5.1. Goodness of Model Fit
 Figures 7–10 display model results for the four chronosequence soils. At all sites, the model shows that SOM1 and SOM2 account for the high percentages of C and N in the upper ∼30 cm, as well as the deviation in contemporary Δ14C from prebomb values in this same depth range. The model also captures the general trend of the data for deep soil Δ14C values. Generally, the model matches the rapid transition in δ13C and δ15N from plant values near the soil surface to relatively stable deep soil values, which are 3–9‰ higher. The goodness of fit in these respects allows the model parameters (Tables 5a and 5b) to be compared across sites.
Figure 7. Measurements (crosses or asterisks) and model results (circles) for each horizon in <3 ky bulk soils. Q and R represent aboveground quadrats and belowground root cores, respectively. (a) 1997 % organic C. (b) 1997 %N. (c) 1949, 1978 and 1997 Δ14C. (d) 1949 and 1997 δ13C. (e) 1997 δ15N.
Download figure to PowerPoint
Figure 8. Measurements (crosses or asterisks) and model results (circles) for each horizon in 200 ky bulk soils. Q and R represent aboveground quadrats and belowground root cores, respectively. (a) 1997 % organic C. (b) 1997 %N. (c) 1949, 1978 and 1997 δ13C. (d) 1949 and 1997 δ13C. (e) 1997 δ15N.
Download figure to PowerPoint
Figure 9. Measurements (crosses or asterisks) and model results (circles) for each horizon in 600 ky bulk soils. Q and R represent aboveground quadrats and belowground root cores, respectively. (a) 1997 % organic C. (b) 1997 %N. (c) 1952, 1978 and 1997 Δ14C. (d) 1952 and 1997 δ13C. (e) 1997 δ15N.
Download figure to PowerPoint
Figure 10. Measurements (crosses or asterisks) and model results (circles) for each horizon in 1–3 My bulk soils. Q and R represent aboveground quadrats and belowground root cores, respectively. (a) 1998 % organic C. (b) 1998 %N. (c) 1949, 1978 and 1998 Δ14C. (d) 1949 and 1998 δ13C. (e) 1998 δ15N. Upper surfaces of hydrologically limiting duripans are marked in Figure 10c.
Download figure to PowerPoint
Table 5a. Model-Derived SOM Parameters
| ||Decomposition Rates||Transfer coefficents||Advective Transport Rates||SOM C:N ratios||Passive Relocation||Root e-fold Depth|
|1/k1 yr||1/k2 yr||1/k3 yr||t2||t3||v1 mm yr−1||v2 mm yr−1||v3 mm yr−1||C:N1||C:N2||C:N3||P gC m−2 yr−1||L cm|
Table 5b. Model-Derived Stable Isotope Parameters
| ||15N Enrichment During||13C Enrichment During|
|N1 N215εt2 ‰||N1 N315εt3 ‰||C1 C213εt3 ‰||C1 C313εt3 ‰||C3 Decomp 13εt3 ‰|
 The <3 kyr soil displays poorer model fits than many other soils, with a deep soil Δ14C discrepancy of ∼10–50‰ and problems with fits for percentages of C and N near the surface (Figure 7). The <3 kyr model fits δ15N well, and δ13C well except for one very low δ13C at depth. Most problems in the model fits at the <3 kyr site and can be attributed to deep root inputs or the uncertain depositional history of this floodplain site.
 The model produces excellent fits for Δ14C, and percentages of C and N in the 200 and 600 kyr sites (Figures 8 and 9), presumably because these sites meet the model's assumption of uniform transport as a function of normalized soil depth. The 200 kyr soil, and particularly the 600 kyr soil, display an increase in the percentages of C and N at depth in argillic horizons, which is not accounted for by the model. The model also fails to represent the deep Δ14C increase from the prebomb to the contemporary soil profiles shown in Figures 8 and 9, indicating that a process moving bomb-14C below the rooting zone is not accounted for. The model fits for δ15N are generally within expected variability for the data, while fits for δ13C are also good but indicate that the model could be improved for this tracer.
