| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Dep. of Forest, Rangeland, and Watershed Stewardship and Graduate Degree Program in Ecology, Colorado State Univ., Ft. Collins, CO 80523
b Dep. of Biology and Center for Environmental Studies, Arizona State Univ., Tempe, AZ 85287
c Rocky Mountain Res. Stn., 240 West Prospect, Fort Collins, CO 80526
* Corresponding author (dan{at}cnr.colostate.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Abbreviations: 
13C, the per mil difference between the carbon-13 content of the sample and the Pee Dee belemnite standard • C3–C, C3–derived carbon • C4–C, C4–derived carbon • CI, confidence interval
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Even when precise soil C measurements are possible, predicting soil C storage is complicated by at least two other factors: the turnover time of different soil C pools, and interactions between soil N accumulation and soil C storage. For example, at the Calhoun site, precise soil C measurements confirmed that these forest soils were a C sink, but 14C measurements showed that C was mainly accumulating in rapidly cycling pools that were only short-term sinks for C (Richter and Markewitz, 2001). Soils beneath N-fixing trees typically accumulate C faster than those under other types of trees (Kaye et al., 2000; Resh et al., 2002), but only a few studies have examined the response of soil C pools to fertilization in forests (Canary et al., 2000; Homann et al., 2001).
We tested two hypotheses about rates of change in soil C and N contents across time in a E. saligna plantation in Hawaii with low spatial variation in soil properties. Eucalyptus plantations cover more than 40 million ha of land in the tropics; intensive silviculture leads to very rapid rates of growth, and 6- to 8-yr rotations (Brown et al., 1997, FAO 1999).
The prior land use on this site was sugarcane agriculture, which provided a 13C signature that allowed changes in soil C to be partitioned to gains of C from Eucalyptus, and losses of older (sugarcane-derived) C. Our first hypothesis was based on the results from a comparison of soils under sugarcane agriculture and Eucalyptus forestry (Bashkin and Binkley, 1998): Total soil C would not change through an 8-yr rotation of Eucalyptus forestry because gains of new C from Eucalyptus would be offset by equal losses of older soil C. The changes in C would also vary by depth, with gains of C derived from Eucalyptus concentrated in the top 30 cm, and losses of older soil C from the entire profile. The second hypothesis was based on changes in soil C under the influence of N-fixing Falcataria moluccana (Miq.) Barneby & J.W. Grimes at a nearby location on the same soil series (Kaye et al., 2000, Resh et al., 2002): Large additions of N from fertilizer would increase total soil C content by increasing the rate of accretion of Eucalyptus-derived C, and decreasing the rate of loss of older soil C.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The control fertilization regime involved fertilization with N, P, K, and Ca in holes at 1- by 1-m spacings, and a broadcast application of the same amount at 7 mo. Total fertilizer application to the control plots (representing current operational rates) was 310 kg ha–1 N as urea, 130 kg ha–1 P, and 125 kg ha–1 Ca as triple-superphosphate, 260 kg ha–1 K as potassium chloride, and 100 kg ha–1 Granusol 2GB5 micronutrient fertilizer (5% Mn, 5% Zn, 5% Mg, 5% Fe, 1.5% Cu, and 0.5% B; American Minerals, King of Prussia, PA). The medium fertilization treatment matched the control, with additional fertilization beginning at age 3 yr, with quarterly applications for 3 yr of 65 kg ha–1 N, 31 kg ha–1 P, 46 kg ha–1 K, and annual additions of 125 kg ha–1 Ca, 58 kg ha–1 S, 23 kg ha–1 Mg, and 100 kg ha–1 Granusol micronutrients. The total fertilizer application for the medium treatment was 700 kg ha–1 N, 300 kg ha–1 P, 580 kg ha–1 K, 350 kg ha–1 Ca, 35 kg ha–1 Mg, and 300 kg ha–1 micronutrients. The high fertilization treatment matched the fertilization of the control plots at 0 and 7 mo, followed by quarterly fertilization for 5 yr, with total applications of 1600 kg ha–1 N, 680 kg ha–1 P, 1320 kg ha–1 K, 830 kg ha–1 Ca, 70 kg ha–1 Mg, and 600 kg ha–1 micronutrients.
