Midmorning Point Sampling May Not Accurately Represent Nitrous Oxide Emissions Following Fertilizer Applications

A common approach for measuring N2O emissions is to collect midmorning or early evening gas samples from experiments utilizing the chamber-based flux methodology. However, due to high spatial and temporal variability, N2O estimates based on midmorning or early evening sampling may not provide accurate estimates of total emissions. This study determined the impact of sampling collection timing on the precision and accuracy of N2O emissions estimates. Nitrous oxide emissions, air and soil temperatures, and soil moisture were measured for 21 d following the application of 224 kg urea-N ha–1 on 20 Sept. 2017, 11 Oct. 2017, and 1 May 2018, at six time intervals (0130–0230, 0530–0630, 0930–1030, 1330–1430, 1730–1830, and 2130–2230 h) over a 24-h period. Based on multiple daily measurements, point samples collected between 0930 and 1030 h (midmorning) were inconsistent in their ability to predict N2O emissions. However, samples collected between 2130 and 2230 h (early evening) were similar to average emissions. The number of randomly collected point samples to be within 20% of the mean 80% of the time over a 21-d period ranged from 13 samples for fertilizer applied on 20 Sept. 2017 to 48 samples for fertilizer applied on 11 Oct. 2017. This research indicates that management and climatic variability affect N2O emissions, and that accurate sampling protocols vary across management and climates. To reduce uncertainty, N2O sampling protocol should be tested under conditions likely to occur and where possible, nearcontinuous measurement systems should be adopted.


I
n crop production, two important greenhouse gases used in carbon footprint calculations are CO 2 and N 2 O. Carbon dioxide is released as soil organic matter is mineralized (Clay et al., 2012), and N 2 O is released during nitrification, denitrification, and co-denitrification (Chang et al., 2016;Selbie et al., 2015).In carbon footprint calculations, all greenhouse gases are converted to an equivalent amount of carbon dioxide using appropriate conversion factors.For example, based on IPCC (2007), 1 kg of N 2 O is equivalent to releasing 298 kg of CO 2 .Carbon footprints are used to document the impacts of human activities on global warming potential and to determine if individual products or services achieve regulatory requirements (California Environmental Protection Agency, 2009;Stone et al., 2012;Lupo et al., 2013;Clay et al., 2012;Butterbach-Bahl et al., 2013).However, the ability to measure greenhouse gas emissions from individual activities has not kept pace with the need for accurate information.
In annual crop production systems, carbon footprints generally increase with the amount of N fertilizer applied (Clay and Shanahan, 2011) and decrease with sequestered carbon (Clay et al., 2012).The large carbon footprint associated with and the impact of N 2 O on the ability of the atmosphere to store energy (Klein et al., 2006;Butterbach-Bahl et al., 2013).Nitrous oxide emissions can be separated into at least two processes: the production of N 2 O by soil organisms and the equilibrium relationships between N 2 O solubility in water and the amount of N 2 O in the air-filled pore space.Both of these processes interact to influence N 2 O emissions from soil.

Nitrous Oxide Production
In aerobic soils, select soil organisms have the capacity to obtain energy from conversion of ammonia to nitrate.This process, called nitrification, is not 100% efficient, and a portion of the nitrified ammonia is released as N 2 O (Khalil et al., 2004).Under anaerobic conditions, some soil organisms have the capacity to use nitrate as the terminal electron acceptor in respiration.End products of this process, called denitrification, include N 2, N 2 O, and NO.The percentage of NO 3 converted to nitrous oxide (N 2 O) varies across management systems and soils (Liu et al., 2016).Nitrification and denitrification are temperaturedependent processes that can be modeled using mathematical equations (Rodrigo et al., 1997;Shurpali et al., 2016).

