Phosphate adsorption coefficient can improve the validity of RothC model for Andosols

Abstract

The conventionally-modified RothC model for Andosols required pyrophosphate-extractable aluminum (Alp) for changing humus decomposition rate of the original RothC model. However, any Japanese soil database, which were derived from national soil survey projects, did not have the Alp dataset, and thus the conventionally-modified RothC model required Alp prediction from soil organic carbon (SOC) content. From this reason, there is a risk of Alp prediction error in the run-up to predict the SOC turnover. Objectives of this study were (1) to explore the alternative soil property for re-modifying the conventionally-modified RothC model and (2) to validate the re-modified model against long-term experimental data sets of Japanese Andosols. Phosphate adsorption coefficient (PAC), which is an indicator of the content of amorphous aluminum (Al) and iron (Fe) compounds, was tested to replace Alp using three Andosols database. A stability factor, H(f), was defined as the factor needed to divide the decomposition rate constant of the humus pool so that the modeled SOC level matched the measured level. Phosphate adsorption coefficient showed positive correlation with the H(f). The H(f) was regressed by the exponential equation using PAC as an independent variable, and its R ² value was higher than in the Alp derived regression. We incorporated the PAC derived regression into the original RothC model as the PAC-modified RothC model. From the comparison of the models validity, the PAC-modified RothC model showed low mean error with low root mean square error in the long-term experimental data sets. These results indicate that PAC can replace Alp for changing the decomposition rate of humus pool in RothC model with accuracy enhancement.

Keywords:

• Andosols database
• humus pool
• modification RothC model
• phosphate adsorption coefficient
• pyrophosphate-extractable aluminum.

Introduction

Modeling soil organic carbon (SOC) turnover at national scale is essential for developing the national strategies of mitigation and adaptation global warming. It is also of local importance as it describes ecosystem function, soil fertility, water holding capacity, stabilizing soil aggregates and other many functions (The GEFSOC project team 2006). The SOC turnover models have also taken on the prediction of the impact of climate change to SOC turnover on a global scale (Jones et al. 2005), spatial mapping of national SOC stock (van Wasemael et al. 2010; Meersmans et al. 2011), assessing the future carbon sequestration at national and watershed scale (Webb et al. 2003; Milne et al. 2007; Schulp et al. 2008; Smith et al. 2009; Yokozawa et al. 2010), proposing agricultural management practices at the field scale (Liang et al. 2008), and modeling stabilization mechanism of soil structure at the micro scale (Malamoud et al. 2009). Beyond this, SOC turnover models are vital tools for describing ecosystem functions, therefore, establishment and/or purpose-upgrading of SOC models to fit specific environmental conditions are an essential part as a bridge between the local scale and global scale.

The RothC model (Coleman and Jenkinson 1996) is one of the most widely used SOC turnover models from local to global scale researches. The RothC model shows high performance on several land use even though it takes simply monthly input parameters (Smith et al. 1997). This model has been properly validated in a cold region (Romanovskaya 2006), a temperate region (Smith et al. 1997), and a tropical region (The GEFSOC project team 2006). In Japan, Shirato and Taniyama (2003) reported that the RothC model performed well on the non-volcanic ash soils, while it showed high error on the volcanic ash soils (refer to Andosols; IUSS-ISRIC-FAO 2006). Shirato et al. (2004) pointed out that the high error of the RothC model on Andosols was derived from soil organic matter stabilization by aluminum (Al)/iron (Fe)-humus complexes. Furthermore, Shirato et al. (2004) explored the factors for adjusting the decomposition rate of humus in Andosols, and they found that pyrophosphate-extractable Al (Alp) could be used for changing humus decomposition rate constant. This modified RothC model (conventionally-modified RothC model) was used for estimating the carbon sequestration potential in Japanese cultivated soil (Yokozawa et al. 2010).

