Trace gas exchange of N2O, NO, and CO2 between soil and the atmosphere was measured with high temporal resolution for 5 years (2004–2008) at the Höglwald Forest, Germany, using a fully automated measuring system. On the basis of these long-term continuous measurements, we calculated the annual budgets of soil-atmosphere trace gas exchange with high accuracy and demonstrated substantial seasonal and interannual variations. The mean annual soil-atmosphere exchange of N2O, NO, and CO2 at our site for the years 2004–2008 was 1.20 ± 0.09 kg N2O-N ha−1 yr−1, 8.64 ± 0.19 kg NO-N ha−1 yr−1, and 7.15 ± 0.08 t CO2-C ha−1 yr−1, respectively. Seasonal patterns of soil N2O fluxes were characterized by event emissions, generally occurring during thawing after longer freezing periods. In contrast to N2O emissions, the seasonal patterns of NO and CO2 soil-atmosphere exchange followed soil temperature changes, although a substantial increase in CO2 emissions was also observed during the freeze and thaw periods. The fact that NO fluxes were higher than N2O emissions during most of the entire observation period indicated that nitrification might have been the primary pathway of N-trace gas production in our study. The extremely high N2O emissions and the substantial interannual variation of N2O flux rates caused by the freeze and thaw effect demonstrate the need for long-term measurements with high temporal resolution in order to come up with more reliable estimates of soil-atmosphere trace fluxes.
 Nitrous oxide (N2O) is one of the main greenhouse gases and contributes at present approximately 6% to the total observed global warming [World Meteorological Organization, 2006]. Although N2O is only present as a trace gas in the Earth's atmosphere, it contributes to reactions that influence atmospheric chemistry and radiative properties. Nitrous oxide not only leads to the chemical destruction of ozone in the stratosphere, but also has a global warming potential of about 300 times of that of carbon dioxide (CO2) on a per-molecule basis [Intergovernmental Panel on Climate Change (IPCC), 2007]. In view of its atmospheric increase of about 0.3% per year and of its atmospheric lifetime of about 150 years it can be expected that the contribution of N2O to global warming will further increase in the future.
 In contrast, nitric oxide (NO) is indirectly involved in global warming and contributes to the net production of radiatively active tropospheric ozone and the formation of acid rain [Williams et al., 1992]. Moreover, NO is also important in controlling the oxidizing capacity of the troposphere, thereby affecting the fate of carbon monoxide, methane and nonmethane hydrocarbons [Liu et al., 1987].
 Soil-atmosphere exchange of trace gases is mainly determined by the simultaneous kinetics of production, consumption and diffusion of the gases in the sequential biochemical reactions. Like all biogenic processes, microbial turnover processes vary largely on spatial and temporal scales, since they are significantly influenced by a number of soil environmental factors and ecological drivers [Gödde and Conrad, 2000; del Prado et al., 2006; Borken and Matzner, 2009]. Among these factors, soil temperature and soil moisture have been identified as major drivers on hourly to interannual timescales for observed temporal changes of N2O, NO and CO2 emission from forest soils [Gasche and Papen, 1999; Papen and Butterbach-Bahl, 1999; Borken et al., 2002; Schindlbacher et al., 2004]. The effect of soil temperature on soil-atmosphere exchange of N2O, NO and CO2 is mostly direct. With increase in temperature, emissions of these gases will increase due to the fact that enzymatic processes and, thus, microbial turnover rates generally increase with temperature as long as other factors are not limiting. In contrast to soil temperature, the effect of soil moisture is more complex. Despite its function as a transport medium for NO3− and NH4+, soil water influences the rate of O2 supply and thereby determines whether aerobic processes such as nitrification or anaerobic processes such as denitrification prevail within the soil [Schindlbacher et al., 2004; Pilegaard et al., 2006]. While N2O emissions are accepted to increase at higher water contents through greater loss from denitrification [Papen and Butterbach-Bahl, 1999; Wolf and Russow, 2000], the maximum NO and CO2 emissions are assumed to occur at low to medium soil water content [Gasche and Papen, 1999; Brümmer et al., 2009].
 In view of the large spatial and temporal variability of environmental factors as well as the limited number and restricted temporal coverage of field measurements, the contribution of temperate forest soils to regional and global budgets of atmospheric N2O, NO and CO2 is still afflicted with a high degree of uncertainty. To more accurately estimate the emission strengths of these trace gases from soils, empirical and process-oriented models [e.g., Parton et al., 1996; Li et al., 2000; Stange et al., 2000] have been developed and are accepted as tools to reduce the associated uncertainties and improve the current estimates [Kiese et al., 2005; Butterbach-Bahl et al., 2009]. However, to expand the applicability and adequately validate these models, in situ trace gas flux measurements at high temporal resolution are essential and urgently needed.
 Therefore, the main objectives of this study were (1) to establish a unique 5 year data set of continuous subdaily N2O, NO and CO2 fluxes from a temperate forest soil, (2) to identify seasonal and interannual variations of N2O, NO and CO2 emissions, (3) to study the relationships between trace gas fluxes and environmental factors and soil parameters, and (4) to quantify annual gas losses from the forest soil with a high degree of accuracy.
