Journal Archive

The study site was used for single cropping of paddy rice located in Tsukubamirai City, Ibaraki Prefecture, Japan (35°58ʹ27ʺN, 139°59ʹ32ʺE, 10 m A.S.L.), where Japan’s Institute for Agro-Environmental Sciences established a free-air CO₂ enrichment (FACE) facility for paddy rice in 2009. The paddy area is in a typical rural area in this region, where the land represents a mosaic dominated by cropland (mostly rice paddies), small forests, and residential areas. The area had a trapezoidal shape, with a length of about 200 to 300 m for the east-west direction and about 600 m for the north-south direction; it consisted of many fields individually managed in terms of their irrigation and drainage. This area is bordered on the east by the Kokaigawa River, on the south and west by small forests and residential areas, and on the north by the Joban Expressway. Our atmospheric monitoring point was in the paddy area, at a minimum distance of about 400 m from the expressway. Our irrigation monitoring plots were adjacent to the atmospheric monitoring point.

The following were general agricultural practices in the paddy area. Basal fertilization was conducted by a single application of organic and inorganic mixed fertilizer containing 69% organic matter (Dokidoki Yuki Ippatsu, Co-op Chemical Co. Ltd., Tokyo, Japan; N-P-K: 12%-7%-5%) at an application rate of 50 to 60 kg N ha^-1, followed by submersion and puddling in late April (the applied fertilizer was therefore incorporated into the soil). Rice seedlings were transplanted in early May. Mid-summer drainage was conducted in mid- to late June. Rice paddies were then re-submerged until drainage in mid- to late August. Paddy rice was harvested in early September. Rice paddies were entirely drained in a fallow season with a mostly bare soil surface with scattered rice residues until the next spring. Two or three tillage events with a mixing depth of approximately 15 cm were conducted in and at the end of the fallow season. A relatively low N fertilization rate is used to enhance the rice quality. The rice yield was approximately 4800 kg ha^-1 as brown rice at a moisture content of 15%. The source of irrigated water was groundwater pumped up near the paddy area and provided to each paddy using pipelines.

We defined the cropping season as the period from submersion of the rice paddies (late April) to the completion of harvesting (mid-September). The study period was 2 years, from 15 September 2010 to 12 September 2012 (i.e., 104 weeks), and was divided into four periods: fallow season 1 (FS1) from 15 September 2010 to 27 April 2011 (32 weeks), cropping season 1 (CS1) from 27 April 2011 to 14 September 2011 (20 weeks), fallow season 2 (FS2) from 14 September 2011 to 25 April 2012 (32 weeks), and cropping season 2 (CS2) from 25 April 2012 to 12 September 2012 (20 weeks). The mean air temperatures were 9.6, 22.9, 9.3, and 22.6°C in FS1, CS1, FS2, and CS2, respectively.

Target chemical species throughout the study period were NH₃, HNO₃, and HNO₂ as gases and pNH₄, pNO₃, and particulate nitrite (pNO₂) as particles. Their weekly mean air concentrations (μg N m^-3 20℃, 1013 hPa), with separate day and night values, were measured at heights of 6 and 2 m above the surface using a filter-pack method with open-face filter holders (NLO, NILU). The day/night separation in the weekly measurements was intended to reduce the systematic error in flux calculation that results from the long averaging time of air concentrations (here, 1 week) particularly for chemical species such as NH₃ and HNO₃ that show large diurnal variation in their concentrations (Hayashi et al., 2013b). No size separation was conducted for the particles.

Three sets of filter packs were prepared at each height to collect the target chemical species: one during the day and one during the night, with active sampling, and one as a field blank (passive sampling, without suction). Active sampling using an air pump (APN-085, Iwaki) was performed for the day and night sampling lines, which were switched at sunrise and sunset. Each filter pack was covered with aluminum foil as a sunshade to reduce degradation of the sampled particles by heat. The amount of each chemical species collected by the filter pack was expressed as the difference in quantities between the exposed filters and the unused filters as a blank. An ion chromatograph (ICS-1600, Thermo Scientific Dionex) was used to quantify the collected amounts. Details of the method used to calculate the weekly mean air concentrations with separation between day and night data are provided by Hayashi et al. (2013b).