In spite of the importance of Alp for the conventionally-modified RothC model, any Japanese soil database, which were derived from national soil survey projects, did not have Alp dataset. Thus, Alp estimation from total SOC concentration [Alp (%) = (SOC%–0.95)/5.96; Shoji et al. 1993] was required for running the conventionally-modified RothC model (Yokozawa et al. 2010). There was not only estimation error of the Alp calculation, but also it had the risk of a circular logic; using carbon to calculate Alp, then using the calculated Alp to calculate carbon. For planning authoritative national agricultural strategy, therefore, an establishment of a soil sampling network to construct the Alp database should be challenged. However this kind of project needs enormous time, cost, and labor. Thus, the conventionally-modified RothC model should be re-modified by a general-purpose soil property as an alternative way.

The objects of this study are (1) to explore the alternative soil property for re-modifying the conventionally-modified RothC model and (2) to validate the re-modified model against long-term experimental data sets of Japanese Andosols as done by Shirato et al. (2004).

Materials and Methods

Description of the conventionally-modified RothC model

The original RothC-26.3 model (Coleman and Jenkinson 1996) splits incoming plant residues into decomposable plant material (DPM) and resistant plant material (RPM); these both decompose to form microbial biomass (BIO), humified organic matter (HUM), and evolved carbon dioxide (CO₂). The clay content of the soil determines the proportion that goes to CO₂ or to BIO + HUM. BIO and HUM both decompose to form more CO₂, BIO, and HUM. The original model also includes an inert pool of organic matter (IOM).

Shirato et al. (2004) set IOM content at zero, and they introduced a stability factor “H(f)” to change the decomposition rate of the HUM pool,

where SOC_modeled and HUM_modeled were derived from running the original RothC model to steady-state (10 000 years) using monthly climate parameters and potential carbon input defined by vegetation type as mentioned later. The H(f) is a factor needed to change the HUM decomposition rate so that the modeled total carbon matched the measured value. The H(f) had a positive correlation with Alp, and it was linearly regressed by Alp (R ²= 0.518). And, Shirato et al. (2004) incorporated the regression equation into the original RothC model changing the decomposition rate of the HUM pool as the conventionally-modified RothC model.

Phosphate adsorption coefficient

As we mentioned earlier, we explore the alternative soil property (general-purpose property) for changing the decomposition rate of the HUM pool. As a general-purpose soil property, a potent one is the phosphate adsorption coefficient (PAC), which is an indicator of the content of amorphous Al and Fe compounds. The PAC increases with increasing contents of amorphous Al and Fe compounds. The PAC is used as a criterion for the Andosol group in the Japanese soil classification system (Classification Committee of Cultivated Soils 1996), and it is also used to determine fertilizer application rate on cultivated Andosol fields in Japan. The PAC is determined by the 2.5% ammonium phosphate method: 25 g of soil (dry weight) is placed in a 100 ml flask, 50 ml of a 2.5% ammonium phosphate (pH = 7) is added, left for 24 hours, then the concentration of residual phosphate in the filtrate is determined by a colorimetric method (Crop Production Division 1979).

Model modification using 33 data sets on Andosols

Table 1 shows 33 data sets, including forest and grasslands, from the Andisol TU Database 1992 (Shoji et al. 1993), the Ando Soils in Japan database (Wada et al. 1986), and Soil Monolith database of the National Institute for Agro-environmental Sciences (http://soilgc.job.affrc.go.jp/monolith/). Each site was standardized to a depth of 20 cm, and the clay content, SOC content, bulk density, Alp, and PAC were weighted accordingly. Shirato et al. (2004) used 32 data sets from the Andisol TU Database to modify the original RothC model. In this study, we selected 19 sites from the Andisol TU Database, in which PACs were determined.

