Estimates of the total new production are similarly uncertain. It is striking that a synthesis of recent estimates of global annual new production (Table 1) shows an upward trend which has doubled the best estimate in the period covered by this review. This trend parallels the refinement and development of technique, accumulation of data, and deepening of understanding which has occurred over the past few years. The temporal and spatial variability of the ocean is so great as to preclude any reasonable expectation that observations alone will yield accurate global estimates of new production. Large scale syntheses must come from models which build on the results of the process and time series studies reviewed here [ Evans and Fasham, 1992; Fasham, 1993]. A coupled model of the physical, biological and nutrient fields observed in the 1989 JGOFS NABE assimilated satellite altimetric estimates of sea surface height to demonstrates the vital role of eddy circulation in driving new production [ Robinson et al., 1993]. Future models will assimilate ocean color observations. The first coupled basin scale models of ocean circulation and biogeochemistry which describe biogeochemical processes including new production explicitly have only been developed in the past 5 years [ Fasham et al., 1993; Sarmiento et al., 1993]. Their model has a simple generic ecosystem structure which nonetheless has over 20 parameters to be evaluated. A scheme based on the allometric scaling of biological parameters may simplify the parameter e valuation process [ Moloney and Field, 1991]. Inverse techniques borrowed from geophysics are now routinely applied to the analysis of biogeochemical systems to recover unmeasured or poorly estimated rates processes (e.g. new production, export, grazing) from observed data [ Jackson and Eldridge, 1992]. The next generation of biogeochemical flux studies must combine new observational and modeling approaches to gain the factor of 2 or greater improvements in precision required to constrain the global carbon budget.
Acknowledgments. I am grateful to the following individuals who provided references, manuscripts and reprints for my review: A. Alldredge, M. Bacon, N. Bates, K. Buesseler, C. Carlson, R. Duce, W. Gardner, J. Goldman, D. Hansell, J. Hedges, B. Huebert, C. Lee, J. Marra, A . Michaels, C. Winn. Supported by NSF OCE9116304 and OPP 9319222.