Nitrous oxide emissions from subtropical horticultural soils: a time series analysis.
Nitrous oxide ([N.sub.2]O) is a long-lived greenhouse gas which has also been linked to ozone depletion (Crutzen and Ehhalt 1997). Soils are one of the major sources of [N.sub.2]O (Bouwman 1998), with 65% of annual global emissions (IPCC 2001) and 6% of the overall global warming effects coming from this source (Dalal et al. 2003). The atmospheric concentration of [N.sub.2]O is increasing at a rate of 0.2-0.3% per year (IPCC 2001). Over 16% of the total national greenhouse gas emissions in Australia are from agriculture, with 80% of annual [N.sub.2]O emissions in Australia from agriculture alone (Australian Greenhouse Office 2001).
Large temporal and seasonal variability in [N.sub.2]O emissions from soil has been reported (Garcia-Mendez et al. 1991; Ruz-Jerez et al. 1994; Breuer et al. 2000; Choudhary et al. 2002; Kiese et al. 2003). Emissions of [N.sub.2]O from soil are mediated by two microbiological pathways, nitrification and denitrification, both heavily influenced by changes in soil temperature and soil water content (Williams et al. 1992; Machefert et al. 2004). Data on [N.sub.2]O emissions and environmental factors such as soil temperature and soil water content are often measured simultaneously to determine temporal relationships. For example, the effect of a single rainfall event on N20 lasted for 1 week in a potato field (Flessa et al. 2002); the highest [N.sub.2]O emissions lasted for >1 week after a heavy rainfall in a dairy-grazed pasture (Saggar et al. 2004); and there was a 2-week deferred effect of soil water content and soil temperature on [N.sub.2]O in a beech forest (Kitzler et al. 2006). The above studies indicate that current soil status might be affected by previous events, and that time-lagged effects of soil environmental parameters on [N.sub.2]O emission should be investigated. Regression models have been developed which relate the annual [N.sub.2]O emissions to soil temperature, soil water content, and N substrate (Bouwman et al. 2002; Sozanska et al. 2002; Roelandt et al. 2005). Linear regression models have been developed over short time periods (i.e. season or part of the growing season), and significant relationships have been found between [N.sub.2]O emissions and soil water content and soil temperature in different land-use systems (Choudhary et al. 2002; Sehy et al. 2003; Wang et al. 2005). However, hypotheses tested and goodness-of-fit statistics from linear regression models for time series data are often not valid because the assumption of independent residuals from the linear regression model is violated. There are few [N.sub.2]O models that have explicitly included terms that describe longer term temporal features of the data, such as auto-correlation and seasonality. Thus, the time-lagged effects on [N.sub.2]O are still largely unknown.
Mechanistic simulation models have also been constructed to estimate [N.sub.2]O emissions. These include DNDC (Li et al. 1992), DAYCENT (Parton et al. 1998; Del Grosso et al. 2000), NGAS (Mosieretal. 1983; Parton et al. 1988), WN MM (Li et al. 2007), and NLOSS (Riley and Matson 2000). Many uncertainties and limitations still exist in these simulation models (Chen et al. 2008), including the inherent difficulty of providing soil, plant, and climate data of high spatial and/or temporal resolution. Collection of these data is both expensive and time-consuming and severely limits the application of these models across a diverse set of environmental conditions, hence the need for simpler modelling approaches which utilise readily available (or easily collected), robust information to infer causal relationships.
The objective of this study was to quantify [N.sub.2]O emissions from soils under a selection of subtropical horticultural crops (custard apple, mango, and pineapple crops) and apply time series models to describe the influence of environmental factors (soil water content and soil temperature) on the emissions of [N.sub.2]O. To our knowledge, this is the first time that the magnitude of [N.sub.2]O emissions has been determined for subtropical horticultural crops, thus providing valuable information for international and national inventories such as Australia's National Greenhouse Gas Inventory (NGGI). By taking into account time-lagged soil water content, time-lagged soil temperature, autoregressive processes and seasonality, the model provides more-detailed information on the nature of the relationship between [N.sub.2]O and the environmental drivers and the effects of temporal resolution on [N.sub.2]O emissions, obtained from fitting the model with weekly or monthly data. The time series model potentially improves the estimation of [N.sub.2]O fluxes and reduces the uncertainty of temporal variability in estimated [N.sub.2]O emissions.
Materials and methods
The study field site was at the Queensland Department of Primary Industry Horticulture Research Station at Nambour (latitude -26.63, longitude 152.95) in South East Queensland, Australia. The region is a major centre for subtropical horticulture (e.g. macadamia nuts, bananas, citrus fruits, custard apples, and pineapples) and sugar production. The majority of rainfall (54%) occurs from December to March, and significant rainfall events are common between December and May. The average monthly rainfall is 49-259 mm with an annual mean rainfall of 1732mm (Bureau of Meteorology 2012). The annual evaporation rate is ~1400mm. The average maximum monthly temperature is 21-29[degrees]C, and the average minimum monthly temperature 8-20[degrees]C. Highest temperatures generally occur in January and February, and the lowest temperature is typically in July.
The soil type at the experimental site was a slightly acidic Dermosol (Isbell 2002) with a sandy clay loam texture (0-20 cm). For further information regarding the study site, the reader is referred to Burgess and Ellis (2006). Average topsoil (0-10cm) organic carbon (C) content (by dry combustion) was 2.6, 4.2, and 3.5% for mango, pineapple, and custard apple field plots, respectively, with a bulk density of 1.3 g/[cm.sub.3]. Field capacity and saturated volumetric water contents were estimated to be 23.5 and 46.8% (vol/vol), respectively. Mineral N inputs equivalent to 445 and 92 kg N/ ha were applied to the pineapple and custard apple plots, respectively. In total, 20 urea applications ranging from 8 to 56 kg N/ha were applied to the pineapple crop at regular intervals during the 12-month study period, resulting in an annual application of 445 kg N/ha. In the custard apple plots, a single application of N (92 kg N/ha) as a compound synthetic fertiliser (Rustica Plus; Campbells Fertilisers Australia, Laverton North, Vic.) containing 12% N, in equal portions of nitrate and ammonium, was applied on 8 November 2006. Sugarcane mulch was applied along the rows of the custard apple crop in late September 2006. No mineral or organic fertiliser had been applied to the mango plots for 4 years and none was applied during the study.