 In the 1 to 3 Myr soils (Figure 10), the model fits the data for percentages of C and N above the duripan, while the failure of the model to account for lower C and N concentrations below the duripan suggests that the modeled transport mechanisms do not effectively penetrate the duripan. The Δ14C data fits the model acceptably at the surface for the contemporary profiles and to ∼30 cm for the prebomb profiles. Since the fit is adequate in the horizons with elevated SOM contents, estimates of parameters for SOM1 and SOM2 are adequate, but confidence is lower in parameters describing SOM3, particularly v3. Given that measured δ15N values are highly uncertain due to low percentage of N below the duripan, the model achieves an excellent fit for δ15N, but fails to describe δ13C as well. Generally, most deviations from the model in the oldest soils can best be explained by the failure of the model to represent the complicated hydrology at this site. Nevertheless, the model sufficiently describes processes controlling the observed C, 14C, 13C, N, and 15N profiles to an extent that the cases in which the model fails to match empirical data provide an opportunity to examine processes not represented in the model. Some processes are examined below. Other processes, which might improve future models, include variation in decomposition rates (k1, k2, and k3) and transport rates (v1, v2, and v3) or transport mechanisms (e.g., diffusion versus advection) with depth.
5.2. Interpretation: Δ14C and Percentages of C and N
 We discuss Δ14C first because this tracer provides the most powerful description of SOM turnover and transport. In particular, the fine depth resolution of sampling in the two uppermost 1997–1998 horizons provides Δ14C data that are extremely sensitive to SOM1 and SOM2 turnover and transport parameters. The approximately 50‰ variation in contemporary A1 horizon Δ14C and >100‰ variation in contemporary A2 horizon Δ14C led to observed formation (t2) and turnover (k2) rates for SOM2 that differ by over a factor of 2 among the soils examined. Other model-derived parameters (t3, v1, v2, v3, L, C:N3, and P) also display large and systematic variation across the soils studied (see Tables 5a and 5b), but several sources of uncertainty must be considered when interpreting model results.
 In Figures 7–10, the percentages of C and N show that the contribution of passive SOM3 to total SOM is substantial throughout the soil profile. Since SOM3 ≅ total SOM below ∼30 cm*, the amount and depth distribution of deep SOM determine the size of SOM3 as well as its turnover rate (k3), formation rate (t3), and transport rate (v3). Therefore, extrapolation of the modeled passive SOM3 profile in the zone with negligible SOM1 and SOM2 plays an important role in obtaining estimates for the dynamics of SOM1 and SOM2 in the surface soil. The linear function of observed Δ14C versus depth below the rooting zone in the 1949–1952 200 and 600 kyr profiles allows the downward transport rate v3 to be set precisely, the 14C age of deep SOM is determined by the time since it was formed near the soil surface (after taking the small contribution of passive relocation into account). Values of t3 control the amount of OM transferred to the passive pool for transport below the rooting zone, while k3 determines the rate at which the observed percentages of C and N decline below the rooting zone.
 Recalling that passive SOM3 is negligible at the soil surface without passive SOM3 relocation, the model's consideration of passive SOM3 relocation improves model fits in all soils (Figures 7c–10c). Without this model component, modeled prebomb Δ14C values for uppermost horizons deviate strongly from the observed data: the model cannot include enough old SOM to match the measured surface Δ14C values. Nevertheless, the uppermost horizons still have lower Δ14C than the model predicts. The nature of passive SOM near the soil surface therefore remains an important source of uncertainty in the model. The discrepancy could be explained if the upper horizons of archived soils (ranging from 9–28 cm thick) do not include material equivalent to the contemporary A1 (0–2 cm) horizon. These samples would then contain less young FPOM than the model predicts. Indeed, this often appears to be the case for thick upper horizons of 1978 soils. For example, in the 600 kyr soil the model overestimates the Δ14C of the thick upper horizon. The measured sample matches the model curve (line) but does not match the model average (circle) for this horizon, which includes considerable high-Δ14C SOC near the soil surface. We assume that this problem accounts for the observed differences between the model and the archived samples for the upper horizons plotted in Figures 7–10.