Table 1 Summary of parameterization data sets for upgrading the conventionally-modified RothC model

Site name	DB_Name	Vegetation	Soil temperature regime†	Lat. N	Long. E	MAAT	C input	PAC	Clay	Alp	SOC	Original model		H(f)
												SOC	HUM
						(^oC)	(Mg C ha^−1y^−1)	(mg P2O5/100g soil)	(%)	(%)		(Mg C ha^−1)
Abashiri1	A_TU	DF	Frigid	43.833	144.268	5.4	1.63	850‡	5.3	0.28	40	22.1	14.7	2.2
Abashiri2	A_TU	DF	Frigid	43.973	144.152	5.5	1.64	1650‡	19.3	0.78	118	28.3	20.3	5.4
Abashiri3	A_TU	F	Frigid	43.834	144.267	5.8	1.67	1367‡	17.0	0.58	86	27.0	19.2	4.1
Erimo1	A_TU	G	Mesic	41.953	143.216	7.1	3.80	1740	46.0	1.11	235	68.1	53.9	4.1
Fujisawa	A_TU	CF	Mesic	39.306	141.118	10.4	2.11	2064	37.3	1.37	103	26.1	19.4	5.0
Fusiko	A_TU	DF	Frigid	42.841	143.098	5.8	1.67	1195	10.3	0.53	33	24.5	16.9	1.5
Kanayama	A_TU	CF	Mesic	37.439	139.557	8.3	1.89	1861	11.6	1.37	148	22.7	15.8	8.9
Kitamotoimatsu	A_TU	NG	Frigid	42.838	143.122	5.8	1.67	849	5.1	0.45	84	47.3	33.8	2.1
Kitayama	A_TU	MS	Mesic	38.755	140.755	9.4	3.49	2011	31.3	1.52	150	44.3	34.6	4.1
Koyama	A_TU	MS	Thermic	35.516	134.176	14.4	3.72	1978	51.0	1.32	99	33.6	26.5	3.5
Meotozaka	A_TU	PF	Mesic	39.148	140.996	9.4	2.00	2180	52.8	1.75	134	29.0	21.6	5.9
Mukaiyama	A_TU	MS	Mesic	38.712	140.707	9.0	3.49	1976	27.0	1.54	130	45.3	35.2	3.4
Ohkura	A_TU	MS	Mesic	38.661	140.207	9.9	3.49	1680	21.6	1.23	158	40.7	31.3	4.7
Ohnodai	A_TU	CF	Mesic	40.187	140.327	9.6	2.03	2299	25.3	2.09	143	27.3	19.9	6.8
Rokuhara	A_TU	CF	Mesic	39.228	141.094	10.4	2.11	2105	47.3	1.54	109	27.3	20.3	5.0
Sasagaharu	A_TU	PF	Mesic	35.360	133.757	13.0	2.41	2363	36.0	2.46	161	23.5	17.4	8.9
Shinjo	A_TU	CF	Mesic	38.817	140.333	9.8	3.00	1695	48.0	1.02	131	38.4	30.3	4.1
Takada	A_TU	S	Mesic	37.452	139.823	10.5	3.00	2050	21.2	1.47	133	31.8	24.4	5.1
Wakami	A_TU	PF	Mesic	39.964	139.915	10.7	2.14	1407	39.9	0.73	106	27.5	20.4	4.8
Erimo2	ASJ	P	Mesic	41.960	143.208	7.0	3.80	1743	33.3		222	67.0	52.6	3.9
Hiroshima	ASJ	F	Mesic	42.949	141.471	6.8	1.75	1349	18.9		109	27.5	19.7	5.1
Imaichi	ASJ	DF	Mesic	36.683	139.767	11.6	2.24	2528	38.7		144	25.0	18.6	7.4
Kanagasaki	ASJ	PF	Mesic	39.133	140.983	9.3	1.99	2075	49.0		131	29.1	21.7	5.7
Kitakami	ASJ	F	Mesic	39.317	141.117	10.6	2.13	1504	58.8		32	27.2	20.3	1.2
Shizukuishi	ASJ	P	Mesic	39.733	141.000	9.2	3.80	1965	32.8		143	50.7	39.6	3.3
Takizawa	ASJ	P	Mesic	39.783	141.117	9.2	3.80	2089	34.8		124	49.4	38.7	2.9
Aso	Monolith	DF	Mesic	33.004	131.013	10.3	2.09	2433	34.0		107	29.4	21.7	4.6
Fujinomiya	Monolith	F	Mesic	35.372	138.650	9.1	1.97	2327	22.0		89	28.7	20.8	3.9
Fujinomiya2	Monolith	CyP	Mesic	35.370	138.637	9.7	2.04	2472	41.0		116	28.9	21.5	5.0
IWATA	Monolith	F	Thermic	34.681	137.839	16.1	2.83	1450	16.4		57	20.9	14.8	3.4
Memuro	Monolith	LF	Mesic	42.888	143.009	6.1	1.69	1273	17.4		43	27.1	19.3	1.8
Tama	Monolith	DF	Thermic	35.640	139.383	13.8	2.51	1333	63.5		43	25.1	18.8	1.9
Yasato	Monolith	CF	Thermic	36.210	140.157	13.8	2.51	2066	42.0		79	25.2	18.7	3.9
MAAT, mean annual air temperature; PAC, phosphate adsorption coefficient; Alp, pyrophosphate-extractable aluminium; SOC, soil organic carbon; HUM humified organic matter; H(f), stability factor.
DB_Name (Database name): A_TU, Andisol TU database; ASJ, Ando Soils in Japan.
Vegetation: DF, Deciduous forest; F, Forest (species unknown); G, Grassland (species unknown); CF, Cedar forest; NG, Natural grassland; MS, Miscanthus sinensis; PF, Pine forest; S, Scrub; P, Pasture; CyF, Cypress forest; LF, Larch forest.
All the soil properties were standardized to a depth of 20 cm. H(f), which is a conversion factor used to change the decomposition rate for the HUM pool, was calculated using the original RothC with the inert pool of organic matter (IOM) content set at zero by the equation: H(f) = [SOCmeasured–(SOCmodeled–HUMmodeled)]/HUMmodeled.
†Derived from “Soil Information Web Viewer”(National Institute for Agro-Environmental Sciences, http://agrimesh.dc.affrc.go.jp/soil_db/), ‡Derived from Hakamada (1977).