Replicate sampling of [N.sub.2]O emissions, topsoil temperature, and water content (0-10 cm) was undertaken on five mango, five custard apple, and six adjacent pineapple plots on a weekly basis between 4 December 2006 and 11 November 2007. The in-field gas sampling system utilised the static closed chamber technique (Hoben et al. 2011) to capture emissions from soil under the three crops. This method uses a single gas-tight chamber (non steady-state, non through-flow) enclosing an area of soil in each replicate plot over a given time interval. Manual sampling chambers consisted of a 200-mm-diameter PVC bucket with the bottom removed and a gas-tight lid. A 100-mm-deep, sharpened steel ring of the same diameter as the chambers was first hammered into the soil then removed; the chambers were then inserted into the groove and left in place for the duration of the study. Chambers did not include plants but were placed as close as possible to the base of the plants. Chamber headspaces were measured several times over the experiment and averaged 8-10 cm over the experimental period. The exact volume was calculated for each chamber at each sampling time to ensure accuracy in the calculation of emissions.
Chamber closure was achieved using a gas-tight lid with a rubber septum that allowed the insertion of a needle and syringe. Fluxes of N20 were measured by collecting four gas samples (including time zero) from the chamber headspace at 15-min intervals over a 1-h closure period. A double-ended syringe was used to extract a 12-mL gas sample into an evacuated glass vial (Exetainer, Labco, UK). Gas samples were collected at the same time (10:00) on the same day once a week. Vials were then stored overnight at room temperature and a 1-mL sample was analysed for [N.sub.2]O using a gas chromatograph (GC-8A, Shimadzu, Kyoto) with an electron capture detector (ECD) with a 5% methane/95% argon carrier gas (flow rate 30 mL/min) and an Alltech Porapak Q 80/100 mesh separation column (Grace Davison Discovery Sciences, Rowville, Vic.). The column and ECD temperatures were set at 100[degrees]C and 350[degrees]C, respectively. Gas loss from the vials during storage was tested using a calibration gas stored with the samples.
Topsoil (0-10cm) water data (vol/vol) were recorded for each manual chamber on each gas sampling occasion and immediately before closure, using a hand-held Hydrosense TDR (Campbell Scientific Australia Pty Ltd, Townsville, Qld) calibrated for the site. Topsoil temperature ([degrees]C) was also recorded at each chamber at this time using a digital thermometer inserted to 5 cm and recorded after 10min. The flux rate was calculated using the procedure outline by Barton et al. (2008). Flux rates were discarded if the coefficient of determination ([R.sup.2]) was <0.80 (Rowlings et al. 2012).
Time series analysis can account for sequential data points, such as [N.sub.2]O emissions, which may have some internal structure such as autocorrelation and/or seasonal variation (Chatfield 2004). In an autoregressive model, each value in a time series is expressed as a linear function of the preceding values (Yaffee and McGee 2000). The autocorrelation function (ACF) and partial autocorrelation function (PACF) can be used as exploratory techniques to confirm any seasonal patterns in the data and diagnose correlation between the observations for different lags. This diagnostic information enables an application-relevant but parsimonious model to be developed (Yaffee and McGee 2000). In an adequate model, the residual autocorrelations should fall within the upper or lower 95% confidence bands around zero (in the plots of ACF and PACF). The Box-Ljung Q statistic can also be used to test the significance of autocorrelations (Yaffee and McGee 2000).
In this study, the data on the fluxes of [N.sub.2]0, soil water content, and soil temperature between 2006 and 2007 were analysed as time series for each of the horticultural crops. Separate time series regression models were used to describe both weekly and monthly mean data for each crop, with [N.sub.2]O emissions as the dependent variable and soil water content, soil temperature, autoregressive process, seasonality, and temporal lags of soil water content and soil temperature as potential exploratory variables. In general, such a model will have the following form:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)
Here, [y.sub.st] is the observed value of [N.sub.2]O at time period t and at experimental plot s; [x.sub.1,s,t] and [x.sub.2,s,t] represent covariates (soil temperature and soil water content) at time period t and at plot s, respectively; [Y.sub.s, t-k, [x.sub.1,s,t-j], and [X.sub.2,s,t-m] represent previous values of y and x at plot s and at time periods t-k, t-j, and t-m, respectively (i.e. time lags k, j, and m); and [y.sub.j] and [[delta].sub.m] are the regression coefficients for soil temperature and soil water content which are lagged by j and m time periods, respectively. In this study, J and M were set equal to 1, so that soil water content at a lag of 1 week or 1 month and soil temperature at a lag of 1 week or 1 month were included in the model.
The autoregressive (AR) term [[phi].sub.kYs,t-k] in Eqn 1 describes the effects of the relationship between sequential [N.sub.2]O values in a series, which expresses that the current [N.sub.2]O value will be influenced by its own behaviour in prior periods, with the coefficient [phi] indicating how strongly the current value depends on the preceding value. The most common forms of autoregressive processes are first-order and second-order autoregressive models, denoted by AR(1) (corresponding to [[phi].sub.1] and AR(2) (corresponding to [[phi].sub.2]), respectively. The AR (1) term describes a relationship between the immediately previous [N.sub.2]O value and the current [N.sub.2]O value in a series. The AR(2) term is concerned with the relationship between the values in the two previous periods and current value in a series. The term A(t) in Eqn 1 represents the seasonal component of the time series. The seasonal relationship between [N.sub.2]O flux and the variables soil water content and soil temperature are considered using four models. In model 1, seasonality was described by a linear combination of sine-cosine functions of different frequencies (Wei 1990). Thus, A(t) represents the harmonic factors a sin(2[pi]t/12)+b cos(2[pi]t/12) for the monthly model and a sin(2[pi]t/52)+ b cos(2[pi]t/52) for the weekly model. In model 2, seasonality was alternately described as a categorical variable (spring, summer, autumn, and winter). In model 3, seasonality was excluded due to the small dataset collected for this study. Model 4 excluded autoregressive processes, lagged effects, and seasonal terms. Thus, [[phi].sub.k] = 0 [for all] k, [[y.sub.j] = 0 [for all] j, [[delta].sub.m] = 0 [for all] m, and A(t) = 0 (i.e. a simple linear regression model).
Although the weekly model provides a more temporally sensitive assessment of the relationship between the dependent and exploratory variables, the amount of missing data at the weekly scale led to concern over the reliability of this model. As an alternative to exclusion of cases with missing data, which would have led to considerable loss of information, weekly missing data were replaced as the mean of the four nearest neighbours (Yaffee and McGee 2000). Cumulative emissions (g [N.sub.2]O-N/ha) were determined by interpolating daily fluxes for each replicate (g [N.sub.2]O-N/ha.day) between sample days over the course of the year. Other methods of imputation were also employed (Yaffee and McGee 2000; Bennett 2001), but led to no substantive differences in inference. The monthly data, on the other hand, were more stable in terms of temporal consistency and completeness of the data.
The goodness-of-fit of the models was assessed by comparing the [R.sup.2], adjusted [R.sup.2], and index of agreement (at) (Willmott 1981) statistics under the four models. A high value of these statistics indicates better performance. All analyses were performed with SPSS 14.0 software (Norus 2006).