Air concentrations of NO and NO₂ at the two heights were measured using a nitrogen oxides analyzer (42iTL, Thermo Scientific) from 15 June 2011 to 15 February 2012, and the difference between total nitrogen oxides and NO was defined as the NO₂ concentration. The data were obtained at 30-minute intervals and converted into the weekly mean concentrations of NO and NO₂, with separate day and night values.

The mean 1-week air concentration (without separation into day and night data) was also calculated for all the target substances as a weighted mean using the day and night lengths as a weighting factor.

Micrometeorological observations were conducted for the following factors. Air temperature and relative humidity were measured using a temperature-humidity sensor (HMP45A, Vaisala) (height of 4.8 m). Global solar radiation was measured using a pyranometer (CM3, Kipp and Zonen). Atmospheric pressure was measured using a barometer (PTB101B, Vaisala). Fluctuations in the three-dimensional wind velocity and virtual temperature at 10 Hz were measured (Model 81000 anemometer, Young) at a height of 6.0 m to determine the mean wind velocity, friction velocity, and Monin-Obukhov length. The 30-min means or sums of these factors were recorded using a data logger (CR1000, Campbell Scientific).

We applied a gradient method to determine the atmosphere-rice paddy exchange flux, in which the flux equals the difference in air concentrations between the two heights multiplied by the diffusion velocity between the two heights. The weekly mean diffusion velocity, with separate day and night values, was calculated by the method of Hayashi et al. (2013b). The weekly mean exchange flux, with separate day and night values, expressed as the flux at the standard temperature and atmospheric pressure was corrected based on the mean air temperature and atmospheric pressure during the week. The weekly mean exchange flux was converted into a weekly total flux, with separate day and night values, by multiplying the mean flux by the day and night length, respectively. The cumulative flux in each of four seasons was then calculated. For NO and NO₂, for which 30-min air concentrations were available, the exchange flux was also calculated on a 30-min basis and then summed to provide the weekly and seasonal values.

To support a discussion of the fluxes determined by the gradient method, we applied the inferential method (Matsuda, 2008) to estimate the dry deposition flux. This method calculates the total resistance that a target substance experiences during its deposition from a target height to the ground surface. The reciprocal of the total resistance denotes the deposition velocity. The dry deposition flux is obtained by multiplying the deposition velocity at a given height by the air concentration at that height. Application of the inferential method to multiple Nr species is still being refined; for example, the calculated deposition velocities of an Nr species vary greatly depending on the chosen parameterization (Flechard et al., 2011). In this study, we calculated the deposition velocities for NH₃. We used the surface resistance parameterization of Wesely (1989) except for the outer surface resistance of NH₃. For the outer surface resistance of NH₃, the non-stomatal resistance of Smith et al. (2000) was used. The weekly mean deposition velocity at a height of 6 m above the surface, with separate day and night data, was calculated. Note that the resistance model of Matsuda (2008) assumes no surface emission of target substances. The dry deposition flux expressed as the flux at the standard temperature and atmospheric pressure was corrected based on the mean air temperature and atmospheric pressure during the measurement.

A wet-only sampler with refrigerated storage (US-330, Ogasawara Keiki) was installed at the study site. Precipitation (mostly rain, but also several snowfalls during the winter) was collected on a weekly basis. The concentrations of inorganic N (ammonium, NH₄-N; nitrate, NO₃-N; and nitrite, NO₂-N) and total N dissolved in the precipitation were determined using an ion chromatograph (ICS-1600, Thermo Scientific Dionex) and a total organic carbon and total N analyzer (TOC-V/TNM-1, Shimadzu), respectively. The difference between the total and inorganic N was defined as organic N (Org-N). The weekly flux in the form of wet N deposition was obtained by multiplying the precipitation amount by the N concentration.