Nitrous Oxide Emissions
Once produced, nitrous oxide is contained within either the soil solution or the air-filled pore space.The amount of soil gases contained within the soil solution can be modeled using Henry's Law, p = K H,T ´ c, where p is the partial pressure of the solute above the solution, c is the concentration in the solution, and K H,T is the temperature dependent Henry's Law gas constant (Montes et al., 2009).This equation indicates that N 2 O solubility decreases with increasing temperature, which in turn should increase its concentration in the air-filled pore space.Henry's Law suggests that temperature induced changes in N 2 O solubility, has the potential to produce emissions patterns that are aligned with soil temperature.However, emissions are also influenced by the ability of the gas to move from the soil gas phase to the atmosphere.
The amount of N 2 O emitted from soil can be estimated using Fick' s Law, F » D s ´ dC/dz, where F is the gas flux, D s is the soil-gas diffusion coefficient, dC is the concentration gradient, and dZ is the change in distance (DeSutter et al., 2008).Increasing soil moisture reduces D s which slows GHG emissions (Moldrup et al., 2000).In addition, D s may also be a function of tillage, which influences the pore size distribution, crusting, and the length of time required for water to percolate through the soil (Moldrup et al., 2000).For example, Jung et al. (2007) reported that over a 2-h period, cumulative water infiltration was 15 mm in three annually cropped soils and 45 mm in three grassland soils.The higher water infiltration rate in the grassland than the cropped soils most likely resulted from an increase in the number of large pores, which has the potential to increase D s (Eynard et al., 2004).
Based on the above discussion, the alignment of the soil temperatures and N 2 O emissions depends on at least the temperature-induced changes in microbial activity and solubility and the management and soil impacts on D s .It makes sense that if the soil temperatures and N 2 O emissions were aligned then sampling at the average soil temperature, as proposed by Parkin and Venterea (2010), Alves et al. (2012), and de Klein and Harvey (2015), may produce accurate GHG emission estimates.However, if emissions are controlled by D s which is impacted by management and climatic conditions, then N 2 O emissions and soil temperature may not be aligned.Under these conditions, point sampling based on the average soil temperature may not accurately estimate GHG emissions and a near continuous sampling approach might be appropriate (Denmead et al., 1979;Williams et al., 1999;Smith and Dobbie, 2001;van der Weerden et al., 2013).
These findings suggest that a one-size-fits-all sampling protocols for GHG sampling may not be appropriate.This problem was noted by Parkin and Venterea (2010), who stated that, "Because of our inability, at this time, to precisely assess the extent of bias associated with a given chamber design and sampling protocol under the range of conditions which might exist, we have adopted our 'best guess' protocol."The hypothesis of this study was that weekly point measurements of N 2 O flux collected at the average air temperature provides the accuracy and precision needed to accurately determine N 2 O fluxes following fertilizer applications.The objective of this research was to determine the impact of sampling time on the precision and accuracy of N 2 O emissions estimates.