Annual [N.sub.2]O emissions from mango, pineapple, and custard apple crops in subtropical South East Queensland were estimated to be 1590, 1156, and 2038 g [N.sub.2]O-N/ha. The [N.sub.2]O-N emitted from the pineapple and custard apple crops was equivalent to 0.3 and 2.2%, respectively, of the applied mineral N. Table 1 outlines summary statistics for the observed data. Across the five experimental plots for mango crop, individual [N.sub.2]O measurements ranged from 0.1 to l l.2g [N.sub.2]O-N/ha.day, with an average of 4.6g [N.sub.2]O-N/ha.day. The mean soil water content over the study period was 25.2%, with individual measurements ranging from 7.0 to 51% across the five mango crop plots. The average soil temperature was 21. 1[degrees]C, with individual readings fluctuating between 12.3 and 28.4[degrees]C across the five mango plots over the study period. Across the six experimental plots for pineapple, the study averages for [N.sub.2]O emission, soil water content, and soil temperature were 3.3 g [N.sub.2]O-N/ha.day, 11.2%, and 22.7[degrees]C. The ranges for individual [N.sub.2]O emissions, soil water content, and soil temperature were 0-16.4 g [N.sub.2]O-N/ha.day, 4.0-33.8%, and 15.7-28.4[degrees]C across the six pineapple plots. In the five experimental plots for custard apple, the overall average for [N.sub.2]O emissions, soil water content, and soil temperature were 5.9g [N.sub.2]O-N/ha.day (individual measurement range 0.0-92.7g [N.sub.2]O-N/ha.day), 12.7% (range 4-36.5%), and 22.4[degrees]C (range 14.4-29.8[degrees]C). The mango crop had higher mean soil water content than the custard apple and pineapple crops. The average soil temperatures in the three crops were similar. The custard apple crop had higher mean [N.sub.2]O emission than both pineapple and mango crops. The [N.sub.2]O flux and soil water content displayed positive skewness in the custard apple and pineapple crops (Table 1).
Emissions of [N.sub.2]O in the custard apple soil tended to be more variable than in the other two crops. Mango had lower average soil temperature and higher average soil water content than the other two crops due to the shading effect of the mature trees. Soil temperature fluctuations were similar in magnitude for pineapple and custard apple during the study period, but the soil water content measurements for custard apple were generally higher than those for pineapple due to presence of sugarcane mulch. The reduced variation of the mean monthly data (relative to the weekly data) was strongly illustrated for [N.sub.2]O emissions. For example, in the custard apple soil, the range in the monthly means (0.8-50.9g [N.sub.2]O-N/ha.day) was almost half that of the weekly means, measured across the five experimental plots.
The mean weekly measurements for [N.sub.2]O emissions, soil temperature, and soil water content over the study period for the three crops are depicted in Fig. 1. No stable pattern of variation in [N.sub.2]O emission with soil water content and soil temperature was immediately evident in the different crops. Greater variation in [N.sub.2]O emissions was evident for custard apple. However, the [N.sub.2]O emissions appeared to be seasonal, with highest emissions in summer and autumn. Highest soil water contents occurred in the humid summer months of November and December. Soil temperature was much less variable than soil water content and changed with season. The minimum soil temperature was in July, and the maximum occurred in January and February.
Pairwise linear associations (Spearman correlation coefficients) were detected between [N.sub.2]O emission and soil water content, soil temperature, soil water content at alag of l week and 1 month, and soil temperature at a lag of 1 week and 1 month respectively. For mango, negative relationships with [N.sub.2]O emissions were observed for soil water content at a lag of I week (P< 0.05) or l month (P< 0.05), and soil temperature at a lag of 1 month (P<0.05); positive relationships were observed between [N.sub.2]O emissions and weekly soil temperature (P < 0.0 l) and soil temperature at a lag of 1 week (P < 0.01). Emissions of [N.sub.2]O had a positive relationship with weekly and monthly soil water content (P< 0.05) and soil water content at a lag of 1 week (P< 0.05) for custard apple. No significant relationships were found between [N.sub.2]O emission and other variables for pineapple. Moreover, the correlation coefficients were ~0.9 for both current soil temperature and soil temperature at a lag of 1 week for the three crops. Additionally, the variance inflation factors for temperature at a lag of 1 week were 16, 10, and 10 for mango, custard apple, and pineapple, respectively. This showed that collinearity was present between weekly temperature and weekly temperature at a lag of 1 week. Hence, the latter variable was not included in the models in order to reduce standard error in the related covariates and overfitting.
Log-transformations were required for [N.sub.2]O emissions and soil water content data due to substantially positively skewed distributions of these data for custard apple and pineapple crops. The values for skewness decreased after log transformation for these crops (Table 1). Plots of the ACF of the residuals from the simple weekly linear regression models (model 4) for the three crops at each study site showed that the assumption of independent residuals was clearly violated, indicating that the simple linear regression model might be not appropriate for these time series data. This was supported by large autocorrelation coefficients, using Ljung-Box Q Statistic test, for residuals at a lags of 1 and 2 months from the simple monthly linear regression models. Hence, first-order AR(1) and second-order AR(2) autoregressive processes were included in the time series models. Moreover, the plots of the ACF for the three crops exhibited slight cosine patterns. This justified consideration of model 1 and model 2.
A comparison of the [R.sup.2], adjusted [R.sup.2], and index of agreement d values obtained from the four models based on the weekly and monthly data for each crop is presented in Table 2. In the simple linear regression models (model 4), adjusted [R.sup.2] values were very low for the six models across the three crops. No significant relationships were found between [N.sub.2]O and soil water content or soil temperature for any crop. In contrast, the adjusted [R.sup.2] values were greatly improved for the rest of the models across the three crops, although they were still not high in the weekly time series regression models for mango and custard apple crops. Model 1 exhibited the best fit for the monthly scale data. The adjusted [R.sup.2] and index of agreement d values for models 1, 2, and 3 were similar for the weekly data for the custard apple and pineapple crops.
Results of the time series regression models (model 1) applied to the weekly and monthly [N.sub.2]O emissions estimates are presented in Table 3. In the weekly models, there was a substantive relationship (P < 0.05) between [N.sub.2]O emissions and soil water content for mango; there was also a weakly significant relationship (P<0.1) between [N.sub.2]O emissions and soil temperature, with adjustment for season and autocorrelation at a lag of 1 week. For custard apple, there was a weakly significant relationship (P< 0.1) between [N.sub.2]O emissions and soil water content after adjustment for autocorrelation at lags of 1 and 2 weeks. For pineapple, [N.sub.2]O emissions were related to soil water content (P< 0.05) and soil water content at a lag of I week (P<0.1) after adjustment for autocorrelation at lags of 1 and 2 weeks.
