Impact of extreme precipitation and water table change on N 2 O fluxes in a bio-energy poplar plantation

Introduction Conclusions References


Introduction
Nitrous oxide (N 2 O) is one of the major greenhouse gases, with a global warming potential ∼300 times higher than CO 2 , and that plays also a role in the destruction of stratospheric ozone (Cicerone, 1989). Approximately two thirds of the atmo-25 spheric N 2 O originates from the biogenic processes of nitrification and denitrification 2058 The mechanisms responsible for N 2 O emission and the environmental controls over its production are complex. For instance, increases in soil water content initially have a positive effect, but further increases have a negative effect on N 2 O emissions; how soil water controls N 2 O fluxes is not yet fully understood (Davidson et al., 2000;Jungkunst et al., 2008;Castellano et al., 2010). Dry, well-aerated soils favor the oxidative process Introduction al., 2010) were not able to unequivocally explain the processes responsible for N 2 O release. Moreover, the processes leading to N 2 O consumption within the soil are still largely unknown (Chapuis-Lardy et al., 2007). Nitrous oxide uptake has been observed in different ecosystems, such as grasslands (Glatzel and Stahr, 2001;Neftel et al., 2007) and forests (Cavigelli and Robertson, 2001;Butterbach-Bahl et al., 2002), and it 5 has been connected to anaerobic microbial denitrification (Zumft, 1997). N 2 O release mostly occurs in short peak emissions connected to fertilization and precipitation events (Wagner-Riddle et al., 2007;Eugster et al., 2007;Jungkunst et al., 2008;Neftel et al., 2010). The short-term nature of the N 2 O release and the difficulties in modeling N 2 O emission generate the need for continuous monitoring to 10 estimate annual N 2 O emission from ecosystems. Unfortunately most studies have been of discontinuous nature (from weekly to monthly) (Kavdir et al., 2007;Neftel et al., 2007;2010;Mammarella et al., 2010) and/or involved the use of soil chambers (Hellebrand et al., 2003;Kroon et al., 2010b;Wu et al., 2010), with associated spatial scaling issues, and with resulting uncertainties in annual estimates of more than 50% 15 (Flechard et al., 2007). Small spatial and discontinuous temporal resolution of chamber measurements prevent the accurate capture of some of these peak events-based N 2 O release. Thus far, few studies have been performed at ecosystem scale with eddy covariance (Neftel et al., 2007;2010;Eugster et al., 2007;Mammarella et al., 2010;Kroon et al., 2010a) or gradient techniques (Wager-Riddle et al., 2007). Neftel et 20 al. (2007) reported that N 2 O emission measured with eddy covariance exceeded that obtained by the chamber technique threefold.
Understanding the impact of climate and soil hydrology on N 2 O emissions is particularly important, because the frequency and magnitude of drought and precipitation events are expected to increase with climate change (Kunkel et al., 2008). Future 25 changes in rainfall patterns are predicted to increase N 2 O emission by 45% even with reduced fertilizer application (Hsieh et al., 2005). As soil water content is believed to be among the most important controls on N 2 O emission, altered precipitation patterns could significantly affect the emission of this greenhouse gas (Davidson et al., 1991; BGD 8,2011 Impact of extreme precipitation and water   McClain et al., 2003;Castellano et al., 2010).
The main objective of this study was to investigate the impact of soil hydrological changes (e.g. WFPS and water table change) on N 2 O emission in a high-density bioenergy poplar plantation, recently converted from cropland and pasture. We hypothesized that increases in water table and WFPS connected to rain events lead to in-5 creases in N 2 O emissions. We also hypothesized that increases in soil temperature stimulate N 2 O production and thus increase N 2 O emissions if adequate water is available in the soil. 10 The research site is located in Lochristi, Belgium (51 • 06 44 N, 3 • 51 02 E), 11 km from the city of Ghent at an altitude of 6.25 m above sea level (Fig. 1). The longterm average annual temperature is 9.5 • C and the average total annual precipitation is 726 mm (Royal Meteorological Institute of Belgium). The soil has a sandy texture with a clay-enriched deeper soil layer. The soil C:N ratio (measured in February-March 2010) 15 in the first 90 cm of the soils was on average 13.3 ± 1.4 (n = 110) and the bulk density was ∼1.482 ± 0.075 g cm −3 . The soil pH was on average 5.51 ± 0.66 (n = 42 water. These ditches were draining water into deeper canals (1.5 m depth) at the outer edges of the field site. As a consequence the soil surface was mostly dry and drainage of standing water was fairly rapid. The previous land uses were pasture and cropland (ryegrass, wheat, potatoes, beets, and most recently monoculture corn with regular fertilization, 200-5 300 kg N ha −1 y −1 liquid animal manure and chemical fertilizers sured on 29 October 2010) was on average 2.37 ± 0.005 mg N l −1 (the sum of NO − 3 -N and NO − 2 -N) and 0.31 ± 0.0416 mg N l −1 (NH + 4 -N).

