Stream runoff was simulated from 1996 to 2000 using XTOP_PRMS (coupling of TOPMODEL and Precipitation Runoff Modeling System) model under Modular Modeling System at Martinelli and Green Lake 4 catchments in Green Lakes Valley, Colorado Front Range. Two flowpaths determined by XTOP_PRMS model, surface flow (infiltration-excess overland flow) and subsurface flow, were compared against the flowpaths determined by mixing model using isotopic and chemical tracers. Three tracers (DOC, K/Si, and ??18O) were used in mixing model to identify four flowpaths, i.e., overland, upper soil horizon, lower soil horizon, and base flow. The results showed that the runoff simulation using XTOP_PRMS model is reasonably successful for Martinelli catchment (8 ha in drainage area). The Nash-Sutcliffe efficiency is 0.76. The t-test of two means for paired sample showed that the difference between the observed and modeled runoff was not significantly different at ??=0.05 at Martinelli catchment (n = 1611, p = 0.6). The flowpaths identified by XTOP_PRMS model matched the flowpaths determined by the tracer-mixing model reasonably well in magnitude, but poorly in pattern. The surface flow primarily occurred in the beginning of snowmelt at Martinelli as illustrated by the tracer-mixing model. Both runoff simulation and flowpath identification using XTOP_PRMS model were relatively poor at Green Lake 4 catchment, which has a drainage area of 220 ha. The runoff peaks observed in May and June were not captured in runoff simulation. The problem may be caused by poor understanding of behaviors of flowpath parameters and insensitivity of snowmelt to daily mean air temperature.