In the monthly models, the results showed that there was a significant relationship between [N.sub.2]O emission and soil temperature at a lag of 1 month after adjustment for autocorrelation at a lag of 1 month and season in the mango crop. Emissions of [N.sub.2]O were significantly associated with soil temperature at a lag of 1 month after adjustment for autocorrelation at a lag of 2 months and season in the custard apple. Emissions of [N.sub.2]O were significantly associated with soil temperature at a lag of 1 month after adjustment for autocorrelation at a lag of 2 months in pineapple.
The simple linear regression models poorly predicted [N.sub.2]O emissions (Fig. 2), whereas the time series regression models (model 1) were able to capture the flux patterns. However, it was obvious that the weekly regression model was not as accurate at predicting [N.sub.2]O fluxes as the monthly models.
The residuals of all models 1 were centred around zero with no clear pattern and little auto-correlation after accounting for autoregressive process, sinusoidal term, and temporal lagged factors. However, the small [R.sup.2] values, especially for the weekly models for mango and custard apple crops, indicated that there was substantial variation in the [N.sub.2]O emissions that was not explained by the variables considered here. Comparing the two time scales analysed in our study, the weekly models 1 explained 20.3%, 25.7%, and 36.3% of the temporal variation of [N.sub.2]O emission for mango, custard apple and pineapple crops, respectively. Nearly 50.4%, 43.7% and 31.2% of temporal variation of [N.sub.2]O emissions was explained by the three monthly models 1 for mango, custard apple, and pineapple crops, respectively. Moreover, the index of agreement d indicated that the monthly models performed better than the weekly models (Table 2), particularly in mango and custard apple.
Magnitude of emissions
This is the first study to measure soil-borne [N.sub.2]O emissions from a range of subtropical horticultural crops (mango, custard apple, and pineapple) over a 12-month period in Australia. Annual emissions ranged from 1.2 to 2.0kg [N.sub.2]O-N/ha, with a daily average of 4.6 g [N.sub.2]O-N/ha, comparable to an average of 5.5 g [N.sub.2]O-N/ha.day emitted from N-fertilised soils under sugarcane in the same region (Huang et al. 2011). The unfertilised mango crop emitted 1.6 kg [N.sub.2]O-N/ha over the year, which is of the same magnitude as emissions from a horticultural tree crop (lychee) grown in the same region (Rowlings 2010). Globally, losses of [N.sub.2]O from horticulture range from 1.3 to 2.9% of applied N (Bouwman 1998), which would indicate that the emission factors (uncorrected for zero N applied) of 0.3% and 2.2% found for the pineapple and custard apple crops, respectively, are consistent with global estimates.
Comparison of time series models
Time series models are recommended for modelling [N.sub.2]O emissions as these data, and the factors that regulate these losses, are collected simultaneously and exhibit significant temporal variability. In the simple linear regression models for the three crops, strong autocorrelation in the residuals was evident, whereas autocorrelation in the residuals almost disappeared in model 1, model 2, and model 3. The adjusted coefficients of determination [R.sup.2] (Table 2) and correlation between residuals indicated that traditional linear regression models that assume independence between observations are inappropriate for explaining temporal variability in [N.sub.2]O emissions from relatively fertile, subtropical horticultural soils. The adjusted [R.sup.2] of the three weekly time series models for custard apple and pineapple crops, and the three monthly time series models for the pineapple crop were similar, with no significant relationships between [N.sub.2]O and seasonality in model 1 and model 2. On the other hand, the weekly and monthly time series models with seasonality included (models 1 and 2) improved the adjusted [R.sup.2] and d values, with statistically significant relationships between [N.sub.2]O and seasonality for weekly or monthly series for mango and monthly for custard apple. However, the results indicated that model 1 could capture the seasonal profile better than model 2, even with these limited data. The explained variance in terms of the adjusted [R.sup.2] was improved by the models which acknowledged the temporal nature of data by including lagged soil water content, lagged soil temperature, seasonality, and autoregressive processes, particularly for the monthly models across the three crops.
Tendency for [N.sub.2]O emissions
The temporal persistence in the emissions data was also confirmed by the ACF plots. There were quite different behaviours in autocorrelation for weekly and monthly models in the three crops. The results indicated that the magnitude of [N.sub.2]O emissions might persist for 1 week for mango and custard apple, and up to 2 weeks for pineapple; and monthly mean emissions might persist for 2 months for custard apple and pineapple. However, there was a strong negative autocorrelation at a lag of 1 month for the mean [N.sub.2]O emissions for mango. It indicated that an increase in current monthly [N.sub.2]O emission on average was associated with a decrease in the next monthly [N.sub.2]O emission. In this study, the changes in soil temperature and soil water content aligned more closely with [N.sub.2]O emissions in the custard apple and pineapple crops than in the mango crop. This in part could be explained by the fact that the mango crop did not receive any additional mineral N sources (to support either nitrification or denitrification) during the study and the emissions were consistently at a lower level compared with both pineapple and custard apple crops, regardless of the fluctuation in soil temperature and soil water content.
Effect of soil water content on [N.sub.2]O emissions
In the weekly time series regression models, soil water content was a significant environmental factor and was positively related to [N.sub.2]O emissions for all three crops. Moreover, there was a weakly significant negative relationship between [N.sub.2]O emissions and soil water content at a lag of 1 week for pineapple. Many previous studies have demonstrated that soil water content has a significant influence on [N.sub.2]O emissions (Sehy et al. 2003; Machefert et al. 2004). The weekly models confirmed that soil water content was a key factor, and was associated with increasing [N.sub.2]O emissions for these three crops, after adjusting for seasonality and autocorrelation. In the three crops, topsoil water content was <31% (vol/vol), which is equivalent to 60% water-filled pore space (WFPS) (data not shown) for the majority of the year. Nitrification is the predominant source of [N.sub.2]O emissions when soil WFPS is <60% (Davidson 1991), and emissions of [N.sub.2]O via the nitrification pathway increase with increasing soil water content (Maag and Vinther 1996).
When WFPS is >60% and the soil atmosphere becomes increasingly anaerobic, losses of N via the identification pathway become apparent (Linn and Doran 1984). The magnitudes of these losses are also dependent on the amount of nitrate in the soil and availability of a labile (readily decomposable) source of C. In the latter stages of the study (October-November), when excessive rainfall did result in the soil water content exceeding 60% for an extended period of time, [N.sub.2]O emissions from both pineapple and mango crops remained static. The relatively low proportion of [N.sub.2]O evolved over the year from the pineapple crop in response to the high fertiliser N input and predominately aerobic conditions (due to the high plant density depleting surface soil water content) potentially confirms that nitrification was the dominant source of [N.sub.2]O emissions over the year in that crop. The nitrification pathway would potentially have been the primary contributor to [N.sub.2]O losses under mango, which had not received any mineral N inputs for 4 years but maintained higher (average) soil water content throughout the year than both pineapple and custard apple crops due to the shading effect of the tree itself. In contrast, we estimate that approximately one-third of total annual [N.sub.2]O emissions from the soil under custard apples could potentially be attributed to denitrification due to the high soil water content and source of labile C (from the fresh cane mulch) in the latter weeks of the study.
