Research effort in IoT network architectures for new technologies is directed towards improving quality of service (QoS) or reducing energy consumption to extend networks lifetime. Simultaneously, the environmental impact of services is of crucial importance. In this paper, we propose a new approach to analyse the data transportation carbon footprint induced by different IoT architectures with the same infrastructure. The proposed method considers an end-to-end energy consumption model as well as the carbon emissions due to data transportation, which depend on the route taken by the data from the access network to the edge, and within the core network. We introduce this method by comparing three existing architectures: a centralized architecture, a semi-distributed architecture and a locally-distributed architecture. The performance evaluation is conducted using emulation tools on three scenarios: Unicast, Geocast and Broadcast. We used two metrics: the carbon footprint and the QoS in terms of delay. The obtained results show the benefits and concerns of each architecture evaluated.