C02Phosphorus Interactions in Tropical Forests

In this section we use a simple model to examine the possible magnitude of phosphorus constraints on the CO, fertilization response of tropical rain forests. The model presented is based on that originally developed by Lloyd and Farquhar (1996) and modified by Lloyd (1999a). In brief, the model consists of a simulation of ecosystem plant growth in response to changing C02 concen trations coupled with the Rothamstead model of soil carbon dynamics (Jenkinson and Rayner, 1977). In Lloyd (1999a), simulations of nitrogen/carbon interactions were undertaken for temperate and boreal forests and the same principles are applied to phosphorus and moist tropical forests here. We consider the effects of the release of adsorbed phosphorus into the soil solution in response to higher rates of P removal, as a consequence of the C02-induced growth stimulation. The possible release of additional phosphorus into the soil solution as a consequence of increased plant growth eventually leading to increases in soil carbon density (Sec. 2.3) is also considered.

There is good evidence that soil microbes can actively compete with plants for soil P (Singh et all, 1989). Thus, even though there is also considerable evidence for wide-ranging C/P ratios in microbial biomass (McGill and Cole, 1981; Gijsman et al, 1996) we also consider the extent to which plant phosphorus availability might be reduced by the phosphorus demands of soil microbial biomass. The soil microbial pool increases its size and activity in these simulations as increased plant growth in response to elevated C02 is rapidly translated into increased litter inputs into the soil.

Based on recent data, some modifications of model parameters used in Lloyd and Farquhar (1996) and Lloyd (1999a) have been made. In particular, based on recently reported foliar and woody respiration rates (Meir, 1996; Malhi et al, 1999) we change our estimates for the proportion of the total plant maintenance respiration for leaves, branches, and boles from 0.36, 0.10, and 0.18 to 0.30, 0.10, and 0.10, respectively. Malhi et al. (1999) also suggested a much lower proportion of Gross Primary Productivity, GP, being lost as plant respiration (0.49) than has been reported for other studies in tropical forests (0.67-0.87: for a summary, see Medina and Klinge, 1983; Lloyd et al., 1995; Lloyd and Farquhar, 1996). The low estimate of Malhi et al. (1999) is, however, made on the unverified assumption that root respiration and root detritus production are approximately in balance. But this means that root respiration would represent only 41% of the total soil respiration and only 22% of GP. Given the respiratory costs associated with the extensive mycorrhizal symbioses in tropical soils (Sec. 3.3), this seems too low. But we must also take into account that a previous estimate of autotrophic respiratory losses accounting for 75% of GP (Lloyd and Farquhar, 1996) could be too high. This is because some estimates of Np giving rise to that number probably ignored fine root production (Lloyd, 1999a). We therefore reduce our estimate of total plant respiration in tropical forests to 65% of GP, but we allocate 50% of the total maintenance respiration to the roots. Overall, this gives a respiration rate for coarse and fine roots which is about 30% of GP.

Also, the tropical forest molar leaf area ratio used in Lloyd and Farquhar (1996) and Lloyd (1999a) of 0.34 itT2 mol-1 C is probably most applicable to the deciduous leaves of drier tropical forests or to the pioneer species of moist tropical forests (Medina and Klinge, 1983: Reich et al, 1995; Raaimakers et al., 1995). For the moist forest climax species of interest here we therefore use an amended value of 0.20 m-2 mol-1 C.

To account for the importance of sorbed phosphorus, we first characterize the relationship between the sorbed phosphorus concentration (ground area basis) and the concentration of P in the soil solution. Rather than using the Langmuir model (Sanyal and DeDatta, 1991), we use.

Sm,JPs sol J

where [Psori,] is the amount of P sorbed on or in the soil particles (ground area basis) Smjx is the maximum absorption, fPS0]l is the concentration of P in the soil solution and is a constant relating to the P binding energy. Values of Ks and Smjx for the simulation here are taken from a fit to the data for a Zimbabwe oxisol presented by Sibanda and Young (1989) for which, assuming an active soil rooting depth of 0.5 m and a bulk density of 1.3 g cm-3, we estimate Smjx = 5 mol m 2 and ks = 1 mmol m 2. Based on soil P data for an adystrophic rainforest in Venezuela (Tiessen el ill., 1994a), we take an initial estimate for [P,0!-i,l of 350 mmol m 2 for which the equivalent [Psoi] = 0.075 mmol m-2, about 1.5 ¿¿M.

