| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a School of Resource Management, the Univ. of Melbourne, Parkville 3052, Victoria, Australia
b CSIRO Land and Water, GPO Box 1666, Canberra 2601, ACT, Australia
c Present address: NR&M, 80 Meiers Rd, Indooroopilly, Brisbane, Qld. 4068, Australia
* Corresponding author (weijin.wang{at}nrm.qld.gov.au).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: DOC, dissolved organic C • DON, dissolved organic N • ka, mineralization rate constant for Na • ks, mineralization rate constant for Ns • LFC, light fraction organic C • LFN, light fraction organic N • MBC, microbial biomass C • MBN, microbial biomass N • Na, active organic N pool • Ns, slow organic N pool • TON, total organic N • WHC, water-holding capacity
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Many kinetic models such as sigmoid, hyperbolic, and single exponential plus linear models have been used to describe inorganic N production from soil or amended organic materials (Juma et al., 1984; Broadbent, 1986). However, few authors have fully explained the theoretical implications of the parameters in these mathematical equations (Ellert and Bettany, 1988). Consequently, first-order kinetic models including the single (Stanford and Smith, 1972) and the double (Molina et al., 1980) exponential models remain widely used.
The double exponential model separates the mineralizable organic N into active and slow pools and can be presented as:
 | [1] |
There have been appeals to link the conceptual pools in soil organic matter models to measurable fractions (Elliott et al., 1996). Mechanistic models usually allow for replenishment of the pools, whereas the kinetic model does not. To relate the kinetically determined pools to soil properties is of theoretical as well as practical interest. Nonetheless, there have been concerns that the estimated pool sizes and their rate constants in Eq. [1] could be affected by the incubation time (Cabrera and Kissel, 1988b; Dou et al., 1996). This led Sierra (1990) to comment, "the assumption of the existence of discrete size pools of mineralizable N is a faulty concept." Similarly, Dou et al. (1996) concluded that the model parameters are merely mathematically defined quantities and do not represent soil N mineralization potentials.
It is generally thought that the pool sizes of soil organic matter are soil specific, while their mineralization rate constants vary with environmental conditions. Thus, net N production under field conditions can be predicted using the soil-specific pool sizes and their temperature- and moisture-dependent rate constants. If the above assumption is valid, one may expect that Na or Ns estimated under different temperatures and moistures should be similar for a given soil. However, few studies, if any, have experimentally examined the impacts of temperature and moisture variations on the kinetically estimated Na, Ns, ka, and ks, in spite of the fact that the incubation conditions used to obtain N mineralization data often differed considerably in different studies (Ellert and Bettany, 1988; Dendooven et al., 1997).
Satisfactory predictions of net N production in the field have been achieved using the kinetic models determined from laboratory incubation (Stanford et al., 1977; Marion et al., 1981; Campbell et al., 1984). However, overestimations have been occasionally reported by others (Verstraete and Voets, 1976; Cabrera and Kissel, 1988a). The overestimation could have been due to several causes including the crushing and air-drying of soil samples before incubation, removal of plant residues with high C/N ratios from incubated samples, plant debris or exudates inputs and/or possible N losses from soil under filed conditions. Besides, the validity of using model parameters estimated from laboratory incubation could be affected by the fluctuations in temperature and moisture in field conditions (Sierra, 2002).
The objectives of this study were to (i) re-appraise the meaningfulness of the active and slow soil organic N pools and their mineralization rate constants obtained with the conventional double exponential model; (ii) examine the temperature and moisture dependence of the model parameters; (iii) modify the structure and fitting procedure of the model to estimate soil-specific pool sizes and temperature- and moisture-dependent mineralization rate constants; (iv) test the feasibility of using laboratory data to predict net N mineralization under field conditions; and (v) relate the kinetically estimated pool sizes to basic soil properties and several physical, chemical, and biological fractions of readily mineralizable organic N.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Laboratory Incubation
The incubation procedures have been described in detail in Wang et al. (2003b). Briefly, triplicate soil samples equivalent to 200 g of oven-dry matter each were incubated in 2-L glass jars at the near optimum temperature (35°C, Stanford and Smith, 1972) and moisture (55–65% WHC) for 41 wk. The jars were normally closed but opened periodically to maintain an aerobic condition for the soils. Water loss in the jars was monitored by weight and replenished every time the jars were opened. No leaching was conducted during the course of incubation for reasons discussed in Wang et al. (2003b). The cumulative amount of inorganic N in soil was determined by sampling at 0, 1, 2, 4, 6, 9, 12, 16, 20, 24, 29, 35, and 41 wk. Net N mineralization during a period of time was calculated by subtracting inorganic N content at time zero from that at the time of sampling.