 The model adequately represents the variation in percentage of C in the upper 50 cm of soil in all soils except the <3 kyr profile. The youngest profile (<3 kyr) shows considerably greater storage at all depths below the uppermost horizon. In this case, deeper inputs from the roots of oak trees may account for the difference in the pattern of SOM storage, although a second possible cause is recent burial of SOM-rich surface soil by alluvial sediments. Indeed, the dramatic increase in C:N ratio in the fourth horizon (A4) sampled from this profile could be indicative of greater root inputs or a buried surface horizon.
 Much of the failure to obtain perfect percentage of C fits can be attributed to inappropriateness of the assumption of exponentially decreasing root inputs with depth. This problem is particularly evident in the <3 kyr soil since roots extend much deeper than the e-folding depth, L, suggests. Depositional history could account for the variation in percentages of C and N at the <3 kyr site, but more realistic root input functions would improve the model at all sites. The most appropriate treatment might be two separate exponential root input functions; one representing roots associated with nutrient uptake concentrated near the surface, and a second extending to greater depths associated with water uptake. The single exponential root input function preserves simplicity and consistency in the model given limited data to constrain the partitioning of belowground litter in this study.
 The percentages of C and N at depths below 50 cm show variation related to soil clay content and hydrologically limiting layers, which the model cannot account for. The increase in percentage of C in the 200 and 600 kyr soils appears to result from increased affinity for SOM in the clay-rich layer. The larger increase in percentage of N observed in these two soils most likely results from both increased affinity for SOM, and ammonia fixation within the clay minerals. The assumption that transport continues through these soil layers appears good based on the agreement between the modeled and observed percentages of C and N, and Δ14C observations below the argillic horizon in the 200 kyr soil.
 In the 3 Myr profile, the model fit for percentages of N and C does not deviate from the data within the upper portions of the very thick argillic horizon (15–66 cm*). But observed percentages of C and N data do decrease by over 50% at the depth (66 cm*), where the duripan limits downward transport. The model does not account for this sharp decline in SOM concentrations; the drop results from the fact that SOM transport mechanisms do not effectively penetrate the duripan. Indeed, the extremely old Δ14C value observed in the upper duripan of the 1998 3 Myr profile (−890‰) indicates a radiocarbon age of nearly 18,000 years. Interestingly, the increasing Δ14C with depths below the hydrologically limiting layer in both the 1949 1 Myr profiles and 1998 3 Myr profile appear to indicate that some SOM is transported through macropores in the duripan and disperses in the deeper soil.
 Since transport does not continue directly through the duripan as the model suggests, the fate of the SOM not moved below the duripan must be considered. In the absence of leaching, much of the water that ponds on the duripan during the rainy season most likely leaves the profile at the beginning of the dry season as evapotranspiration. The resulting upward transport of soil solution likely carries with it dissolved and colloidal OM. The combination of downward transport early in the rainy season, and upward transport at the beginning of the dry season, would have a net result resembling diffusion. Indeed, the linear alignment of contemporary Δ14C in this 3 Myr profile supports the concept of diffusive mixing of SOM as modeled by Elzein and Balesdent .
 While the model describes many aspects of the soil profile well, uncertainty in the Δ14C value of plant inputs (Table 4b) must be considered when quantitatively interpreting model parameters. The model assumes that plant inputs possess the Δ14C value of the winter atmosphere derived from clean air sites. This assumption is not problematic for aboveground inputs except at the <3 kyr site where inputs from oaks may have higher Δ14C values because the trees access the water table, allowing them to remain photosynthetically active during the dry summer period when atmospheric Δ14C is 10–20‰ higher. This difference is not accounted for in the model. Moreover, belowground >2-mm material measured at the study sites differs from inferred atmospheric values by up to ∼50‰. This difference would be consistent with a lag time for belowground inputs, or a fraction of 10- to 30-year-old coarse POM in upper soil horizons. In addition, contributions of 14C-enriched CO2 during the midwinter months cannot be ruled out as extensive burning of prunings from nearby orchards occurs concurrently with atmospheric inversions.
