calibration of the rothamsted organic carbon turnover model (rothc ver. 26.3), using measurable soil...

© CSIRO 2004 10.1071/SR03013 0004-9573/04/010079

www.publish.csiro.au/journals/ajsr Australian Journal of Soil Research, 2004, 42, 79–88

CSIRO PUBLISHING

Calibration of the Rothamsted organic carbon turnover model(RothC ver. 26.3), using measurable soil organic carbon pools

J. O. SkjemstadA,B, L. R. SpouncerA, B. CowieC, and R. S. SwiftD

ACSIRO Land and Water, PMB 2, Glen Osmond, SA 5064, Australia.BCRC for Greenhouse Accounting.

CDepartment of Natural Resources and Mines, Locked Mail Bag 1, Biloela, Qld 4715, Australia.DFaculty of Natural Resources, Agriculture and Veterinary Science, The University of Queensland, Gatton,

Qld 4343, Australia.

Abstract. A fractionation scheme that provided the measurement of a labile pool (particulate organic carbon), acharcoal-carbon pool, and a humic pool by difference was tested as a means of initialising the Rothamsted organiccarbon turnover model version 26.3. Equating these 3 fractions with the resistant plant material, inert organic matter,and humic pools of the model, respectively, gave good agreement between measured and modelled data for 2long-term rotation trials in Australia using a soil depth of 30 cm. At one location, Brigalow Research Station inQueensland, there were 3 distinct soil types, two clays and a duplex soil, in a semi-arid, subtropical climate. At thissite, continuous wheat with some sorghum was established after clearing land under brigalow (Acacia harpophylla)and continued for 18 years. The second location was near Tarlee, South Australia, and was established on existingagricultural land. One soil type (red brown earth) with 2 rotations (continuous wheat and wheat–fallow) wereavailable over a period of 8 years.

The modelled and measured data were in good agreement for both locations but the level of agreement wassubstantially improved when the resistant plant material decomposition rate was reduced from 0.3 to 0.15/year. Noother modifications were required and the resulting values provided excellent agreement between the modelled andmeasured data not only for the total soil organic carbon but also for the individual pools. Using this fractionationscheme therefore provides an excellent means of initialising and testing the Rothamsted model, not only inAustralia, but also in countries with similar soil types and climate.

For the first time, the work reported here demonstrates a methodology linking measured soil carbon pools witha conceptual soil carbon turnover model. This approach has the advantage of allowing the model to be initialised atany point in the landscape without the necessity for historical data or for using the model itself to generate an initialequilibrium pool structure. The correct prediction of the changing total soil organic carbon levels, as well as thepool structure over time, acts as an internal verification and gives confidence that the model is performing asintended.SR03013J. O. Skjems tadet al .Cali br ati on of t he Rothams ted model

IntroductionSoil organic carbon (SOC) turnover simulation models havebeen widely used to predict SOC change with changingenvironmental and management conditions. These modelsinclude the Rothamsted (Jenkinson 1990), Century (Partonet al. 1987), and APSIM (McCown et al. 1996) models. Allare based on several conceptual SOC pools turning over atvarying rates through the microbial biomass. Each of themodels is similar in construct in that they utilise acombination of pools with a rapid turnover (annual),moderate turnover (decadal), and slow turnover (millennial)or inert.

Each of these pools is conceptual in nature and generallynot measured directly. An assessment of pool structure can,

however, be derived in several ways. The models can be runto equilibrium for a particular scenario, thereby giving thepool structure that should be present under those conditions.For the Rothamsted and APSIM models, which contain aninert pool that is isolated from the flow of decomposablecarbon through the remaining pools, this approach is notapplicable. To overcome this difficulty, the Rothamstedmodel can accept OC radiocarbon dating information, fromwhich a pool structure can be derived. In the APSIM model,an assumption is made that inert C is relatively constant foreach soil horizon in a profile and that nearly all of the soil Cat depth is inert. The total organic carbon (TOC) at depth ina profile is thereby used as an estimate of inert C in thesurface horizons.

80 Australian Journal of Soil Research J. O. Skjemstad et al.

The versatility of a model could be greatly improved ifthe SOC pools, or at least the major pools, could bedirectly measured. This would have several advantages.First, it would allow the model to be initialised at any pointin the landscape without the necessity for historical data,and second, the correct prediction of the changing totalSOC levels as well as the pool structure over time wouldact as an internal verification, providing confidence thatthe model was performing as intended. Linking directlymeasured with conceptually modelled soil organic carbonpools, however, is not straightforward and it has beensuggested that modelling the measurable may be moresuccessful than attempting to measure the modellable(Elliot et al. 1996). In this context, recent work hassuggested that physically isolated fractions may be moreuseful in this regard than the more classical chemicalapproaches (Cambardella 1998; Sohi et al. 2001).

