1.Nutrient over‐enrichment increasingly threatens global water resources. Stressor‐response studies specifically designed to identify levels of nutrients strongly associated with undesirable ecological conditions are needed to inform numeric nutrient criteria that protect inland waters. 2.Diatoms are important components of aquatic life, which support higher trophic levels and are sensitive to nutrient enrichment. We tested a framework that relies on stressor‐response modelling of phosphorus (P) enrichment and stream diatom assemblages across many field locations, multiple years and seasons within years, and under controlled experimental conditions to inform nutrient criteria development. 3.Diatom species composition was nonlinearly correlated with total phosphorus (TP) throughout the two‐year field study. This occurred despite temporal shifts in species composition between two hydrologically distinct years and over eight seasons. 4.Species assemblages on rocks transplanted from a low P stream to mesocosms representing a P enrichment gradient (8, 20 and 100 μg L−1) shifted into two groups over time. Species composition on rocks in low (20 μg L−1) and high (100 μg L−1) P mesocosms was consistent with assemblages at P‐enriched field sites, whereas rocks in control (8 μg L−1) mesocosms had significantly different species composition, consistent with low P field sites. Species composition on rocks transplanted from high P streams did not shift as dramatically and were not significantly different after exposure to different P treatments in mesocosms. 5.Threshold Indicator Taxa Analysis identified synchronous declines in several diatom species that culminated in assemblage thresholds associated with TP concentrations > 20 and 25 μg L−1 for 2006 and 2007 respectively. 6.Synthesis and applications. Diatom assemblages show consistent responses to nutrient enrichment despite temporal shifts associated with confounding factors common in stream ecosystems. Regulators should include diatom assemblage responses when developing numeric nutrient criteria. We present a framework that includes spatial, temporal and experimental components, and has broad applicability for use in different ecological settings to evaluate ecological endpoints and set limits for a variety of contaminants threatening freshwater ecosystems throughout the world.