The uncertainty in simulations by a Global Biome Model (BIOME3) to alternative parameter values

The sensitivity of a global biome model (BIOME3) to uncertainty in parameter values was investigated by testing the model's sensitivity to minimum and maximum parameter values obtained from an extensive literature search. Simulations were conducted replacing the default parameter value by each of the maximum and minimum values determined from the literature. In doing so, the aim was to identify those parameters where the use of an alternate (observed) value leads to a significant change in the simulation of plant functional types at a global scale, in order to identify those which are functionally important to the model. BIOME3 was found to be insensitive to changes in the majority of its parameters, providing a generally sound foundation for confidence in model simulations. However, there was considerable sensitivity shown to over a quarter of the parameters. Three main types of parameters led to a change in plant functional types distribution relative to the control simulation: (i) parameters affecting the photosynthesis parameterization; (ii) parameters affecting the evapotranspiration parameterization; and (iii) root distribution which affected both parts of the model. The main causes of sensitivity were changes in the photosynthesis parameters leading to differential changes in plant functional type's net primary productivity. This caused a shift in the competitive balance between specific plant functional types or between C3 and C4 plant types, and a consequent change in their global distribution. Changes to the evapotranspiration parameters and root distribution similarly affected net primary productivity and soil moisture, and often led to shifts in the competitive balance between grass and trees. Changes in the value for several poorly known parameters produced substantial changes in the distribution of plant functional types, and reduced the κ‐statistic to a large degree, indicating areas of potential uncertainty in the model. This suggests that great care must be taken in prescribing values to these parameters and provides guidance on which parameters need further attention in observational work.