Monthly average air temperature (MT) and monthly precipitation (MP) were obtained from the Japan Meteorological Agency (2002). The potential evapotranspiration was calculated using the Thornthwaite method (Thornthwaite 1948).

Since actual plant input carbon from vegetation is unknown, we estimated them from MT for each forest site using a meteorological model (Seino 1990).

We set carbon concentration of the forest product as 45%. Annual organic matter input for grassland was set to 9.5 Mg ha^–1 and 40% carbon concentration resulted in 3.8 Mg ha^–1yr^–1 of carbon input under grass vegetation (Shirato et al. 2004). Annual organic matter yield of natural grassland (Miscanthus sinensis) in Kitayama, Mukaiyama, and Ohkura were set at 8.7 Mg ha^–1yr^–1, and in Koyama set at 9.3 Mg ha^–1yr^–1 following the report of the National Grassland Research Institute (1993). The carbon concentration of Miscanthus sinensis is about 40%, so the annual input carbon of 3.49 Mg ha^–1yr^–1 and 3.72 Mg ha^–1yr^–1 were used, respectively. Annual carbon input for scrub was set to 3.0 Mg C ha^–1yr^–1 (Shirato et al. 2004). A DPM/RPM ratio of 1.44 was used for grassland, 0.67 for Miscanthus sinensis and scrub, and 0.25 for forest. Soils were assumed to be covered with vegetation throughout the year. We set the IOM to zero for all sites (Shirato et al. 2004).