Soil Pool Changes
Soils were sampled four times during the 8-yr rotation. Three permanent sampling locations were established in each of the 18 plots at the time of planting (April 1994). The first soil sampling involved excavation of small pits to a nearly 40-cm depth, with samples taken by trowel from the 0- to 15-cm depth, and the 15- to 30-cm depth. Bulk density (oven dry mass of soil per volume) was determined from samples of known volume taken from the midpoint of each depth. The soil is almost rock-free, and no correction for rock content was necessary. The upper 30 cm of the soil was a loose, Ap horizon as a result of >50 yr of plowing. The 30- to 45-cm depth was also an Ap horizon, similar but was more compacted as subsoil plowing occurred only once every 5 to 10 yr during the past 20 yr.
Subsequent soil samplings at age 2.5, 5.5, and 8.0 yr were done with a Giddings soil corer (i.d. 53 mm; Giddings Machine Co., Windsor, CO), with a plastic sleeve lining to allow for precise separation of soil depths (0–15, 15–30, and 30–45 cm) and determination of bulk density. Six samples (30 by 30 cm) were taken annually of the O horizon and composited within plots. Soil sampling occurred about 3 mo after the most recent fertilizer application (for fertilized plots).
Soils were air dried and 2-g subsamples ground and acidified with 10 mL of 0.2 M HCl to remove any residual carbonates from previous fertilization or liming treatments, and neutralized with drops of 1 M NaOH. Changes in total C and N were based on the concentrations measured in the carbonate-free samples (determined with a LECO 1000 CN analyzer), and bulk density. The ratios of C isotopes were used to estimate rates of gain of C from Eucalyptus, and loss rates of older soil C derived from sugarcane. Carbon isotopes were measured with a VG isochrom-NA stable isotope ratio mass spectrometer (VG, Middlewich, UK) at Colorado State University for the 0- and 2.5-yr sampling, and by the Stable Isotope Facility at the University of California at Davis for the 5.5- and 8.0-yr sampling. A quality assurance experiment showed high precision in 13C (the per mil difference between the 13C content of the sample and the Pee Dee belemnite standard) between the laboratories, with no bias. The proportion of total soil C derived from sugarcane was calculated (after Vitorello et al., 1989) as
 |
is the 13C of the soil sample, o is the 13C of soil samples with no C from C4 plants, and c is the 13C of C4 (sugarcane) plant material. The value for soil samples with no C input from C4 plants was taken as –25.5
, based on values from native Hawaiian forests within a few kilometers of this site (Bashkin and Binkley, 1998). The 13C of sugarcane material was taken as –11.5 (Bashkin and Binkley, 1998). The percentage of the soil C derived from C3 plants (including old C from the presugarcane period and new Eucalyptus C) was calculated as:
 |
We expected minimal changes across 8 yr with the soil C derived from native C3 vegetation more than 80 yr ago before the site was converted to agricultural use, so the change in C3–C should be determined almost solely by the dynamics of C derived from the Eucalyptus trees.
Incubation Experiment
We measured the effects of fertilization and soil temperature on the decomposition of C derived from sugarcane and Eucalyptus trees using 6-mo laboratory incubations. Surface soils (0 to 15 cm) collected at age 5.5 yr (December 1999) from the control and high fertilization plots from both planting densities were stored at 4°C for 6 wk before the incubation. For sugarcane and tree isotopic endpoints, we collected surface soils from one active sugarcane field adjacent to our plantations, and one Eucalyptus plantation that had never been cultivated (i.e., native forest was directly converted to Eucalyptus plantations). These isotopic endpoints had no site or plot replicates, and any reported variance comes from lab replicates (we incubated two subsamples from each endpoint site). We composited all subsamples from a given plot, dried (105°C) a 10-g subsample of the composite to constant mass to determine water content, and incubated a 75-g (fresh mass) subsample of the composite at field capacity and 21.5°C. From control plots, we incubated two additional subsamples, one augmented in the lab with the equivalent of 3.3 g m–2 N as aqueous NH4NO3 immediately before bringing the soils to field capacity, and a second (with no N addition) incubated at 25°C. The lab fertilization rate was equivalent to two field fertilizations of 0.005 kg N m–2, assuming 30% of that N stays in the surface mineral soil. The soils placed in plastic cups and sealed in airtight 1-L jars fitted with Leur valves for headspace sampling. Approximately 20 mL of deionized water were placed in the bottom of each jar to prevent soil drying. Every two weeks this water was changed and the soil brought to field capacity with deionized water.