Monitoring of the irrigation water for NH₄-N, NO₃-N, NO₂-N, and total N contents was conducted in a paddy field adjacent to the atmospheric monitoring point from 27 April to 19 August 2011 in CS1 and from 26 April to 27 August 2012 in CS2. The field was 32.6 m×100 m (0.326 ha) with two water inlets at the short southern side and one water outlet at the short northern side. This field was flooded continuously from the beginning of submergence to the final drainage. The water flows were recorded using a measuring weir with a water gauge (SE-TR/WT500, TruTrack) at each inlet and outlet. The water quality was monitored weekly by sampling and analyzing water at the inlets and outlet.

The inorganic N concentrations were determined using the abovementioned ICS-1600 ion chromatograph and the total N was determined using the abovementioned TOC-V/TNM-1 total organic carbon and total N analyzer. The input with irrigation and the output with outflow for each N species were calculated by multiplying the water flow rate by the concentration of each N species, with linear interpolation applied to account for temporal variation in the concentration of each N species. The N inflow and outflow were further divided by the area of the field (0.326 ha) to obtain the N flux per unit area. The mean water quality for the target substances throughout the irrigation period was also calculated by dividing the total flow of the substance by the total water flow.

Fig. 1 shows the weekly mean air concentrations of Nr during the day and night, and Table 1 shows the corresponding seasonal mean concentrations. The following are results for the main target species, i.e., NH₃, HNO₃, pNH₄, and pNO₃. NH₃ had the highest concentration among the target Nr, except for higher NO and NO₂ concentrations in January 2012. The mean NH₃ concentrations (6 m above the surface) during the study period were 3.8 and 3.1 μg N m^-3, respectively, during the day and night. The NH₃ concentrations were higher at a height of 2 m than at a height of 6 m, particularly in the day during CS1 and CS2. HNO₃ showed very low concentrations, except during the day in the warm season (which mostly corresponds to CS1 and CS2); the mean HNO₃ concentrations (6 m above the surface) were 0.41 and 0.08 μg N m^-3, respectively, during the day and night (Fig. 1). Both pNH₄ and pNO₃ existed in the atmosphere, but at lower concentrations than those of NH₃ (Fig. 1). The mean concentrations (6 m above the surface) were 1.1 and 1.3 μg N m^-3 during the day and night, respectively, for pNH₄, and were 1.3 and 1.3 μg N m^-3, respectively, for pNO₃. pNO₂ was not detected.

Fig. 1. 
Weekly mean air concentrations of gaseous nitrogen (ammonia [NH₃], nitric acid [HNO₃], nitrous acid [HNO₂], nitrogen oxide [NO], and nitrogen dioxide [NO₂]) and particulate nitrogen (ammonium particles [pNH₄] and nitrate particles [pNO₃]) at two heights (6 and 2 m above the surface), with separate values for the day and night. FS and CS denote the fallow and cropping seasons, respectively; FS1, 32 weeks from 15 September 2010 to 27 April 2011; CS1, 20 weeks from 27 April to 14 September 2011; FS2, 32 weeks from 14 September 2011 to 25 April 2012; and CS2, 20 weeks from 25 April to 12 September 2012.

Fig. 4 shows the weekly exchange fluxes, with positive and negative values representing net deposition and net emission, respectively. Uncertainty in these calculated fluxes is discussed in Section 4.4. NH₃ showed a tendency towards net deposition in the night, particularly in FS1 and FS2, and net emission in CS1 and CS2, particularly during the day; the net emission in CS1 and CS2 was occasionally large. The maximum weekly NH₃ flux was net deposition of 0.45 kg N ha^-1 wk^-1 and net emission of 1.0 kg N ha^-1 wk^-1. HNO₃ generally showed net deposition during the day in CS1 and CS2; the maximum flux was 0.17 kg N ha^-1 wk^-1 of deposition. pNH₄ and pNO₃ tended to show net deposition. The net emission that occasionally occurred during the day might be attributable to some data artifacts (Section 4.4), because there were no direct emission sources for particles at the paddy surface. The maximum weekly net deposition of pNH₄ and pNO₃ was 0.16 and 0.22 kg N ha^-1 wk^-1, respectively.