MATErIALS AND METHODS
A field study was conducted at the Aurora Research Farm near Aurora in South Dakota (44°18¢20.57²N, 96°40¢14.04²W).The soil type was a Brandt silty clay loam (Fine-silty, mixed, superactive, frigid Calcic Hapludolls), with percentages of clay, silt, and sand of 28, 65, and 7%, respectively.The initial soil organic C for the surface 15 cm was 36 Mg ha -1 , and the carbon mineralization kinetics were previously discussed in Clay et al. (2015).The soil parent materials were loess (0-60 cm) over glacial outwash.The soil water contents at the field capacity and the wilting point were 0.315 and 0.177 cm 3 cm -3 , respectively, and the soil bulk density (0-15 cm) at the beginning of the study was 1.29 g cm -3 .A more complete description of these soils are available in Kim et al. (2008) and Clay et al. (1994Clay et al. ( , 1995Clay et al. ( , 1996Clay et al. ( , 2015)).
Two long-term opaque chambers (8100-104, LI-COR, Lincoln, NE), that were designed to measure GHG emissions were installed according to LI-COR protocol.Each chamber consisted of a base and cover that pivots to cover the base during sample collection.The bases were only covered when the samples were collected.All chambers were within an area that had the dimensions of 10 m by 10 m.During sample collection, the air temperatures within the chambers were measured with a thermistor, a vent was used to equalize the chamber and atmospheric pressures, and the gas within the chamber was mixed.The chambers were sampled sequentially in a designated sequence, and corrections were applied individually to account for the air volume enclosed between the soil surface and the chamber cover.Gas samples were collected from each chamber every 4 h.At each sampling time, samples were collected at 1-s intervals for 15 min from a 317-cm 2 area.A total of 900 samples were collected and analyzed for N 2 O by a Picarro Cavity Ringdown Spectrometer (model G2508; Picarro Inc., Santa Clara, CA).Nitrous oxide flux was calculated from 45 to 900 s after chamber closing using the LI-COR SoilFluxPro software (v.4.0.1)(LI-COR, Lincoln, NE).Soil moisture and temperature for the surface 5 cm were measured using soil moisture (LI-COR 8100-204) and temperature (LI-COR 8150-203) probes.The Picarro Cavity Ringdown Spectrometer factory calibration was checked with N 2 O standards at the beginning and end of each experiment.The standards were purchased from Airgas Specialty Gases (Airgas USA LLC, Cinnaminson, NJ), and they had N 2 O concentrations of 0.378 and 149 ppm.The equation between the standard and the factory calibrations, conducted pre-and postexperiment was y = 0.02 + 1.013(standard), r 2 = 0.99; p < 0.01.
In this experiment, urea (224 kg N ha -1 ) was dissolved in 10 mL water and was applied to the soil surface within the long-term chambers.Nitrous oxide-N emissions following the application of urea was measured six times daily (0130-0230, 0530-0630, 0930-1030, 1330-1430, 1730-1830, and 2130-2230 h) for 21 d.These times were selected to match the average, minimum, and maximum soil and air temperatures.This experiment was based on prior research that showed that N 2 O emission increase for a relatively short period followed fertilizer applications (Clay et al., 1990a;Omonode et al., 2015;Fujinuma et al., 2011;Fernández et al., 2016;Thomas and Hao, 2017).The N treatments were applied on 20 Sept. 2017, 11 Oct. 2017, and 1 May 2018.Within a fertilizer application date, each treatment was applied to two chambers.Across the three application dates, fertilizer materials were applied to given plot only once.
At the site, soybeans (glycine max) was seeded into the notillage field in May 2018.These plants were removed from the study site prior to the start of the study.During the experiment, between 50 and 60% of soil surface was covered with plant residue.For the fertilizer applied on 20 Sept. 2017, emissions were measured from 20 Sept. to 11 Oct. 2017.For the fertilizer applied on 11 Oct. 2017, emissions were measured from 11 Oct. to 1 Nov. 2017.For the fertilizer applied on 1 May 2018, emissions were measured from 5 May to 22 May 2018.
Prior to the start of each fertilizer application date, soil samples from two soil depth increments (0-15 and 15-30 cm) were collected from non-treated areas located adjacent to the chambers.At the completion of the study, two soil samples (0-15 and 15-30 cm) were collected from within the chambers.Soil samples were dried, ground, sieved and analyzed for inorganic N (Kim et al., 2008).Bulk densities of the sampling depths were determined.A comparison between the N 2 O measurements for two chambers were conducted for the three fertilizer application dates.This comparison showed that for the 20 September, 11 October, and 1 May experiments, N 2 O measurements in the two chambers were highly correlated (p < 0.01), and the correlation coefficients between the chambers were 0.44, 0.73, and 0.59, respectively.The complete data set used in this experiment is available in Thies (2018).