The decomposition of sugarcane and tree C was estimated by capturing CO2 in the headspace of the incubation jars. Incubation jars were flushed with CO2–free air at 1, 6, 20, 41, 48, 62, 92, 123, and 159 d until the headspace contained <15 µmol CO2 mol–1 air and sealed for 1 to 5 d (longer at the end of the incubation). The concentration of CO2 in the headspace was then determined using an infrared gas analyzer (LICOR-6200, Lincoln, NE). The headspace was sampled by first mixing with a 35-mL syringe and then sampling 2 mL with a 10-mL syringe. Three sealed jars without soil were used as blanks to ensure that all changes in headspace CO2 were from microbial activity. Atmospheric pressure, air temperature, jar volume, sampled gas volume, and dry soil mass were used to convert headspace CO2 concentration to g C. The headspace CO2 was also analyzed at 1, 6, 20, 48, and 92 d for 13C. The percentage of headspace C derived from sugarcane or trees was calculated with the equations above. The total C, C4–C, and C3–C evolved from the soil during the incubation period (159 d for total C, and 92 d for C3– and C4–C) was interpolated between flux rate measurements by fitting exponential decay curves to plots of C flux rate vs. incubation time (r2 
0.95 for all curves).
Statistical Analyses
Changes in total soil C, C derived from C3 plants, C derived from C4 plants, and total soil N were evaluated as a repeated measures ANOVA using fixed effects of soil depth, sample year, tree spacing, and fertilization (with compound symmetry for covariance structure, and Kenward-Roger degrees of freedom method, P = 0.05; SAS version 8.2, Proc Mixed). We also tested the value of the repeated sampling for determining the effects of spacing and fertilization at the end of the rotation (Year 8), with an ANOVA of Year 8 values with Year 0 values as covariates.
The incubation experiment was also analyzed by repeated-measures ANOVA using fixed effects of fertilization level (control, fertilized in the field, and fertilized in the laboratory), incubation temperature (two levels: 21.5°C for all fertilization treatments, and 25°C for control soils), and incubation time. Initially, we also analyzed for the effect of tree spacing, but this effect was never statistically significant (P > 0.18) and was not a component of our incubation hypotheses, so we pooled data from both planting densities before incubation analyses. The total respired C, respired C3–C, and respired C4–C were also analyzed by a standard ANOVA with N fertilization level and incubation temperature as main effects. Pairwise comparisons among means for factors with significant F values were made using Fisher's LSD tests.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
The pool of C derived from C3 plants increased by an annual rate of 0.136 kg m–2 yr–1 (P < 0.0001, 95% CI = ±0.030 kg m–2 yr–1). The C3 pool also differed among soil depths, and depth interacted with year (P < 0.0001), with large increases in the 0- to 15-cm depth across time, little change in the 15- to 30-cm depth, and a slight decrease in the 30- to 45-cm depth. The pool of C derived from C4 plants declined at an annual rate of 0.144 kg m–2 yr–1 (P < 0.0001, 95% CI = ±0.030 kg m–2 yr–1), and again the depth x year interaction was significant (with greater losses from the 0- to 15- and 30- to 45-cm depths than the 15- to 30-cm depth). As with total soil C, no effects (or interactions) of tree spacing or fertilization were significant (P > 0.2).
Surprisingly, soil N content was not affected by fertilization (P = 0.99). The average rate of change in the mineral soil N pool was about 0.5 g N m–2 yr–1 (95% CI = ±4.2 g m–2 yr–1 N) with no overall change through the rotation. This rate of change is small compared with the amount of N fertilizer added (ranging from 31 g m–2 N for the control treatment to 160 g m–2 N for the high fertilization treatment), as well as the approximate N content of the trees (on the order of 40 g m–2 N). Within the rotation, the only significant trend across time was an increase of about 22.6 g m–2 N between Years 2.5 and 5.5, followed by a decline of 20.8 g m–2 N by Year 8.0. The O horizon increased to a total N content of 26 g m–2 N at age 8 yr, a rate of increase of 3.2 g m–2 yr–1 N. We expect that most of the fertilizer N leached from the soil into deeper horizons or to streams; a few measurements of gaseous N loss demonstrated relatively high rates relative to other soils (on the order of 0.3 g m–2 yr–1; P. Matson, C. Giardina, 1999, unpublished data), but very low rates relative to the dominant fluxes in the N cycle in this forest.
The second statistical analysis of soil pool sizes used a nonrepeated measures ANOVA with only the end-of-rotation 8-yr values, testing whether Year 0 values were useful covariates in detecting the effects of depth or treatments. The Year 0 covariate significantly improved the model for C4–C (P = 0.002) and N (P = 0.02), but not for total C or C3–C.