 Tests of model sensitivity to Δ14C lag times of up to 5 years between atmospheric CO2 and active SOM do not improve model fits (not shown) and suggest that k2, k3, t1, and t2 do not react significantly to lag times. However, the same tests suggest that k1 must be considered highly uncertain. The low sensitivity of most parameters to the isotope ratio of plant inputs results from the presence of the passive C relocation term (P), which allows fitting of model Δ14C estimates to measured Δ14C data for the upper ∼10 cm*. Following adjustment of P, values of k2, k3, t1, and t2 react sensitively to the Δ14C difference between the A1 and A2 horizons, but not the absolute Δ14C value of these horizons. Generally, the A1/A2 Δ14C difference reflects the high proportion of FPOM in the A1 horizon and mineral-associated SOM fractions in the A2 and deeper horizons [Baisden et al., 2002].
 In summary, the model appears to successfully fit percentages of C and N, and Δ14C data in the 200 and 600 kyr profiles. The uniform soil texture above the argillic horizons in these profiles appears to fit the model assumptions of uniform transport rates quite well, allowing the model to be fit to observations with a high degree of precision. The 3 Myr and <3 kyr soils provide more difficult cases for model fitting, and suggest cases in which soil hydrology, depositional history, and differences in plant input functions may alter SOM turnover and transport. Given uncertainty in the transport mechanism and depth distribution of belowground plant inputs, some parameters such as transport velocities (v1, v2, and v3) and the e-folding depth of root inputs (L) may not be quantitatively comparable between all sites. Nevertheless, these <3 kyr and 3 Myr profiles are interpretable due to the power of the Δ14C as tracer, good sampling resolution, and the opportunity for comparison to results obtained from the more uniform 200 and 600 kyr profiles.
5.3. Interpretation: δ13C and δ15N
 The SOM turnover and transport model, once calibrated to percentages of C and N, and Δ14C, offers an opportunity to use observed δ13C and δ15N of SOM to interpret decomposition processes within soil profiles. This is particularly true in the 200 and 600 kyr soils where the model fits for δ13C and δ15N are excellent, providing a semiindependent indication that the model represents SOM dynamics with fidelity.
 In a few cases for δ13C, a small but highly uncertain fractionation factor, 13ε3, suggests fractionating losses from the passive SOM3 pool. A similar fractionation factor, 15ε3, was not included for N losses from SOM3. It is plausible that either (1) the variability in nitrogen isotope data prevents detection of an N isotope fractionation (15ε3) associated with marginally significant observed C-isotope fractionation (13ε3); or (2) there is no significant fractionation in the loss pathway of N (and possibly C) from SOM3. No observed fractionation associated with loss from SOM3 implies either a nonfractionating pathway of loss (such as DOM) or a pathway of loss in which the rate-limiting mechanism is downstream of SOM3 by one or more short-lived reservoirs. Fractionation during decomposition from SOM3 could occur since losses from this pool are most likely highly incremental and the “reactant” remains preserved in soil. Nevertheless, fractionation during decomposition of passive SOM appears to be small and uncertain.
 As Figures 7–10 show, the δ13C and δ15N model fits were highly successful at the 200 and 600 kyr sites, and also successful for δ15N at the <3 kyr and 3 Myr sites. The model appears to be appropriate at the <3 kyr site for δ13C, except for one data point, which appears to result from root inputs or a buried surface horizon. At the 1 to 3 Myr site, δ13C displays a linear trend with the increase in δ13C extending much deeper than the model predicts. This difference between the model and observations at the 1 to 3 Myr site could be accounted for by the same diffusive mixing of dissolved and colloidal OM, which appears to cause deviations from the models predicted behavior for Δ14C.
 The success of the model for stable isotopes supports the conceptual model presented by Baisden et al.  that 13C and 15N enrichment occurs during the preservation of C and N in mineral-associated (stabilized and passive) SOM. The model-derived fractionation factors (Table 5b) suggest that the greatest fractionation occurs during passive SOM formation while lesser and variable fractionation factors apply to stabilized SOM formation. Presuming that passive SOM is more intimately associated with mineral surfaces than stabilized SOM, the relative size of these two fractionation factors agrees with the conclusion of Baisden et al.  that 13C and 15N enrichment is positively correlated with the degree of mineral-association.