Effect of soil temperature on N20 emissions
Temperature has been shown to be strongly associated with [N.sub.2]O emission (Kamp et al. 1998; Lu et al. 2005; Zhu et al. 2005; Koponen et al. 2006; Saggar et al. 2007; Yan et al. 2008). At the monthly scale, we found that soil temperature at a lag of 1 month was significantly and positively associated with [N.sub.2]O emissions rather than current soil temperature for all three crops. At the study location, the average monthly soil temperature fluctuated between 13 and 28[degrees]C across the year of the study, and the average range was only 5[degrees]C for the four seasons for the three crops. It is well known that soil temperature varies with heat flux into and out of soil. The statistically significant effects of soil temperature at a lag of 1 month for [N.sub.2]O might be explained by the observation that thermal capacity and conductivity are more stable than water content in the soil, particularly at lower depths. Surface heat fluxes are more than an order of magnitude more sensitive to soil water content than is the temperature profile within the soil (McCumber and Pielke 1981). It might also be due to the reduced variation in air temperature (compared with soil water content) throughout the year. Jury et al. (1982) concluded that due to the large lag phase caused by gaseous diffusion, particularly in wet soils (as is the case at our study site), monitoring may be required for several weeks after [N.sub.2]O production ceases to ensure that all emissions are captured. Thus, some emissions may be more predictable based on the lagged effects of soil parameters. The time-lagged effects of soil parameters might influence current soil conditions, such as pore continuity, microbial activity, and so on. There could also be other factors that influence the release of [N.sub.2]O, which are independent in their own right or for which the above variables are surrogates. More research is required to investigate these phenomena further. For example, covariates might need to be collected from different soil layers. Using a time series GARCH model, Kitzler et al. (2006) found a 2-week deferred effect of soil water content and soil temperature on [N.sub.2]O in beech forest, and indicated that the time series model could explain a high percentage of variation in [N.sub.2]O emissions. The present study supports the fact that a time series model can more effectively reveal temporal variation in [N.sub.2]O emissions, and relationships with environmental variables, and that lagged air temperature might be an important factor associated with [N.sub.2]O emission in the three crops, particularly at the monthly scale.
Occasional high [N.sub.2]O fluxes cannot always be related to current environmental variables such as soil temperature, soil water content, and nitrate concentration (Nishina et al. 2009). Our weekly models confirmed that soil water content was a key variable and was positively associated with increasing [N.sub.2]O emissions for the three crops, after adjusting for seasonality and autocorrelation. The pattern of [N.sub.2]O fluxes from soil also fluctuates with climate, agricultural management, and crop growth. The inclusion of seasonal terms, such as the sine cosine term used in this study, can help to capture the temporal nature of [N.sub.2]O emissions and thus decrease the uncertainty of temporal resolution of [N.sub.2]O emission from soil.
Some limitations of this study should be acknowledged. First, while the period of study is a single year, the study site has a consistent long-term record of achieving near-average rainfall, thus providing a solid basis for longer term studies examining environmental drivers of [N.sub.2]O emissions in subtropical soils; for ~100 years of complete climate records, annual rainfall has been within 20% of the mean in the majority of years. Second, many additional factors could affect [N.sub.2]O emissions (e.g. spatial variability in soil texture and gas diffusivity), and this is borne out by the low coefficient of determination obtained for the models constructed in this paper. Lastly, very high emissions of [N.sub.2]O often occur over very short time intervals. Soil water content will respond rapidly to heavy rainfall and/or high evaporation rates. Our results indicated that the weekly models did not adequately capture the weekly emissions of our dataset. We acknowledge that a finer timescale, such as daily or hourly measurements, may be necessary to more accurately capture temporal variability in [N.sub.2]O in a future study. However, one of the purposes of our study was to examine the predictive utility of relatively coarse (temporal) scale, but readily collectable, emissions and environmental data.
Annual [N.sub.2]O emissions for subtropical soils under mango, pineapple, and custard apple were 1590, 1156, and 2038g [N.sub.2]O-N/ha, respectively, with the majority of emissions potentially due to nitrification. The [N.sub.2]O-N emitted from the pineapple and custard apple crops was equivalent to 0.26 and 2.22%, respectively, of the applied mineral N. Our results reflect the highly variable nature of [N.sub.2]O emissions over time over different crops within the same landscape. The change in soil water content was the key variate for describing [N.sub.2]O emissions at the weekly scale; soil temperature at a lag of 1 month exhibited a strong effect on [N.sub.2]O emissions at the monthly scale. Time series regression models could explain a higher percentage of the temporal variation of [N.sub.2]O emission than simple regression models using soil temperature and soil water content as environmental drivers. Confining our analyses to a small number of easily measurable covariates (soil water content and soil temperature) at a relatively coarse temporal scale has provided valuable insights into more complex climate, soil, and plant interactions. Finally, we suggest that taking into account seasonal variability and temporal persistence in [N.sub.2]O associated with soil water content and soil temperature may lead to a reduction in the uncertainty surrounding estimates of [N.sub.2]O emissions based on limited sampling designs.
We thank the Queensland University of Technology and the Australian Research Council for funding this study. We also thank David Rowlings (1SR) and Bevan Zischke (Queensland Department of Employment, Economic Development and Innovation) for assistance in data collation and analysis. We also thank Associate Editor Louise Barton and three reviewers for their precise comments and helpful suggestions.
Received 2 May 2011, accepted 12 October 2012, published online 13 November 2012
Australian Greenhouse Office (2001) 'National Greenhouse Gas Inventory 1999 with Methodology Supplements.' (Australian Greenhouse Office, Commonwealth of Australia: Canberra, ACT)
Barton L, Kiese R, Garter D, Burterbach-Bahl K, Buck R, Hinz C, Murphy DV (2008) Nitrous oxide emissions from a cropped soil in a semi-arid climate. Global Change Biology 14, 177 192.
Bennett D (2001) How can I deal with missing data in my study? Australian and New Zealand Journal of Public Health 25, 464-469.
Bouwman A (1998) Nitrogen oxides and tropical agriculture. Nature 392, 866-867. doi: 10.1038/31809
Bouwman A, Boumans L, Batjes N (2002) Modelling global annual [N.sub.2]O and NO emissions from fertilized fields. Global Biogeochemical Cycles 16, 1080-1088. doi:10.1029/2001GB001812
Breuer L, Papen H, Butterbach-Bahl K (2000) [N.sub.2]O emission from tropical forest soils of Australia. Journal of Geophysical Research 105, 26 353-26 367. doi: 10.1029/2000JD900424
Bureau of Meteorology (2012) Basic Climatological Station Metadata, Nambour Bowling Club. Available at: www.bom.gov.au (accessed 27 July 2012).