Environmental variables
A complete set of meteorological variables were recorded continuously from the beginning of June 2010 to the present day. Soil water content was measured at different 20 depths (0-30 cm, 0-20 cm, 0-10 cm in different locations), and across a vertical transect (at 1 m, 60 cm, 40 cm, 30 cm, and 20 cm) in the proximity of the eddy covariance mast using 8 Time Domain Reflectometry (TDR, model CS616 Campbell Scientific, Logan, UT, USA) moisture probes. Soil water content was then converted to water-filled pore space (WFPS) according to Wu et al. (2010). Soil temperature was recorded 25 by temperature probes which provided the average temperature of a soil layer of 8 cm BGD 8,2011 Impact of extreme precipitation and water and model CR1000, Campbell Scientific, Logan, UT, USA) and each environmental variable was read once every 0.1-10 s and the 30 min averages are output to a PC.

Eddy covariance measurements
An eddy covariance mast was installed at the beginning of June 2010 and it was continuously been operated to the present day. The eddy covariance mast was positioned in 5 the northeast part of the plantation (Fig. 1) including areas with both previous land use types (cropland and pasture). The eddy covariance mast included a sonic anemometer for the measurement of the three-dimensional wind components, wind speed, wind direction, and the energy fluxes (Model CSAT3, Campbell Scientific, Logan, UT, USA), and several fast-response analyzers, among them a closed-path Los Gatos N 2 O/CO analyzer (model 908-0014, Los Gatos Research, Mountain View, LGR, CA, USA) and a closed path CO 2 /H 2 O infrared analyzer (LI-7000, LI-COR, Lincoln, NE, USA). The sonic anemometer and the inlet of the sampling lines were positioned at 5.8 m above the surface. The mast location was chosen according to the prevalent wind direction (from southeast, Fig. 1), to maximize the footprint of the tower. The sonic anemome- 15 ter was oriented to 175 • from true north. The large majority of wind directions were between 198 • and 252 • from north ( Fig. 1). The Los Gatos N 2 O analyzer employs a cavity enhanced laser absorption technique in which an optical cavity is used as the measurement cell. This allows for a longer optical pathlength (400 ± 10 m) compared to conventional laser absorption techniques, 20 resulting in increased sensitivity. The analyzer utilizes a room temperature mid-infrared quantum cascade laser and detector at a specific narrow band (4.6 µm). The internal pressure of the optical cell is fixed at 10 kPa. The analyzer has a 1s 1σ precision of 0.3 ppbv for both N 2 O and CO. A scroll pump (model XDS-35i, Edwards, MA, USA) was used to draw air through the N 2 O analyzer. A two-meter long vacuum tubing was used 25 to dampen the air flow and pressure in the air stream. The flow rate in the sampling line of the N 2 O analyzer was ∼25 l min −1 . 8,2011 Impact of extreme precipitation and water The N 2 O analyzer was calibrated at the LGR Company on 6 July 2010 using a NOAA primary standard at 322.24 ppbv N 2 O in air (uncertainty less than 0.1 ppbv). The linearity of the analyzer was then tested by diluting a higher concentration bottle (440 ppbv N 2 O) by known amounts and measuring the analyzer response. This dilution test proved that the accuracy of the instrument was better than 1% over the range 5 of 40-440 ppbv N 2 O (R. Provencal, Los Gatos Research, personal communication, 2010). We calibrated the N 2 O analyzer again on 31 August 2010 with 733 ppbv (ultra high purity ≥99.997 vol% with 10% accuracy, limited by dilution system).