2. Materials and Methods
2.1. Site and Soil
 The experimental site, the “Höglwald,” lies in the hilly landscape of Southern Bavaria, Germany, about 50 km northwest of Munich (11°11′ E, 48°30′ N) at an elevation of 540 m above sea level. The climate is suboceanic with a mean annual precipitation of 850 mm and a mean annual temperature of 7.6°C [Rothe et al., 2002]. The study described here was performed in a more than 90 year old Norway spruce (Picea abies) plantation. The soil at the experimental site is a Typic Hapludalf (FAO: dystric cambisol), strongly acidified in the topsoil and weakly aquic in the argillic horizon [Kreutzer and Weiss, 1998]. The soil pH value (measured in CaCl2) is 2.9–3.2 in the organic layer and 3.6–4.0 in the uppermost mineral soil layer [Kreutzer, 1995]. Main characteristics of the site and the forest soil investigated are summarized in Table 1. The site is characterized by N saturation due to long-term heavy atmospheric nitrogen input, with an N load in the throughfall of more than 35 kg N ha−1 yr−1 (NH4+: NO3− ratio of the throughfall equals ∼2:1) and NO3− leaching via seepage being in the range of 20–30 kg N ha−1 yr−1 [Kreutzer et al., 2009]. Further information about the Höglwald site was given in detail by Kreutzer  and Kreutzer and Weiss .
Table 1. Main Characteristics of the Site and Soil Investigated at Höglwald Foresta
2.2. In Situ Measurements of N2O, NO and CO2 Fluxes
 For continuous trace gas emissions measurements, fully automated measuring systems were established at the Höglwald spruce site [Butterbach-Bahl et al., 1997; Gasche and Papen, 1999; Papen and Butterbach-Bahl, 1999]. Five chambers of the closed type for N2O flux determinations, and five dynamic measurement chambers and one dynamic reference chamber for NO and CO2 flux measurements were installed (dimensions: 0.5 m × 0.5 m × 0.15 m, length × width × height). In contrast to the measuring chambers, the reference chamber had a gas-tight bottom made of perspex. The temporal resolution of flux measurements was 2 h for N2O fluxes, i.e., 1 h closure time followed by 1 h with the chamber lids open, and 1 h for NO and CO2 fluxes, i.e., 6 min closure time followed by 6 min with the chamber lids open for five cycles. Each of the five NO/CO2 measuring chambers was sampled once per hour for 6 min, and the reference chamber was sampled for 6 min between each measuring chamber, i.e., five times per hour. If rainfall occurred chambers were opened automatically [Butterbach-Bahl et al., 1997]. N2O was determined using a gas chromatograph (GC-14A, Shimadzu, Duisburg, Germany) equipped with a 63Ni electron capture detector. Air samples from each closed chamber were taken automatically every 15 min for 3 min by a membrane pump at a rate of 200 ml min−1. Water vapor and CO2 in sample air were removed by a permapure dryer (Ansyco, Karlsruhe, Germany) and an ascarite precolumn, respectively. N2O was separated on a stainless steel column packed with Hayesep N (3 m, 1/8 inch, 60/80 mesh), with oven and detector temperatures at 60°C and 340°C, respectively, and with N2 as carrier gas at a flow rate of 20 ml min−1. Drift of the gas chromatograph was monitored by sampling reference gas (0.4 ppm N2O in synthetic air, Messer Griesheim, Germany) twice every 2 h. N2O emission rates were calculated from the linear increase of N2O concentrations with time within the chambers during closure.
 Measurements of NO and CO2 flux rates were performed in the same chambers by using the dynamic chamber method. During sampling, ambient air was sucked at a constant rate (50 L min−1) through the chambers by a sampling pump and transferred via PTFE tubing (inner diameter: 10 mm; length 20 m, internal volume: 1.6 L) to the analyzers. NO and CO2 concentrations at the inlet and the outlet of the chambers were determined using a highly sensitive NOx analyzer (chemoluminescence detector CLD 770 AL ppt and photolysis converter PLC 760, Eco Physics AG, Switzerland) and a differential infrared (IR) CO2 analyzer (BINOS 100, Rosemount, Hanau, Germany), respectively. Ozone concentrations in sample air were measured continuously by an infrared ozone analyzer (TE49C, Thermo Environmental Instruments, Incorporated, Franklin, Massachusetts, United States) to correct the measured NO and NO2 concentrations for the reaction of NO with O3 prior to the calculation of NO and NO2 flux rates [Butterbach-Bahl et al., 1997; Gasche and Papen, 1999]. Calibration of the NOx analyzer was performed weekly with a multigas calibrator (Eco Physics AG, Dürnten, Switzerland) using standard gas (1 ppm NO in synthetic air, Messer Griesheim, Germany), which was diluted with synthetic air to a final NO concentration of 10 ppb. Determination of NO, NO2 and O3 concentrations in ambient air and at the outlet of the chambers, calculation of NO flux rates and efficiency of photolytic cleavage of NO2 into NO were described in detail by Butterbach-Bahl et al. . The calculation of CO2 flux rates was in principle similar to NO fluxes, with the difference that the CO2 analyzer detected the difference between reference and measuring chambers directly. The CO2 analyzer was calibrated regularly with two reference gases (400 and 450 ppm CO2 in synthetic air, Messer Griesheim, Germany) by applying one gas simultaneously to both cuvettes of the instrument for setting the zero point, and the two gases for setting the span. All flux rates were corrected for temperature and air pressure. Detailed descriptions of the automated measuring systems including design of chambers, gas chromatographic conditions and modes of calculation of flux rates can be found in previous publications [Butterbach-Bahl et al., 1997; Gasche and Papen, 1999; Rosenkranz et al., 2006a]. Gaps originating from instrumental failure were filled by linear interpolation between measured fluxes for calculation of cumulative annual emissions.