Soil 15 (Kandosol in Australian classification or Palexeralf in U.S. taxonomy, 220 g clay kg–1, 14 mg TOC kg–1, and 1.1 mg TON kg–1) and Soil 17 (Chromosol in Australian classification or Rhodoxeralf in U.S. taxonomy, 120 g clay kg–1, 21 mg TOC kg–1, and 1.6 mg TON kg–1) were incubated under four different temperatures (5, 15, 25, or 35°C) in factorial combination with four moisture levels (7, 11, 15, or 19% w/w) for 29 wk to study the effects of temperature and moisture on N production kinetics (Wang et al., 2003b). Net N mineralization in soil was determined by intermittent sampling as above.
Field Incubation
Soils 15 and 17 were used for this experiment. Air-dried sample equivalent to 100 g of oven-dry soil was weighed into twenty 200-mL polystyrene jars for each soil. The soils were then moistened to 55% WHC (19% w/w) with water. The jars of each soil were placed in a plastic box (40 x 30 x 25 cm3) that was then covered with a lid drilled with 15 evenly distributed holes (5 mm dia.) and half-buried into the ground in a birdcage. A wooden cover was placed about 30 cm above each box to shield from direct sunlight and rainfall. The temperature in each box was measured with six thermocouples randomly placed in soil at different locations within the box and recorded every 15 min with a data logger (Wesdata 692, Australia). Soil moisture was monitored by weight every 3 to 5 d, depending on the rate of moisture loss. The moisture in different jars was maintained at the same level by adding water, and controlled within the range of 8 to 19% (w/w) to avoid denitrification and drying-wetting effects on mineral N accumulation. At various time intervals, triplicate jars were destructively sampled for mineral N determination. This experiment was undertaken from 6 Sept. 1999 to 10 Jan. 2000 (the soils were incubated in the laboratory for the last 17 d during Christmas holiday).
Model Fitting and Statistical Analysis
Equation [1] was fitted with the Fit Curve procedures of SigmaPlot 7.1 (SPSS Inc., IL) that uses the Marquardt–Levenberg algorithm and an iterative process to find the parameter values that minimize the residual sum of squares. The resultant pool sizes and their mineralization rate constants, particularly ka, could be sensitive to the initially assigned parameter values and the iteration step size. Generally, the automatically estimated initial parameters resulted in acceptable parameter values. In some cases, manual adjustment of the initial parameter values and the step size was needed to obtain sensible results (e.g., ka should not be greater than a few orders of magnitude; and the pool sizes should not be negative).
Stepwise multiple regression was performed using SigmaStat, Version 2.03 (SPSS Inc., Chicago, IL) to relate the model pool sizes with the physical, chemical, and biological properties of soil. Analysis of variance was performed using the procedure of GenStat 6th ed., Release 6.1 (Payne, 2002). The difference between treatment means was tested using the least significant difference (LSD) at levels of P < 0.05 and P < 0.01.
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
The sum of Na + Ns initially given by the non-linear regression procedure exceeded the TON for Soils 6, 8, 10, 11, and 17 (data not shown), which was similar to the findings of Dendooven et al. (1997). When a constraint of Na + Ns 
TON was applied in the curve-fitting procedure, the values of R2 and residue mean square (RMS) were not significantly affected. The standard error for a parameter could be multiples of the parameter value per se, suggesting that the model parameters obtained from the non-linear regression were not necessarily unique for a data set. Furthermore, the mineralization rate constants varied widely across soils (ka = 0.065–37 wk–1; ks = 0.0007–0.108 wk–1). At times, ks for one soil was greater than ka for another. Therefore the active or slow pool as mathematically estimated by the regression process had different meanings for different soils, and no single parameter can be used as an indicator of soil N mineralization capacity.