The laboratory incubations showed rapid initial rates of CO2 production, dropping to less than half the initial rates within 4 wk (Fig. 2). The rate of CO2 production for the high-temperature incubation exceeded the rates for all other incubations, and the effects of N addition (either from the N-fertilized field soil, or the N addition in the laboratory to control soils) were not significant. The 13C of the CO2 declined across time for all treatments (and even slightly for the endpoint soils), dropping by about 5. The C4–C contributed about 75% of the CO2 production at the beginning of the incubation in the Year 5.5 soils, declining by about half after 2 mo. The cumulative total loss of C did not differ among the control, field-fertilized, and laboratory-fertilized treatments (Fig. 3), but the higher temperature incubation increased the total C loss by about 40%. The lack of effect of N treatment was apparent for the loss of C derived from both C3 and C4 plants.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
If the rate of gain and loss for the entire C pool from 0 to 45 cm averaged 0.14 kg C m–2 yr–1, the turnover time of the C in this soil would be on the order of 100 yr. We expect the actual turnover rate for the entire soil C pool is somewhat longer than 100 yr, as 80+ years of sugarcane agriculture in this area depleted the soil pool of C3–C by only 40% (Bashkin and Binkley, 1998).
Our second hypothesis was strongly refuted by the results of both the field measurements of soil pools across time, and by the laboratory incubations. The addition of N fertilizer had no effect on total soil C, the accretion of new C3–C, or the loss of older C4–C. The laboratory incubations showed no effect of N (added over time in the field, or in a single application in the laboratory) on the release of CO2 from pools derived from either C3 or C4 plants. These results contrast sharply with the effects of the N-fixing Falcataria trees, where N accumulated at a rate of 13.5 g m–2 yr–1 for 17 yr, with associated accretion of C of 0.118 kg m–2 yr–1 (P < 0.01, Kaye et al., 2000). The experimental design in the present study would have detected a significant fertilizer effect on C storage if the effect were even half as large as the N-fixer effect. About 2/3 of the increase in soil C under the N-fixer was derived from new C added by the trees, and about 1/3 resulted from greater retention of older C derived from sugarcane. The contrasting effects of N fixation and N fertilization (on the same soil series, at locations within 3 km) indicates that N supply was not the apparent driver of the effect of the N-fixing trees on soil C. We expect that the increased storage of C under the influence of the N-fixing trees resulted from the indirect effect of the trees on the soil communities; the soils under the N-fixing trees had several-fold more worms (Zou, 1993), as well as greater bacterial biomass and lower fungal biomass (Garcia-Montiel and Binkley, 1998), than the soils under Eucalyptus. Canary et al. (2000) and Homann et al. (2001) looked for changes in soil C accumulation in response to large additions of N fertilizer in Douglas-fir [Pseudotsuga menziesii (Mirb.) Franco], but high spatial variability limited detectability to >15% change in C, which exceeded any effect of the N fertilizer.
How do our changes in soil C compare with other forest soils? Bashkin and Binkley (1998) measured soil C in locations near the present study site, contrasting sugarcane agriculture with Eucalyptus plantations. They found no significant changes in total soil C with afforestation (+0.020 kg C m–2 yr–1, P = 0.78); this lack of change resulted from offsetting losses of C4–C (–0.148 kg C m–2 yr–1, P = 0.03), and gains of C3–C (0.162 kg C m–2 yr–1, P = 0.01). Gains of C3–C were greater in the upper soil layers, and losses of C4–C were distributed more evenly with depth (from 0–55 cm). The apparent reforestation effect from the land-use comparison was remarkably similar in both rates and vertical pattern to the within-site changes we measured in the present study. Both studies found very low and nonsignificant changes in total soil C, and the rates of change in C pools derived from C4 and C3 plants were well within the CI of the soil C changes in the present study. Post and Kwon (2000) summarized the available literature on rates of change in soil C with afforestation, and found an average increase of about 0.034 kg C m–2 yr–1. The minimum detectable change (95% CI) in our study across 8 yr was 0.057 kg C m–2 yr–1, so our site appears to be consistent with the global average from Post and Kwon (2000). However, two-thirds of the reforestation studies cited by Post and Kwon (2000) sampled to 30 cm or less, so their average increase would just about match the upper bound of our 95% CI for 0- to 30-cm depth. Moreover, a one-tailed hypothesis that soil C increased at our site by 0.034 kg C m–2 yr–1 in the 0- to 30-cm depth (based on Post and Kwon's average) would be rejected (P < 0.01), so we conclude that the rate of C change in our site was less than the global average summarized by Post and Kwon (2000). Guo and Gifford's (2002) metaanalysis of estimates of change in soil C with afforestation found an 18% (95% CI from 9 to 27%) increase in soil C, far exceeding our 95% CI (0–45 cm) of 3.3%, again indicating that our rate of change was unusually low compared with other sites. Most of the studies used by Guo and Gifford had both lower total soil C and lower precision than the present study; the lower percentage change in soil C could derive from either a truly lower (near 0) rate of change in the present study, or from unwarranted assumptions in the metaanalysis.