 Most importantly, the good fits obtained for δ13C and δ15N based on a well-constrained multipool Δ14C-transport model allow some modifications and clarification of the concepts presented in previous models for stable isotopes in SOM [Brenner et al., 2001; Evans and Ehleringer, 1993; Feng et al., 1999; Martinelli et al., 1999; Nadelhoffer and Fry, 1988]. Generally, these models assume a single homogenous pool of SOM and cannot constrain the transport of OM within the soil profile. Most models invoke Rayleigh distillation, in which a “batch” of SOM (the reactant) undergoes 13C and 15N enrichment due to loss of a 13C and 15N depleted product. As a result, no distinction exists for SOM preserved in stabilized or passive SOM pools. The more detailed model presented here is no longer consistent with a “batch” distillation; all SOM pools have simultaneous inputs and losses, and SOM1 has several loss pathways. This distinction may be minor from the point of view of many modeling exercises, but from a mechanistic modeling standpoint the distinction can be important. If fractionation occurs during the transfers to preserved pools, then neither a single fractionation factor nor source δ15N values can be inferred based on a simple Rayleigh fractionation relationship, commonly plotted as the logarithm of concentration versus δ values [e.g., Evans and Ehleringer, 1993]. Here, the observed δ15N and δ13C values for the soils studied are approximately linear with respect to the logarithm of concentration, but the model structure suggests that this relationship is merely a coincidence. Nevertheless, one pool SOM models may still represent simplified ecosystems appropriately when a principal purpose of the model is to describe the plant-soil difference in δ values [Brenner et al., 2001], a difference dominated by the 15N and 13C enrichment of SOM2 and SOM3, which are large relative to SOM1 in most soils.
 It is also important to note that deviation of real transport processes (i.e., advective versus diffusive) from those modeled can cause incorrect estimation of fractionation factors [Amundson and Baisden, 2000]. Additionally, interannual variation in plant input δ13C or δ15N and depth-dependence of δ13C or δ15N in root inputs could cause errors in estimated fractionation factors. Given plausible differences in transport processes between sites, and limited observations of up to ∼2‰ interannual and depth-dependent variation at the sites, we do not compare the fractionation factors (Table 5b) across sites.
 Nevertheless, the success of the model for stable isotopes suggests that the information captured in stable isotopes can be represented successfully in three-pool models of SOM dynamics, perhaps allowing constraint of the depth-dependence of SOM pools. There is one caveat however: the model suggests that most fractionation occurs during the creation of the large passive SOM3 pool, while much smaller differences appear in δ13C and δ15N between the two faster cycling pools, which govern most ecosystem processes. Nevertheless, the passive SOM3 pool is large and poorly understood. Given the large differences inferred in δ13C and δ15N between passive and faster cycling SOM pools, stable isotopes could be useful for investigating the potential for changes in this poorly understood SOM pool.
5.4. Implications for SOM Dynamics as a Function of Soil Age
 Application of an SOM turnover and transport model to a soil chronosequence presents the opportunity to examine the relationship between the model-derived parameters and soil age, as well as results obtained from a semiindependent method. Figure 11a confirms that the transport model (presented here) and the semiindependent method presented in the companion paper [Baisden et al., 2002] yield similar turnover rates and pools sizes for three pools of SOC. Using Δ14C from archived and contemporary samples permits particularly good constraint of decadal stabilized (C2) SOM turnover rates and pool sizes based on incorporation of the bomb 14C-spike. Pool sizes for active and passive SOM pools appear reliable, but turnover rates for these pools must be considered approximate.
Figure 11. (a) Comparison of pool sizes and stabilized pool residence times derived from the transport model (TM) and density fractionation (DF) methods for surface soil horizons. (b) Total soil N in active, stabilized and passive pools from the transport model to infinite depth. (c) Total soil organic C in active, stabilized and passive pools from the transport model to infinite depth. For Figures 11a–11c, all pool sizes are based on 1997–1998 profiles only.