BGD
The LI-7000 (LI-COR, Lincoln, NE, USA) was used to measure CO 2 and H 2 O fluxes. A vacuum pump was positioned at the outlet of the LI-7000 analyzer, generating a flow 10 of ∼22 l min −1 . Two buffer volumes of 0.5 l each respectively were positioned between the pump and the outlet of the analyzer to dump the fluctuations of the pump. Here we describe only calibration procedure for the H 2 O fluxes as they were used to correct the N 2 O fluxes (see following sections). CO 2 fluxes are presented and discussed elsewhere. The H 2 O vapor was calibrated every week using ultra-high purity nitrogen for 15 the zero, and a dew point generator (LI-610, LI-COR, Lincoln, NE, USA) to produce an air stream with a known water vapor dew point (typically 7 • C lower than the ambient air temperature) for the H 2 O span.
Fluxes of H 2 O, N 2 O, and momentum were measured using eddy covariance, a micrometeorological method that quantifies the net exchange of a scalar between the 20 biosphere and the atmosphere (Swinbank, 1951;Desjardins and Lemon, 1974;Baldocchi, 2003).
Teflon tubing (∼15 m long and 8 mm inner diameter) was used for two separate sampling lines for the LI-7000 and for the N 2 O analyzer. The two inlets were positioned 10 cm from the center of the sonic anemometer. A 1 µm teflon filter (Gelman) was used 25 at the inlet of the sampling line of the LI-7000 analyzer. A stainless steel Swagelok™ filter (60 µm pore size SS-4FW4-60) was positioned at the inlet to protect the sampling line of the N 2 O analyzer. Another stainless steel Swagelok™ filter (2 µm pore size, SS-4FW4-2) was also present at the input of the sampling line to prevent dust from BGD 8,2011 Impact of extreme precipitation and water entering the sample cell. The H 2 O, N 2 O fluxes, and sonic wind components were recorded at 10 Hz using a data logger (model CR 5000, Campbell Scientific, Logan, Utah, USA). All the analyzers, the data loggers, and the PC were positioned inside a wooden cabin maintained at a stable temperature (21 • C).
A two-components rotation was applied to set mean vertical (w) and lateral (v) veloc-10 ity components to zero. Time delays (on average 1.6 s for N 2 O and 1.8 s for H 2 O) were calculated using a cross-correlation function of the scalar fluctuation and the vertical wind velocity. A frequency response correction was applied to the eddy covariance fluxes following Moore (1986) and using theoretical attenuation functions and Kaimal model spectra to account for high frequency and low frequency fluctuations in signal 15 losses (Kaimal et al., 1972). We also applied a correction for density change (WPL) according to Webb et al. (1980). We only applied the water vapor term of the WPL correction as we assume the long tube attenuated the temperature fluctuation. sonic anemometer (reported by the diagnostics of the CSAT-3D), when failing the stationarity test with a threshold of 30% as suggested by Foken and Wichura (1996). A footprint model was applied to the data (Klijun et al., 2004) indicating that the 90% of the fluxes were coming from the first ∼200 m upwind of the eddy mast. Data with wind direction between 285 • and 135 • (from the north, back, and from the right of the tower) 5 were removed. Only N 2 O fluxes are being presented in this manuscript, but CO 2 fluxes were used to derive u* (defined as √ u w ) threshold, then applied to the N 2 O fluxes. CO 2 fluxes for a solar radiation < 10 Wm −2 were regressed with u* and a u* threshold was set to 0.15 m s −1 (data not shown (1) 15 Where BR is the basal respiration and Q 10 describes the response of respiration to temperature (soil T ) increase.

Statistical analyses
General linear modeling (GLM) was used to identify the most important predictors of N 2 O fluxes (Systat version 13, Systat Software Inc., 2002, Chicago, IL, USA)). A single 20 variable and a forward stepwise multiple regression approach were used to discriminate among and rank the most important variables (surface temperature, soil temperature at 0-8, 20,30,40,and 60 cm,20,30,40,and 60 cm,water  Over the entire month of August 2010 the total precipitation at our site was 185 mm (187 mm at Ukkel, classified as an "exceptional event" by the Royal Meteorological Institute of Belgium, a denomination used for events that occur once every 30 yr). The 15 record high total monthly precipitation at Ukkel was measured in 1996 (231 mm).
This extreme precipitation event led to a steep increase in water table and WFPS (Fig. 2). Prior to the precipitation event the water table was at ∼136 cm below the surface and it was below 80 cm for the entire summer season (Fig. 2). The heavy rain on 16-17 August caused flooding of the field site (in several locations there was standing 20 water) and overflowing of the ditches. The weekly total precipitation from 20 June to 16 August 2010 was on average 13 ± 11 mm while from 16 August to 3 October it was on average 37 ± 27 mm, not allowing the shallower soil layers (0-10 cm) to become drier than ∼60% WFPS after 17 August (Fig. 2). 8,2011 Impact of extreme precipitation and water