Using a similar approach, Skjemstad and Janik (1996)showed that the pool structure of the Rothamsted modelversion 26.3, henceforth referred to as the RothC model,could be approximated by using a soil fractionation schemein which soil organic carbon was partitioned into 3 fractions.These were the >53 µm particulate organic carbon (POC)fraction (Camberdella and Elliot 1992), the <53 µmcharcoal-C fraction, and the fraction <53 µm that was notcharcoal. These fractions correspond well with the resistantplant material (RPM), inert organic matter (IOM) and humic(HUM) pools of the RothC model, respectively, althoughearlier results showed that some modification to the rateconstants may be necessary.

By using this approach, Skjemstad and Janik (1996)predicted the change in TOC with time at 20 long-term sitesvarying in soil type, climate, and management, some with asmany as 4 different rotations or management practices. TheRothC model was initialised using the initial soil carbonvalues, which may or may not be prior to clearing orcultivation, and the soil C pool structure obtained from theabove fractionation scheme. The cultural practices within thetrials covered a range from low input systems such aswheat–fallow to high carbon input systems such as fertilisedpasture. The model was run for the duration of each trial(ranging from 11 to 73 years) using soil clay content, climatedata, and inputs from residues obtained from the trial site.The final predicted soil C was compared with the actual finalsoil C as measured in the laboratory. The predicted v.measured results showed reasonable agreement for the0–10 cm soil horizon, considering the limited data available.More recently, Skjemstad et al. (2001) demonstrated that thisfractionation scheme worked very well for 2 Vertisols fromsouth-east Queensland using the 0–30 cm horizon whenaveraged long-term climate and yield data were used.

For most of the sites described by Skjemstad and Janik(1996), the full range of data required to initialise and run themodel over a duration >10 years was not available. For

example, at many sites, only average yield or climate datawere available or soil bulk density had not been measuredover the course of the trial. Based on the preliminary resultsobtained from the RothC model discussed above, 2long-term field trials were selected where both the requireddata to initialise and run the model and archived soil sampleswere available. The soil samples and site data were used toinitialise and run the RothC model and model outputs werethen compared with measured soil data in subsequent years.

Model description

The RothC model, version 26.3 (Jenkinson 1990; Jenkinsonand Coleman 1994) can be used for calculating the rate ofcarbon loss or sequestration for specific rotations inagricultural soils. It can also be used to predict long-termchanges in carbon due to changing climate. The model wasinitially developed for grassland, forest, pasture, and cropsunder temperate European conditions and carbon fluxes buthas since been tested and used under Australian conditions.Figure 1 is a schematic representation of the RothC modelconstruct and depicts plant residues entering the soilenvironment, undergoing decomposition by the soil microbialbiomass to form several pools with the evolution of CO2.

As shown in Fig. 1, the model consists of 5 pools: (a)easily decomposable plant material (DPM), (b) RPM, (c)microbial biomass (BIO), (d) HUM, and (e) IOM highlyresistant to microbial decomposition. The BIO pool isactually represented by 2 sub-pools (BIO slow and BIO fast)to account for microbial biomass in either a protected or anunprotected environment. Both DPM and RPM decomposeto form CO2, BIO, and HUM, with subsequent furtherdecomposition of the BIO and HUM to produce more CO2,BIO and HUM. The amount and nature of the plant residues(e.g. crops, trees, grass, legumes), clay content, soiltemperature, rainfall, evaporation rate, and thedecomposition rate constant for each pool affect themodelled carbon balance in the soil and are important inputsto ensure effective use of the model.

Fig. 1. Schematic representation of the RothC model (modifiedfrom Jenkinson 1990)

Calibration of the Rothamsted model Australian Journal of Soil Research 81

The amount of carbon (Y) that decomposes from anyactive pool in a particular month follows an exponentialdecay function:

that uses the initial amount of carbon in the particular pool(Y0), a series of rate-modifying factors for temperature (a)(through the monthly average temperature), soil watercontent (b), plant retainment (c), the decomposition rateconstant for that particular compartment (k), and a constant(t) = 1/12 to convert k to a monthly time step. Inputs directlyrelated to the exponent variables in Eqn 1 thereforecontribute exponentially to the soil carbon remaining at theend of each month.

The driving mechanism for carbon loss from the soil isthrough microbial decomposition processes, which areaffected by the available soil water content and by thetemperature of the soil system. The way in which therate-modifying factor for soil water is used by the model isthrough a soil moisture deficit (SMD), which is dependenton the clay content of the soil, monthly average rainfall, andpan evaporation. The maximum SMD is related through aquadratic function to clay content, and the accumulatedSMD is calculated monthly from the balance in soil waterbetween rainfall and evaporation. The plant retainmentfactor further modifies the soil water balance for the extentof plant cover.