2.3. Soil Environmental and Climatic Measurements
 Daily precipitation and air temperature at 2 m above ground level from 2000 to 2008 were obtained from the German Weather Service station Augsburg-Mühlhausen, which is about 20 km northwest from the Höglwald Forest site. Soil temperatures at various depths (organic layer, 5, 10, 15 and 20 cm) were measured every minute by PT100 probes (IMKO GmbH, Ettlingen, Germany) in close vicinity to the chambers. Hourly soil moisture measurements were carried out with horizontally installed TDR probes (IMKO GmbH) at 10 cm soil depth from 2000 to 2004. Due to instrumental failure and removal of the soil moisture sensors, in situ soil moisture measurements during 2005 to 2008 were not available at our site. To fill this gap, a novel machine-learning technique, called support vector machine (SVM), was employed. The SVM method is based on a statistical learning algorithm [Cristianini and Shawe-Taylor, 2000] and has been successfully applied to bioinformatics, data mining, surface temperature prediction and hydrology [Furey et al., 2000; Chen and Yu, 2007; Anandhi et al., 2009]. The fundamental principle of SVM and its formulation were described in detail by Kecman . Daily meteorological input data, required as model drivers, that is, minimum and maximum temperature, mean relative humidity, total precipitation, sunshine duration, mean degree of cloud cover and mean wind force from 2000 to 2008, were either obtained from continuous measurements at the Höglwald Forest site or from the German Weather Service station Augsburg-Mühlhausen. Since for soil moisture determination not only observations of the very day are important, meteorological variables were aggregated to 8 different additional cumulative sums (i.e., 1, 3, 5, 7, 14, 21, 28 and 35 days previously) for creating a training data set, which contains 92 variables in total. The best 10 variables, which could explain more than 95% of the variance in the data set from 2000 to 2004, were used to predict soil water content for the years 2005–2008. Field-measured and model-simulated volumetric soil moisture values were transferred into water-filled pore space (WFPS) values by the following equation:
where WFPS is the water-filled pore space value (%), Vol is the volumetric soil water content (%), BD is the bulk density (g cm−3) and 2.65 is the density of quartz (g cm−3).
2.4. Statistical Analyses
 The Kolmogorov-Smirnov test was performed to test for normal distribution of the data. For normally distributed data the t test was applied to identify significant differences between data sets, for nonnormally distributed data the nonparametric Mann-Whitney test was performed instead. Linear regression analysis, nonlinear regression analysis, and Pearson correlation were used to examine relationships between N2O, NO and CO2 fluxes and the measured environmental parameters. Multivariate nonlinear regression analysis was performed with SigmaPlot 2000 (SPSS, Incorporated, Chicago, Illinois, United States) to evaluate the conjunct influence of soil temperature and soil moisture on trace gas fluxes and the ratios among these fluxes.
3.1. Soil Moisture Modeling
 To validate the performance of the SVM modeling, we compared measured and simulated values of WFPS of 2000 and 2004 for the Höglwald Forest site (Figure 1). For the mineral soil at 10 cm depth, the SVM approach was able to capture the seasonal trend of changes in WFPS as well as the magnitude of WFPS with high precision (r2 = 0.95, n = 1638). However, the SVM approach resulted in a slight relative underestimation of soil moisture values of ∼7%. We used the SVM modeling approach for gap-filling of times without soil moisture measurements, i.e., to create a 5 year data set of continuous daily soil moisture values (2004 to 2008) for further analysis of temporal dynamics of C and N trace gas fluxes at the Höglwald Forest site (Figure 2b).
3.2. Seasonal and Interannual Variations of N2O, NO and CO2 Emissions
 In Table 2 and Figure 2, annual means of air temperature, soil temperature (organic layer) and WFPS at 10 cm depth, and precipitation at the Höglwald Forest are given for 2004–2008. Compared to the long-term annual mean air temperature (7.6°C), the period of 2004 to 2008 was much warmer, especially in 2007. Precipitation in the relatively dry years (2004 and 2008) was much lower than in the relatively wet years (2005 and 2007), resulting in large variations among these 5 years. The 5 year mean annual precipitation was 754 mm for the observation site. There were no significant annual differences in air and soil temperature (Figure 2a). Obvious seasonal patterns were observed for air and soil temperatures as well as precipitation. Air and soil temperatures increased gradually from January to July and then decreased until December. Near half of the annual precipitation occurred from May to August (Figures 2 and 3).
