We examine the sensitivity of the seasonal cycle of heterotrophic respiration to model estimates of litterfall seasonality, herbivory, plant allocation, tissue chemistry, and land use. As a part of this analysis, we compare heterotrophic respiration models based solely on temperature and soil moisture controls (zero-order models) with models that depend on available substrate as well (first-order models). As indicators of regional and global CO2 exchange, we use maps of monthly global net ecosystem production, growing season net flux (GSNF), and simulated atmospheric CO2 concentrations from an atmospheric tracer transport model. In one first-order model, CASA, variations on the representation of the seasonal flow of organic matter from plants to heterotrophs can increase global GSNF as much as 60% (5.7 Pg C yr(-1)) above estimates obtained from a zero-order model. Under a new first-order scheme that includes separate seasonal dynamics for leaf litterfall, fine root mortality, coarse woody debris, and herbivory, we observe an increase in GSNF of 8% (0.7 Pg C yr(-1)) over that predicted by the zero-order model. The increase in seasonality of CO2 exchange in first-order models reflects the dynamics of labile litter fractions; specifically, the rapid decomposition of a pulse of labile leaf and fine root litter that enters the heterotrophic community primarily from the middle to the end of the growing season shifts respiration outside the growing season. From the perspective of a first-order model, we then explore the consequences of land use change and winter temperature anomalies on the amplitude of the seasonal cycle of atmospheric CO2. Agricultural practices that accelerate decomposition may drive a long-term increase in the amplitude, independent of human impacts on plant production. Consideration of first-order litter decomposition dynamics may also help explain year-to-year variation in the amplitude.

In recent years, the chief approaches used to describe the terrestrial carbon sink have been either (1) inferential, based on changes in the carbon content of the atmosphere and other elements of the global carbon cycle, or (2) mechanistic, applying our knowledge of terrestrial ecology to ecosystem scale processes. In this study, the two approaches are integrated by determining the change in terrestrial properties necessary to match inferred change in terrestrial carbon storage. In addition, a useful mathematical framework is developed for understanding the important features of the terrestrial carbon sink. The Carnegie-Ames-Stanford Approach (CASA) biosphere model, a terrestrial carbon cycle model that uses a calibrated, semimechanistic net primary production model and a mechanistic plant and soil carbon turnover model, is employed to explore carbon turnover dynamics in terms of the specific features of terrestrial ecosystems that are most important for the potential development of a carbon sink and to determine the variation in net primary production (NPP) necessary to satisfy various carbon sink estimates. Given the existence of a stimulatory mechanism acting on terrestrial NPP, net ecosystem uptake is expected to be largest where NPP is high and the turnover of carbon through plants and the soil is slow. In addition, it was found that (1) long-term, climate-induced change in heterotrophic respiration is not as important in determining long-term carbon exchange as is change in NPP and (2) the terrestrial carbon sink rate is determined not by the cumulative increase in production over some pre-industrial baseline, but rather by the rate of increase in production over the industrial period.

Testing estimates of year-to-year variation in global net primary production (NPP) poses some challenges. Large-scale, multiyear records of production are not readily available for natural systems but are for agricultural systems, We use records of agricultural yields at selected sites to test NPP estimates produced by CASA, a global-scale production model driven by both meteorological data and the satellite-derived normalized difference vegetation index (NDVI). We also test estimates produced by the Miami model, which has underlain several analyses of biosphere response to interannual changes in climate. In addition, we test estimates against tree ring data for one boreal site for which data from both coniferous and deciduous species were available. The agricultural tests demonstrate that CASA can reasonably estimate interannual variation in production. The Miami model estimates variation more poorly. However, differences in NDVI-processing algorithms substantially affect CASA's estimates of interannual variation. Of the four versions tested, the FASIR NDVI most closely reproduced yield data and showed the least correlation with changes in equatorial crossing time of the National Oceanic and Atmospheric Administration satellites, One issue raised is the source of the positive trends evident in CASA's NDVI-based estimates of global NPP. The existence of these trends is consistent with potential stimulation of terrestrial production by factors such as CO2 enrichment, N fertilization, or temperature warming, but the magnitude of the global trends seen is significantly greater than suggested by constraints imposed by atmospheric fluxes.