Download figure to PowerPoint
 The estimated stabilized (C2) SOM turnover rates agree nearly perfectly in the 200 kyr soil, and display a similar pattern across the 600 kyr and 3 Myr soils but differ between methods by up to 8 years. Stabilized SOM resides longest in the 200 kyr soil, with increasing rates of stabilized SOM turnover in the 600 kyr and 3 Myr soils. Comparison of values cannot be made for the <3 kyr soil since the density fractionation procedure was not performed for Δ14C. Based on the transport model only, the youngest soil (<3 kyr) displays intermediate stabilized SOM residence times, and a different partitioning of SOM, with much more stabilized SOM in the surface horizons. Within the older three soils, the size of the stabilized C2 pool appears to be roughly constant within the ∼0- to 10-cm depth increment (Figure 11a), but clearly decreases with soil age on a whole soil basis (Figure 11c).
 Inspection of Figure 11c suggests that observed changes in the whole soil C turnover rate (flux/pool size; based on Figures 2b and 2c) appear to result from changes in the partitioning of SOM between pools with different residence times. Within the 200 kyr to 3 Myr age range, older soils tend to have a smaller proportion of active and stabilized SOM relative to passive SOM. Despite this, at least one reservoir, the stabilized SOM pool displays decreasing residence times as soil age increases. Therefore, examination of the sizes and turnover rates of the three SOM pools can be more appropriate than discussion of the apparent decreases in whole soil C turnover rates.
 The results presented above and in Tables 5a and 5b suggest systematic variation in a number of model-derived parameters as a function of soil age, particularly within the 200 kyr to 3 Myr age range. While some model-derived parameters (ε's, P, and L) are too uncertain given the poorer 1 to 3 Myr model fits to permit substantial comparison across soil age, certain parameters do vary as expected given increases in clay content and development of water-limiting horizons with soil age. In particular, systematic decreases in transport rates (v1, v2, and v3), and increases in stabilized and passive SOM formation rates (t2 and t3) may be linked to increases in clay content (Table 1 and Appendix A).
 Conversely, increases in the modeled turnover rates of stabilized SOM (k2) (Table 5a) and decreases in stabilized SOM pool size (Figures 11b and 11c) with increasing soil clay content appears to conflict with the accepted understanding of the role of clay in SOM storage and turnover [Burke et al., 1989; Oades, 1995; Parton et al., 1987]. The 200 kyr site, which seems ideal for plant growth in many respects, has a large pool of SOM2 that turns over twice as slowly as the SOM2 pool at the oldest site (k2 values in Tables 5a and 5b). Yet at the oldest site, where soil pH dips below 4 (Appendix A), soil texture and hydrology seem more prone to create a large SOM2 pool with a very slow turnover rate, but do just the opposite. Could the increased turnover and decreased pool size of stabilized SOM2 result from increased demand for nutrients in “impoverished” older soils, as suggested by the conceptual model for stabilized SOM turnover in the work of Baisden et al. ?
 Estimating the turnover of a limiting nutrient is one of the most critical controls in many ecosystem biogeochemistry models, so the use of the Δ14C model to obtain independent empirically derived estimates of N turnover has considerable value. SOM turnover implies the mineralization of organic N into plant available forms. In turn, plant available N often constrains plant productivity and therefore the magnitude of inputs to the SOM pool. We calculate the N turnover rates presented in Figure 12 as the result of multiplying the pool sizes (calculated from equations (A4)–(A6) in Appendix A; with modifications for N as per Tables 3a and 3b) by their turnover rates, k1, k2, and k3. As required by mass balance, the total N turnover is equal to the values measured as inputs from quadrats (Table 4a and Figure 2c). More importantly, the youngest and oldest sites show higher proportional contributions to N availability from the slower cycling SOM2, than do middle aged sites. Given the hypothesis that stabilized SOM can be decomposed to release nutrients following the disruption of the soil structure stabilizing this SOM pool by growing plant roots or filamentous microorganisms [Baisden et al., 2002], the model-derived results indicating faster SOM2 in older sites could plausibly be explained if greater demand for N exists in older sites.
Figure 12. Nitrogen turnover (mineralization) calculated as the size of each N pool multiplied by its turnover rate. Turnover from N3 is negligible (1% of total or less). Pool sizes are based on 1997–1998 profiles.