Table 2. Annual Means of Air Temperature, Soil Temperature, WFPS, and Precipitation at Höglwald
Air Temperature (±SD) (°C)
Soil Temperature (Organic Layer) (±SD) (°C)
WFPS (10 cm Depth) (±SD) (%)
Precipitation (±SD) (mm)
Simulated values by support vector machine approach. WFPS, water-filled pore space.
 The temporal variations of daily means of N2O, NO and CO2 emission rates are given in Figures 2b–2d. Annual cycles of NO and CO2 fluxes at the Höglwald site showed clear seasonal patterns, mainly following the annual courses of air and soil temperatures (Figures 2 and 3). A seasonal, temperature-driven change in magnitude of fluxes was also evident for N2O, but was less pronounced (Figures 2b and 3). In winter 2005 and 2006, a large short-term increase of N2O emission rates during thawing periods after long-term freezing was observed (Figures 2b and 4).
 The continuous measurements of N2O, NO and soil CO2 fluxes performed at high temporal resolution allowed us to identify interannual variations in mean annual flux rates (Table 3) and to calculate cumulative annual trace gas emissions for each of the observation years with high accuracy (Table 4 and Figure 5). Though the basic seasonal patterns of N2O, NO and CO2 fluxes were quite similar among these 5 years, the magnitude varied significantly from year to year (p < 0.01).The mean annual N2O emission rate in 2004 at Höglwald was 2.3 ± 0.1 μg N2O-N m−2 h−1, which was ∼3 times lower than in 2007 and ∼5 times lower than in 2005 and 2008, respectively, and more than 10 times lower (p < 0.001) than the mean annual N2O emission rate of 2006 (33.6 ± 5.3 μg N2O-N m−2 h−1). These large interannual fluctuations of N2O fluxes were mainly due to peak emissions during thawing after long-term frost periods in 2005 and 2006. The annual mean N2O loss from the Höglwald soil for the entire observation period 2004–2008 was 12.4 ± 1.0 μg N2O-N m−2 h−1 (Table 3). The mean annual NO emission rate in 2004 at Höglwald was 106.4 ± 4.5 μg NO-N m−2 h−1 and therefore ∼1.3 times higher (p < 0.001) as compared to annual means calculated for 2005 and 2008 (77.2 ± 3.2 μg NO-N m−2 h−1 and 85.8 ± 4.6 μg NO-N m−2 h−1, respectively). The highest and lowest mean annual NO fluxes were observed in 2006 (137.7 ± 6.1 μg NO-N m−2 h−1) and 2007 (69.5 ± 2.8 μg NO-N m−2 h−1), respectively. The annual mean NO emission rate from Höglwald soil for the entire observation period 2004–2008 was 96.3 ± 2.1 μg NO-N m−2 h−1 (Table 3). In contrast to mean annual N2O and NO emissions, there were no pronounced differences in the magnitude of the mean annual CO2 emission rates among the different years of observation, though statistical analysis showed significant differences over all 5 years (p < 0.01). The highest and lowest mean annual CO2 flux rates were observed in 2006 (95.1 ± 1.5 mg CO2-C m−2 h−1) and 2008 (68.8 ± 2.0 mg CO2-C m−2 h−1), respectively. The annual mean CO2 emission rate from the Höglwald soil for the entire observation period 2004–2008 was 82.1 ± 0.9 mg CO2-C m−2 h−1 (Table 3).
Table 3. Mean Annual and Maximum and Minimum Mean Daily Fluxes of N2O, NO, and CO2 With the Coefficient of Variation at Höglwald From 2004 to 2008a
CV, coefficient of variation; N, number of valid N2O, NO, and CO2 fluxes.
