The concept of biome, or plant functional type, is used frequently in earth system models to incorporate discrete local measurements of vegetation properties into the model grid framework. This approach has the underlying assumption that all plants within the same biome have the same behaviour in response to environmental factors. We test this assumption using a global process based model of leaf phenology based on the hypothesis that leaf gains and loses are a strategy for optimal carbon gain, which describes leaf seasonal cycles as a function of temperature solar radiation and soil moisture. To this purpose we use three different model parametrisations by setting parameter values to be either location or biome specific or a combination of the two. We then fit all three different models to five years of leaf area index (LAI) data from the MODIS satellite instrument using a Bayesian fitting algorithm. We show that the biome-wide parametrisation has a much higher error than the fully local model and is only able to explain 20% of the spatial variation in LAI seasonality compared to the local model which explains over 90% of the variation. However, a model parametrisation which has some biome wide parameter values and some location specific ones shows a much better fit, explaining up to 60% of the spatial variation in LAI mean and amplitude. This indicates that while certain phenological characteristics are common at the biome scale others are location specific. We attribute this specificity to local variations in nutrient availability which have not been explicitly included in the model and show that the biome assumption is only valid when taking into account such variations in the local conditions.