Furthermore, the magnitudes of Na, Ns, ka, and ks varied considerably with incubation time for most soils. Cabrera and Kissel (1988b) found that the pool sizes increased, whereas their rate constants decreased as the incubation time increased. Nonetheless, the opposite results were also obtained for some soils in the present study (e.g., Soils 2 and 3; Table 1). For several soils (e.g., Soils 5, 6, 8, 16, and 18), changes in the magnitudes of Na, Ns, ka, and ks with time had no consistent trend, although close fitting to N mineralization data was always achieved (R2 > 0.99 in most cases).
|
Influences of Temperature and Moisture on Model Parameters
The dynamics of net N mineralization for Soils 15 and 17 at different temperature and moisture combinations were given in Wang et al. (2003b). The cumulative net N mineralization increased consistently with increasing temperature and moisture within the tested ranges. The double exponential model closely fitted the experimental data from 0 to 29 wk under all the temperature-moisture combinations for both soils (R2 = 0.945–0.999; median 0.995). Unlike the cumulative net N mineralization, however, changes in Na and Ns with temperature and moisture were largely inconsistent and unpredictable. Poor trends were also observed for the values of ka and ks (data not presented).
Changes in the kinetically determined pool sizes, and inconsistent or little changes in the rate constants with temperature and soil moisture contradicted the general assumption in modeling that the sizes of soil organic matter pools should be unaffected by, whereas their mineralization rate constants should be a function of the environmental factors (Stanford et al., 1977; Parton et al., 1987; Jenkinson, 1990; Goncalves and Carlyle, 1994). The temperature dependence of the kinetically estimated pool size rather than the rate constant was interpreted by Zogg et al. (1997) as the result of alternation in microbial composition, wherein dominant populations at higher temperatures have the ability to mineralize substrates that are not used by microbes at lower temperatures. Similarly, Dalias et al. (2003) concluded that the substrate pools should be defined not only by their chemical quality but also by the temperature under which decomposition takes place.
In the present study, the patterns of mineral N accumulation appeared linear at lower temperatures, but curvilinear at higher temperatures (Wang et al., 2003b). This is probably because demand for the readily mineralizable organic N by microbes was so small at lower temperatures that substrate supply was not the primary limiting factor for their activity throughout the period of incubation; while at higher temperatures the readily mineralizable organic matter was consumed faster and N mineralization gradually slowed down due to the limit of substrate availability. The values of the model parameters are affected by net N mineralization patterns and by errors in the experimental data. However, we hypothesize that the meaninglessness of the estimated model parameters and their inconsistent changes with time, temperature and moisture were largely because Eq. [1] has too many (four) unknown parameters to be derived by the iterative curve-fitting procedure and these parameters (particularly Na vs. ka and Ns vs. ks) are interdependent.
Model Modification
The conventional double exponential model (Eq. [1]) was modified following the approach of Christensen and Olesen (Christensen and Olesen, 1998). Based on the ka values given by Eq. [1] and visual inspection of the mineral N accumulation dynamics (Fig. 1), a half-life (t1/2) of 1 wk was used to define the turnover rate of Na that was responsible for the initial net N mineralization flush. As 1/2Na = Nae–kat1/2, the mineralization rate constant for Na would therefore be 
a = ln 2/t1/2 = 0.693 wk–1. The mineralization rate constant for Ns was defined as s = 0.054 wk–1, an average value obtained by Stanford and Smith (1972) after removal of the first two weeks' data. So the conventional double exponential model (Eq. [1]) was modified into a model with two unknown parameters (Na and Ns), each with a standard mineralization rate constant:
 | [2] |
Moderate variation in the empirically defined ka and ks values would not significantly affect the goodness of fit to data; a higher rate constant would normally result in a lower pool size, and vice versa. The ka and ks values defined in this study are considered appropriate for the incubation conditions (at 35°C for 41 wk). Indeed, there have been no universally agreed values for the empirically defined active and slow pools. The rate constants for a namely synonymous pool in different mechanistic models could differ by many folds (Molina and Smith, 1998), mainly depending on the time scale of prediction. If an incubation is conducted for a period much longer than in the present study, the rate constant for the slow pool may need to be adjusted, or a resistant pool with a lower rate constant can be added in the model.
