The Influence of Nutrient Availability on Soil Organic Matter Turnover Estimated by Incubations and Radiocarbon Modeling

Abstract

We investigated the decomposability of soil organic matter (SOM) along a chronosequence of rainforest sites in Hawaii that form a natural fertility gradient and at two long-term fertilization experiments. To estimate turnover times and pool sizes of organic matter, we used two independent methods: (1) long-term incubations and (2) a three-box soil model constrained by radiocarbon measurements. Turnover times of slow-pool SOM (the intermediate pool between active and passive pools) calculated from incubations ranged from 6 to 20 y in the O horizon and were roughly half as fast in the A horizon. The radiocarbon-based model yielded a similar pattern but slower turnover times. The calculation of the ¹⁴C turnover times is sensitive to the lag time between photosynthesis and incorporation of organic C into SOM in a given horizon. By either method, turnover times at the different sites varied two- or threefold in soils with the same climate and vegetation community. Turnover times were fastest at the sites of highest soil fertility and were correlated with litter decay rates and primary productivity. However, experimental fertilization at the two least-fertile sites had only a small and inconsistent effect on turnover, with N slowing turnover and P slightly speeding it at one site. These results support studies of litter decomposition in suggesting that while plant productivity can respond rapidly to nutrient additions, decomposition may respond much more slowly to added nutrients.

Appendix: 14C Approach for Modeling the Turnover Time of the Slow (Intermediate) C Pool

We estimated the turnover time of slow-pool SOM with a stock-flow model of passive, slow, and active C pools that kept track of C and ¹⁴C in annual time steps (Trumbore 1993). In Equation, (A1) the radiocarbon content of SOM in a given year is the content in the previous year plus the difference between the ¹⁴C added in plant inputs and the ¹⁴C lost in decomposition and radioactive decay.

$$ F_t = {F_{t - \Delta t} C_{t - \Delta t} + \Delta t\left( F_{I,t} I - F_t C_t \left( {1 / \tau + \lambda} \right) \right) \over {C_t }} $$

(A1)

where F = radiocarbon content of soil organic C (¹⁴C fraction modern); C = soil organic C (g C); Δt is one year; F _I = radiocarbon content of plant inputs (¹⁴C fraction modern); I = inputs (g C y⁻¹); τ = turnover time of SOM (y); and λ = ¹⁴C radioactive decay constant (0.000121 y⁻¹). Using an explicit formulation (i.e., decay and decomposition are a function of F_t−ΔtC_t−Δt rather than F _t C _t ) had a negligible effect on results.

$$ F_{OM}={{F_a C_a + F_s C_s + F_p C_p} \over{C_a + C_s + C_p }} $$

(A2)

where the symbols are as defined above with subscripts for C pool as follows: a = active, s = slow, and p = passive. In modeling each site and horizon, we solved for the turnover time of the slow pool at which the model-predicted Δ¹⁴C for the whole soil matched what we measured. The model requires parameterization of the amounts and turnover times of active and passive carbon. We used the following parameters: For all sites, the turnover time of the active SOM pool was set to 1 y (Parton and others 1987); the amount of active SOM was set to 3%, which was between the incubation-based and extraction-based estimates. As described for incubations, the amount of passive carbon was estimated as twice the mineral-stabilized C. For the turnover time of passive C, we used the turnover time of the deepest horizon at the site (Harrison and others 1993) estimated at these sites by Torn and others (1997) up to a maximum of 11,000 y. A sensitivity analysis using turnover times of 810, 2,000, and 11,000 y (bracketing the fastest and slowest values among sites) showed only a small effect except at the 20 ky site, which had the highest mineral content and an estimated 14% passive C.

Radiocarbon modeling requires estimating the ¹⁴C content of the C inputs to the soil, which is determined by the age of inputs (that is, the year that photosynthesis fixed the C from atmosphere relative to the year the inputs enter the soil). For example, if inputs to the O horizon have an age of 2 y, it means that inputs to that soil horizon in 1996 are assumed to have the ¹⁴C value of the atmosphere in 1994. In modeling with ¹⁴C, we took into account the lag time between photosynthesis and C input to SOM. We use the average residence time of all inputs to a given horizon, henceforth called the average age of inputs, in modeling.