The model also adjusts for clay content by altering thepartitioning between evolved CO2 and (BIO + HUM) poolsformed during decomposition. This adjustment is madeusing the exponential equation (Eqn 2):

where x is the ratio of CO2/(BIO + HUM) and BIO and HUMare the corresponding biomass and humic pools formedinitially by incoming plant material. Incoming plant materialis partitioned into a highly labile DPM pool and a moreresistant RPM pool. The ratio DPM/RPM is used directly inthe model to allow for changes in the ‘quality’ ordecomposability of the input from vegetation. The change inSOC on a monthly basis is then calculated as the sum of the5 pools.

Materials and methods

For calibration, sites needed to be identified where detailed soil, crop,and climate data were available. The essential data requirements arelisted in Table 1. Detailed analyses of the soil samples for their RPM,HUM, IOM, and clay contents were essential and these required at least50 g of soil material <2 mm. Two trials that met the necessaryrequirements were located at the Brigalow Research Station,Queensland, and the Tarlee Rotation Trial, South Australia.

Calibration sites

Brigalow catchment study

This site is situated in semi-arid, subtropical Queensland, latitude24.83°S, longitude 149.78°E, originally supporting open forestcommunities dominated by brigalow (Acacia harpophylla).

The Brigalow trial consisted of 3 forested catchments of 12–17 ha.Two catchments were cleared in 1982; one was planted to buffel pasture(Cenchrus ciliaris) and the other to wheat (Triticum aestivum) andoccasional sorghum (Sorghum bicolor). The remaining catchment wasleft under native brigalow forest. Within each of the catchments, 3monitoring sites were established in recognition of 3 soil types (2 claysand 1 duplex). Each of the soil types were treated as separate replicatesand soil sampling was carried out to reflect these differences.Therefore, 3 separate sets of soil samples were collected within eachcatchment at each sampling, one on each of the clay soils and one onthe duplex soil.

Archival samples had been maintained at Brigalow Station frombefore the clearing of the trial site in early 1982 and at intervals up tothe year 2000. Six of these sets of samples were obtained for analysis.They were from samplings 1, 3, 5, 7, 9, and 11 corresponding to March1982, September 1983, October 1985, October 1987, April 1994, andMay 2000, respectively. Sampling 1 was before clearing. Sampling 3took place 17 months after clearing and 10 months after burning thefelled timber in situ.

Two sets of samples were collected from the field: surface samples(0–10 cm) and profiles to 200 cm. The top layers of the profile weresampled at intervals of 0–10, 10–20, and 20–30 cm. The surfacesamples at each site were the bulk of 80 individual cores while 5individual cores were bulked to produce representative 10–20 and20–30 cm samples.

Climate data from an on-site weather station were converted toaverage monthly temperature (°C) and monthly rainfall and panevapouration (mm water). Yield data for wheat and sorghum wereavailable for all years. The yield data are given in Table 2. Climate dataon a monthly time step for this and the Tarlee site are available inSkjemstad and Spouncer (2003) or can be accessed on the AustralianGreenhouse Organisation (AGO) website (http://www.greenhouse.gov.au/ncas).

( )0 (1)abckteY = Y 1 −−

( )0.0786*%Clay e= 1.67 1.85 +1.60 (2)x −

Table 1. Data and other requirements for calibration sites for the duration of RothC testing

Data and other requirements

Comments

Soil samples Representative soil samples at least from the beginning and end of a period >10 years to a depth of 30 cm. In all cases, archival samples were required

Soil bulk density Measured at the time of sampling using core weight/volume

Crop yields Yield of grain for each of the years to be modelled

Management Details of individual crops, rotations, fallow periods, stubble burning and incorporation. If grazing occurred, estimates of consumption and return from animals

Climate Details of average monthly temperature, rainfall and pan evaporation


Tarlee rotation trialThis site is situated in South Australia near Tarlee in a

Mediterranean climate, latitude 34.28°S, longitude 138.77°E. Thiswheat–pasture trial was established on previously farmed land.

The Tarlee rotation trial was established in 1977 by the SouthAustralian Research and Development Institute (SARDI) on agradational Red Brown Earth to monitor the long-term effects ofintensive and traditional rotations on soil properties and cropproduction. Continuous wheat, wheat–pasture, and wheat–fallowrotations were set up with 3 stubble treatments; burning, incorporation

and surface retention. Three nitrogen levels 0, 40, and 80 kg/ha werealso applied to the wheat phases. The trial was discontinued after 1996and so only archival soil samples were available. Soil samples coveringthe entire top 30 cm of the profile were obtained for the years 1979,1985, and 1996 for the 0 N treatment from all 3 rotations. Thestubble-retained management rotations were used. Good agronomicdata were available along with wheat yields and BD to 30 cm. Pasturecuts from the wheat–pasture rotation were not made however, and againthis rotation was not included in the calibration process. The yield dataare given in Table 3.

