Similar interactions lead to the water--nitrogen--NPP correlations in Biome-BGC and TEM, although in these models the N inputs and outputs are not simulated explicitly as functions of precipitation.
In Biome-BGC and Century, allocation and hence biomass turnover (litterfall/biomass) varies allometrically between grid cells depending upon the amount of vegetation carbon, causing scatter within vegetation types.
Biome-BGC's NPP is the most highly correlated with MAP ([R.sup.2] = 0.95); the correlations in TEM and Century are not as strong ([R.sup.2]s of 0.78-0.79), indicating the tighter linkage of NPP to hydrology in Biome-BGC relative to the other models.
Differences at the CPER occur because the Century and Biome-BGC groups calibrated their models to represent an assumed areally dominant landscape position (unproductive uplands, NPP [approximately equals] 120 g/[m.sup.2]) whereas the TEM group assembled a carbon budget for calibration using the most complete data sets, collected in productive sites (NPP [approximately equals] 200 g/[m.sup.2]).
Total Annual NPP carbon storage Model ([10.sup.12] g C/yr) ([10.sup.15] g C) Biome-BGC 3772 118 Century 3125 116 TEM 3225 108 The VEMAP grid contains 3168 land grid cells, and so in displays like those shown in Fig.