N2O Fluxes (μg N m−2 h−1)
Annual mean (±SE)
2.29 ± 0.08
11.78 ± 0.86
33.62 ± 5.29
6.63 ± 0.23
10.84 ± 0.54
12.4 ± 0.95
NO Fluxes (μg N m−2 h−1)
Annual mean (±SE)
106.36 ± 4.5
77.2 ± 3.15
137.65 ± 6.1
69.51 ± 2.79
85.84 ± 4.64
96.33 ± 2.1
CO2 Fluxes (mg C m−2 h−1)
Annual mean (±SE)
73.78 ± 2.12
91.46 ± 2.25
95.11 ± 1.5
75.41 ± 1.97
68.81 ± 1.95
82.12 ± 0.93
Table 4. Annual Cumulative Fluxes of N2O, NO, and CO2 at Höglwald From 2004 to 2008
N2O (kg N ha−1 yr−1)
NO (kg N ha−1 yr−1)
CO2 (t C ha−1 yr−1)
0.20 ± 0.01
9.38 ± 0.40
6.99 ± 0.20
1.02 ± 0.07
7.34 ± 0.30
7.90 ± 0.19
3.24 ± 0.51
12.34 ± 0.55
8.37 ± 0.13
0.61 ± 0.03
6.27 ± 0.25
6.62 ± 0.17
0.95 ± 0.05
7.87 ± 0.43
5.87 ± 0.17
1.20 ± 0.09
8.64 ± 0.19
7.15 ± 0.08
 During the entire observation period, N2O emissions showed the strongest interannual variations, followed by NO fluxes, whereas soil respiration had the weakest interannual variation. The annual coefficient of variation (CV) of N2O fluxes in 2004, 2007 and 2008 was in general < 100% (Table 3). Owing to the long-term frost periods in 2005 and 2006, CV values increased up to 137.7% and 256.0%, respectively, reflecting the large amplitudes of N2O emission rates. Though the CV values of NO emissions in all the observation years were in general < 85%, the highest value was observed in 2008 reflecting the relatively large temporal and seasonal fluctuations. CV values for CO2 flux rates did not change markedly over the years (<50%), and the mean CV value for the entire observation period 2004–2008 was 43.3%.
3.3. Trace Gas Emissions During Freeze and Thaw Periods
 In winter 2005 and 2006, high N2O emissions occurred after long-term frost periods during thawing of the soil (Figure 2b, and in more detail in Figure 4). From the end of January until 25 March 2005, the soil temperature of the organic layer was around 1.0°C, while air temperature was below freezing point most of the time. In the middle of February 2005, the air temperature increased from below freezing to positive values and then dropped back again, resulting in a relative small N2O emission peak at the same time. From 12 March 2005, N2O emission rates increased from 3.7 to 122.8 μg N2O N m−2 h−1 within 5 days and reached the maximum (173.5 μg N2O N m−2 h−1) on 18 March 2005. This maximum occurred at the same time when air temperature showed a first maximum above 10°C (Figure 4). Thereafter, N2O emission rates decreased gradually to values < 10 μg N2O N m−2 h−1, though both air and soil temperatures in the organic layer were still increasing.
 The freeze and thaw effect on N2O emissions in 2006 was more significant than that in 2005. From the middle of January until the middle of February 2006, the temperature of the organic layer was below 0°C. Though the soil was frozen, N2O emissions increased gradually and reached a first maximum (156.7 μg N2O N m−2 h−1) on 8 February 2006, when the temperature of the organic layer reached the freezing point and part of the snow cover (∼5 cm) was melting (Figure 4). Thereafter, N2O emissions rates decreased again with decreasing air and soil temperatures. However, with the second onset of thawing of the organic layer (increase in soil temperature from −1.0°C to +0.5°C) and a first increase in air temperature up to 5°C, a second, much higher maximum (487.3 μg N2O N m−2 h−1) of N2O emission was observed (Figure 4). Thereafter, similar to 2005, N2O emission rates decreased gradually to values < 20 μg N2O N m−2 h−1 despite the increase of air and soil temperatures.
 Beside the outstanding N2O emissions during the freeze and thaw periods in 2005 and 2006, also relatively high CO2 emission peaks were observed during these periods (Figure 2d, and in more detail in Figure 4). The highest CO2 fluxes during the freeze and thaw periods in 2005 and 2006 were 129.2 and 151.3 mg CO2 C m−2 h−1, respectively. It is worthy to mention that these relatively high CO2 emission peaks usually occurred several days earlier than the N2O emission peaks and were more correlated to the changes in air temperature than soil temperature. In both years, the elevated CO2 fluxes during freeze and thaw periods were in most cases accompanied by a sharp increase of air temperature beyond freezing point (Figure 4). Although NO also showed a weak response during/after thawing, which can be seen especially for the years 2005, 2006 and 2007 (Figure 2), in contrast to N2O the changes of NO fluxes due to the freeze and thaw effect were negligible.
3.4. Annual Cumulative Precipitation and Trace Gas Flux Rates
 Annual cumulative precipitation at the Höglwald Forest site was in the range of 628.5 to 889.5 mm for 2004–2008, and mean monthly precipitation sums were higher in July and August than during the rest of year (Table 2 and Figure 3). The cumulative precipitation from January to February in 2004 was significantly higher than in other observation years, although the lowest amount of annual cumulative precipitation was also observed in 2004 (Figure 5).
 Annual cumulative N2O emissions ranged from 0.20 ± 0.01 to 3.24 ± 0.51 kg N ha−1 yr−1 showing significant differences between years during the observation period 2004–2008 (Table 4). The larger annual cumulative N2O emissions found in 2005 and 2006 were primarily due to extremely high fluxes during the soil freeze and thaw periods (Figure 5). N2O emissions during the periods of soil freezing and thawing contributed 24.4% and 73.3% to the total annual N2O emission in 2005 and 2006, respectively. The lowest annual cumulative emission of N2O was found in 2004. For the same year also the lowest annual cumulative precipitation was observed, of which about 40% occurred in the cold season from January to April (Figure 5). The mean annual cumulative N2O flux rate over all 5 years was 1.20 ± 0.09 kg N ha−1 yr−1.
