The effects of aggregating sub-grid land surface variation on large-scale estimates of net primary production

The use of large grid cell databases (1/2° to 5°) to drive nonlinear ecosystem process models may create an incompatibility of scales which can often lead to biased outputs. Global simulations of net primary production (NPP) often assume that bias due to averaging of sub-grid variations in climate, topography, soils, and vegetation is minimal, yet the magnitude and behavior of this bias on estimates of NPP are largely unknown. The effects of averaging sub-grid land surface variations on NPP estimates were evaluated by simulating a 1° × 1° land surface area as represented by four successive levels of landscape complexity, ranging from a single computation to 8,456 computations of NPP for the study area. Averaging sub-grid cell landscape variations typical of the northern US Rocky Mountains can result in overestimates of NPP as large as 30 %. Aggregating climate within the 1° cell contributed up to 50 % of the bias to NPP estimates, while aggregating topography, soils, and vegetation was of secondary importance. Careful partitioning of complex landscapes can efficiently reduce the magnitude of this overestimation.