Equation [2] fitted the net N mineralization data from 0 to 41 wk very well (Fig. 1); the R2 values were in the range of 0.958 to 0.998, which were identical to or only slightly lower (mostly in the third decimal) than those for Eq. [1]. In contrast to the model with four unknown parameters (Eq. [1]), use of fixed 
a and s in fitting Eq. [2] resulted in Na or Ns values that were remarkably similar at different lengths of incubation after 20 wk (partly shown in Table 1). Therefore, the modified double exponential model successfully solved the problem of time-dependence in Na and Ns values that has been a concern to many researchers (Cabrera and Kissel, 1988b; Sierra, 1990; Dou et al., 1996; Dendooven et al., 1997).
Equation [2] also eliminated the confounding effect of the otherwise variable rate constants on pool size estimates. This allowed soil N mineralization capacity to be expressed explicitly with pool sizes. A high Na value should indicate a large initial N mineralization flush, and a high Ns value should suggest a large long-term N mineralization rate. The use of fixed a and s values provides a basis toward a more standardized concept of active or slow pool, that is, Na or Ns for different soils in different studies would be comparable, providing the incubation conditions are standardized or corrected (e.g., 35°C and approximately 55% WHC or field capacity).
Temperature and Moisture Effects on Rate Constants of the Modified Model
Because the pool sizes in Eq. [2] were insensitive to the length of incubation after 20 wk, the Na 
and Ns 
were estimated for Soils 15 and 17 using the N mineralization data under 35°C and 55% WHC (19% w/w for both soils) from 0 to 29 wk, a duration close to that used by Stanford and Smith (1972) and many others. Assuming that Na (14 mg kg–1 for Soil 15 and 36 mg kg–1 for Soil 17) and Ns (199 mg kg–1 for Soil 15 and 247 mg kg–1 for Soil 17) do not change with temperature and moisture, Eq. [1] was modified to fit the data of net N mineralization under other temperature and moisture treatments by using fixed pool sizes and variable rate constants as below:
 | [3] |
 | [4] |
Equations [3] and [4] closely described net N mineralization dynamics under different temperature and moisture conditions (R2 = 0.915–0.998 and median = 0.985 for Soil 15; R2 = 0.881–0.998 and median = 0.994 for Soil 17). Compared with the single exponential model with a fixed N0 and an unknown k (Wang et al., 2003b), better fittings to the experimental data were achieved in many cases with Eq. [3] and [4], as indicated by higher R2 and lower RMS values (data for RMS not presented).
In contrast to the results obtained with Eq. [1], the mineralization rate constants of the slow pool for Soil 15, and of both the active and slow pools for Soil 17, showed clear trends of linear increases with increasing moisture (Fig. 2). The lack of consistent relation between ka and soil moisture for Soil 15 was probably because the size of Na (14 mg kg–1) was so small that the moisture effect was obscured by experimental errors. The moisture effect on both ka and ks was insignificant for the treatments at 5°C for both soils (P > 0.05).
|
(% w/w) and temperature T (°C) as below:
 | [5] |
 | [6] |
 | [7] |
 | [8] |
and fT are moisture and temperature correction factors, respectively; max is the optimum soil moisture for net N mineralization.
Prediction of Net Nitrogen Mineralization under Controlled Field Conditions
The temperature during the 126-d field incubation of Soils 15 and 17 fluctuated considerably (2.5–35.8°C; Fig. 3a,d). The soil moisture was operationally controlled within a range of >8% (w/w) to avoid the drying-rewetting effect and <55% WHC to avoid N loss from denitrification (Fig. 3b,e).
|
To predict net N mineralization under fluctuating temperature and moisture using the model parameters estimated under constant conditions, ka and ks were adjusted using Eq. [5] and [6] for Soil 15 and using Eq. [7] and [8] for Soil 17 in both hourly and daily time intervals. The temperature during each time interval was calculated as the arithmetic mean of the measurements, and the moisture was estimated by linear interpolation between two sequential measurements. The amount of N mineralized from Na during the first time interval (Nt1) was calculated as below:
 | [9] |
 | [10] |
The cumulative amounts of N mineralized from Ns were calculated using the same algorithm. The sum of the predicted net N mineralization from both Na and Ns was used as the prediction of the amount of N mineralized from the soil.