 The seasonal patterns of annual cumulative NO fluxes showed a sigmoidal curve shape (Figure 5). The highest annual cumulative emission of NO (12.34 ± 0.55 kg N ha−1 yr−1) was observed in 2006, which was approximately twofold higher than that in 2007 (6.27 kg ± 0.25 N ha−1 yr−1). The mean annual cumulative NO flux over all 5 years was 8.64 ± 0.19 kg N ha−1 yr−1, indicating a much stronger emission of NO than N2O at the Höglwald site. Interannual differences of annual cumulative soil respiration were smaller than those of N2O and NO emissions, but similar to N2O and NO the highest value of annual cumulative soil respiration (8.37 ± 0.13 t C ha−1 yr−1) was also observed in 2006 (Table 4). The mean annual cumulative soil respiration over all 5 years was 7.15 ± 0.08 t C ha−1 yr−1.
3.5. Correlation of Trace Gas Flux Rates With Soil Temperature and WFPS
 To identify the effect of changes in soil temperature and WFPS on the magnitude of trace gas flux rates, correlation analyses were performed. The correlations between N2O emission, soil temperature and WFPS at 10 cm soil depth were generally weak. This was mainly due to the extremely high N2O emissions during the soil freeze and thaw periods in 2005 and 2006, which overrode the normal effect of soil temperature and moisture. These correlations could be significantly improved by excluding the extremely high N2O emissions during the soil freeze and thaw periods from the regression analysis (Figures 6 and 7). The relationships could be described best by quadratic functions (p < 0.001), i.e., increased values of soil temperature led to an increase in N2O emissions, while the correlation function for N2O fluxes versus WFPS had its maximum at WFPS values between 40 and 45% (Figure 7). A contour plot analysis showing the joint effect of soil temperature and WFPS on N2O emissions is displayed in Figures 8a and 8b. In general, the dominant environmental factor affecting the magnitude of N2O gas fluxes at Höglwald was soil temperature rather than soil moisture, but soil moisture could become a crucial regulator during freeze and thaw periods. The positive effect of WFPS on N2O emissions was most pronounced for values > 40% during the freeze and thaw periods and at high temperatures. In contrast, a stronger correlation between N2O emissions and soil temperature could be obtained when N2O fluxes during the soil freeze and thaw periods were excluded (Figure 8b). Within the range between 2.5°C to 10°C, soil temperature had only little influence on the magnitude of N2O emissions. However, N2O emissions increased significantly when soil temperatures exceeded 10°C (Figures 6, 8a, and 8b).
 During the entire measuring period, a quadratic relationship was also found for NO fluxes and soil temperature at our site (Figure 6). Like in the case of N2O, the correlation between NO fluxes and soil moisture was generally weak but significant (Figure 7). Figure 8c shows NO fluxes plotted against soil temperature and WFPS. NO emissions were positively correlated with WFPS up to a soil temperature of 15°C, but at soil temperatures above 15°C, highest NO emissions were found at lowest WFPS values. In contrast, for N2O the soil moisture dependency increased with increasing soil temperature except for the freeze and thaw periods.
 Both soil temperature and WFPS showed a pronounced effect on CO2 fluxes in our study. A significant positive correlation between CO2 emissions and soil temperature was observed during the entire observation period (Figure 6). Also for the relationship between CO2 emissions and WFPS, a quadratic regression function resulted in a better coefficient of determination than linear regression (Figure 7). The conjunct effect of soil temperature and WFPS on CO2 flux rates is shown in Figure 8d. The dependency of soil respiration rates on soil temperature was greater than that on soil moisture. A significant positive correlation between CO2 emissions and soil temperature was observed when temperature values exceeded 10°C (r2 = 0.16, p < 0.0001). In contrast, CO2 emission rates were negatively correlated with increasing WFPS in the range of WFPS observed in this study (Figure 7).
 The ratio of NO:N2O was calculated to assess the relative importance of nitrification and denitrification in producing NO and N2O (Figure 9). During the entire observation period, the NO fluxes were significantly higher than N2O fluxes (p < 0.001), resulting in a relatively high value of NO:N2O emission ratio (mean value for 2004–2008 was 24.11 ± 3.63, n = 1270, ± SE). Except for the freeze and thaw periods, NO emissions dominated N2O fluxes. The dominant environmental factor affecting the NO:N2O ratio was soil moisture rather than soil temperature. A significant positive correlation between soil temperature and NO:N2O ratio was only observed when WFPS values were below 45%. The highest NO:N2O ratio was observed when soil temperatures were around 15°C and WFPS values < 40% (Figure 9a), which were more favorable to nitrification. Similar to the NO:N2O ratio, the ratio between CO2 and N2O fluxes was also mainly affected by soil moisture. Here, increased values of WFPS led to a decrease of the ratio; that is, N2O emissions increased faster with increasing values of WFPS as compared to CO2 emissions. The highest CO2:N2O ratios were observed when soil temperatures were around 5°C and WFPS values < 35% (Figure 9b). The contour plot analysis showing the joint effect of soil temperature and WFPS on CO2:NO ratios demonstrated that both soil moisture and soil temperature had a significant positive effect on the CO2:NO ratio at low and medium values of soil temperatures and WFPS (Figure 9c). At WFPS values < 45% and soil temperatures below 10°C, the CO2:NO ratio increased with increasing soil temperature and WFPS. However, the effect of soil temperature and moisture on the CO2:NO ratio became weak or even negative when soil temperature and WFPS values exceeded these levels.