Burgess J, Ellis R (2006) 'Soil and landscape attributes of the Maroochy River catchment.' (Department Natural Resources, Mines and Water, Land Resources: Brisbane, Qld)
Chatfield C (2004) 'The analysis of time series, an introduction.' 6th edn. (Chapman & Hall/CRC Publishing: Boca Raton, FL)
Chen DL, Li Y, Grace P, Mosier AR (2008) [N.sub.2]O emissions from agricultural lands: a synthesis of simulation approaches. Plant and Soil 309, 169-189. doi:10.1007/sl 1104-008-9634-0
Choudhary M, Akramkhanov A, Saggar S (2002) Nitrous oxide emissions from a New Zealand cropped soil: tillage effects, spatial and seasonal variability. Agriculture, Ecosystems & Environment 93, 33-43. doi: 10.1016/S0167-8809(02)00005-1
Crutzen P, Ehhalt D (1997) Effects of nitrogen fertilizers and combustion on stratospheric ozone layer. Journal of the Human Environment 6, 112-117.
Dalal R, Wang W, Robertson G, Parton W (2003) Nitrous oxide emission from Australian agricultural lands and mitigation options: a review. Australian Journal of Soil Research 41, 165-195. doi: 10.1071/ SR02064
Davidson E (1991) Fluxes of nitrous oxide and nitric oxide from terrestrial ecosystems. Methane, nitrogen oxides, and halomethanes. In 'Microbial production and consumption of greenhouse gases'. (Eds JE Rogers, WB Whitman) (American Society for Microbiology: Washington, DC)
Del Grosso S J, Parton W J, Mosier AR, Ojima DS, Kulmala AE, Phongpan S (2000) General model for [N.sub.2]O and [N.sub.2] gas emissions from soils due to denitfification. Global Biogeochemical Cycles 14, 1045-1060. doi: 10.1029/1999GB001225
Flessa H, Ruser R, Schilling R, Loftfield N, Munch JC, Kaiser EA, Beese F (2002) [N.sub.2]O and [CH.sub.4] fluxes in potato fields: automated measurement, management effects and temporal variation. Geoderma 105, 307-325. doi:l 0.1016/S0016-7061(0l)00110-0
Garcia-Mendez GJ, Mass JM, Matson PA, Vitousek PM (1991) Nitrogen transformations and nitrous oxide flux in a tropical deciduous forest in Mexico. Oecologia 88, 362-366. doi: 10.1007/BF00317579
Hoben J, Gehl R, Millar N, Grace P, Robertson G (2011) Nonlinear nitrous oxide ([N.sub.2]O) response to nitrogen fertilizer in on-farm corn crops of the US Midwest. Global Change Biology 17, 1140-1152. doi: 10.1111/ j. 1365-2486.2010.02349.x
Huang X, Grace P, Mengersen K (2011) Spatio-temporal variation in soil derived nitrous oxide emissions under sugarcane. The Science of the Total Environment 409, 4572-4578. doi:10.1016/j.scitotenv.2011.07. 044
IPCC (2001) Technical summary. In 'Climate Change 2001: The scientific basis. Contribution of Working Group I of the lntergovemmental Panel on Climate Change'. (Eds JT Houghton, Y Ding, DJ Griggs, M Noguer, PJ Van der Linden, X Dai, K Maskell, CA Johnson) (Cambridge University Press: Cambridge, UK)
Isbell RF (2002) 'The Australian Soil Classification.' Revised edn (CSIRO Publishing: Melbourne)
Jury WA, Letey J, Collin T (1982) Analysis of chamber methods used for measuring nitrous oxide production in soil. Soil Science Society of America Journal 46, 250-256. doi: 10.2136/sssaj1982.036159950046 00020007x
Kamp T, Steindl H, Hantschel R, Beese F, Munch J (1998) Nitrous oxide emissions from a fallow and wheat field as affected by increased soil temperatures. Biology and Fertility of Soils 27, 307-314. doi:10.1007/ s003740050438
Kiese R, Hewett B, Graham A, Butterbach-Bahl K (2003) Seasonal variability of [N.sub.2]O emissions and CH4 uptake by tropical rain forest soils of Queensland, Australia. Global Biogeochemistry 17, 1043-1054. doi: 10.1029/2002GB002014
Kitzler B, Zechmeister-Boltenstern S, Holtermann C, Skiba U, Butterbrach-Bahl K (2006) Nitrogen oxides emission from two beech forests subjected to different nitrogen loads. Biogeosciences 3, 293-310. doi: 10.5194/bg-3-293-2006
Koponen HT, Duran CE, Maljanen M, Hytrnen J, Martikainen PJ (2006) Temperature responses of NO and [N.sub.2]O emissions from boreal organic soil. Soil Biology & Biochemistry 38, 1779-1787. doi:10.1016/j.soil bio.2005.12.004
Li C, Frolking S, Frolking TA (1992) A model of nitrous oxide evolution from soil driven by rainfall events: I. Model structure and sensitivity. Journal of Geophysical Research 97, 9759-9776. doi: 10.1029/92JD 00509
Li Y, Chen DL, White RE, Zhang JB, Li B, Zhang YM, Huang YF, Edis R (2007) A spatially referenced Water and Nitrogen Management Model (WNMM) for (irrigated) intensive cropping systems in the North China Plain. Ecological Modelling 203, 395-423. doi: 10.1016/j.ecolmodel. 2006.12.011
Linn D, Doran J (1984) Effect of water-filled pore space on carbon dioxide and nitrous oxide production on tilled and non-tilled soils. Soil Science Society of America Journal 48, 794-799. doi:10.2136/sssaj1984. 03615995004800040019x
Lu Y, Huang Y, Zheng X (2005) [N.sub.2]O emission factor for agricultural soils. Ying Yong Sheng Tai Xue Bao 16, 1299 1302.
Maag M, Vinther FP (1996) Nitrous oxide emission by nitrification and denitrification in different soil types and at different soil moisture contents and temperatures. Applied Soil Ecology 4, ~14. doi: 10.1016/0929-1393 (96)00106-0
Machefert S, Disc N, Goulding K, Whitehead P (2004) Nitrous oxide emissions from two riparian ecosystems: key controlling variables. Water, Air, and Soil Pollution 4, 427-436. doi:I0.1023/B:WAFO.000 0028369.86947.9f
McCumber MC, Pielke RA (1981) Simulation of the effects of surface fluxes of heat in a mesoscale numerical model 1. Soil layer. Journal of Geophysical Research 86, 9929 9938. doi:10.1029/JC086iC10p 09929
Mosier AR, Parton W J, Hutchinson GL (1983) Modeling nitrous oxide evolution from cropped and native soils. Ecological Bulletins 35, 229-241.