N 2 O fluxes
The spectral analysis showed that the co-spectra of w T s , and of w N 2 O presented a reasonable comparison, demonstrating the good performance of the instruments (Fig. 3). The co-spectra of w N 2 O showed a slight loss at the high frequencies (typical for closed path analyzers) (Fig. 3). The energy budget closure for the presented data 5 averaged 85%. During the days immediately following the large rainfall event and the steep increase in water table and WFPS on 16-17 August, a steep increase in N 2 O emission from the plantation was observed (Fig. 4). This large N 2 O emission started on 19 August when the water table and WFPS progressively decreased ( Fig. 4 and Fig. 5). From 10 19 to 22 August the N 2 O emission presented a pronounced diurnal trend following the daytime increase in soil temperature (Fig. 5), and wind speed (or u*) (Fig. 6). From 23 to 25 August when the wind speed was generally > 2 m s −1 (and the u* was mostly > 0.3 m s −1 ) N 2 O emissions did not present a diurnal pattern any more (Fig. 6).  (Table 2). During this period surface temperature and the shallow soil temperature (0-8 cm) explained 25 56% and 54% of the N 2 O fluxes, respectively (Table 2). At this time wind speed and u* were also important but presented lower explanatory power (29% and 33% respectively) than temperature ( 19-22 August) and soil temperature (0-8 cm) was exponential and exhibited a Q 10 of 3 (Fig. 7). The multi-variable model that presented the highest explanatory power of the N 2 O fluxes from 19-22 August included surface temperature, soil T (60 cm depth), and wind speed, and it was able to explain 68% of the variability in N 2 O emissions (F-ratio 67, p < 0.001).

5
N 2 O emissions between 23 and 25 August did not present a diurnal trend (Fig. 5). During 23-25 August, soil and surface temperature were no longer significant predictor of N 2 O fluxes (Table 3). During these days, wind speed and u* were the variables with the highest explanatory power of N 2 O fluxes and they explained 38% and 42% of the variability in N 2 O fluxes, respectively, (Table 3). A multi-variable model that included 10 u*, WFPS (20 cm), and WFPS (40 cm) was able to explain 79% of the variability in N 2 O fluxes (F-ratio 148, p < 0.001). Interestingly, we noticed that the WFPS at intermediate depth in the soil profile (20 cm) was sometimes lower than in the shallower and deeper layers. The water content in the shallower layers increased due to the mist and light rainfall (a clear example is shown on 7 July, Fig. 2 when WFPS at 0-10 cm increased 15 right after a small rainfall, even while WFPS at 20 cm increased later only after a larger rainfall event).
The low turbulence at night (u*< 0.1 m s −1 ) and moderately turbulent conditions during daytime (u* ∼0.5 m s −1 ), during the first four days (19-22 August), led to N 2 O concentration increases at night, ranging from ∼325 ppb to ∼340 ppb over a few hours pe-20 riod (Fig. 6). The last three days (23-25 August) exhibited higher turbulent conditions, with u* spanning from ∼0.3 m s −1 at night to ∼0.8 m s −1 during daytime, and presented a lower variability of N 2 O concentration with no marked diurnal cycle (Fig. 6). The daily total N 2 O-N emission from 19-22 August was fairly stable, on average 0.26 ± 0.01 (SD) kg N 2 O-N ha −1 , while the total daily emission from 23-25 August was 25 also fairly stable (e.g. on average 0.13 ± 0.014 kg N 2 O-N ha −1 ). We also estimated the approximate N present in the soil water using the average sum of NO   Rain events that occurred after 25 August 2010 led to similar increases and decreases in water table (and WFPS), but did not lead to N 2 O emissions of the same 5 magnitude of the one observed on 19-25 August (Fig. 4). Overall, N 2 O fluxes before and after the peak emissions of 19-25 August, were mostly close to zero.