4.1. Comparison With Other Studies
 To our knowledge this is the first paper to report 5 years of continuous data of soil-atmosphere exchange of N2O, NO, and CO2 with a subdaily temporal resolution for a terrestrial ecosystem, here a temperate forest. The range, magnitude, and spatial variability of N2O and NO emissions observed in this study for the entire observation period at the Höglwald spruce site are in agreement with previous observations at this site, which covered the period from 1994 to 1997 [Gasche and Papen, 1999; Papen and Butterbach-Bahl, 1999]. However, compared with other published N2O and/or NO fluxes from temperate coniferous forest soils [Schmidt et al., 1988; Davidson and Kingerlee, 1997; Stark et al., 2002; Kitzler et al., 2006; Rosenkranz et al., 2006a, 2006b], rates in this study are much higher. The annual mean N2O fluxes calculated by Ambus and Christensen  and Borken and Brumme  from long-term measurements at temperate spruce forest sites were approximately 0.8 and 0.6 kg N2O-N ha−1 yr−1, respectively, i.e., comparable to those we found at our site (1.2 kg N2O-N ha−1 yr−1). Uptake of atmospheric N2O by the forest soil, as, e.g., observed by Rosenkranz et al. [2006b] over longer time periods for a Mediterranean pine forest soil, was not observed at our site. By summarizing three observations, Davidson and Kingerlee  estimated the mean NO flux from N-affected temperate forest was 2.7 kg NO-N ha−1 yr−1 (ranging from 1.1 to 5.0 kg NO-N ha−1 yr−1). However, two of the three estimates were obtained from deciduous forests, which have been shown to have much lower NO emissions than coniferous forests [Davidson and Kingerlee, 1997; Gasche and Papen, 1999]. Fenn et al.  reported NO fluxes from a mixed conifer forest soil in California of 2.56–6.81 kg NO-N ha−1 yr−1, i.e., a range close to our results (8.64 kg NO-N ha−1 yr−1). The elevated N2O and NO emissions observed in our study could be partly explained by the fact, that the investigated forest was subject to high atmospheric N input (>35 kg N ha−1 yr−1), associated with N saturation and strong acidification of the upper soil layers [Kreutzer, 1995; Kreutzer et al., 2009]. The results of Fenn et al. , Butterbach-Bahl et al.  and Pilegaard et al.  also showed a close relationship between atmospheric N deposition and N-trace gas emissions. Furthermore, the hitherto published estimates of mean annual N2O emissions are subject to a high degree of uncertainty and tend to underestimate N2O losses from temperate forest soils because of the high temporal variability and low frequency of measurements (weekly to monthly). This conclusion is supported by the extremely high N2O emissions during long-term freeze and thaw periods in 2005 and 2006 (Figures 2b and 4), which would have been missed without continuous measurements and, in consequence, would have led to a dramatic underestimate of annual N2O losses from our site. With regard to soil CO2 emissions, the observed values are consistent with the results from previous studies for temperate forest soils [Raich and Schlesinger, 1992; Raich and Tufekcioglu, 2000; Rosenkranz et al., 2006a], but are higher than the values reported by Kitzler et al.  for an alpine forest site at the German-Austrian border. The reasons may lie in the lower annual mean temperature (6.5°C) and higher annual precipitation (1733 mm) at the study site of Kitzler et al.  compared with those at Höglwald, since lower temperature and increased precipitation could result in decreased soil respiration due to lower microbial activity and limited oxygen availability.
4.2. Seasonal and Interannual Variations
 The results obtained in this study demonstrate large seasonal and interannual variations in N2O, NO and CO2 soil-atmosphere fluxes at our study site. The seasonal variations of trace gas emissions mainly followed the annual changes in air and soil temperature and soil moisture, which is consistent with earlier publications [Butterbach-Bahl et al., 1997; Gasche and Papen, 1999; Papen and Butterbach-Bahl, 1999; Kitzler et al., 2006]. The previous studies at Höglwald, including measurements in beech and spruce stands, and one study in a limestone Alps spruce forest soil also showed a distinct seasonal pattern of trace gas fluxes, with higher emissions in summer and lower emissions in winter [Gasche and Papen, 1999; Papen and Butterbach-Bahl, 1999; Kitzler et al., 2006]. However, also some low background N2O fluxes without seasonal pattern were reported for temperate forest soils [Brumme et al., 1999; Lamers et al., 2007]. According to the classification of emission patterns introduced by Brumme et al. , our results of N2O emissions showed a typical “event emission pattern,” characterized by low seasonal fluxes but extremely high emission events (such as rewetting following a dry period and during frost/thaw periods). The fact that there was also a pronounced N2O peak in summer 2008 at our site, but not so in the other years, although soil temperature and moisture conditions were comparable, could not be elucidated unambiguously.