Data Analysis
A fast Fourier transform (FFT) was conducted to identify the dominant frequencies (cycles per day) for the air temperatures and N 2 O-N emissions (Chang et al., 2017).This transformation converts data that follows a periodic function from the time to the frequency domain.The transformation method was summarized by Klingenberg (2005).
Each point measurement was converted to time adjusted values that represented a 4-h block.The N 2 O emissions from the two chambers were combined into a single data set that consisted when the samples were collected and time adjusted N 2 O emission values.This data set was sorted, based on the hour and minute of sample collection, and separated into six time periods (0130-0230, 0530-0630, 0930-1030, 1330-1430, 1730-1830, and 2130-2230 h).Total emissions were determined by summing each of the time adjusted point measurement over the 21 d.Means, variance, confidence intervals, and the sampling requirements for each of the six sampling intervals for the two soil chambers were determined.The average emissions for the six time periods were compared with the average emission over the 21-d experiments.
Within a fertilizer application date, a t test was used to compare the average N 2 O emissions at each time-period with average emissions over the study period.An F statistic was used to compare the variances of the different sampling periods with the variance across all sampling intervals.The estimated random sampling requirements were determined with the Stein equation (Stein, 1945), n = (t 2 p ´ s 2 )/d 2 , where t 2 p was the student t value associated with a specific probability level and degrees of freedom, s 2 was the variance, and d was 1/2 the total desired range of the mean.Using this equation, the sampling requirement for producing N 2 O measurements ±20% and ±30% of the mean 80% of the time (a = 0.2) were calculated.This equation was previously used to estimate soil sampling requirements (Black, 1992;Skopp et al., 1995;Clay et al., 1997Clay et al., , 2002;;Nolan et al., 2006).
The average air chamber temperatures during gas sample collection were used to determine the average air temperature during the heating and cooling periods.Soil temperatures generally increased from 0600 to 1600 h and decreased from 1600 to 0600 h.The average temperature during the warming period was identified as "T ave,heat " (midmorning) and the average temperature during the cooling period was identified as "T ave,cool " (early evening).

Soil Inorganic Nitrogen
For the 20 Sept. 2017 fertilizer application date, the initial amount of inorganic N in the soil prior to applying the fertilizer for the 0-to 30-cm depth was 48.9 kg ha -1 (Table 1).At the completion of the study, the amount of inorganic N in the surface 30 cm was 123 kg N ha -1 .For the 20 Oct. 2017 experiment, prior to applying the fertilizer the 0-to 30-cm soil depth contained 65 kg ha -1 of inorganic N and at the completion of this experiment, these plots contained 156 kg N ha -1 .For the 1 May 2018 experiment, the inorganic N in the surface 30 cm prior to the experiment was 39 kg N ha -1 .When the experiment was completed, the soil contained 107 kg N ha -1 .The difference between the initial plus the applied N (224 kg N ha -1 ) and the amount of N remaining in the soil at the completion of the experiment suggests that during the experiment, NH 3 volatilization, N 2 O emissions, nitrate leaching, microbial immobilization, and/or fixation may have combined to reduce the concentration of inorganic N in the soil matrix (Clay et al., 1990a(Clay et al., , 1990b)).

rainfall and Soil Moisture
Rainfall between 20 Sept. and 11 Oct. 2017 was 14.3 cm and rainfall between 11 Oct. and 1 Nov. 2017 was 0.51 cm (Table 1).Rainfall between 5 May and 20 May 2018 was 1.63 cm.Between 20 September and 11 October, the volumetric soil moisture contents for the surface 5 cm ranged from 0.35 (68.2% water filled pore space) to 0.50 cm 3 cm -3 (98% water-filled pore space).Between 11 October and 1 November, the volumetric water content for the surface 5 cm ranged from 0.18 (35% water-filled pore space) to 0.42 cm 3 cm -3 (82% water-filled pore space), and from 5 May to 22 May, soil moisture ranged from 0.28 (55% water-filled pore space) to 0.37 cm 3 cm -3 (72% water-filled pore space).Additional information on soil moisture contents are available in Table 1 and Thies (2018).

Chamber Temperature
The average soil (0-5 cm) and air temperatures decreased as the fall progressed and were 13.5 and 15.2°C between 20 September and 11 October, respectively.Between 11 October and 1 November the average air and soil temperatures were 7.3 and 9.3°C, respectively.In the following spring, average air and soil temperatures from 5 May to 22 May were 15.6 and 15.6°C, respectively.Additional information on soil temperatures is available in Table 1 and Thies (2018).
The T ave,heat value represents the half-way point between the minimum and maximum temperatures, during a portion of the day when air temperatures were increasing.This value represents the midmorning value as specified by Parkin and Venterea (2010).From 20 Sept. to 11 Nov. 2017, T ave,heat occurred at approximately 1030 h (Table 1), which was similar to Chang et al. (2016).For the October 2017 and May 2018 experiments, the T ave,heat values were 0924 and 1012 h, respectively.
The T ave,cool values represents the half-way point between the maximum and minimum values during the time of day when air temperatures were decreasing.For the September, October, and May experiments, T ave,cool occurred at 2218, 2200, and 2212 h, respectively.In the USDA-ARS GRACnet protocols (Parkin and Venterea, 2010) the T ave,cool values should represent the early evening value.