Table 2 shows the seasonal and annual exchange fluxes of gases and particles. The cumulative NH₃ fluxes were net deposition in FS1 and FS2 (2.0 to 3.4 kg N ha^-1) and net emission in CS1 and CS2 (1.7 to 3.2 kg N ha^-1). The annual NH₃ flux was 1.2 kg N ha^-1 of net emission in year 1 and 1.7 kg N ha^-1 of net deposition in year 2. HNO₃ showed net deposition in every season, but the amount was negligibly small in FS1 and FS2. The annual HNO₃ deposition flux in years 1 and 2 was 0.55 and 0.30 kg N ha^-1, respectively. pNH₄ showed net deposition in FS1 and FS2 (1.0 to 1.1 kg N ha^-1) and net deposition in CS1 (0.35 kg N ha^-1) or relatively small net emission in CS2 (0.16 kg N ha^-1). The annual pNH₄ flux showed net deposition (0.87 to 1.4 kg N ha^-1). pNO₃ showed net deposition in every season, and the annual pNO₃ deposition flux in years 1 and 2 was 2.2 and 0.74 kg N ha^-1, respectively. The seasonal exchange fluxes as Nr, excluding NO and NO₂, showed net deposition in FS1 and FS2 (4.4 to 4.8 kg N ha^-1) but net emission in CS1 and CS2 (1.2 kg N ha^-1 in both cases) due to the high NH₃ emissions; thus, the annual Nr exchange fluxes were net deposition (3.2 to 3.6 kg N ha^-1).

Table 5 summarizes the seasonal and annual N fluxes as wet deposition, gas and particle exchange, irrigation inputs, and outflows. In FS1 and FS2, the wet deposition (4.0 to 4.1 kg N ha^-1) and the exchange as net dry deposition (4.4 to 4.8 kg N ha^-1) were similar, totaling 8.4 to 8.9 kg N ha^-1 of N deposition. In CS1 and CS2, irrigation inputs (11.1 to 14.5 kg N ha^-1) were the largest component, followed by wet deposition (4.0 to 5.8 kg N ha^-1), and the gas and particle exchange showed net emission (both 1.2 kg N ha^-1); the sum of these inputs was 19.1 and 13.9 kg N ha^-1 in CS1 and CS2 respectively. The annual N input in years 1 and 2 was 27.5 and 22.9 kg N ha^-1, respectively. The contribution of each component to the annual N input was in the following sequence: irrigation>wet deposition> gas and particle exchange.

In the present study, NH₃ emissions were occasionally observed during the day in mid-summer (Fig. 4), i.e., the cropping seasons with well-grown rice plants. Here, we discuss possible causes of the NH₃ emissions. One possible cause is emissions from the surface of floodwater in rice paddies, and the other is emissions from rice plants.

Fig. 4. 
Exchange fluxes of gaseous nitrogen (ammonia [NH₃], nitric acid [HNO₃], nitrous acid [HNO₂], nitrogen oxide [NO], and nitrogen dioxide [NO₂]) and of particulate nitrogen (ammonium particles [pNH₄] and nitrate particles [pNO₃]) as weekly values, with separate values for the day and night. The positive and negative values represent dry deposition and emission, respectively. FS and CS denote the fallow and cropping seasons, respectively; FS1, 32 weeks from 15 September 2010 to 27 April 2011; CS1, 20 weeks from 27 April to 14 September 2011; FS2, 32 weeks from 14 September 2011 to 25 April 2012; and CS2, 20 weeks from 25 April to 12 September 2012.

