The spatial-temporal scales on which environmental observations are made can significantly affect our perceptions of ecological patterns in nature. Understanding potential mismatches between environmental data used as inputs to predictive models, and the forecasts of ecological responses that these models generate are particularly difficult when predicting responses to climate change since the assumption of model stationarity in time cannot be tested. In the last four decades, increases in computational capacity (by a factor of a million), and the evolution of new modeling tools, have permitted a corresponding increase in model complexity, in the length of the simulations, and in spatial-temporal resolution. Nevertheless, many predictions of responses such as shifts in range boundaries are often based on coarse spatial and temporal data, for example monthly or yearly averages. Here we model the effects of environmental change on the physiological response of an ecologically and commercially important species of mussel, the fitness of which can have a cascading influence on ecosystem levels. Using a Dynamic Energy Budget (DEB) model integrated with climatic data produced from IPCC-A1B scenarios, we investigated the effect of temporal resolution of physical data on predictions of the growth and reproductive output of the mussel Mytilus galloprovincialis. We ran models using five different temporal scales, 6, 4, 3, 2 and 1. h (derived by interpolating between 6. h points), at 5 Italian locations in the Central Mediterranean Sea, for the period ranging from 2006 to 2009. Results from these models were further compared against the results from a DEB model that used hourly environmental data recorded at the five locations as inputs. Model outputs included estimates of life history traits relevant to ecological performance as well as parameters related to Darwinian fitness. Results showed that predictions of maximum theoretical shell length were similar regardless of which source of environmental data was used. However, while the DEB model using 1-h modeled data produced predictions of reproductive output very similar to those obtained using recorded (hourly) environmental data from the same time period, results using coarser resolution modeled data greatly underestimated reproductive output. Thus, the use of modeled weather data can yield predictions similar to those generated from measured data, but only when data are provided at relatively high frequency. Our results suggest that metrics of model skill can diverge significantly when physical outputs of climate models are applied to biological questions, and that the temporal resolution of environmental data can strongly alter predictions of biological responses to environmental change. © 2014 Elsevier B.V.

cited By 14