Nishina K, Takenaka C, lshizuka S (2009) Spatiotemporal variation in [N.sub.2]O flux within a slope in a Japanese cedar (Cryptomeria japonica) forest. Biogeochemistry 96, 163-175. doi: 10.1007/s 10533-009-9356-2
Novas M (2006) 'SPSS 14.0 guide to data analysis.' (Prentice-Hall Inc. Press: Upper Saddle River, N J)
Parton W, Stewart J, Cole C (1988) Dynamics of C, N, P, and S in grassland soils: a model. Biogeochemistry 5, 109-131. doi:10.1007/BF02180320
Parton W J, Hartman M, Ojima DS, Schimel DW (1998) DAYCENT and its land surface submodel: description and testing. Global and Planetary Change 19, 35-48. doi: 10.1016/S0921-8181 (98)00040-X
Riley W J, Matson PA (2000) NLOSS: a mechanistic model of denitrified [N.sub.2]O and [N.sub.2] evolution from soil. Soil Science 165, 237-249. doi: 10.1097/00010694-200003000-00006
Roelandt C, Wesemael B, Rounsevell M (2005) Estimating annual [N.sub.2]O emissions from agricultural soils in temperate climates. Global Change Biology 11, 1701-1711. doi:10.1111/j.1365-2486.2005.01025.x
Rowlings D (2010) Influence of historic land use change on biosphereatmosphere-exchange of C and N trace gases in the humid, subtropical region of Queensland. PhD Thesis, Queensland University of Technology, Australia.
Rowlings D, Grace P, Kiese R, Weier K (2012) Environmental factors controlling temporal and spatial variability in the soil-atmosphere exchange of C[O.sub.2], [CH.sub.4] and [N.sub.2]O from an Australian subtropical rainforest. Global Change Biology 18, 726-738. doi: 10.1111/j.1365-24 86.2011.02563.x
Ruz-Jerez BE, White RE, Ball PR 0994) Long-term measurement of denitrification in three contrasting pastures grazed by sheep. Soil Biology & Biochemistry 26, 29-39. doi:10.1016/0038-0717(94)90 192-9
Saggar S, Andrew RM, Tate KR, Hedley CB, Rodda N J, Townsend JA (2004) Modelling nitrous oxide emissions from dairy-grazed pasture. Nutrient Cycling in Agroecosystems 68, 243-255. doi:10.1023/B: FRES.0000019463.92440.a3
Saggar S, Giltrap D, Li C, Tate K (2007) Modelling nitrous oxide emissions from grazed grasslands in New Zealand. Agriculture, Ecosystems & Environment 119, 205-216. doi: 10.1016/j.agee.2006.07.010
Sehy U, Ruser R, Munch J (2003) Nitrous oxide fluxes from maize fields: relationship to yield, site-specific fertilization, and soil conditions. Agriculture, Ecosystems & Environment 99, 97-111. doi:10.1016/ SO 167-8809(03)00139-7
Sozanska M, Skiba U, Metcalfe S (2002) Developing an inventory of [N.sub.2]O emissions from British soils. Atmospheric Environment 36, 987-998. doi: 10.1016/S1352-2310(01)00441-1
Wang Y, Xue M, Zheng X, Ji B, Du R, Wang Y (2005) Effects of environmental factors on [N.sub.2]O emission from and CH4 uptake by the typical grasslands in the Inner Mongolia. Chemosphere 58, 205-215. doi:10.1016/j.chemosphere.2004.04.043
Wei W (1990) 'Time series analysis.' (Addison Wesley Longman: New York)
Williams E, Hutchinson G, Fehsenfeld F (1992) NOx and [N.sub.2]O emissions from soil. Global Biogeochemical Cycles 6, 351-388. doi:10.1029/ 92 GB 02124
Willmott CJ (1981) On the validation of models. Physical Geography 2, 184-194.
Yaffee RA, McGee M (2000) 'Introduction to time series analysis and forecasting with applications of SAS and SPSS.' (Academic Press: USA)
Yan Y, Sha L, Cao M, Zheng Z, Tang J, Wang Y, Zhang Y, Wang R, Liu G, Wang Y, Sun Y (2008) Fluxes of CH4 and [N.sub.2]O from soil under a tropical seasonal rain forest in Xishuangbanna, Southwest China. Journal of Environmental Sciences 20, 207-215. doi:10.1016/S1001-0742(08) 60033-9
Zhu R, Sun L, Zhao S, Xie Z, Liu X, Yin X (2005) Preliminary studies on nitrous oxide emissions from the ornithogenic soils on Xi-sha atoll, South China Sea. Journal of Environmentul Sciences 17, 551-556.
Xiaodong Huang (A), Peter Grace (A,B), Keith Weier (B), and Kerrie Mengersen (A,B,C)
(A) Science and Engineering Faculty, Queensland University of Technology, GPO Box 2434, Brisbane, Qld 4001, Australia.
(B) lnstitute for Future Environments, Queensland University of Technology, GPO Box 2434, Brisbane, Qld 4001, Australia.