Discussion
The emission of N 2 O differed dramatically between the week following the first extreme rain event and the rest of the study period. The low N 2 O emission observed before the 10 large rainfall on 16-17 August, could be related to the fact that under normal conditions well aerated sandy-loam soils are unlikely to develop the large number of anaerobic micro-sites necessary for N 2 O production by denitrification (Skiba et al., 1993).
In contrast, a first extreme rain event induced production and release of substantial amount of N 2 O. The maximum N 2 O emission observed after the large rain fall in this 15 study was several orders of magnitude higher than what is usually observed (Pilegaard et al., 2006;Davidson et al., 2000;Schaufler et al., 2010) Boeckx and Van Cleemput, 2001).
Peak N 2 O emission with re-wetting of dry soils has been observed in several ecosystems (Sexstone et al., 1985;Wagner-Riddle et al., 1996;Hsieh et al., 2005;Wagen-Riddle et al., 2007). The large release in N 2 O emissions observed on 19-25 August 5 may have been connected to multiple mechanisms. The flooding of the land could have transported NH + 4 and NO − 3 from the ditches or from surrounding agricultural fields to the plantation at a rate that exceeded the uptake of plants and microorganisms, leading to significant rates of denitrification and N 2 O emission. It is likely that the extreme rain event probably also caused the reactivation of water-stressed bacteria following the dry 10 period, which decomposed and mineralized the labile organic matter fraction, suddenly available in the soil (Birch, 1964). Additionally, the prolonged drier conditions before 16 August could have led to death of the microbial population in the shallower soil layers and the release of nitrogen in the soil, emitted as N 2 O once the intense rain event suddenly increased moisture availability. 15 The observed lag between the rain event (16-17 August) and N 2 O emission (19-25 August) was probably related to the rate of water infiltration through the soil profile (Fig. 5). The sustained high N 2 O emission that we observed for a week was accompanied by the drop in the water table from the surface until about 60 cm below the surface from 16 to 23 August 2010 (Fig. 4). We observed the highest N 2 O emission when 20 the soil profile became less anoxic (e.g. WFPS 0-10 cm between 60-72%, Fig. 5) preventing the complete reduction of N 2 O into N 2 (Davidson, 1991), but leaving sufficient anaerobic micro-sites available for denitrification (Rolston et al., 1982;Sexstone et al., 1985). Deeper soil layers presented a stable and higher WFPS (80%, Fig. 5) where probably N 2 production was dominant instead (Davidson, 1991;Davidson et al., 2000).

25
While N 2 O fluxes from 19-22 August presented a pronounced diurnal trend with increased emission during daytime, the following days (23-25 August) presented stable emissions and no diurnal trend. This could be related to the interaction of different processes responsible for the N 2 O emission. The observed diurnal pattern in N 2 O BGD 8,2011 Impact of extreme precipitation and water  (Chang et al., 1998;McBain et al., 2004). The increase in transpiration with increasing wind speed, with abundant water in the soil and high stomatal conductance (Campbell and Norman, 1998), could explain the dependence of N 2 O fluxes from wind speed at this time.