Table 2. Total dry matter (TDM) and grain yield data (t/ha) from the Brigalow Station sitew, Wheat; s, sorghum

Year Plot 64 Plot 65 Plot 66Phase TDM Grain Phase TDM Grain Phase TDM Grain

1982 Cleared Cleared Cleared1983 Fallow Fallow Fallow1984 s 3.82 1.53 s 2.43 0.97 s 2.37 0.951985 w 7.72 2.86 w 8.00 2.96 w 4.45 1.651986 w 6.35 2.35 w 7.43 2.75 w 6.05 2.241987 w 12.03 4.45 w 11.73 4.34 w 8.82 3.261988 w 6.89 2.55 w 6.65 2.46 w 6.32 2.341989 w 5.91 2.19 w 6.20 2.29 w 4.85 1.801990 w 4.36 1.61 w 4.36 1.61 w 2.83 1.051991 w 4.50 1.67 w 4.34 1.61 w 4.34 1.601992 w 5.43 2.01 w 3.73 1.38 w 6.02 2.231993 Fallow Fallow Fallow1994 w 8.81 3.26 w 7.27 2.69 w 7.19 2.661995 s 0.56 0.22 s 0.08 0.03 s 0.69 0.281995 s 4.38 1.75 s 5.63 2.25 s 0.95 0.381996 w 2.60 0.96 w 2.62 0.97 w1997 s 9.64 3.86 s 7.98 3.19 s 9.03 3.611998 w 4.76 1.76 w 5.41 2.00 w 4.21 1.561999 s 2.31 0.92 s 1.59 0.64 s 2.35 0.94

Table 3. Total dry matter (TDM) and grain yield data (t/ha) for two rotations from the Tarlee sitew, Wheat; f, fallow; n.a., not available

Year Wheat–wheat Wheat–fallowPhase TDM Grain yield Phase TDM Grain yield

1979 w 3.75 1.54 w 8.05 3.151980 w 1.18 0.55 f1981 w 3.47 1.76 w 5.11 2.631982 w 1.05 0.47 f1983 w 4.21 1.80 w 6.71 2.891984 w 3.11 1.33 f1985 w 2.58 1.06 w 7.28 3.11986 w 2.97 1.16 f1987 w 1.72 0.65 w 4.86 1.891988 w 1.19 n.a. f1989 w 2.69 n.a. w 3.96 n.a.1990 w 2.60 1.17 f1991 w 3.87 2.23 w 7.81 3.231992 w 2.51 1.13 f1993 w 1.20 0.54 w 3.77 1.621994 w 1.31 0.59 f1995 w 2.00 0.90 w 4.44 1.911996 w 1.69 0.76 f


Sample preparation

All samples were received as <2 mm. In all cases, replicate cores werebulked to give one 0–10, 10–20, and 20–30 cm sample for eachsampling time. These samples were then further bulked taking intoaccount differences in bulk density to give one 0–30 cm sample for eachsampling time and for each rotation.

Total organic carbon

TOC was determined on finely ground soils and some fractions using aLeco combustion furnace (C-144) as described by Merry and Spouncer(1988). All samples were tested with 1 M HCl for the presence ofcarbonate. Samples shown to contain carbonate were weighed (1.0 g,<100 µm) into Leco combustion boats with nickel foil liners to preventloss of soluble carbon. Soils were then treated with excess sulfurousacid (6% SO2 solution) on a hotplate set at 100°C until the reactionceased. Samples were further heated until dry. These samples were thenanalysed by Leco furnace as for non-calcareous samples except that asmall amount of Zn wool was introduced to the top of the first dryingtube to reduce the amount of SO2 gas entering the analysis train.

Particulate organic carbon

The procedure used for POC was similar to that described byCambardella and Elliot (1992) for particulate organic matter. Soilsamples (10 g, <2 mm) were shaken overnight with 40 mL of 5%(wt/vol) sodium hexametaphosphate. The samples were then passedthrough a nest of 2 sieves (200 and 53 µm). The material remaining inthe sieves was gently worked with a spatula to ensure that no aggregateswere retained in the particulate fractions, and the suspension of <53 µmmaterial was made to 500 mL. The material remaining on the sieves wasoven-dried (75°C) and weighed, and TOC was measured by Lecocombustion furnace (C-144). The POC content was calculated from thesum of the OC from both sieves. Calcareous samples were pretreated

with sulfurous acid and washed before the addition ofhexametaphosphate.

