Fracture network connectivity fundamentally controls subsurface fluid flow and rock mass behavior across spatial scales, yet determining the representative elementary volume (REV) remains a core challenge in geological system characterization. This study investigates scale-dependent connectivity through systematic analysis of natural outcrop data and artificial discrete fracture networks (DFNs). We implement a novel connectivity metric, C<inf>t</inf>, integrating both intra-cluster connectivity and inter-cluster interactions, and propose the Standard Deviation Stability Criterion (SDSC) for objective REV determination using second-order statistical measures. Analysis of 63 natural outcrop maps and various artificial DFN configurations reveals several key findings. First, fracture network connectivity exhibits pronounced scale-dependence with REV values approaching the same order of magnitude as the investigated systems, with mean REV values of 0.586 for natural outcrops and exceeding 0.2 for artificial networks. Second, preferential orientations increase REV requirements, particularly under stress conditions where only critically stressed fractures remain permeable, with fracture clustering further amplifying this effect. Third, in-situ stress conditions substantially increase REV requirements, with values nearly doubling when only critically stressed fractures remain active. Complete sealing creates the most challenging REV determination due to orientation selectivity, while partial sealing provides intermediate behavior by preserving orientation diversity. These findings demonstrate that obtaining representative volumes through conventional sampling presents fundamental limitations and provide critical insights for enhancing predictive models in subsurface engineering and environmental applications.