Diurnal and Season Patterns
For the three fertilizer application dates, air temperatures and N 2 O emissions followed diurnal cycles.Air temperatures and N 2 O emissions data and the fast Fourier Transformation (FTT) of this data were shown in Fig. 1 and 2. In the FFT, the frequency is the number of cycles per day, and the magnitude represents the dominance of that frequency.The FFT is an analytical approach for inspecting data for repeating cycles.Frequencies with large magnitudes indicate that frequency has a large impact on the measured data.For example, in Fig. 1 and 2, large magnitude associated with the one-cycle-per-day frequency indicated that the chamber air temperature and N 2 O-N emissions measurements were influenced by a diurnal cycles.However, data sets can also contain low and high frequency cycles.The low frequency temperature and N 2 O emissions FFTs were attributed to gradual temperature and emissions decreases as the season progressed.
Table 1.Mean median, standard deviation, and confidence interval of air temperatures, rainfall and inorganic N (0-30 cm depth).Also included in this table are the average soil moisture and temperatures content for the surface 5 cm during three time periods following the application of urea fertilizer.The average temperature during the warming period was identified as "T ave,heat " (midmorning) and the average temperature during the cooling period was identified as "T ave,cool " (early evening).Times when T ave,heat and T ave,cool occurred are given in decimal format (e.g., 10.54 = 10 + 32/60 = 1032 h).

Emission Signatures
For fertilizer applied on 20 Sept. 2017, the maximum N 2 O-N emission and air temperatures generally occurred between 1330 and 1430 h (Table 2).However, there were important exceptions when the temperature and N 2 O emissions were not aligned (Fig. 3 and 4).For example, non-alignment occurred between 28 and 30 September (Fig. 3).As predicted by Fick's Law, nonalignment can result from high soil water contents and/or soil crusting (Clay et al., 1990a;DeSutter et al., 2008;Balaine et al., 2013).The non-alignment between temperature and emissions peaks were not permanent and by 2 Oct. 2017, temperatures and emissions were aligned.The reasons for alignment of these cycles between the 2nd and 6th of October were not identified.
It is likely that the length of time that the temperatures and emissions are not aligned depends on many factors including the source of the GHG, microbial activity, soil temperature and texture, and the soils water flow characteristics.In well drained soils with high water infiltration rates, non-alignment may be very short.For example, in research reported by van der Weerden et al. (2013) on well drained perennial grassland that had a low bulk density (0.96 g cm -3 ), they recommended that sampling at the average temperature between 1000 and 1200 h had zero bias.
For the 20 Sept. 2017 fertilizer application date, the median T ave,heat value was at 0936 h (Table 1), and if samples were collected between 0930 and 1030 h total emissions would have been underestimated by 18% (Table 2).For the 11 Oct. 2017 fertilizer application date, the N 2 O emission peak occurred between 1730 and 1830 h, and the minimum value occurred between 0930 and 1030 h.Point sampling at T ave,heat would have underestimated total N 2 O emissions by 31%.
In the second example, the impact of soil moisture on N 2 O emissions for spring applied N was explored (Fig. 4).This relationship showed that emissions decreased rapidly with an increase in soil water content from 0.32 to 0.36 cm 3 cm -3 on 11 May 2018.This decrease was attributed to a decrease in diffusivity resulting from an increase in the water-filled pore space and/or a decrease in N 2 O production.Emissions remained relatively low until 13 May 2018, and from 13 May to 16 May, water contents gradually decreased to 0.32 cm 3 cm -3 , and diurnal N 2 O emission cycles were again evident in the data set.These results highlight the importance of considering that emissions are the result of interactions between different components of a complex system.
For the 1 May 2018 fertilizer application date, maximum emissions occurred between 1330 and 1430 h, and that samples collected between 1330 and 1430 h overestimated emissions by 26% (Table 2).Minimum emissions occurred between 0130 and 0230 h and underestimated emissions by 22%.It is important to point out that for this application date, gas samples collected between 0930 and 1030 h, which included T ave,heat were similar to total emissions.Temperature dependence of N 2 O emission patterns is not new and has been reported by others (Burzaco et al., 2013;Chang et al., 2016).
These findings suggest that midmorning sampling was inconsistent in the measured values to accurately predict total emissions.Our findings were conceptually in agreement with Williams et al. (1999) and Blackmer et al. (1982).Both of these studies indicated that samples collected at prescribed times may not provide accurate indicators of total flux.Our findings also differ from van der Weerden et al. ( 2013), who evaluated N 2 O emissions in a moderately well drained soil supporting perennial plants.

