An accurate quantification of forest vertical transpiration (T) is essential for sustainable forestry in water-limited areas. This study established 72 temporary plots of larch plantations to characterise the site and vegetation attributes (i.e., elevation, slope aspect, stand age, and stand density) and to measure the stratified (i.e., tree layer, shrub layer, and herb layer) leaf area index (LAI). Additionally, three permanent plots were established to monitor stratified transpiration during the growing season of 2021 and 2022, together with reference evapotranspiration (ETo), relative soil water content (RSWC), and LAIs of each vertical layer. The results showed that the developed stratified LAI models, coupling elevation, slope aspect, age, density, and upper shading, could effectively capture layer relationships and site-stand influences. Accordingly, stratified T models incorporating these LAI effects were further developed. Integrating these models enabled quantification of vertical stand structure effects on T. Simulations across three permanent plots with varying site and vegetation characteristics revealed that maintaining an identical stand T (e.g., 0.8 mm & sdot;d- 1) required different stand densities across plots due to divergent site and stand attributes. Under climate change (e.g., 15 % ETo rise), site-specific LAI stratification became essential to maintain the target T. Although a 31 %-32 %, 27 %-28 %, and 15 %-16 % reduction in stand density achieved LAI control under current, 25 % reduced, and 50 % reduced soil water conditions, respectively, the optimal vertical distribution of LAI still varied significantly across plots, underscoring the need for precise, location-specific management strategies. This study elucidates how dynamics of vertical stand structure modulate forest transpiration under different site/vegetation conditions and provides a theoretical basis for the site- and stand-specific forest-water management.