Charcoal carbon

Charcoal content was determined by the method described bySkjemstad et al. (1999). TOC was determined on the 500 mLsuspensions (Heanes 1984) remaining after the POC analysisprocedure, and from this, the volume necessary to give ~2.5 mg OC wascalculated. This aliquot was placed in a quartz tube, made to ~20 mL,and photo-oxidised for 2 h as described by Skjemstad et al. (1999). Thephoto-oxidised suspensions were transferred to centrifuge tubes and thesamples centrifuged at 800g for 20 min. The clear supernatant wasdiscarded and the washing procedure repeated twice more with water toensure any soluble OC remaining had been removed. The contents ofeach tube was then transferred to 75-mL digestion tubes and the OCcontent remaining in the photo-oxidised samples determined using amodified dichromate method with glucose as a standard (Heanes 1984).Each of these measurements was determined in duplicate.

A second set of photo-oxidations (36–72 individual oxidations)were carried out on each sample, and after centrifugation, these werecombined into one bulked sample. These samples were further treated9 times with 2% HF as described by Skjemstad et al. (1994) to removeFe and to concentrate the OC in the fraction. After washing with water3 times to remove excess HF, the samples were freeze-dried andanalysed by nuclear magnetic resonance.

Cross-polarisation with magic angle spinning 13C nuclear magnetic resonance spectroscopy

The 50.309 MHz cross-polarisation with magic angle spinning(CP/MAS) 13C nuclear magnetic resonance (NMR) spectra of finelyground (<0.2 mm) HF-treated samples were obtained on a Varian Unity200 spectrometer with a 4.7 T wide-bore Oxford super-conductingmagnet. Weighed samples (0.2–0.6 g) were spun at 5 kHz in

RothC Model Calibration

Soil sampling

from calibration

sites

Fractionate soils

Initialise model

with soil data

Run model using

soil and other

data

Climate and

residue data

Test final soil C

Modelled

v.

Measured

Model

verification

Reset RPM and

HUM pool rate

constants

Data do

not agree

Data

agree

Fig. 2. Calibration procedure for RothC model and order of events.


7-mm-diameter zirconia rotors with Kel-F caps in a Doty ScientificMAS probe. All spectra were attained with a 1 ms contact time and a300 ms recycle time. The number of transients required for anacceptable signal/noise (S/N) ranged from 3 to 150 k. Using thestandard Varian pulse sequence, the free induction decays wereacquired over a sweep width of 40 kHz over an acquisition time of15 ms in a 1214-point database, with a constant signal gain betweensamples. All spectra were obtained with 32 k zero filling and 50 HzLorentzian line broadening and a 0.01-s Gaussian broadening.Chemical shift assignments were externally referenced to the methylresonance of hexamethyl benzene at 17.36 ppm. The proportionalcontribution of the functional groups ketonic/aldehyde (220–190 ppm),carbonyl (190–165 ppm), O-aryl (165–140 ppm), aryl (140–110 ppm),O-alkyl (110–45 ppm), and alkyl (45–0 ppm) was calculated byintegration. The limiting chemical shift values for each region were notstrictly adhered to, particularly where large and small bands occurredadjacent to one another. In this case, ‘natural valleys’ between the bandswere judged as more appropriately representing the chemical shiftboundaries (Wilson 1987).

The content of charcoal-C was calculated as g C/kg of soil using thearyl content of photo-oxidised samples. The aryl-C was corrected forpossible lignin contribution and the underestimation of highlycondensed aryl structures such as charcoal as described by Skjemstadet al. (1999).

Calibration procedure

The overall procedure used for calibration is given in Fig. 2. All of thebulked horizons were analysed for TOC by Leco C furnace (carbonateremoved with SO2 solution where necessary), POC by sieving, andcharcoal-C (Char-C) by photo-oxidation and 13C NMR analysis. TheHUM pool was determined as (HUM-C) = (TOC) – (POC) – (Char-C).13C NMR analyses were also carried out on the unfractionated soilsfollowing HF treatment as a separate check on the charcoal analysis.

All soil pool data were converted from g C/kg soil to t C/ha for usein the model. This was accomplished for 0–30 cm bulked samples usingthe formula:

t C/ha = g C/kg × BD × 3 (3)

where BD varied, it was essential to ensure that the same weight of soilper ha was used throughout. Following clearing, agricultural activityusually leads to an increase in BD as a result of a combination oforganic matter loss, compaction through livestock, or mechanicaltraffic. This compaction, which is usually greatest at the surface, resultsin the 30-cm core taken from a compacted site containing more soil byweight for the same volume than its pair taken from a non-compactedsite. Effectively, this is equivalent to sampling the cultivated site to adepth beyond 30 cm.

