From Miami to Madison: Investigating the Relationship Between Climate and Terrestrial Net Primary Production

The ``Miami Model'' (Lieth, 1975) was the first global scale empirical model of terrestrial net primary productivity (NPP). This model used sparse observations to calibrate NPP based on the minimum of temperature and precipitation functions. The simplicity and relative accuracy of this model has led to its continued use. The development of improved techniques to measure NPP in the field (Gower et al. 1999, 2001; Malhi et al. 2002, Clark et al., 2001), and the expanded spatial and temporal range of observations have prompted this study, which reexamines the relationship of ecophysiological variables to NPP. A large reference dataset (n = 2268) of NPP field observations, including many from the Global Primary Production Data Initiative (Olson et al., 2001), were compiled for calibration and parameter optimization. We developed a rigorous statistical test for several paired climatic variables in order to investigate their relationship to terrestrial NPP. In addition to temperature and precipitation, we chose more robust pairs of independent climatic variables based on their ability to represent plant stressors such as heat and water stress. These variables included growing degree-days base 0, a water stress index, and average incident photosynthetic active radiation (PAR) for each day the average temperature is greater then 0°C. The water stress and PAR indices used in the study were calculated using a simple surface energy and water balance model (Prentice et al., 1993; Foley, 1994; Haxeltine and Prentice, 1996) using climate data from New (2002). Both linear and sigmoidal functional forms were chosen to relate the climatic variables to NPP. The model coefficients were determined by minimizing the least squared error between the observations and simulations. Calculated annual global NPP ranged from 50 - 60 Pg, well within the estimates of previous studies (i.e. Cramer et al. 1999). Spatial patterns of NPP were compared using biome averages as well as latitudinal averages, and the differences in model performances compared to observed data is discussed.