We present a detailed investigation of the gross C-12 and C-13 exchanges between the atmosphere and biosphere and their influence on the delta(13)C variations in the atmosphere. The photosynthetic discrimination Delta against C-13 is derived from a biophysical model coupled to a general circulation model [Sellers et al., 1996a], where stomatal conductance and carbon assimilation are determined simultaneously with the ambient climate, The delta(13)C of the respired carbon is calculated by a biogeochemical model [Potter et al., 1993; Randerson et al., 1996] as the sum of the contributions from compartments with varying ages, The global flux-weighted mean photosynthetic discrimination is 12-16 parts per thousand, which is lower than previous estimates. Factors that lower the discrimination are reduced stomatal conductance and C-4 photosynthesis. The decreasing atmospheric delta(13)C causes an isotopic disequilibrium between the outgoing and incoming fluxes; the disequilibrium is similar to 0.33 parts per thousand for 1988. The disequilibrium is higher than previous estimates because it accounts for the lifetime of trees and for the ages rather than turnover times of the biospheric pools. The atmospheric delta(13)C signature resulting from the biospheric fluxes is investigated using a three-dimensional atmospheric tracer model. The isotopic disequilibrium alone produces a hemispheric difference of similar to 0.02 parts per thousand, in atmospheric delta(13)C, comparable to the signal from a hypothetical carbon sink of 0.5 Gt C yr(-1) into the midlatitude northern hemisphere biosphere, However, the rectifier effect, due to the seasonal covariation of CO2 fluxes and height of the atmospheric boundary layer, yields a background delta(13)C gradient of the opposite sign. These effects nearly cancel thus favoring a stronger net biospheric uptake than without the background CO2 gradient. Our analysis of the globally averaged carbon budget for the decade of the 1980s indicates that the biospheric uptake of fossil fuel CO2 is likely to be greater than the oceanic uptake; the relative proportions of the sinks cannot be uniquely determined using C-12 and C-13 alone. The land-ocean sink partitioning requires in addition, information about the land use source, isotopic disequilibrium associated with gross oceanic exchanges, as well as the fractions of C-3 and C-4 vegetation involved in the biospheric uptake.

We characterized decadal changes in the amplitude and shape of the seasonal cycle of atmospheric CO2 with three kinds of analysis. First, we calculated the trends in the seasonal cycle of measured atmospheric CO2 at observation stations in the National Oceanic and Atmospheric Administration Climate Monitoring and Diagnostic Laboratory network. Second, we assessed the impact of terrestrial ecosystems in various localities on the mean seasonal cycle of CO2 at observation stations using the Carnegie-Ames-Stanford Approach terrestrial biosphere model and the Goddard Institute for Space Studies (GISS) atmospheric tracer transport model, Third, we used the GISS tracer model to quantify the contribution of terrestrial sources and sinks to trends in the seasonal cycle of atmospheric CO2 for the period 1961-1990, specifically examining the effects of biomass burning, emissions from fossil fuel combustion, and regional increases in net primary production (NPP). Our analysis supports results from previous studies that indicate a significant positive increase in the amplitude of the seasonal cycle of CO2 at Arctic and subarctic observation stations. For stations north of 55 degrees N the amplitude increased at a mean rate of 0.66% yr(-1) from 1981 to 1995. From the analysis of ecosystem impacts on the mean seasonal cycle we find that tundra, boreal forest, and other northern ecosystems are responsible for most of the seasonal variation in CO2 at stations north of 55 degrees N. The effects of tropical biomass burning on trends in the seasonal cycle are minimal at these stations, probably because of strong vertical convection in equatorial regions. From 1981 to 1990, fossil fuel emissions contributed a trend of 0.20% yr(-1) to the seasonal cycle amplitude at Mauna Loa and less than 0.10% yr(-1) at stations north of 55 degrees N. To match the observed amplitude increases at Arctic and subarctic stations with NPP increases, we find that north of 30 degrees N a 1.7 Pg C yr(-1) terrestrial sink would be required. In contrast, over regions south of 30 degrees N, even large NPP increases and accompanying terrestrial sinks would be insufficient to account for the increase in high-latitude amplitudes.

To provide a common currency for model comparison, validation and manipulation, we suggest and describe the use of impulse response functions, a concept well-developed in other fields, but only partially developed for use in terrestrial carbon cycle modelling. In this paper, we describe the derivation of impulse response functions, and then examine (i) the dynamics of a simple five-box biosphere carbon model; (ii) the dynamics of the CASA biosphere model, a spatially explicit NPP and soil carbon biogeochemistry model; and (iii) various diagnostics of the two models, including the latitudinal distribution of mean age, mean residence time and turnover time. We also (i) deconvolve the past history of terrestrial NPP from an estimate of past carbon sequestration using a derived impulse response function to test the performance of impulse response functions during periods of historical climate change; (ii) convolve impulse response functions from both the simple five-box model and the CASA model against a historical record of atmospheric delta(13)C to estimate the size of the terrestrial C-13 isotopic disequilibrium; and (iii) convolve the same impulse response functions against a historical record of atmospheric C-14 to estimate the C-14 content and isotopic disequilibrium of the terrestrial biosphere at the 1 degrees x 1 degrees scale. Given their utility in model comparison, and the fact that they facilitate a number of numerical calculations that are difficult to perform with the complex carbon turnover models from which they are derived, we strongly urge the inclusion of impulse response functions as a diagnostic of the perturbation response of terrestrial carbon cycle models.