Corrections were therefore made to all soil TOC and carbon poolvalues to take variations in BD into account. For the 0–30 cm horizons,the equation:

t C/ha (corrected) = t C/ha × (BD virgin/BD cultivated) (4)

was used. This equation results in a slight over-correction because itassumes that OC in each pool is equally distributed throughout the30 cm horizon.

At the Tarlee site, the trial was initialised on existing agriculturalland. Here, the same calculations were applied, using the profile withlowest BD as the starting point for calculations.

Residue inputs were measured on some occasions. When these werenot available, inputs were calculated from grain yield data using aharvest index (HI) derived from data collected for each crop at each site.

HI = grain yield (t/ha)/total above-ground dry matter (t/ha) (5)

For the Brigalow site, the HI for wheat and sorghum was 0.37 and0.40, respectively. For the Tarlee site, the average HI for the wheat andwheat–fallow rotations was 0.45 and 0.43, respectively. Thecontribution from roots was calculated as:

Roots (t/ha) = total above-ground dry matter (t/ha) × 0.40 (6)

All residues were assumed to have a carbon content of 45%.

Results and discussion

Soil organic carbon and pool structure

Brigalow catchment study

Table 4 gives the TOC, POC, IOM, and HUM (bydifference) in t C/ha for 3 soil types from the croppedBrigalow catchment. A consistent decrease in TOC over the18 years after clearing for all soil types by ~30% of the initialvalue was evident (Table 4). POC, on the other hand, showeda much more dramatic decrease by ~70%, most of whichoccurred in the initial clearing/fallow period. The HUM poolshowed a slow decline, whereas the IOM pool remainedessentially constant.

Tarlee rotation trial

The TOC, POC, IOM, and HUM reported in t C/ha for the2 rotations are given in Table 5. The continuous wheatrotation showed a decline in TOC of ~10% over the 8 years,whereas the wheat–fallow showed a slightly higher decline

Table 4. Total organic carbon (TOC), particulate organic carbon (POC), humic (HUM), and inert organic carbon (IOC, Char-C) in

t C/ha for the bulked (0–30 cm) soil samples from the Brigalow Station site

Year POC HUM IOM TOC

Soil 64

1982 21.48 30.06 12.23 63.771984 14.47 28.28 11.38 54.131986 16.04 31.36 11.76 59.161988 10.64 29.99 11.51 52.141994 9.13 29.34 13.63 52.092000 5.72 27.27 12.28 45.26

Soil 65

1982 25.02 26.47 11.60 63.091984 19.61 31.07 11.87 62.551986 18.41 28.59 13.07 60.071988 13.40 29.51 15.03 57.941994 9.58 31.27 14.09 54.952000 8.54 29.33 12.99 50.85

Soil 66

1982 23.03 32.81 10.05 65.891984 13.55 32.75 7.79 54.101986 14.91 29.91 8.98 53.791988 10.07 29.71 7.93 47.701994 6.74 30.70 9.49 46.922000 6.20 28.12 9.89 44.21


of 12%. In both rotations, the relative decline in POC wasmuch faster than the HUM pool, whereas IOM showed nochange.

Application of the RothC model

The data presented in Tables 2–5 and the climate data wereused to initialise the model. POC was used as a surrogate forRPM, Char-C was used as IOM, and HUM was calculated asTOC – POC – IOM. DPM, Bio-F (biomass with fastturnover), and Bio-S (biomass with slow turnover) wereinitially set at zero but were quickly generated by the model.Setting these 3 pools to small realistic values, at the expenseof the RPM or HUM pools, made no significant difference tothe output after the first year and since all runs were >2 years,the setting of these pools to initial values of zero was usedthroughout all modelling runs.

Figure 3 shows the results from the model runs (lines)compared with the measured data (symbols) for the 3

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0 (a)

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

C (

t/ha

)

(b)

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

1982 1987 1992 1997

Year

(c)

Table 5. Total organic carbon (TOC), particulate organic carbon (POC), humic (HUM), and inert organic carbon (IOC, Char-C) in t C/ha for the bulked (0–30 cm) soil samples from the Tarlee site

Year POC HUM IOM TOC

Continuous wheat

1979 8.19 26.97 4.04 39.201985 6.61 25.10 4.09 35.801997 6.63 25.21 4.06 35.90

Wheat–fallow rotation

1979 7.09 27.41 4.10 38.601985 5.93 27.32 4.34 37.601997 4.13 25.76 3.91 33.80

Fig. 3. Predicted change in TOC and pools (lines) compared withmeasured TOC and pools (symbols) at Brigalow Station for the 3 soils:(a) 64, (b) 65, and (c) 66. Original model settings. Modelled TOC(—), HUM (—), IOM (—), RPM (- -); measured TOC(�), HUM (�), CHAR (�), POC (�).

