Straw retention has been shown to reduce carbon dioxide (CO_{2}) emission from agricultural soils. But it remains a big challenge for models to effectively predict CO_{2} emission fluxes under different straw retention methods. We used maize season data in the Griffith region, Australia, to test whether the denitrificationdecomposition (DNDC) model could simulate annual CO_{2} emission. We also identified driving factors of CO_{2} emission by correlation analysis and path analysis. We show that the DNDC model was able to simulate CO_{2} emission under alternative straw retention scenarios. The correlation coefficients between simulated and observed daily values for treatments of straw burn and straw incorporation were 0.74 and 0.82, respectively, in the straw retention period and 0.72 and 0.83, respectively, in the crop growth period. The results also show that simulated values of annual CO_{2} emission for straw burn and straw incorporation were 3.45 t C ha^{−1} y^{−1} and 2.13 t C ha^{−1} y^{−1}, respectively. In addition the DNDC model was found to be more suitable in simulating CO_{2} mission fluxes under straw incorporation. Finally the standard multiple regression describing the relationship between CO_{2} emissions and factors found that soil mean temperature (SMT), daily mean temperature (
Atmospheric
There are many factors affecting
Straw retention has been adopted worldwide to increase crop production. It has been shown to reduce
Field experiments were conducted on a commercial farm in New South Wales (NSW) Australia, 34°30′S, 146°11′E, located approximately 30 km southeast of Griffith. Mean annual precipitation is 432 mm, and mean maximum and minimum temperatures are 23.0 and 10.3°C, respectively (measured at the nearest recording station, Leeton). The soil (0–20 cm) is classified as a Typic Natrixeralf and Mundiwa clay loam with clay particle content in 53.11% [
Input parameters used in the DNDC model (0–20 cm).
Parameter  Soil bulk density 
pH  Field capacity 
Wilting point 
Clay fraction 
SOC in the surface soil 
C/N  Initial 
Initial 

Data  1.37  5.5  38.01  10.22  53.11  0.03  10.90  6.30  3.32 
The field experiment began on 11 May 2010 (day 1) and ended on 10 May 2011 (day 365). There were two maize straw treatments in the field experiment. A randomized block design with three replicates was used in the 12 plots. The maize straw treatments were (1) application of 300 kg N ha^{−1}, maize straw burnt and left on the field (300Nburn), and (2) application of 300 kg N ha^{−1}, maize straw mulched (the amount of maize straw was 6750 kg ha^{−1}) and incorporated into soil (5 cm) soon after harvest (300Nincorporated). The 300Nburn treatment used 6 plots and 300Nincorporated treatment used another 6 plots. The result of each treatment was the mean value. The two straw retention methods lasted for one maize season. Fertilizer was applied three times: 90 kg N ha^{−1} as
Irrigation times and amount of water used in each irrigation.
Data  Irrigation 
Data  Irrigation 

28/10/2010  197  26/11/2010  200 
18/12/2010  127  27/12/2010  90 
5/1/2011  82  13/1/2011  102 
28/1/2011  61  15/2/2011  74 
24/2/2011  76  3/3/2011  76 
The
In this study the DNDC model (version 9.5;
The DNDC model was used to simulate
ME compares the squared sum of the absolute error with the squared sum of the difference between the observations and their mean value. It compares the ability of the model to reproduce the daily data variability with a much simpler model that is based on the arithmetic mean of the measurements. ME values close to 1 indicate a “nearperfect” fit [
Five continuous longterm measurement factors were considered for the statistical analysis, namely, daily maximum temperature
Data were analyzed by correlation analysis and path analysis using SPSS 13.0. Path analysis can be used for the analysis of multiple variables and the linear relationship between variables. It was a development of regression analysis [
The straw retention period and crop growth period were studied separately because the sources of the
The simulated and observed values of daily
Comparison of observed and simulated
300Nburn
300Nincorporation
The DNDC model was also used to simulate the daily
Comparison of observed and simulated
300Nburn
300Nincorporation
The correlation coefficients between the observed and simulated values of
The observed values of
The observed and simulated annual CO_{2} emission for the maize season.
300Nburn 
300Nincorporation  

CO_{2}observed values  4.7  3.5 
CO_{2}simulated values  3.45  2.13 
Fixed factors (continuous longterm measurements) were used for the sensitivity analysis. Because the DNDC model was more suitable for the simulation of
The
Correlation coefficients between CO_{2} and same soil variables.



WFPS  SMT  CO_{2}  


1.0000  0.7967**  0.9592**  0.5525**  0.7259**  0.5681** 

1.0000  0.9350**  0.6952**  0.7123**  0.5114*  

1.0000  0.6494**  0.912**  0.5125**  
WFPS  1.0000  0.6307**  0.5366**  
SMT  1.0000  0.6729**  
CO_{2}  1.0000 
Path analysis was used to analyze the relationship among these five factors (Tables
The standard multiple regression coefficients.
Unstandardized coefficients  Standardized coefficients  

Model 

Std.error  Beta 

Sig.  
1  (Constant)  −3.832  2.089  —  1.835  0.069 
Stemper  0.316  0.018  0.573  7.430  0.000  







3  (Constant)  −34.113  7.434  —  4.098  0.000 
SMT  0.8067  0.030  0.452  3.360  0.000  

0.6392  0.152  0.681  2.135  0.000  
WFPS  0.4014  0.237  0.339  2.672  0.005 
Path coefficient of each factor on CO_{2} emission.
Soil factor 
Direct path coefficient  Indirect path coefficient  




Total  Error path coefficient  

0.8067  1  0.5830  0.2531  1.1016  0.2363 

0.6392  0.7357  1  0.2607  1.2499  

0.4014  0.5088  0.4151  1  1.0915 
The observed and simulated
The correlation coefficient between simulated and observed values and ME values implies that the DNDC model can be used to simulate daily
Both the observed and simulated values for the treatment 300Nburn were higher than those for the 300Nincorporation. The result was the same as the straw retention method. Straw decomposition rate varies with the depth of incorporation [
The correlation coefficient between simulated and observed values and ME values indicates that the DNDC model was more suitable for simulating
SMT,
The DNDC model can be used to simulate
The authors thank Dr. Christopher Ogden (formerly of Weill Cornell Medical College in Qatar) for his check of English and comments on this paper. They also wish to express their thanks to anonymous reviewers for providing useful comments to improve the paper. This study was supported by the Australian Government Department of Agriculture, Forestry and Fisheries, and the Australian Research Council.