Sampling requirement
For fertilizer applied on 20 Sept. 2017, samples collected between 0930 and 1030 h had variances that were similar or lower than those collected across all sampling periods.For this fertilizer application date, the Stein (1945) equation predicted that 13 randomly collected samples over 21 d were required to estimate N 2 O emissions to within 20% of the mean 80% of the time.For fertilizer applied on 11 Oct. 2017, 38 and 17 samples were required to be within 20 and 30% of the mean 80% of the time, respectively.The higher sampling requirement for fertilizer applied on 11 October than 20 September was attributed to the calculation approach, which resulted in smaller denominator [(0.3 ´ mean) 2 ] for fertilizer applied on 11 October than 20 September.For fertilizer applied on 1 May 2018, 13 and 6 samples were required to be within 20 and 30% of the mean 80% of the time, respectively.These findings suggest that a large number of samples are required to provide precise emission estimates.
Our results differ from Parkin (2008), where sampling every 3 d resulted in flux measurements that were within 10% of expected value.Differences between our results and Parkin (2008) were attributed to the data processing method.Parkin (2008) determined daily emissions by averaging the four values collected at 0600, 1200, 1800, and 2400 h over a 24-h period.It is important to note that because samples were not collected between 1330 and 1430 h, total daily emissions may have been underestimated.Parkin (2008) then resampled this constructed data set to estimate the seasonal sampling requirement.We tested the Parkin (2008) analysis approach using the data from our study.As expected, our analysis showed that by averaging the six values collected over a 24-h period, reduced calculated sampling requirement by 38%.
Other factors that should be considered when designing sampling protocol include the expected variance structure.For example, the different sampling dates may have different variances (Thomas and Hao, 2017;Woodley et al., 2018).In our experiment, samples collected for the 21 d after the 20 Sept. 2017, 11 Oct. 2017, and 1 May 2018 had a variances of 6557, 1497, and 13336 (Table 2), respectively.In addition, if the total annual emissions over a season is estimated using the equation Regardless of the approach used to aggregate temporal data, it is important to be aware of underlying assumptions.Many statistical methods assume homogeneity of variance and there are consequences of having non-constant variances (Ott and Longnecker, 2016).These include inaccurate parameter estimates, invalid statistical tests, and interval estimates that are for a sampling time interval is different from the average emissions across all sampling times at the 0.05% level ** Reported value for a sampling time interval is different from the average emissions across all sampling times at the 0.01% level.† Reported value for a sampling time interval is different from the average emissions across all sampling times at the 0.10% level.
either too wide or narrow depending on an experimental region with the low or high variances.