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0 (a)

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

C (

t/ha)

(b)

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

1982 1987 1992 1997

Year

(c)

Fig. 4. Predicted change in TOC and pools (lines) compared withmeasured TOC and pools (symbols) at Brigalow Station for the 3 soils:(a) 64, (b) 65, and (c) 66. Original model settings except that the RPMrate was changed from 0.3 to 0.15/year. Modelled TOC (—), HUM(—), IOM (—), RPM (- -); measured TOC (�), HUM (�),CHAR (�), POC (�).


cropped soils 64, 65, and 66 at Brigalow Station using themodel with all of the original settings. In all cases, the RPMpool declined faster than the measured POC pool. The rateconstant for the RPM pool was therefore decreased, througha series of iterations, from 0.3 to 0.15/year in accordancewith the methodology described in Fig. 2. This approach isjustified considering that the measured and conceptual poolswill not be the same. If, however, the size and turnover ratesare comparable, then relatively small changes in turnoverrate of the conceptual pool to fit the measured rate will affectneither the structure of the model nor its overallperformance. Figure 4 shows model runs with the RPM rateset at 0.15/year, which result in a much better fit between themodelled and measured pool and TOC data.

Of the 3 soils, soil 65 shows the greatest differencesbetween modelled and measured data. Some site variation

may be expected at the Brigalow site, where there was nosignificant ploughing activity for several years after clearing.Soil descriptions of the sites indicate the presence of somegilgai, which may tend to be evened out by tillage. Otherissues include the effects of disturbance caused by clearingand burning of pulled vegetation at the site, prior toestablishment of the crops. The data for the first samplingpoint (model initialisation) are therefore critical for thesubsequent performance of the model. The highly consistentcharcoal (IOM) pool shows that burning at establishmentcontributed little charcoal to the soil, suggesting that thispool had accumulated over a long period of time and washighly stable to decomposition over the period of the trial.

To examine the effects of the initialisation point, the 3soils were started at the second measurement point (1984),which followed initial preparation of the sites and the firsttillage event. Model runs using the 0.15/year rate constantfor the RPM pool are shown for the 3 cropped soils in Fig. 5.The fit of the model to the measured points for TOC and thesoil C pools remains excellent for soils 64 and 66 but is nowsignificantly improved for soil 65. These results suggest thatinitial disturbance has probably affected the soil carbonvalues. Such a disturbance would cause changes in soilcarbon through movement of soil material rather than by

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0 (a)

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

C (

t/h

a)

(b)

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

1984 1989 1994 1999

Year

(c)

Fig. 5. Predicted change in TOC and pools (lines) compared withmeasured TOC and pools (symbols) at Brigalow Station for the 3soils: (a) 64, (b) 65, and (c) 66 but initialising the model at 1984instead of 1982. Original model settings except that the RPM rate waschanged from 0.3 to 0.15/year. Modelled TOC (—), HUM(—), IOM (—), RPM (- -); measured TOC (�), HUM (�),CHAR (�), POC (�).

0.0

10.0

20.0

30.0

40.0

50.0

C (

t/ha)

(a)

0.0

10.0

20.0

30.0

40.0

50.0

1979 1984 1989 1994

Year

(b)

Fig. 6. Predicted change in TOC and pools (lines) compared withmeasured TOC and pools (symbols) at the Tarlee Site for (a)continuous wheat and (b) wheat–fallow. Original model settings.Modelled TOC (—), HUM (—), IOM (—), RPM (- -);measured TOC (�), HUM (�), CHAR (�), POC (�).


decomposition and could therefore not be predicted byRothC or any equivalent soil carbon turnover model.

For the Tarlee site, the RothC model was initialised usingthe TOC and carbon pool structure for the year 1979 and runfor 19 years. The model runs for the 2 rotations and themeasured points for TOC and carbon pools are given inFig. 6. Again, the model was run using the rate constant of0.3/year for the RPM pool, and once again, the rate of declineof the RPM pool was faster than the measured POC pool.Reducing the rate to 0.15/year improved the fit of measuredagainst modelled data for both the RPM and TOC forcontinuous wheat (Fig. 7). For the wheat–fallow rotation, themodel suggests a slightly lower loss of TOC and POC thanthe measured values.

Conclusions

Considering the good agreement between measured andmodelled TOC and pool data across 2 very differentenvironments, 3 soil types, and 3 distinctly differentrotations, we suggest that the pool structure as described herecan be used to initialise RothC. As a result of the goodagreement between measured and modelled TOC and soilOC pool structure, we recommend only one modification to

the model for use in Australia, changing the rate constant ofthe RPM pool from 0.3 to 0.15/year. This fractionationscheme appears to provide a useful means of initialising andtesting the RothC model, not only in Australia but also incountries with similar soil types and climate.