10
As N 2 O emission from poplar leaves has been observed only under extremely high soil N 2 O concentration (McBain et al., 2004), it was probably connected to the observed nighttime decrease in wind speed (and u*) and increase in N 2 O concentration (Fig. 6). This decrease in turbulence probably led to high N 2 O concentration in the soil during 19-22 August. On the other hand, the high diffusion rates (Chang et al., 1998) and 15 pressure pumping between 23-25 August prevented an N 2 O concentration increase (Fig. 6), and probably an increase in concentration in the soil profile, thus reducing the importance of N 2 O emission through leaves.
From 23-25 August, soil temperature lost its importance in explaining N 2 O fluxes and the main environmental variables controlling N 2 O release were wind speed (or 20 u*) in combination with moisture content in deeper soil layers (WFPS at 20 and 40 cm depth). At this time the main mechanism of N 2 O emission was probably mass flow through the soil layers, not transpiration through the poplar leaves anymore. The wind pumping effect (Gu et al., 2005) probably pushed N 2 O from deeper soil layers (where temperature was more stable, not presenting a diurnal trend) into the atmosphere, thus 25 reducing the residence and the travelling time of N 2 O in and from deeper soil profiles, and preventing its reduction to N 2 . The occurrence of a more aerobic layer between 20 and 40 cm depth into the soil may have been either the site of production or storage of N 2 O, that was released once the wind speed (or u*) increased (Fig. 5). 8,2011 Impact of extreme precipitation and water The Q 10 values reported for N 2 O emission in laboratory incubations of sandy-loamy soils span a very wide range, 1. 9-8.9 (Maag and Vinther, 1996), 1.4-5.2 (Vicca et al., 2009), 12.4 (Vinther, 1992, and up to 23 (Christensen, 1983). Possible mechanisms behind these very high Q 10 values include the increase in size of the existing anaerobic micro-sites and the generation of new ones (Dowdell and Smith, 1974), connected 5 to the increased respiratory oxygen consumption with increasing temperature (Tiedje et al., 1984). On the other hand, high Q 10 could be related to the confounding effect of changes in microbial population size and/or substrate availability (Davidson et al., 2006). As the R 2 of the Q 10 function in this experiment from 19-22 August was only 51% and could not explain the release between 23-25 August (see Fig. 7) we believe 10 that several complex mechanisms were responsible for the N 2 O emission. The stomatal transport would provide a possible mechanism for this release in the first peak emission days while increased mass flow with higher wind speed probably explained the high emission on 23-25 August.

BGD
Successive rain events after 25 August and associated fluctuation of water table and 15 WFPS which were of same magnitude as those on 16-25 August, did not lead to peak N 2 O emissions (Fig. 4). N 2 O fluxes after the 25 August were very low, and did not respond to temperature increase, water table, or WFPS fluctuations. The lack of large N 2 O emission events after the ones of 19-25 August would indicate that the large soil nitrogen pool was probably completely used (either emitted as N 2 O, immobilized by the recovering microbial population, or taken up by the vegetation) or leached to some other location. Moreover, surrounding agricultural fields were no longer fertilized after mid-August. This observed lack of response after successive rainfall events was previously explained as nitrate or carbon limitation (Sexstone et al., 1985;Wagner-Riddle et al., 1996), as a very specific combination of water content and nutrient availability 25 is necessary to produce denitrification peak fluxes (Grundmann et al., 1988). Not only the observed extreme rainfall event is important but the rainfall pattern over extended period is extremely important in influencing N 2 O emissions. This result is confirmed by the higher N 2 O emission with the same total rainfall, but a longer dry period observed BGD 8,2011 Impact of extreme precipitation and water in previous studies (Rolston et al., 1982;Smith and Patrick, 1983). The intense precipitation event after a long dry period could also have been responsible for higher NO − 3 leaching than the one occurring under lower and more frequent precipitation events (Rolston et al., 1982).

5
Intense precipitation events after extended dry periods could have a large impact on N 2 O emission; weekly or monthly monitoring schemes of N 2 O fluxes could largely underestimate these emissions. In this study we showed that water Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | the eddy covariance data analysis and insight in the data processing, the Royal Meteorological Institute of Belgium for providing climate data, Frans Fierens and the ECMWF (www.ecmwf.int) for the boundary layer data, Ann Cools and Tom Van der Spiet for the water sample analysis, Toon De Groote for help with the meteorological data analysis, and Sara Vicca for help with the Q 10 analysis, John King for the revision of the manuscript. 8,2011 Impact of extreme precipitation and water  Chang, C., Janzen, H. H., Cho, C. M., and Nakonechny, E. M.: Nitrous oxide emission through plants, Soil Sci. Soc. Am. J., 62, 35-38, 1998. Chapuis-Lardy, L., Wrage, N., Metay, A., Chotte, J. L., and Bernoux, M.: Soils, a  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Kramer, K. J., Moll, H. C., and Nonhebel, S.: Total greenhouse gas emissions related to the Dutch crop production system, Agr. Ecosyst. Environ., 72, 9-16, 1999. Kristensen, H. L., Gundersen, P., Callesen, I., and Reinds, G. J.: Throughfall nitrogen deposition has different impacts on soil solution nitrate concentration in European coniferous and deciduous Forests, Ecosystems, 7, 180-192, 2004. 5 Kroeze, C., Mosier, A., and Bouwman, L.: Closing the global N 2 O budget: A retrospective analysis 1500-1994, Global Biogeochem. Cy., 13, 1-8, 1999. Kroon, P., Hensen, A., van