Download figure to PowerPoint
 Ultimately, greater plant demand for N in older soils must result from increases in N losses or decreases in rates of N supply from the atmosphere. Figure 13 presents several lines of evidence for changes in N supply and loss as a function of soil age in study area. Evidence for increased N losses is limited, but straightforward. Figure 13b presents lysimeter data collected near the study sites [White, 1995], which shows increasing soil solution NO3 concentrations with soil age. The large pools of NO3 in the older soils are subject to loss via leaching or denitrification [Paul and Clark, 1996]. In support of this, loss parameters calculated for each soil in Baisden and Amundson  (based on model parameters in Table 5a) suggest the highest rate of loss in the oldest soil.
Figure 13. Relationships between N, P and soil age. (a) Counts of mineral apatite (the primary P mineral) and pH versus soil age from Harden . Low pH may increase P occlusion in Fe oxide minerals. (b) Lysimeters in the vicinity of the field sites display decreasing phosphate and increasing nitrate soil solution concentrations. (c) Greenhouse fertilization experiments conducted using similar soils from Madera County (∼20 km south of Merced) display limitation by N alone in young soils, but increasing levels of N&P colimitation in older soils. Experimental data from Ripperdan, San Joaquin and Redding soils by J. Vlamis reported by Weir . Fertilization equivalent to 220, 330 and 110 kg/ha of N, P and K, respectively.
Download figure to PowerPoint
 Examination of changes in N input rates requires knowledge of P dynamics, which exists for chronosequences of comparable length [Chadwick et al., 1999; Vitousek et al., 1997; Vitousek and Farrington, 1997; Walker and Syers, 1976]. Figure 13 presents available data summarizing P dynamics in the study area, while Figure 14 presents hypothetical relationships among P, SOM dynamics, and soil age based on Figure 13 and the model of Walker and Syers  modified to include atmospheric deposition [Newman, 1995; Vitousek and Farrington, 1997; Walker and Syers, 1976].
Figure 14. Hypothesized long-term control of SOM turnover and storage by P dynamics. Modified from Walker and Syers  to include atmospheric deposition of P, and consider observed reduction in P availability in very old soils due increased occlusion at low pH. See text for discussion.
Download figure to PowerPoint
 Researchers have hypothesized that long-term limitation by rock-derived nutrients such as P may limit symbiotic N-fixation, particularly in older soils [Schlesinger, 1997; Walker and Adams, 1958]. Symbiotic N-fixers are present in California annual grassland ecosystems and respond vigorously to experimental P and S additions [Arkley, 1962; Jones and Woodmansee, 1979]. Figure 13a shows that stocks of weatherable P (apatite) in the chronosequence soils decrease markedly during the first ∼100 kyr of soil development. Much of the apatite remaining in >200 kyr soils may be contained within more slowly weathering minerals such as feldspars [Syers and Walker, 1969; White, 1995]. Additionally, low pH in very old soils may permit consumption of available P by occlusion within iron oxides, which are abundant in older soils [Harden, 1987; McBride, 1994; White et al., 1996].
 Given decreased rates of P input from weathering (Figure 14a) and increased rates of P occlusion, decreased P availability can be expected in older soils (Figure 14c) provided that P inputs from atmospheric sources is small or constant (Figure 14a) [Newman, 1995]. In support, data from fertility experiments (Figure 13c) performed on chronosequence soils collected just south of the intensively studied area suggests that N availability alone limits plant production early in soil development, but N and P availability colimit production in older soils [Weir, 1956]. Recent conceptual models based on investigations of fixer/nonfixer competition [Vitousek and Field, 1999] and plant-mycorrhizal interactions [Grogan and Chapin, 2000] elucidate mechanisms that likely lead to limitation of symbiotic N-fixation, and therefore N availability, by P availability in older soils.