SUMMAry
The hypothesis for our paper was that weekly point measurements of N 2 O flux collected at the average air temperature provides the precision needed to accurately determine fluxes following fertilizer applications.In this study, sampling between 0930 and 1030 h, which contained the median T ave,heat value for the three fertilizer application dates, underestimated N 2 O-N emissions for two of the three application dates.These results were attributed to many factors including, reduced diffusivity resulting from increased soil moisture and/or that the surface soil was dispersed and/or this sampling protocol did not adequately represent the complexity of annually cropped soil system.These results highlight the importance of considering that N 2 O emissions can be separated into at least two components, N 2 O production and solubility that interact to influence emissions.Each of these components can be described mathematically and are impacted by temporal changes in temperature, moisture, and microbial activity.The research findings indicate that, in cropped systems, sampling at a designated time may lead to uncertainty in calculated emissions.Obtaining accurate nitrous oxide emissions measurements is critical to accurately calculating carbon footprints.
This study also considered how often samples need to be collected.Protocols suggested by Parkin and Venterea (2010), Alves et al. (2012), van der Weerden et al. ( 2013), and de Klein and Harvey (2015) suggest that sample timing could range from daily to several times a week.Based on the Stein (1945) equation, GHG sampling twice a week would have resulted in six samples collected over the 21-d experiments, which was less than the calculated sampling requirements (Table 2).Additional characteristics that should be considered when assessing reliability of the calculated measurement are provided by Rochette and Eriksen-Hamel (2008).These considerations include: the absence of absolute reference gas, the lack of consistent chamber deployment time, variation in the number of air samples collected, and differences in mathematical model used to estimate emissions.
Alternative sampling approaches to improve reliability might include the adoption of near continuous measurements of temporal and spatial changes in GHG, temperature, water filled pore space and GHG diffusivity (Balaine et al., 2013;Venterea et al., 2009).Advantages of near continuous measurements are that: (i) emission delays or decreases resulting from low soil temperatures and high soil water contents can be more adequately resolved; and (ii) complex relationships among soils, climate, management, GHG emission and plants can be defined (Shurpali et al., 2016).
If continuous measurement systems are not adopted, locally based point sampling protocols need to be tested for precision and accuracy.Our results suggest that it will be difficult to conduct a meaningful meta-analysis of the published findings, if all studies use unique sampling protocols and do not provide detailed information on temporal changes in soil moisture and temperature (Balaine et al., 2013).For example, in South Dakota, Lehman and Osborne (2016) collected gas samples at approximately 1000 h, whereas in Indiana, Owens   (2017) collected samples between 1000 and 1200 h.In Colorado, Halvorson and Del Grosso (2012) collected samples in midmorning, and in Minnesota, Venterea and Coulter (2015) collected samples between 1000 and 1300 h.
In conclusion, N 2 O-N emissions and air chamber temperatures generally followed diurnal cycles.However, this was not the case following rainfall events.These results were important because they suggest that during a time-period of high N 2 O emissions, temperature-based sampling protocols may or may not produce accurate estimates of total emissions.Accurate N 2 O estimates require a sampling system that accounts for temperature induced diurnal cycles, as well as changes in GHG fluxes following climate and management perturbations.To interpret emission signatures, detailed information on temporal changes in soil moisture, bulk density, and temperature are needed.Many studies do not provide this information.This lack of information when combined with flux differences between point and continuous measurement systems results in uncertainty in the published findings (Weitz et al., 1999;Smith and Dobbie, 2001).Our findings are conceptually in agreement with Barton et al. (2015).

Fig. 1 .
Fig. 1.An example of the mathematical conversion of air temperature (a) to the frequency domain (b) using a fast Fourier Transformation (FFT).The data used in this example were collected between 22 Sept. and 1 Oct. 2017.The FFT shows that the data set contains both long (short frequencies) and diurnal cycles (frequency at 1 cycle per day).Short frequencies or long cycles were attributed to the gradual cooling of the soil.The magnitude in the FFT represents the importance of that frequency.The average volumetric soil moisture content for this time period was 0.42 cm 3 cm -3 .
likely that the resulting estimate may not account for non-constant variances.

Fig. 3 .
Fig. 3. Nitrous oxide emissions, air temperatures, and soil moisture for urea applied on 20 Sept. 2017.Values in the upper panels (a, b) are data from 28 to 30 September.Values in the lower panels (c, d) are data collected from 2 to 7 October.In September (b), the N 2 O peak was delayed relative to chamber air temperatures.In October (d), the air temperature and N 2 O peaks were aligned.

Fig. 4 .
Fig. 4. Nitrous oxide emissions and volumetric soil water contents from 9 to 18 May 2018.rainfall increased the soil water content from 0.32 to 0.36 cm 3 cm -3 on 11 May.A decrease in N 2 O emissions was associated with increases in soil water content.