Model validation against long-term experiments

Comparison of model validity was conducted using four long-term field experiments in the same manner as the previous report (Shirato et al. 2004). Tables 2 and 3 show a summary of four long-term experimental sites. Soil sampling for PAC measurement was conducted at Kitami and Shiojiri in 1999, and it was also undertaken at Fujisaka and Osumi in 2000 and 2002, respectively. In the four sites, soil sampling was done in each treatment. Phosphate adsorption coefficient was measured using the earlier mentioned method, and average PAC of all treatments were used to calculate H(f) values for running the PAC-modified RothC model in each site. The Alp values in Table 2 were predicted using the following equation (Shoji et al. 1993):

where SOC is an initial concentration of SOC at each site. The predicted Alp values were used to calculate H(f) values for running the conventionally-modified RothC model in each site. Detailed information about treatments, average carbon input in each site, and years of SOC measurement for model validation are listed in Table 3.

Table 2 Summary of four long-term experiments on Andosols selected for model validation

				Initial			PAC-modified model			Conventionally-modified model
Site	MAAT	MAP	Clay	SOC	IOM^†	C input^‡	PAC	H(f)	C input	Predicted Alp	H(f)	C input
	(^oC)	(mm)	(%)	(Mg ha^−1)		(Mg ha^−1y^−1)	(mg P2O5/100g soil)		(Mg ha^−1y^−1)	(%)		(Mg ha^−1y^−1)
Kitami	5.2	786	18.0	110	10.4	6.5	1216	2.68	3.09	0.86	4.24	1.96
Fujisaka	9.8	1115	15.3	85	7.8	6.4	1532	3.35	2.71	0.85	4.45	2.04
Shiojiri	10.5	1170	38.0	85	7.8	5.9	1768	3.95	2.11	0.60	4.45	1.83
Osumi	16.4	2542	18.1	92	8.5	12.7	1895	4.32	3.91	1.09	4.85	3.37
C, Carbon; H(f), stability factors; IOM, inert pool of organic matter; MAAT, Mean annual air temperature; MAP, Mean annual precipitation; PAC, phosphate adsorption coefficient; SOC, soil organic carbon.
†IOM for original model calculated by equation of Falloon et al. (1998).
‡C input needed to produce initial SOC content at equilibrium.

Table 3 Average carbon input from crop residues, farm-yard manure (FYM), and years of soil organic carbon (SOC) measurement in each treatment for long-term experiments

		Average carbon input
Site	Treatments	Residue (Mg ha^–1yr^–1)	FYM (Mg ha^–1yr^–1)	Years of SOC measurement
Kitami started from 1959	Continuous fallow	0.00	0.00	1970, 1976, 1982, 1991, 1999
	Continuous potato	0.07	1.12	1970, 1976, 1982, 1991, 1999
	Continuous sugar beet	0.00	1.12	1970, 1976, 1982, 1991, 1999
	Continuous wheat	0.51	0.67	1970, 1976, 1982, 1991, 1999
	Continuous oat	0.17	0.00	1970, 1976, 1982, 1991, 1999
	Continuous soybean	0.12	0.00	1970, 1976, 1982, 1991, 1999
	Continuous kidneybean	0.10	0.00	1970, 1976, 1982, 1991, 1999
Fujisaka started from 1936	Continuous wheat	0.56	0.25	1971, 1995
	Continuous potato	0.15	25.00	1971, 1995
	Continuous soybean	0.08	0.25	1971, 1995
	Continuous maize	0.59	0.25	1971, 1995
	NPK^†	0.11	0.00	1948, 1955, 1965, 1977, 1995, 2000
	NPK + FYM^†	0.29	0.38	1948, 1955, 1965, 1977, 1995, 2000
	NPK + FYM + lime^†	0.34	0.38	1948, 1955, 1965, 1977, 1995, 2000
Shiojiri^‡ started from 1938	None fertilizer application	0.02	0.00	1949, 1961, 1975, 1989, 1997, 1999
	–N	0.09	0.00	1949, 1961, 1975, 1989, 1997, 1999
	–P	0.01	0.00	1949, 1961, 1975, 1989, 1997, 1999
	–K	0.05	0.00	1949, 1961, 1975, 1989, 1997, 1999
	NPK	0.13	0.00	1949, 1961, 1972, 1975, 1989, 1997, 1999
	FYM	0.09	0.87	1949, 1961, 1975, 1989, 1997, 1999
	–N + FYM	0.17	0.87	1961, 1975, 1989, 1997, 1999
	–P + FYM	0.07	0.87	1961, 1975, 1989, 1997, 1999
	–K + FYM	0.21	0.87	1961, 1975, 1989, 1997, 1999
	NPK + FYM	0.26	0.87	1949, 1961, 1972, 1975, 1989, 1997, 1999
Osumi^§ started from 1988	–N	0.41	0.00	1988(2), 1989, 1992, 1993–2001(2), 2002
	NPK	0.63	0.00	1988(2), 1989, 1992, 1993–2001(2), 2002
	NPK + pig manure	0.72	2.29	1988(2), 1989, 1992, 1993–2001(2), 2002
	Pig manure	0.65	5.35	1988(2), 1989, 1992, 1993–2001(2), 2002
	Cattle manure	0.61	10.66	1988(2), 1989, 1992, 1993–2001(2), 2002
	Chicken manure	0.60	2.86	1988(2), 1989, 1992, 1993–2001(2), 2002
Nitrogen, N; phosphorus, P; potassium, K.
^†Rotation of potato–maize–soybean.
^‡Maize: 1938–1966, vegetables: 1967–2000.
^§Barley or wheat in winter, sweet potato in summer.