The work reported here demonstrates a uniquemethodology, bringing together measured soil carbon poolswith a conceptual soil carbon turnover model. This approachhas the advantage of allowing the model to be initialised atany point in the landscape without the necessity for historicaldata or for using the model itself to generate an initialequilibrium pool structure. Correct prediction of bothchanging TOC levels and the pool structure over time, acts asan internal verification and gives confidence that the modelis performing as intended.

Acknowledgments

The authors wish to thank Sonia Grocke for technicalassistance with soil fractionation. This work was madepossible through funding from the Australian GreenhouseOffice.

References

Cambardella CA (1998) Experimental verification of simulated soilorganic matter pools. In ‘Soil processes and the carbon cycle’. (EdsR Lal et al.) pp. 519–526. (Lewis Publishers: Boca Raton, FL)

Cambardella CA, Elliot ET (1992) Particulate soil organic matterchanges across a grassland cultivation sequence. Soil ScienceSociety of America Journal 56, 777–783.

Elliot ET, Paustian SD, Frey SD (1996) Modeling the measurable ormeasuring the modelable: A hierarchical approach to isolatingmeaningful soil organic matter fractions. In ‘Evaluation of soilorganic models using long-term datasets’. (Eds DS Powlson et al.)pp. 161–179. NATO ASI Series I: Global Environmental Change,Vol. 38. (Springer-Verlag: Berlin)

Heanes DL (1984) Determination of total organic-C in soils by animproved chromic acid digestion and spectroscopic procedure.Communications in Soil Science and Plant Analysis 15, 1191–1213.

Jenkinson DS (1990) The turnover of organic carbon and nitrogen insoil. Philosophical Transactions of the Royal Society of London B329, 361–368.

Jenkinson DS, Coleman K (1994) Calculating the annual input oforganic matter to soil from measurements of total organic carbonand radiocarbon. European Journal of Soil Science 45, 167–174.

McCown RL, Hammer GL, Hargreaves JNG, Holzworth DP, FreebairnDM (1996) APSIM: a novel software system for modeldevelopment, model testing and simulation in agricultural systemsresearch. Agricultural Systems 50, 255–271. doi:10.1016/0308-521X(94)00055-V

Merry RH, Spouncer LR (1988) The measurement of carbon in soilsusing a microprocessor-controlled resistance furnace. Communi-cations in Soil Science and Plant Analysis 19, 707–720.

Parton WJ, Schimel DS, Cole CV, Ojima DS (1987) Analysis of factorscontrolling soil organic matter levels in Great Plains grasslands. SoilScience Society of America Journal 51, 1173–1179.

Skjemstad JO, Clarke P, Taylor JA, Oades JM, Newman RH (1994) Theremoval of magnetic materials from surface soils. A solid state 13 CCP/MAS n.m.r. study. Australian Journal of Soil Research 32,1215–1229.

0.0

10.0

20.0

30.0

40.0

50.0

C (

t/h

a)

(a)

0.0

10.0

20.0

30.0

40.0

50.0

1979 1984 1989 1994

Year

(b)

Fig. 7. Predicted change in TOC and pools (lines) compared withmeasured TOC and pools (symbols) at the Tarlee Site for (a)continuous wheat and (b) wheat–fallow. Original model settingsexcept that the RPM rate was changed from 0.3 to 0.15/year.Modelled TOC (—), HUM (—), IOM (—), RPM (- -);measured TOC (�), HUM (�), CHAR (�), POC (�).


http://www.publish.csiro.au/journals/ajsr

Skjemstad JO, Dalal RC, Janik LJ, McGowan JA (2001) Changes inchemical nature of soil organic carbon in Vertisols under wheat insouth-eastern Queensland. Australian Journal of Soil Research 39,343–359. doi:10.1071/SR99138

Skjemstad JO, Janik LJ (1996) Climate change: determining thepotential for carbon sequestration in Australian soils. Final Reportto the Rural Industries R & D Corporation, CSO-5A.

Skjemstad JO, Spouncer LR (2003) Integrated soils modelling for theNCAS. National Carbon Accounting System Technical Report No.36. Australian Greenhouse Office, Canberra.

Skjemstad JO, Taylor JA, Smernik RJ (1999) Estimation of charcoal(char) in soils. Communications in Soil Science and Plant Analysis30, 2283–2298.

Sohi SP, Mahieu N, Arah JRM, Powlson DS, Medari B, Gaunt JL(2001) A procedure for isolating soil organic matter fractionssuitable for modeling. Soil Science Society of America Journal 65,1121–1128.

Wilson MA (1987) ‘NMR techniques and applications in geochemistryand soil chemistry.’ (Pergamon Press: Oxford, England)

Manuscript received 20 January 2003, accepted 23 September 2003

calibration of the rothamsted organic carbon turnover model (rothc ver. 26.3), using measurable soil...

Documents