Paddy rice prefers to use ammonium as an N source (Ishiyama et al., 2004). Fertilizers applied to rice paddies, therefore, consist mainly of ammoniacal N, including urea, whose hydrolysis releases NH₃. Surface broadcast of N fertilizers leads to a large NH₃ volatilization loss from rice paddies, whereas the incorporation of N fertilizers into the soil greatly reduces NH₃ volatilization (e.g., Hayashi et al., 2008, 2006). According to Hayashi et al. (2008), surface broadcast of urea with a rate of 30 kg N ha^-1 showed 20.9% of volatilization loss as NH₃, whereas the incorporation of urea with a rate of 50 kg N ha^-1 into the plowed layer resulted in 2.1% of volatilization loss. The incorporation of controlled-release fertilizers in the study area would inhibit NH₃ volatilization more effectively by reducing the excess NH₄-N in the soil solution. In fact, no remarkable NH₃ emissions were detected in late April after the fertilization at that time (Fig. 4).

Stomata of plant leaves are a channel of atmosphereplant exchange of gases, e.g., carbon dioxide, water vapor, and so on including NH₃. Stomatal NH₃ emissions to the atmosphere occur when the stomatal cavity’s NH₃ concentration becomes higher than the NH₃ concentration in the ambient air. In the opposite case, stomatal NH₃ absorption from the atmosphere takes place, which represents a form of dry deposition. The threshold concentration between emission and deposition can be defined as the NH₃ compensation point (Farquhar et al., 1980). It is known that graminoid upland crops such as wheat and barley emit NH₃ to the atmosphere at a rate that depends on their N nutrition status and plant phenology (e.g., Schjoerring and Mattsson, 2001; Husted et al., 1996).

In the present study, no supplemental fertilization was conducted in mid-summer when large NH₃ emissions were occasionally observed (Fig. 4). In this case, the surface of rice paddies was hard to be a strong NH₃ emitter. Therefore, it is expected that rice plant emission is the most possible pathway of NH₃ loss into the atmosphere. Some field data (Hayashi et al., 2011b, 2008) supports the possibility of NH₃ emission from rice plants. Meanwhile, a laboratory experiment (Miyazawa et al., 2014) showed that the NH₃ compensation point of rice (cv. Koshihikari), which ranged from 0.4 to 0.6 nmol mol^-1, was one order of magnitude lower than those of upland crops; therefore, rice leaves were not an active NH₃ emitter. This range corresponds to 0.23 to 0.35 μg N m^-3 of NH₃, which was less than one-tenth of the mean NH₃ concentrations in the day during the cropping season at times when large NH₃ emissions occurred (Fig. 1); thus, NH₃ emission from rice leaves does not appear to be important in the study area. Further research will be needed to unravel the cause(s) of the NH₃ emissions during the day in the cropping seasons.

Among the input processes, irrigation accounted for the largest annual N input, followed by wet deposition (Table 5). However, the gas and particle exchange, which represents the difference between the gross dry deposition and emissions, was also significant. The tendency towards NH₃ emission, particularly in CS1 and CS2 (Fig. 4, Table 2), masks the NH₃ input by dry deposition. As shown in FS1 and FS2, the NH₃ exchange fluxes were similar quantity of wet NH₄-N deposition (Tables 2 and 3). For reference, the annual dry NH₃ deposition estimated by the inferential method was 3.6 and 3.4 kg N ha^-1 in years 1 and 2, respectively. These values were larger than the annual NH₃ exchange fluxes (Table 2) by 4.8 and 1.7 kg N ha^-1 in years 1 and 2, respectively. Therefore, dry deposition is an important process that quantity is comparable to wet deposition. Atmospheric deposition, as the sum of wet and dry deposition, can therefore be an N source larger than irrigation water, depending on the local water and air quality.