There is good evidence that high-affinity phosphorus uptake by plants can be well represented by Michaelis-Menton kinetics (McPharlin and Bieleski, 1989; Jungk et al„ 1990). And, as discussed in Sees. 3.3 and 4.2, there are good reasons to suppose that, irrespective of the mechanism of phosphorus mineralization and uptake, phosphorus uptake rates per unit line root density are maintained as ambient C02 concentrations increase. We therefore allow the maximum phosphorus uptake rate to increase linearly with increases in fine root density and thus write the rate of plant phosphorus uptake as

where Up is the rate of uptake of P by the vegetation (mol m-2 year-1), y is the maximum (P saturated) rate per unit fine root density, Mtr is the fine root carbon density, and Ku is a constant. There is very little information on the kinetics of P uptake by the roots of tropical plants and so it is hard to determine a priori a reasonable value for Ku. Working mostly with crop species, researchers have reported values between 1 and 10 /xM (Marschner, 1995; Schachtman et al„ 1998; Raghothama, 1999). Given that we expect the trees adapted to these nutrient-poor tropical soils to have developed high-affinity P uptake systems, we assume a value of 2.0 /U.M.

Here, we are mostly interested in calculating "generic" values from which phosphorus fluxes and their dependence on soil availability and internal plant physiological status can be quantified. Because of the effects of soil fertility on phosphorus concentrations discussed above, we consider only moist tropical forests growing on the more abundant but lower nutrient oxisol/ultisol soils. This is because the [C02]/phosphorus interaction we are seeking to model is likely to be most marked on these low-fertility soils. Moreover, compared to other soil types, they tend to dominate the moist tropics (Sanchez, 1976).

Based on the discussion in Sec. 3.2 we take the following phosphorus concentrations for use in the model simulation:

In order to estimate the phosphorus fluxes in the various plant tissues, we assume that 68% of leaf and fine root P is retranslo-cated prior to abscission (Vitousek and Sanford, 1986). The distribution of P between the various tissues is achieved using a scheme similar to that used for nitrogen in temperate and boreal forests (Lloyd, 1999a). Phosphorus is first distributed according to assumed C/P ratios in branches, boles, and roots. The variable remainder is then allocated to the leaf tissue. From Figure 2, the P concentration of leaves of moist tropical rainforest species is already very low and probably strongly limiting for photosynthesis (Sec. 4.1). Moreover, on a canopy basis it seems reasonably constant, despite variations in total leaf carbon density (Vitousek and Sanford, 1986). The model therefore assumes that foliar P concentrations are maintained at 0.5 mmol mol-1, with variations in the total foliar P content being reflected in changes in the total leaf carbon density (and hence leaf area), rather than by differences in the photosynthetic rate per unit leaf area. As was discussed in Sec. 4.1, it seems likely that some plants in moist tropical forests may not be capable of significantly enhanced growth responses to increased phosphorus availability. Variations in leaf area but not leaf photosynthetic capacity per unit leaf area are therefore also assumed when total canopy phosphorus content is either increasing or decreasing. The latter turns out to be the case in some scenarios examined below.

The rate of change in the concentration of P in the soil solution is written as

dt dt a[psorb]

dt where I is the atmospheric input, kL is a constant relating the rate of leaching to the soil solution phosphorus concentration, Lv is the rate of P input into the soil through litterfall, [Porp] is the concentration of organic phosphorus in litter and soil, and Up is the rate of plant P uptake as given in Eq. (3).

In the simulations here, there are two components to I: atmospheric deposition, taken here as 1.5 mmol m-2 year-1, and canopy leaching, which is calculated on the basis of the canopy P content assuming a rate of 6 mmol -2 year-1 in 1730 (see Sec. 3.1.3). To estimate kL in our standard cases, we assume that the input of phosphorus through atmospheric deposition was exactly balanced by leaching losses in 1730. This is somewhat at odds with several observations suggesting that much of the atmospherically derived P deposited onto tropical forests is retained rather than being leached out of the system (Sec. 3.1.2). Nevertheless, as was discussed in Sec. 3.1.1 almost all of this atmospherically derived P

probably comes from the forests themselves and thus does not represent a net positive P input.

The term c)[P ]/c)t is calculated assuming that the concentration of phosphorus in all decomposing litter is 0.16 mmol mol-1. This is based on the 68% retranslocation of P from leaves and fine roots and the average branch, bole, and coarse root P concentrations (Sec. 3.2). Where the sensitivity of the model to P accumulation in the microbial carbon pool is tested, based on data summarized by Gijsman et al. (1996) we use a tissue P concentration for microbes of 6.4 mmol P mol-1 C. In all simulations, it is assumed that soil phosphorus mineralization proceeds with a rate constant of 0.5 year-1, with phosphorus mineralization proceeding independently of carbon mineralization. This is on the basis of the evidence discussed in Sec. 2.1. Indeed, inflexible soil carbon pool C/P ratios which effectively link phosphorus mineralization rate to the carbon mineralization rate in models such as CENTURY (Parton et al., 1988) have been strongly criticized by some tropical soil chemists (Gijsman et al., 1996).

