The number of variants can become high when discovering real-world processes. We observed, e.g., more than 5500 process variants discovered from a clinical process event log and 20000 variants for different product configurations in a manufacturing setting. When mining process variants, not only are the variants themselves of interest, but also their root causes, i.e., the reason why a specific variant emerges from a baseline or reference process model. These reasons range from the existence of infrequent behavior to variants that emerge due to the allocation of specific resources, e.g., an additional check task must be conducted if the initial check was performed by an assistant clerk. This work focuses on the discovery of such resource-driven root causes of process variants from process event logs based on association rule mining combined with an ILP-based rule selection approach. The output is a set of rules that describe the relation between resource allocation and a change pattern, e.g., inserting a task. Depending on the resource allocated to a task, a change pattern is applied to the baseline process, resulting in the associated process variant. This helps to identify root causes for process variants and leads to a de-cluttering of the discovered process models. The approach is prototypically implemented and applied to artificial and real-world event logs.