Large exchange fluxes as net deposition, comparable to NH₃, were observed for NO and NO₂ in FS2, particularly at night (Fig. 4), though these gases were excluded from our evaluation of the annual N input due to the limited duration of their monitoring. The air concentrations of NO and NO₂ were also higher in FS2 than in CS1 (Fig. 1), which implies frequent inflows of these gases into the paddy area. The expressway running from east to west at least 400 m north of the atmospheric monitoring point is a possible major source of the NO and NO₂. Fine weather with relatively strong northwesterly winds prevails in this region in winter. The frequency of west-northwest and northwest winds at the atmospheric monitoring point was 28% in FS1 and 30% in FS2, versus 5% in CS1 and 6% in CS2. Consequently, the large dry deposition of NO and NO₂ might be caused by the following mechanism: the near-surface atmospheric layer becomes stable on clear nights in the fallow season due to radiative cooling, which restricts near-surface convection so that plumes containing vehicle exhaust from the expressway can be transported downwind directly to the atmospheric monitoring point, leading to high rates of dry deposition. We should therefore pay attention to such local Nr sources and the balance between convection and advection conditions when we calculate the N budget for a site.

First, the magnitude of atmospheric N deposition in the present study is compared with other data. Ban et al. (2016) evaluated atmospheric N deposition at 8 remote sites in Japan from 2013 to 2012. Target N species in their study were NH₄-N and NO₃-N for wet deposition, and NH₃, HNO₃, pNH₄ and pNO₃ for dry deposition. The total N deposition at the 6 of 8 remote sites ranged from 9.7 to 12.9 kg N ha^-1 yr^-1, except Rishiri, an island in the northernmost Hokkaido (7.0 kg N ha^-1 yr^-1) and Ogasawara, remote ocean islands (3.0 kg N ha^-1 yr^-1). This range is similar to our data, 11.7-13.0 kg N ha^-1 yr^-1 as the sum of wet deposition and gas particle exchange (Table 5). It is noted that our data denote the exchange fluxes as the difference between dry deposition and emission; where, the gross dry deposition fluxes were unknown. Air concentration measurements with fine time resolution for major Nr species is needed to evaluate gross rates of both dry deposition and emission.

A question in this study is how the N input to rice paddies via atmospheric deposition and irrigation compares with the N fertilization rate and the N demand of the rice. We used the exchange fluxes to evaluate the contribution of dry deposition. Accordingly, the N input is expressed as the sum of wet deposition, gas and particle exchange, and irrigation fluxes. The atmosphere-rice paddy exchange of NO and NO₂ was excluded from these fluxes due to the limited data duration. The 2-year monitoring showed a total N input of 22.8 to 27.5 kg N ha^-1 yr^-1 (Table 5), which corresponds to 38 to 55% of the N fertilizer application rates (50 to 60 kg N ha^-1). This N input also corresponded to 53 to 67% of the N uptake in brown rice for cv. Koshihikari (a staple cultivar in this region), which ranged from 41 to 43 kg N ha^-1 these values were calculated from the mean yield of farmers in the study area (4800 kg ha^-1, at 15% moisture content) and the brown rice N contents in the same area (1.01 to 1.06% w/w) (Zhang et al., 2013). Thus, N deposition and irrigation contributed greatly to the N inputs.

Uncertainty of the determined fluxes in the gas and particle exchange results from errors in measurements and in the flux calculations. The measurement errors can be divided into two categories: air concentrations and diffusion velocities. The flux calculation errors can also be divided into two categories: the averaging time used to determine air concentration differences and the methodological assumptions.

In terms of the precision of the filter packs and their ability to determine air concentrations, the coefficients of variation (CV) of NH₃ and pNH₄ at an upland field in a previous study were 7.1 and 15.1%, respectively (Hayashi et al., 2009a); there are comparable to the CVs of HNO₃, pNH₄, and pNO₃ at four CASTNet sites, with values of 6.2, 3.0, and 11.4%, respectively (Sickles et al., 1999). The filter pack method has a possible artifact that results from the evaporation of particles trapped on the filter in the first stage of a filter pack (Keck and Wittmaack, 2005; Pathak and Chan, 2005; Anlauf et al., 1985). This can result from heating of filter packs by sunlight and reactions of the trapped particles with acidic gases (e.g., NaNO₃+HCl→NaCl+HNO₃). In the present study, wrapping the filter pack with aluminum foil to prevent the direct effect of sunlight should have reduced the magnitude of this artifact. Pathak and Chan (2005) compared a filter pack with an annular denuder, and found that although the latter method is relatively complicated, it eliminates this artifact; the loss of acidity in the first stage of the filter that trapped particles under ammonium-rich conditions (with an ammonium/sulfate ratio greater than 1.5) and re-evaporation of pNO₃ and particulate chloride also occurred in the filter pack method.