In order to estimate the last term of Eq. (4), we differentiate Eq. (2) and then write

atPoJ dt

3[Ch dt

3 [Chun dt

This gives an alternative to Eq. (6) but now with a dependence on

Smax ks

The second term in the denominator is typically around 5000 and represents the "buffering" effect of the sorbed P. That is, in the presence of an appreciable sorbed P pool, the soil solution P concentration is extremely insensitive to the rate of removal of phosphorus into or from it. This is because increased rates of removal of P are almost totally balanced by desorption. Likewise, increased rates of P input result in large increases in [Psorb], but with very little change in [Pso|]. That latter case represents, of course, the tropical soil phosphorus fertilizer "fixation" problem discussed in Sec. 2.2.

As was discussed in Sec. 2.3 there are some indications that increases in soil carbon density could release adsorbed phosphorus into the soil solution, and available evidence suggests that Smax declines and /cs increases with increasing soil carbon density (Sibanda and Young, 1989). A precise understanding of the mechanisms involved is still lacking, as is any general quantitative description of the nature of the relationship. So based on the data of Sibanda and Young (1989) we simply assume for a simple sensitivity study Smax = a/[Chum] and ks = j8[Chum], where a and ¡3 are fitted constants and [Chum] is the modeled soil humus carbon density. We then write

The value of [Chum] in 1730 is used in conjunction with Smax = 5 mol m-2 and Ks = 1 mmol-2 to determine a and ¡3.

The simulations that have been undertaken are as follows:

(A) No consideration is given to plant phosphorus requirements (i.e., the growth rate of tropical forests is affected only by the atmospheric C02 concentration).

(B) Plant phosphorus uptake is dependent on [Psoi] but the sorption and desorption of P are ignored (Eq. (4), but with3[Psorb]/at=0).

(C) Plant phosphorus uptake is dependent on [Psol]. Sorption and desorption of P in response to changes in [Pso]] are considered (Eq. (6).

(D) Plant phosphorus uptake is dependent on [Pso]]. Sorption and desorption of P in response to changes in [Pso|] and [Chuml are considered (Eq. (8)).

(E) Plant phosphorus uptake is dependent on [Pso|]. Sorption and desorption of P occur in response to changes in [Pso!] and [Chum] (Eq. (8)). Increases in microbial biomass remove phosphorus from the soil solution.

Results of these simulations are shown in Table 1. For Scenario A (essentially as in Lloyd, 1999a, but with minor changes as discussed above) the simulated increases in GP and NP are substantial, being 35 and 44% higher for the period 1981-1990 (average value), respectively. This large modeled stimulation of productivity arises mostly because of the high sensitivity of GP to C02 concentrations at warmer temperatures that occur in moist tropical forests (Lloyd and Farquhar, 1996). As must happen with finite turnover times for plant and soil carbon, increases in rates of litterfall always lag behind the CO,-induced increase in GP and NP and the soil respiration rate always lags behind litterfall (Lloyd and Farquhar, 1996; Lloyd 1999a). Thus a substantial sink of carbon is modeled to be occurring in this situation: 8.3 mol m-2 year-1. Spread across the moist tropics (ca. 12 X 1()12 m-2), such uptake would be significant: 0.1 Pmol year-1 or about 50% of most current estimates of the terrestrial carbon sink (Lloyd, 1999b).

But this modeled stimulation of enhanced productivity in response to increasing [CO,] can be greatly modified when variations in phosphorus availability are considered. This is shown in Table 1 for Scenario B. In the absence of a resupply of phosphorus

TABLE 1 Effect of Phosphorus Availability Model Assumptions on Simulated Gross Primary Productivity, Net Primary Productivity, Litterfall, Soil Respiration Rate, Rate of Net Ecosystem Carbon Accumulation, Plant Carbon Density, and Soil Carbon Density for 1730 and 1981 - 1990 (Average Value)

Gross Primary

Net Primary

Rate of Ecosystem

Plant Carbon

Soil Carbon




Soil Respiration

C Accumulation




(mol C m-2 year-1)

(mol C m-2 year-1)

(mol C m'"- year

' ) (mol C m 2 year ')

(mol m 2 year ')

(mol C m 2)

(mol C m 2)

1730: No P constraints








A: 1981- 1990: No P constraints








B: 1981-1990: No P sorption








or desorption

C: 1981-1990: With P sorption








and desorption (Eq. 6)

D: 1981 -1990: With P sorption








and desorption + soil

C effect (Eq. 8)

E: 1981-1990: With P sorption








and desorption + soil C effect

+ P sequestration

and desorption in microbes

Note. See text for a full description of model structure and Scenarios-A to -E.

