Marine ecosystem indicators are sensitive to ecosystem boundaries and spatial scale

Time series indicators are widely used in ecosystem-based management. A suite of indicators is typically calculated for a static region or multiple subregions and presented in an ecosystem assessment (EA). These are used to guide management decisions or determine environmental status. Yet, few studies have examined how the spatial scale of an EA influences indicator behavior. We explore this question using the Northwest Atlantic continental shelf ecosystem (USA). We systematically divided the ecosystem at six spatial scales (31 unique units), covering spatial extents from 250,000 km(2) to 20,000 km(2). The same 22 indicators were calculated for each unit, assessed for trends, and evaluated as 31 independent EAs. We found that the detected signals of indicator trends depended on the spatial scale at which the ecosystem was defined. A single EA for the whole region differed by 23% (in terms of the 22 indicator trend tests) relative to ones for spatially nested 120,000 km(2) subunits, and by up to 36% for EAs at smaller scales. Indicator trend disagreement occurred because (most common) a localized trend was perceived as widespread, (common) a local trend was obscured by aggregating data over a large region, or (least common) a local trend switched direction when examined at a broader scale. Yet, there was variation among indicators in their scale sensitivity related to trophic level. Indicators of temperature, chlorophyll-a, and zooplankton were spatially coherent: trends portrayed were similar regardless of scale. Mid-tmphic level indicators (fish and invertebrates) showed more spatial variation in trends. We also compared trend magnitude and indicator values to spatial extent and found relationships consistent with scaling theory. Indicators at broad scales produced subdued trends and values relative to indicators developed at smaller spatial scales, which often portrayed `hotspots' of local abundance or strong trend. Our results imply that subsequent uses of indicators (e.g., determining environmental status, risk assessments, management decisions) are also sensitive to ecosystem delineation and scale. We suggest that indicators and EAs should be done at multiple spatial scales and complimented with spatially explicit analysis to reflect the hierarchical structure of ecosystems. One scale is not best, but rather we gain a new level of understanding at each scale examined that can contribute to management decisions in a multiscale governance framework characterized by goals and objectives with relevance at different scales.