For errors in the diffusion velocity determined by micrometeorological measurements, Miyata (2001) found that the maximum error of the diffusion velocity was 20% in the day and 100% at night. Hayashi et al. (2013b) evaluated that the maximum errors of weekly mean fluxes determined by the gradient method were 131% for NH₃ and 124% for HNO₃ due to the combined effects of the CV of the filter packs and the maximum error of the diffusion velocity.

The problem of the averaging time originates from a correlation between the air concentration differences and the diffusion velocities. The diffusion velocity usually exhibits a daily pattern, with large values during the day and small values at night (Hayashi et al., 2013a). When the air concentration difference has a daily pattern, the flux (which equals the mean concentration difference multiplied by the mean diffusion velocity) induces a systematic error; that is, the method will underestimate the flux of a substance with larger concentration differences during the day than at night. Our separation of the day and night values in the air concentration measurements should reduce this systematic error in the weekly monitoring; our previous research suggests that this reduced the overestimation of NH₃ fluxes for the gradient method by 121%±128% at the study site during the cropping season (Hayashi et al., 2013b).

In terms of assumptions, the gradient method assumes that there are no changes in the chemical form of a target substance between the two measurement heights, but gas-to-particle conversions (e.g., NH₃+H₂SO₄→NH₄HSO₄) and gas-particle interconversions (e.g., NH₃+HNO₃↔NH₄NO₃) occur in the air (Nemitz et al., 2004). These reactions induce errors in the calculated flux via their effects on the air concentration differences between the two measurement heights. The remarkable emission fluxes of pNH₄ and pNO₃ that we occasionally observed in the daytime (Fig. 4) might be partly attributable to these effects, but may also be an artifact of the evaporation of particles trapped on the filters, because direct surface emissions of particles should rarely occur. Even so, the gradient method offers an advantage, namely that the effect of the changes in the forms of the species on the flux does not affect the flux as the sum of relevant gas and particle fluxes (e.g., ammoniacal N, NH₃+pNH₄), because the diffusion velocity, a function of aerodynamics, is the same regardless of the chemical form.

Enhancing the temporal resolution of the air concentration data can effectively reduce the uncertainty of the determined fluxes. Obtaining flux data with fine temporal resolution can also approximate the gross flux of dry deposition or emission. Continuous monitoring of multiple Nr species, including gaseous and particulate forms, would be expensive, and accurate in-situ measurements would be challenging at rice paddies, particularly in hot and humid summers. Such measurements are, however, worth trying, given the importance of atmosphere-rice paddy Nr exchanges in ecosystem N budgets.

Fluxes derived from the gradient and inferential methods can be compared. Ideally, the fluxes derived from the two methods should be the same for a substance with no surface emissions, such as sulfur dioxide (SO₂). The ratios of SO₂ fluxes obtained using the gradient method to those obtained using the inferential method at our study site were 1.05 (day) and 1.22 (night) during the cropping season and 1.79 (day) and 2.84 (night) during the fallow season (Hayashi et al., 2013b). This result suggests that the gradient method overestimates or the inferential method underestimates SO₂ fluxes (or a combination of both), particularly during the fallow season. Estimated deposition velocities in the inferential method differ greatly among parameterizations (Flechard et al., 2011); thus, the parameterization used in the present study might have underestimated the deposition velocities. This result does not allow the emission flux to be approximated by subtracting the exchange flux estimated using the gradient method from the dry deposition flux estimated using the inferential method. A study that compares a gradient method with an inferential method would be effective and important, since the accuracy of the gradient method could be enhanced by measurements with high temporal resolution, and the parameterization of the deposition velocity of Nr species in the inferential method could be improved by data with high temporal resolution from the gradient method.