In response to growing energy costs and carbon emissions, High Performance Computing (HPC) operators are looking for new strategies to reduce their footprint while maintaining operational efficiency. A direction is to optimize job scheduling in terms of energy and resource usage through the integration of new techniques such as speed scaling or power capping. However, the carbon footprint of an HPC platform can highly vary depending on when the jobs are executed because of the fluctuations of \(CO_2\) rates, which depends on the available energy resources and current demands. In this work, we explore the potential impact of incorporating \(CO_2\) rates in scheduling decisions. We introduce several policies that adjust dynamically in response to \(CO_2\) rate variations and expected job energy requirements. We propose two prediction models to help anticipate future \(CO_2\) rates evolution and potential energy consumption of submitted jobs. We evaluate these policies within a simulated HPC environment with real life job traces. Then, we compare them to the mostly used baseline scheduling policy First Come First Served (FCFS) with backfilling. We also compare them to other efficient policies in the literature. The resulting outcomes indicate that the proposed carbon aware scheduling techniques can achieve measurable reductions in carbon emissions (up to 15%) without compromising the workload throughput or resources usage. Furthermore, their transparency and simplicity makes them easy to implement in any HPC system with a high chance of acceptance by the users.