In modeling each set of experimental data, it was first necessary to run the model to produce an initial SOC content that was the same as that originally present in the soil. The amount of carbon input from plant residue was calculated by running the model inversely, assuming that SOC had been at equilibrium when the experiment started. A DPM/RPM ratio of 1.44 was used for the Kitami and Osumi sites, and 0.25 for the Fujisaka and Siojiri sites, according to Shiato et al. (2004). The IOM for the original RothC model was set using the following equation (Falloon et al. 1998).

IOM was set to zero for the conventionally-modified RothC model and the re-modified RothC model. The original vegetation of Kitami, Fujisaka, Shiojiri, and Osumi before the experiment started were grassland, forest, pine forest, and upland, respectively. Soil was assumed to be covered with vegetation throughout the year.

Once the starting SOC content had been established, the model was run with carbon input and soil cover information for each treatment. The amount of carbon input from farm-yard manure and crop residue (e.g. roots and stubble) was set to the same values as Shirato et al. (2004). The DPM/RPM ratio of 1.44 was used for all kinds of incoming crop residues.

Statistical analysis

Descriptive statistical analysis, one-way analysis of variance (ANOVA), and regression were conducted using JMP 8.0 (SAS 2008). The Turkey–Kramer HSD (Honestly Significant Difference) test was carried out to evaluate statistical significance of difference in the modeled/measured SOC ratio.

Results and Discussion

Exponential regression or linear regression were employed to determine the relationship among H(f), Alp, and PAC. Higher R ² values were observed in the exponential regression, H(f) = a × e ^{b ×PAC(or Alp)} than in the linear regression (data not shown). In the Andisol TU Database, the higher R ² was observed in the PAC equation (0.57) than in the Alp equation (0.51) (Fig. 1). Thus, this result suggests that PAC is an available soil property to change humus decomposition rate constant in the RothC model. Since, F value and P value of the regression analysis using PAC were smaller in the whole dataset than in the Andisol TU Database use only, we incorporated the whole dataset regression into the original RothC model as the PAC-modified RothC model.

Figure 1 Results of regression analysis. The dependent variable is the stability factor, H(f), and the independent variables are pyrophosphate-extractable aluminum (Alp) (a), and phosphate adsorption coefficient (PAC) (b) and (c). Parts (a) and (b) were derived from the Andisol_TU database, and (c) was derived from the whole dataset. The F value was obtained by dividing the explained variance by the unexplained variance; Y stands for H(f) value; N, number of samples using regression analysis; P, probability in statistical significance testing; R² is the square of the sample correlation coefficient between the outcomes and their predicted value.

