Rainfall volume is one of the important data in the hydraulic and disaster engineering. Inefficient rainfall data will induce a error estimation of rainfall volume. This study uses the Kriging model of regionalized variable to estimate the best adding position of the rain gauges. Assuming the best adding place is the place in which the minimum Global Kriging Variance (GKV) reaches after adding a virtual rain gauge. This study uses Chen-Yu-Lan stream to verify above assumption. At first, this study collected the historical rainfall data of every rainfall gauge in the watershed, and calculated the GKV before adding rain gauge; then the watershed is divided into equivalent mesh with the total 112 and the virtual rain gauge installs in the center of mesh to calculate the GKV after adding rain gauge. Finally, TM2 coordinates (232000, 2612000) is the best adding position owing to the GKV is minimum comparing with the other meshes in the watershed.