Mitigating climate change and its impacts is one of the sustainable development goals (SDGs) required by United Nations for an urgent action. Increasing carbon emissions due to human activities is the root cause to climate change. Telecommunication networks that provide service connectivity to mobile users contribute great amount of carbon emissions by consuming lots of non-renewable energy sources. Beyond the improvement on energy efficiency, to reduce the carbon footprint, telecom operators are increasing their adoption of renewable energy (e.g., wind power). The high variability of renewable energy in time and location; however, creates difficulties for operators when utilizing renewables for the reduction of carbon emissions. In this paper, we consider a heterogeneous network consisted of one macro base station (MBS) and multiple small base stations (SBSs) where each base station (BS) is powered by both of renewable and non-renewable energy. Different from the prior works that target on the total power consumption, we propose a novel scheme to minimize the carbon footprint of networks by dynamically switching the ON/OFF modes of SBSs and adjusting the association between users and BSs to access renewables as much as possible. Our numerical analysis shows that the proposed scheme significantly reduces up to 86% of the nonrenewable energy consumption compared to two representative baselines.