The average age of C inputs to the O horizon [Eq. (A3)] and the A horizon [Eq. (A4)] were calculated as the mass-weighted average for the C coming from leaf, root, and dissolved organic carbon (DOC) sources.

$$ A_{I,O} = {{(A_L + A_{LL} )I_{LL} + (A_R + A_{RL} )I_{RL,O} } \over {I_{LL} + I_{RL,O} }} $$

(A3)

$$ A_{I,A}={{(A_D )I_D + (A_R + A_{RL} )I_{RL,A} } \over {I_D + I_{RL,A} }} $$

(A4)

where A = age (y) of the specified C input (that is, the number of years C spends in the specified C reservoir); I = inputs (g C m² y⁻¹) of the specified input; O = O horizon, A = A horizon; L = leaves, LL = leaf litter; R = roots, RL = root litter, D = DOC). As shown in Eqs. (A3) and (A4), the age of leaf or root inputs is the sum of the time spent as live tissue plus as litter before decaying into SOM. The amount of time spent as litter (A_LL or A_RL) before entering SOM is estimated as one-half litter turnover time [where litter turnover time is 1/k, where k is decay rate (y⁻¹)]. We were able to build upon empirical measurements of biomass, lifetime, and decay rates for both leaves and roots for many of the study sites (Crews and others 1995; Herbert and Fownes 1999; Ostertag and Hobbie 1999; Hobbie and Vitousek 2000a). These data and our estimate of input age are summarized in Table 3. The root biomass data comes from cores that sampled the combined O and A horizons (Ostertag and Hobbie 1999). We apportioned root biomass inputs into each horizon in proportion to the horizon’s relative depth interval. The age of leaf litter used in Eq. (A3) is half the litter turnover time in Table 3 (that is, we assume leaf litter is more than half-decayed by the time it is counted as OM rather than litter). For the second and fifth site on the age sequence, we had little of the needed data and interpolated between the closest (in age and location) sites.

For the O horizon of most of the sites, there were two turnover times that would predict SOM with the observed ¹⁴C value, because of nonlinearity in the atmospheric ¹⁴C record of the past 5 decades (Trumbore 2000). Where this occurred, the two turnover times differed by at least twofold and consequently implied very different C stocks and respiration rates (since flux = stock/τ). By comparing the implied stocks and fluxes to measurements, we were able to rule out one of the possible turnover times (Table 6). First, we compared the measured stock and NPP inputs to the implied stock. Second, we compared the predicted respiration from the O horizon to the measurement of total soil respiration at that site. In all but one case, the comparisons eliminated one of the possible turnover times.

Table 6 Turnover Times of SOM Estimated by ¹⁴C Model

In the A horizon, carbon is input via roots, DOC, and bioturbation. We had no data on the amount or age of DOC moving down from the surface, but we constructed scenarios to explore the implications of reasonable assumptions about the amount and age of DOC inputs. Bioturbation was not treated. The scenarios were as follows:

• 1 DOC inputs to the A horizon were (a) 10% or (b) 30% of C mineralized in the O horizon.

• 2 The age of DOC inputs to A horizon was (a) the age of material entering the O horizon SOM from leaves and roots or (b) the age of material leaving the O horizon (that is, the age of litter inputs plus the turnover time in the O horizon).

• 3 The O horizon turnover time was the (a) faster of the two possible turnover times or (b) slower of the two possible turnover times.

For the O horizon, the lag time could be calculated from empirical estimates of the root and leaf lifetimes, litter decay rates, and input rates.

For the A horizon, the age of inputs was not well constrained; assuming 30% DOC input and the slowest O horizon turnover time yielded input ages on the order of 20 y. Conversely, assuming only 10% input from DOC and fast O horizon turnover, input ages to the A horizon ranged from 8 to 11 y, depending on the site. Considering all scenarios and sites, the estimated age of inputs to the A horizon varied from 4.7 to 32.6 y. We identified an intermediate estimate (“best estimate”) based on a scenario with 1b, 2b, and the O horizon turnover times that gave the best fit to the data, according to the method described two paragraphs above and shown in Table 6. The reported estimate of turnover time for the A horizon (Figure 4 and Table 6) is based on this “best estimate” for age of inputs. We also calculated turnovers times based on an input age of 2 y for a sensitivity analysis and comparison with previously published radiocarbon studies (Table 6).

Using an age calculated as the mass-weighted average of the age of all input sources to estimate the ¹⁴C of inputs is a simplification. The complete solution is to convolve, annually, the ¹⁴C content and mass flux of each carbon input to the ecosystem. The simplification was adequate for this sensitivity analysis and for scenarios in which the dates of origin are in the period of roughly linear yearly change in atmospheric ¹⁴C content (that is, do not extend earlier than about 20 y).