition ET and investigate the land use change effects on ET components through a space-for-time substitution approach and an optimised isotopic tracing method on China's Loess Plateau. The soil samples up to 20 m deep were collected from farmland (F) and two vegetation types converted from F, that is, 13-year-old peashrub (P13) and 51-year-old apricot (A51). By using multiple isotopes (delta 2H, delta 18O and 3H), ET was efficiently partitioned to discuss the effects of land use change on soil water balance (SWB) components. Following the conversion of F to P13 and A51, soil water storage declined by 44% and 39%, respectively, while deep drainage was completely eliminated in both cases. ET partitioning revealed that P13 and A51 exhibited significantly higher ET (511.7 and 411.6 mm year-1, equivalent to 131% and 107% of mean annual precipitation) than F (372.5 mm year-1, equivalent to 96% of mean annual precipitation). Further, E slightly decreased in P13 and A51 by 10% and 7% (7.0 and 4.6 mm year-1); whereas T significantly increased by 48% and 16% (146.2 and 49.7 mm year-1), respectively. T dominated ET under the three land uses with T/ET ranging from 82% to 89%, which is explained by lower precipitation and greater tree age in the sample plots of this study area. The quantified SWB well revealed the mechanisms by which land use change affects soil hydrological processes. The findings can provide valuable insights for ecological restoration and water resource protection in arid and semi-arid regions.