(C) Corresponding author. Email: email@example.com
Table 1. Summary statistics of nitrous oxide ([N.sub.2]O) emissions (g [N.sub.2]O-N-ha.day), soil water content (%), and soil temperature ([degrees]C) measured weekly in five mango crop plots, five custard apple crop plots, and six pineapple crop plots under subtropical horticultural crops at the Maroochy Research Station, Nambour, Queensland (2006-07) Crop Variable Mean s.d. Min. Max. Mango [N.sub.2]O flux 4.6 1.8 0.1 11.2 Soil water content 25.2 8.0 7.0 51 Soil temperature 21.1 4.2 12.3 28.4 Custard apple [N.sub.2]O flux 5.9 7.2 0.0 92.7 Soil water content 12.7 5.3 4.0 36.5 Soil temperature 22.4 3.5 14.4 29.8 Log-transformed 1.75 0.5 0.0 4.5 [N.sub.2]O flux Log-transformed soil 2.5 0.4 1.4 3.6 water content Pineapple [N.sub.2]O flux 3.3 1.9 0.0 16.4 Soil water content 11.2 4.5 4.0 33.8 Soil temperature 22.7 3.2 15.7 28.4 Log-transformed 1.4 0.4 0.0 2.9 [N.sub.2]O flux Log-transformed soil 2.4 0.3 1.4 3.5 water content Crop Variable Skewness Mango [N.sub.2]O flux 0.8 Soil water content 0.1 Soil temperature -0.2 Custard apple [N.sub.2]O flux 8.5 Soil water content 1.9 Soil temperature -0.3 Log-transformed 1.1 [N.sub.2]O flux Log-transformed soil 0.4 water content Pineapple [N.sub.2]O flux 2.0 Soil water content 2.9 Soil temperature -0.3 Log-transformed 0.0 [N.sub.2]O flux Log-transformed soil 1.1 water content Table 2. Statistical model comparison describing nitrous oxide emissions using soil water content and soil temperature from soils under subtropical horticultural crops at the Maroochy Research Station, Nambour, Queensland (2006-07) Sinusoidal term: Seasonality was written as sin2[pi]t/T and cos2[pi]t/ T; temporal lagged factors: soil water content at a lag of 1 week/ month and soil temperature at a lag of 1 week/month; autoregressive process: AR(1) and AR(2); seasonal factors: seasonality was described as a categorical variable (spring, summer, autumn, and winter); d: index of agreement Model Parameter Crop Weekly data [R.sup.2] Adj. [R.sup.2] 1 Soil water content Mango 0.226 0.203 Soil temperature Custard apple 0.278 0.257 Sinusoidal term Pineapple 0.378 0.363 Temporal lagged -- factors Autoregressive process 2 Soil water content Mango 0.2 0.18 Soil temperature Custard apple 0.277 0.259 Seasonal factor Pineapple 0.378 0.365 Temporal lagged -- factors Autoregressive process 3 Soil water content Mango 0.181 0.164 Soil temperature Custard apple 0.276 0.261 Temporal lagged Pineapple 0.373 0.362 factors -- Autoregressive process 4 Soil water content Mango 0.03 0.022 Soil temperature Custard apple 0.042 0.035 Pineapple 0.012 0.005 Model Parameter Crop Weekly Monthly data data d [R.sup.2] 1 Soil water content Mango 0.603 0.585 Soil temperature Custard apple 0.639 0.529 Sinusoidal term Pineapple 0.732 0.405 Temporal lagged -- factors Autoregressive process 2 Soil water content Mango 0.573 0.496 Soil temperature Custard apple 0.639 0.401 Seasonal factor Pineapple 0.733 0.378 Temporal lagged -- factors Autoregressive process 3 Soil water content Mango 0.549 0.487 Soil temperature Custard apple 0.639 0.297 Temporal lagged Pineapple 0.729 0.378 factors -- Autoregressive process 4 Soil water content Mango 0.013 Soil temperature Custard apple 0.089 Pineapple 0 Model Parameter Crop Monthly data Adj. d [R.sup.2] 1 Soil water content Mango 0.504 0.854 Soil temperature Custard apple 0.437 0.828 Sinusoidal term Pineapple 0.312 0.754 Temporal lagged -- factors Autoregressive process 2 Soil water content Mango 0.412 0.807 Soil temperature Custard apple 0.302 0.746 Seasonal factor Pineapple 0.294 0.736 Temporal lagged -- factors Autoregressive process 3 Soil water content Mango 0.416 0.803 Soil temperature Custard apple 0.199 0.65 Temporal lagged Pineapple 0.307 0.736 factors -- Autoregressive process 4 Soil water content Mango -0.022 Soil temperature Custard apple 0.067 Pineapple -0.029 Table 3. Time series regression parameters for estimating weekly and monthly nitrous oxide emissions from soils under subtropical horticultural crops at the Maroochy Research Station, Nambour, Queensland (2006-07) Parameters Weekly models Coefficients s.e. P values Mango Constant -0.584 1.932 0.763 [[phi].sub.1] 0.271 0.065 0.000 [[phi].sub.2] 0.011 0.066 0.868 [[beta].sub.1] 0.136 0.081 0.094 [[gamma].sub.1] -- -- -- [[beta].sub.2] 0.059 0.026 0.022 [[delta].sub.1] -0.02 0.024 0.415 [alpha] -0.033 0.406 0.935 b -0.891 0.318 0.005 Custard apple Constant 0.449 0.555 0.419 [[phi].sub.1] 0.428 0.065 0.000 [[phi].sub.2] 0.131 0.066 0.049 [[beta].sub.1] 0.004 0.021 0.843 [[gamma].sub.1] -- -- -- [[beta].sub.2] 0.156 0.093 0.094 [[delta].sub.1] -0.065 0.093 0.481 [alpha] -0.003 0.08 0.607 b -0.065 0.093 0.971 Pineapple Constant 0.089 0.382 0.816 [[phi].sub.1] 0.499 0.058 0.000 [[phi].sub.2] 0.148 0.057 0.010 [[beta].sub.1] 0.013 0.014 0.354 [[gamma].sub.1] -- -- -- [[beta].sub.2] 0.166 0.076 0.031 [[delta].sub.1] -0.13 0.075 0.084 [alpha] -0.015 0.52 0.767 b -0.058 0.046 0.207 Parameters Monthly models Coefficients s.e. P values Mango Constant -0.648 4.318 0.131 [[phi].sub.1] -0.321 0.109 0.005 [[phi].sub.2] 0.125 0.112 0.269 [[beta].sub.1] 0.023 0.124 0.854 [[gamma].sub.1] 0.474 0.134 0.001 [[beta].sub.2] 0.041 0.026 0.116 [[delta].sub.1] 0.004 0.026 0.867 [alpha] -2.549 0.959 0.011 b -2.377 0.772 0.004 Custard apple Constant -3.983 2.364 0.100 [[phi].sub.1] 0.03 0.151 0.843 [[phi].sub.2] 0.606 0.151 0.000 [[beta].sub.1] -0.057 0.065 0.388 [[gamma].sub.1] 0.286 0.061 0.000 [[beta].sub.2] -0.145 0.21 0.492 [[delta].sub.1] -0.177 0.231 0.448 [alpha] -1.338 0.372 0.001 b -0.704 0.427 0.106 Pineapple Constant -0.122 1.297 0.925 [[phi].sub.1] 0.137 0.114 0.234 [[phi].sub.2] 0.479 0.107 0.000 [[beta].sub.1] -0.032 0.031 0.301 [[gamma].sub.1] 0.089 0.035 0.013 [[beta].sub.2] -0.009 0.155 0.952 [[delta].sub.1] -0.351 0.212 0.144 [alpha] -0.292 0.191 0.132 b -0.208 0.186 0.268
|Printer friendly Cite/link Email Feedback|
|Author:||Huang, Xiaodong; Grace, Peter; Weier, Keith; Mengersen, Kerrie|
|Date:||Oct 1, 2012|
|Previous Article:||Soil selenium in a forested seabird colony: distribution, sources, uptake by plants, and comparison with non-seabird sites.|
|Next Article:||Interactive effects from combining inorganic and organic fertilisers on phosphorus availability.|