Note. See text for a full description of model structure and Scenarios-A to -E.

from the labile pool, the increased P uptake required to sustain the extra growth in response to increasing [C02 j results in soil solution phosphorus concentrations being rapidly depleted, being reduced by about 25% for 1981-1990 compared to 1730 in this scenario. This offsets to some degree the increased intrinsic phosphorus uptake ability of the larger trees (due to more fine roots) and thus the C02-induced increases in G,. and Nv are only about half the magnitude of the no-P-constraint case (Scenario A). Accordingly, the rate of net carbon accumulation by the ecosystem is only 3.6 mol -2 year-', less than half the value for Scenario A.

This picture of a substantial phosphorus constraint is drastically altered when the presence of the inorganic labile (i.e., sorbed) phosphorus pool is taken into account (Scenario C). Desorption of phosphate occurs in response to increased rates of removal from the soil solution. Consequently, the reduction in soil solution phosphorus concentration over 1730 levels is only 9% for 1981-1990. This contrasts with the 25%) reduction in Scenario B. Consequently, the enhancements of Gv and Nv are more similar to the no-P-constraint case, though a full expression of the C02-in-duced growth response is still not possible. Accordingly, the rate of net carbon accumulation by the ecosystem is 6.5 mol m-2 year-1, substantially more than Scenario B, but about 20% less than what is modeled to be the case if no phosphorus limitations to plant production occurred.

When the potential positive feedback between increased soil carbon densities and the desorption of phosphorus is considered (Scenario D), all phosphorus constraints on the C02-induced growth response or the rate of net ecosystem carbon accumulation disappear. Indeed, the rates of GP and Nv are actually slightly higher than those in Scenario A. Indeed, according to Scenario D, the average soil solution phosphorus concentration for 1981-1990 is actually higher than the 1730 value. Moreover, including some sequestration of P into the increasing soil microbe pool (Scenario E) has absolutely no effect. This is because the rate of desorption is substantially greater than the rate of sequestration into this pool. But in the absence of the positive feedback between increased soil carbon densities and the desorption of phosphorus, sequestration of P into the microbe pool does have a small effect, reducing fluxes by about 5% for 1981 -1990.

Thus, the simulations here suggest that for tropical soils such as oxisols and ultisols which contain an appreciable pool of labile sorbed phosphorus, the transfer of phosphorus from this pool to the soil solution in response to increased rates of P uptake by plants serves to more or less maintain soil phosphorus concentrations. This allows the increased rates of phosphorus uptake by faster growing vegetation to continue.

Kirschbaum et al. (1998) used a somewhat different modeling approach to simulate the effects of phosphorus availability on temperate forest C02-induced growth responses. But similar to the results here, they concluded that the presence of the "secondary" (labile) pool means that, in the short term, phosphorus availability should not constrain the ability of these forests to respond to [C02]. They also concluded, however, that marked phosphorus constraints should become apparent on a time scale of centuries. This is of course also possible here, as the size of the sorbed phosphorus pool is not infinite. A second uncertainty is related to the degree to which sorption is indeed a reversible process on the time scales of interest here. A traditional view had been that sorption of phosphorus is a more or less irreversible process. But as discussed by Barrow (1983, 1999) and Sanyal and De Datta (1991), this apparent irreversibility more likely reflects the relatively slow time frame over which desorption occurs. The degree to which the sorption of phosphorus onto tropical soils occurs is a truly reversible process remains an important research issue (Gijs-man et al., 1996).

As discussed in Sec. 2.3, there are some indications that increased soil carbon density should act to release previously adsorbed phosphorus, making it available for plant growth. The simulation here suggests that this effect is potentially quite important (Scenario D in Table 1). Indeed, as our parameterization of the dependence of canopy photosynthetic rate on canopy phosphorus content is conservative (being modulated solely by changes in leaf area without any changes in the rate per unit leaf area) it is quite possible that this effect may be even more potent than modeled here. Indeed, it is not inconceivable that a "runaway" positive feedback could occur. This would involve C02-induced increases in tropical forest plant growth giving rise to increases in soil carbon content, which in turn liberates previously sorbed phosphorus, which then gives rise to yet more increased plant growth. In that context, the very high rate of ecosystem carbon sequestration observed by Malhi et al. (1998) for a mature moist tropical forest near Manaus in Brazil, 49 mol m-2 year-1, may not be as unex-plicably high as it first seems.

0 0

Post a comment