The large-scale integration of variable renewable energy sources into the energy system presents techno–economic challenges. Long–term energy system optimization models fail to adequately capture these challenges because of the low temporal resolution of these tools. This limitation has often been addressed either by direct improvements within the long–term models or by coupling them to higher resolution power system models. In this study, a combined approach is proposed to capitalize on the advantages and overcome the disadvantages of both methods. First, the temporal resolution of an energy model was enhanced by approximating the joint probability of the electricity load and the supply of intermittent sources. Second, the projected electricity mix was simulated by a power model at an hourly resolution. This framework was used to analyze mid–century deep decarbonization trajectories for Colombia, subject to future uncertainties of hydroclimatic variability and the development of the bioeconomy. The direct integration method is found to consistently reduce the overestimation of the feasible penetration of VRES. However, its impact is marginal because of its inability to assess the short–term operation of the power system in detail. When combined with the soft–linking method, the reliable operation of the power system is shown to incur an additional overhead of 12–17% investment in flexible generation capacity, 2–5% of the annual energy system cost, and a 15–27% shortfall in achieving the aspired GHG mitigation target. The results obtained by combining both methods are found to be closer to the global optimum solution than using either of these methods individually.