 If N-fixation in the oldest soils is limited by decreasing P availability, and N losses are equal to or greater than those in other sites, the hypothesis of increasing N-demand fueling increasing rates of stabilized SOM turnover in older sites appears reasonable. Despite changes in N demand, and stabilized SOM turnover, total SOM storage may remain relatively constant or even increase considering that C:P ratios in SOM and losses from the SOM pool are not thought to remain relatively constant as is the case for C:N ratios [Neff et al., 2000; Schlesinger, 1997; Tiessen et al., 1982]. Overall, while the interpretation of P dynamics across this chronosequence is based primarily on published results from other chronosequences, and constrained by limited data from the field area, P availability appears to represent an important control on ecosystem C and N dynamics, which merits further examination.
5.5. Implications for General Understanding of SOM
 The transport model points toward two areas in which improved understanding of SOM dynamics are needed. First, uncertainty remains in the estimates of passive SOM3 near the soil surface; the model does not exactly match measured Δ14C values in the uppermost horizons of 1949–1952 profiles. The difference between the prebomb Δ14C data and the model estimate could be accounted if nonuniform sampling did not include the OM-rich surface soil (∼0–2 cm) completely. Alternatively, the difference between modeled and measured 1949–1952 surface horizon Δ14C could be accounted for by charcoal in these soil layers. Owing to the importance of partitioning SOM between pools with different residence times, better understanding of the sources of passive SOM near the soil surface is needed.
 Second, the apparent contribution of bomb-derived C to depths well below 50 cm suggests that DOC transport represents an important process in annual grassland ecosystems. Many researchers have sensibly dismissed the importance of DOC transport in nonhumid systems [e.g., Feng et al., 1999]. Yet mechanisms for DOC production exist that may actually be specific to seasonally dry systems: in response to osmotic shock during rapid wetting events, bacteria may either burst or release intracellular proteins, sugars and amino acids into soil solution to reduce intracellular water potential [Halverson et al., 2000; Kieft et al., 1987].
 Considering the large reservoir of deep SOC [Jobbagy et al., 1999] and the fact that DOM represents an important fate for nutrients in undisturbed systems [Hedin et al., 1995], these estimates support suggestions that decadally cycling SOM could represent significant fluxes in long-term ecosystem C and N budgets [Neff and Asner, 2001]. Raymond et al.  support these suggestions with evidence that riverine DOC has a decadal residence time in terrestrial ecosystems. If as suggested by results for the 200 kyr soil derived from Figure 5, ∼24% of the C in the 50- to 150-cm depth interval turns over on timescales as fast as 20 years, this would represent a flux of ∼23 g m−2 yr−1. Extrapolating to the global area of temperate grasslands (8 × 108 ha) [Schlesinger, 1997], this becomes a flux of 0.2 Pg C yr−1. Further extrapolating across all terrestrial ecosystems, the natural cycling of deep SOM could approach the magnitude of uncertainty and imbalance in the global C budget (1–3 Pg C yr−1). Given the extent of land-use change and alteration of the global N cycle, both of which could alter rates of DOC production [Aitkenhead and McDowell, 2000; Currie et al., 1996; Goodale et al., 2000], the resulting disequilibrium in deep soil C could warrant consideration in global and regional C budgets. Clearly, this result merits further consideration and research.
 Despite these remaining uncertainties, the turnover and transport model we developed for multiple isotope systems in SOM was highly successful, particularly in soils with uniform soil texture and straightforward hydrology. The model allows the empirical calculation of residence times and transport rates for three pools of SOM in a manner that is consistent with most ecosystem biogeochemistry simulation models, and may be useful for including soil depth a working model component. The model agrees with estimates of SOM turnover times obtained from soil density fractions isolated from the same soils, and appears to be consistent with the finding that SOM residence times in the slow pool are less variable than C:N ratio or stable isotope ratios [Baisden et al., 2002]. Finally, the model suggests that three pools of SOM permit a realistic treatment of soil δ13C and δ15N, and that observed SOM 13C and 15N enrichment results primarily during the transfers of SOM from the fast cycling SOM1 to passive SOM3, and to a lesser and more variable extent the slow cycling SOM2. Thus stable isotope fractionation appears to occur primarily during preservation of stabilized and passive SOM. Overall, uniform treatment of complimentary information obtained from all available C and N isotopes within a single multipool transport model permits improved understanding of each isotopic system, and SOM turnover as whole.