The experimental design of this project provided high precision in our tests for changes with treatments and across time. If we tested only for the effects of treatment (fertilization and thinning) on soil C and N after 8 yr, with no prior sampling, the precision of tests would have been lower for some variables than with the repeated sampling that included a Time 0 value for each plot. The detection of treatment effects for total N and C4–C were improved by using Year 0 pool sizes as covariates, although the Year 0 covariates were not helpful for total C or C3–C. The use of prior plot averages as covariates is often more critical when dealing with high-variation soils. For example, Rothe et al. (2002) tested the 19-yr effects of N-fixing red alder (Alnus rubra Bong.) on steep slopes in Oregon; a repeated-measures design with the initial plot average of soil N was able to detect a significant increase in soil N under the influence of alder, whereas a simple ANOVA on the post-19-yr plot averages could not detect the alder effect. They found that inclusion of the initial plot averages reduced the minimum detectable treatment effects by one-third (for C) to two-thirds (for N).
Was the soil in the Eucalyptus plantation C saturated? Six et al. (2002) reviewed the processes underlying the accumulation of C in soils. They suggested that physicochemical characteristics inherent to soils define the maximum protective capacity of these pools, which limits increases in soil C. The physicochemical processes include binding of organic molecules onto mineral particles, the occlusion of organic matter inside organomineral aggregates, and simple biochemical recalcitrance of the organic matter. Bashkin and Binkley (1998) estimated that the soils in this area experienced a net loss of about 1.72 kg C m–2 (13% of total C) as a result of land use conversion from native forest to sugarcane agriculture. The physicochemical processes that allowed the accumulation of this larger amount of soil C before agriculture do not appear to currently allow substantial net increases in soil C; such a hysteresis effect may cloud discussions of C saturation of soils.
The influence of fertilizer N on the accumulation of C in nonforest soils appears to vary substantially among studies and sites. Gregorich et al. (1996) found that three decades of N fertilization increased the labile pool (light fraction) of C in a soil under corn (Zea mays L.) agriculture by twofold, with no increase in nonlabile pools, and no change in the turnover rate of the C. Gregorich et al. (2001) also found that rotations of corn with legumes showed much larger increases in soil C (relative to unfertilized corn monocultures) than fertilization of corn monocultures. Ludwig et al. (2003) found that fertilized plots cropped with corn showed a 2.5 kg C m–2 increase in total soil C; only 14% of this extra C was derived from corn, indicating a substantial reduction in turnover rate of older (more than 40 yr old) soil C. In contrast, Reicosky et al. (2002) found that soil C did not differ between control and fertilized fields of corn after 30 yr of treatment, and Campbell et al. (1991) also found a similar lack of response of soil C to fertilization of winter wheat (Triticum aestivum L.) systems. In an alpine soil, Neff et al. (2002) found that N fertilizer increased stable C storage in an alpine soil and decreased C stored in labile pools. The diverse results of these N addition experiments indicate that no single model (or set of processes) is likely to describe all situations.
We conclude that experimental tests of rates of change in forest soil C can provide high-precision estimates of rates of change on sites with relatively low spatial variance in soil C, including the effects of forest management practices during moderately short time periods. This site showed a precise, near-zero rate of change in total soil C that resulted from relatively rapid rates of loss of older soil C and accumulation of new soil C. The balance between C loss and C gain appears to be malleable, but the drivers of the balance need to be examined more thoroughly. Addition of inorganic N fertilizer did not affect the rate of either C loss or C gain, in contrast to the large effect of N-fixation on reducing C loss and increasing C gain (Kaye et al., 2000; Resh et al., 2002). We suspect that the difference between N fertilization and N fixation may lie in the differences in effects on soil biology more than soil chemistry, and this suspicion needs substantial experimentation before we can describe with confidence the role of soil biology in explaining variable rates of C loss and gain in forest soils (Binkley and Giardina, 1998). Precise conclusions about rates of change in forest soil will depend on our commitment to well-designed, long-term studies (Richter and Markewitz, 2001).
|
ACKNOWLEDGMENTS |
|---|
Received for publication July 11, 2003.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Journal of Natural Resources and Life Sciences Education |
Vadose Zone Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
13C for a nearby Eucalyptus plantation that was never cropped with sugarcane (providing a pure C3 signal), and a site that was currently in sugarcane land use (providing a C4 endpoint).