The predicted dynamics of mineral N in the soils during the 126-d field incubation agreed satisfactorily with the measured data (Fig. 3c,f). The results predicted with daily mean temperature were lower than those with hourly mean temperature and the measured data for Soil 15 (P < 0.01). The underestimation obtained with daily mean temperature was primarily because of the non-linear response of mineralization rate constants to temperature (Fig. 2), which should increase with the increase in the Q10 coefficient and the amplitude of the temperature fluctuation (Das et al., 1995). However, the error associated with the use of daily mean temperature was minor for Soil 17 (Fig. 3f). Sierra (2002) found that use of daily mean temperature induced considerable underestimation of net N mineralization from an Oxisol that had a Q10 of 4. The results of this study confirmed that in situ N mineralization dynamics could be satisfactorily predicted with the parameters determined in the laboratory.
Relationships of the Kinetically Estimated Pools to Chemical, Physical, and Microbial Organic Nitrogen Fractions
The use of defined rate constants for different soils made it possible to explicitly assess the relationships between Na or Ns and measurable soil organic matter fractions and other soil properties without the confounding effect of ka or ks. Stepwise multiple linear regression was conducted to relate Na and Ns to clay content, TON, C/N ratio, LFC, LFN, KCl-N, NaOH-N, HCl-N, DOC, DON, MBC, and MBN at time zero (MBC0 and MBN0), MBC and MBN at the end of the first-week incubation (MBC7 and MBN7), microbial biomass C after 42-d incubation (MBC42), net N mineralization during the first 14 d (N0–14), and net N mineralization from Day 14 to 28 (N14–28). The results suggested that Na could be estimated by:
 | [11] |
Substantial increases in the amount of soluble organic matter after air-drying and flushes of C or N mineralization during subsequent incubation of rewetted dry soils have been observed in many studies (Davidson et al., 1987; Lundquist et al., 1999; Franzluebbers, 1999). In the present study, the size of Na was greater than the amount of DON for most soils, particularly those with Na > 30 mg kg–1. Thus, DON should not be regarded as a direct measure of Na; its value is affected by the extraction procedures such as temperature, vigor and time of shaking, and filter pore size.
Ns was not correlated with any single organic C and/or N fraction described above, but could be approximately estimated by the following equation:
 | [12] |
The use of fixed rate constants is not meant to suggest that the quality of soil organic matter in Na or Ns be identical in different soils. Also, Na and Ns estimated with fixed mineralization rate constants do not represent the absolute amounts of mineralizable organic N in each pool, but rather indicate the N mineralization ability of soil in different time scales, which is interactively determined by soil physical, chemical, and biological properties. The Na indicates the magnitude of N mineralization flush during the early stages of incubation, and is mainly determined by the readily mineralizable organic N content in the original soil (Eq. [11]). The Ns pool is largely affected by the magnitude of net N mineralization after the initial flush, which is determined by the long-term capacity of soil to replenish bio-available substrate. The process of replenishment can be interactively regulated by substrate quantity and quality, and soil biological and physical properties in complicated manners. The amount of N mineralized from Day 14 to 28 (N14–28) integrated the impacts of all regulating factors, hence provides better estimation for Ns than any other single index tested.
Equations [11] and [12] are not only useful for depicting the nature of Na and Ns, but also shed light on the possibility of estimating the organic N pools and thus predicting in situ net N mineralization with relatively less time-consuming techniques. Further evaluation of these equations with results from field studies would be worthwhile. However, attention should be paid to the fact that mineral N accumulation in the field is subject to possible influences from several N-consuming processes in addition to N mineralization/immobilization. Such processes should be quantified to properly assess the performance of the empirical model.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
Received for publication November 12, 2003.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  W. L. Berndt Double Exponential Model Describes Decay of Hybrid Bermudagrass Thatch Crop Sci., November 24, 2008; 48(6): 2437 - 2446. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. B. Mallory and T. S. Griffin Impacts of Soil Amendment History on Nitrogen Availability from Manure and Fertilizer Soil Sci. Soc. Am. J., May 16, 2007; 71(3): 964 - 973. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. L. Vourlitis, G. Zorba, S. C. Pasquini, and R. Mustard Chronic Nitrogen Deposition Enhances Nitrogen Mineralization Potential of Semiarid Shrubland Soils Soil Sci. Soc. Am. J., April 5, 2007; 71(3): 836 - 842. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 | |||