Figure 2 shows the examples of model validation using long-term field experiments. The original RothC model tended to predict much lower SOC content than the measured SOC data. As mentioned in Shirato et al. (2004), the original model could not explain the resistant character of humus decomposition on Andosols. By contrast, SOC content, which were predicted by the conventionally-modified RothC model and the PAC-modified RothC model were very close to those measured data during the 50 years after the experiments started. Since, the H(f) in the PAC-modified RothC model was lower than in the conventionally-modified RothC model (Table 2) in all four sites, the conventionally-modified RothC model predicted lower SOC content than the PAC-modified RothC model.

Figure 2 Examples of model validation using long term field experiment. Upper figures show the results of non-farm-yard manure (FYM) application plots, and lower ones show the results of FYM application plots. The lines show the predicted soil organic carbon (SOC) dynamics following each scenario of field management. Nitrogen, N; phosphorus, P; potassium, K.

Table 4 Comparison of the models validity

	Kitami (n = 35)		Fujisaka (n = 26)		Shiojiri (n = 59)		Osumi (n = 138)
	RMSE	ME	RMSE	ME	RMSE	ME	RMSE	ME
Original model	21.94	20.84	25.85	24.11	25.12	24.20	21.24	14.94
Conventionally-modified model	5.92	–2.85	9.10	–4.26	11.77	–6.73	13.14	–3.49
Phosphate adsorption coefficient-modified model	6.57	4.02	8.06	0.88	10.06	–4.18	12.97	–2.37
RMSE, root mean square error; ME, mean error.

Table 4 shows the comparison of the model validity among the original RothC model, the conventionally-modified RothC model, and the PAC-modified model. The original model shows high mean error (ME) with high root mean square error (RMSE) in the four experimental sites. RMSE was lower in the PAC-modified RothC model than in the conventionally-modified RothC model except for the Kitami site. Moreover, negative ME values were observed in the conventionally-modified RothC model at all four sites. This result indicates that the conventionally-modified RothC model tended to underestimate the humus decomposition rate, or it tended to overestimate the SOC content. The conventionally-modified RothC model has a risk for Alp prediction error using SOC content, and it might be suggested that calculation H(f) in the conventionally-modified RothC model was over estimated. Moreover, absolute value of ME in the PAC-modified RothC model was lower than in the conventionally modified RothC model. This result shows that the PAC-modified RothC model is a less-bias prediction model of SOC dynamics.

PAC is a familiar soil property for Andosols covering the agricultural production area in Japan. About 45% of the Japanese upland cultivated area is covered with Andosols (Takata et al. 2011). The PAC-modified RothC model could be a promising new tool for proposing agricultural management practices at farmland level, estimation of carbon sequestration, spatial mapping of SOC stock, and the like in those areas. As a conclusion of this study, PAC can replace Alp for changing the decomposition rate of HUM pool in RothC model with accuracy enhancement.

Acknowledgments

The authors would like to thank Professor David Jenkinson (IACR-Rothamsted) for advice on RothC; Dr Harushi Kikuchi (Fujisaka branch of the Aomori Agricultural Experiment Station), Dr Hiroyuki Siga and Dr Kazuo Konno (Kitami Agricultural Experiment Station), Dr Toshifumi Murakami and Dr Seishi Yoshida (Nagano Chu-sin Agricultural Experiment Station), and Dr Hidemi Wakikado and Dr Hiroharu Furue (Kagoshima Agricultural Experiment Station) for assisting with soil sampling and data collection; Dr Takashi Kusaba (National Agricultural Research Center) for helping with the collection of long-term experimental data sets; and Miss Asuka Iwano (Toyohashi University of Technology) for arranging long-term experimental data. This work was supported by the Research Project Fund of the Ministry of Agriculture, Forestry, and Fisheries, Japan (Development of mitigation and adaptation techniques to global warming in the sectors of agriculture, forestry, and fisheries) from 2010 to 2014.