A link connects two points. Examples would be an Ethernet interface (host to switch, switch to switch) or a similar WiFi link (host to access point). Link simulations explore the performance of that link. The simulations may need to be sophisticated. For example, in the case of wireless links, you might have to consider multipath distortion, MIMO, this sort of thing.
At the system level, you are considering the behavior of traffic flow among many nodes, splitting and merging, each flow using links to move from node to node. Now you would consider routing algorithms that prevent formation of loops, congestion at merge points and how this is controlled, jitter and latency end to end, and so on.
If I understand the question correctly, we aren't talking about two simulation "methods," but rather, we are talking about simulating entirely different things.
The colleague Albert explained it well. I only stress here that your question needs elaboration. While the communication link is well defined as an interconnection between two communicating nodes, the communication system in the physical sense is wide. Sometimes one refers to one link as a communication system composed of two transceivers and a communication medium.
Here you mean by a communication system communicating nodes interconnected by links according to certain interconnect pattern.A communication system like that is a communication network.
One of the very proper descriptions is the layered model of the network. In this network layered model there is the physical layer, then comes the link layer, then comes the network layer where network routing devices are added to that of the link layer.Then comes the higher layers which are called the software layers or the application layers. Every layer has its functions and protocols.