Environmental dimensionality simplifies biological predictability under complex global change

Predicting biological responses to global change is difficult because organisms experience simultaneous shifts in multiple environmental drivers. Although interactions among drivers can strongly influence biological performance, it remains unclear whether such complexity fundamentally limits predictability under highly multidimensional environmental change. Here, we test whether population growth can be predicted from either the number of environmental drivers or the single most detrimental driver, using a large-scale experiment with the marine diatom Phaeodactylum tricornutum. Populations were exposed to 127 environments differing from a standard condition by one to seven drivers, including warming, elevated CO₂, high light, nutrient limitation, and metal stress. Population growth declined systematically with increasing driver number, while extinction risk increased sharply in high-dimensional environments. Interactions dominated growth responses at low driver numbers, but predictability increased as environmental complexity rose. Across the full driver space, growth was best explained by the strongest individual driver present (cadmium), outperforming additive and multiplicative models. Our findings reveal emergent simplicity in population-level responses to complex environmental change and demonstrate that identifying dominant stressors provides a robust, scalable basis for forecasting biological performance under global change.