 The high emission events are the reason for the substantial interannual variation of trace gas fluxes at Höglwald. All mean annual flux rates of N2O, NO and CO2 in 2005 and 2006 were higher than those in other observation years except for NO emission in 2005. In addition, the seasonal and interannual variability of precipitation might be another driving factor for the remarkable interannual variation of N2O and NO flux rates in our study. However, interannual variation in precipitation showed no or only little effect on annual soil respiration rates at our site, which was in line with the observations reported by Borken et al. . Large interannual variations of trace gas fluxes have also been observed in other investigations of temperate forest soils [Brumme, 1995; Borken and Brumme, 1997; Gasche and Papen, 1999; Papen and Butterbach-Bahl, 1999], agricultural land [Kusa et al., 2006; Mu et al., 2008] and other ecosystems [Brümmer et al., 2008; Song et al., 2009], even though these data sets are by far not as detailed and comprehensive as the one presented here. Our study clearly shows that long-term continuous measurements over several years with high temporal resolution are needed to estimate the annual soil-atmosphere trace gas exchange more reliably, especially for N2O, but also for soil NO and CO2 fluxes.
4.3. Trace Gas Emissions During Freeze and Thaw Periods
 Soil freeze and thaw periods contributed 24.4% and 73.3% to annual N2O emissions in 2005 and 2006, respectively. This, as well as the length of the freeze and thaw periods, was in the range found by Papen and Butterbach-Bahl , Teepe et al. , Groffman et al.  and Goldberg et al. . However, the quantitative importance of freeze and thaw events for annual N2O budgets is still highly uncertain, may show large interannual variations (Groffman et al.  and this study), and differs among ecosystems [Teepe et al., 2000]. The maximum and total amount of N2O emissions at Höglwald during freeze and thaw periods were 173.5 μg N2O N m−2 h−1 or 0.25 kg N2O N ha−1 in 2005, and 487.3 μg N2O N m−2 h−1 or 2.37 kg N2O N ha−1 in 2006, respectively. One possible reason for these large differences between the 2 years could be the lower soil temperature and the longer duration of soil frost in 2006 (Figures 2b and 4), possibly leading to a larger die-back of microbes during the frost period and, hence, to a higher substrate availability during thawing. This assumption is supported by earlier studies that also showed that temperature and duration of soil frost could have a marked effect on the amount of N2O released during freeze and thaw periods [Koponen and Martikainen, 2004].
 Similar to the previous field observations at the Höglwald Forest site [Gasche and Papen, 1999], and a laboratory soil core incubation study [Koponen et al., 2006], there was no significant thawing-related increase of NO emission during the entire 5 year observation period. Koponen et al.  suggested that the factors involved in the regulation of NO and N2O emissions at low temperatures are different. NO emissions are suspected to have a maximum in soils with good aeration, a condition favoring nitrification, whereas N2O emissions dominate at high soil water content through greater loss from denitrification [Wolf and Russow, 2000; Schindlbacher et al., 2004; Pilegaard et al., 2006]. The increased soil water content due to the melting of the snow could favor denitrification after soil thawing. Thus, during soil thawing periods the main process in the soil producing N2O was very likely denitrification rather than nitrification, as indicated by the simultaneously observed low NO:N2O ratios and as found in other studies [e.g., Mørkved et al., 2006].
 In accordance with results obtained by others [Schimel and Clein, 1996; Teepe et al., 2001; Mørkved et al., 2006], we also observed a substantial increase in CO2 emissions accompanying elevated N2O emissions during freeze and thaw periods. Similar to N2O, the mechanisms of physical release and/or enhanced biological activity due to increased nutrient supply could also be the reason for the elevated CO2 emissions. The abundant nutrients supplied by easily degradable substrate during thawing periods might have significantly promoted microbial activity, which in turn could have resulted in higher CO2 production, as well as higher oxygen consumption, mineralization, nitrification and denitrification rates. According to the “hole-in-the-pipe” model of Davidson , enhanced mineralization, nitrification and denitrification rates caused by stimulated microbial activities will induce higher CO2 and N2O production. However, freeze and thaw events had a smaller effect on CO2 than on N2O release in our study, as thawing influenced N2O release (denitrification) more than microbial CO2 production. A laboratory study by Koponen and Martikainen  supported our hypothesis that denitrification benefits more from the freeze-and-thaw-induced substrate increase than soil respiration during mild frost. In both 2005 and 2006, the elevated CO2 fluxes during freeze and thaw periods were highly correlated with the changes in air temperature (Figure 4), which may be due to autotrophic root respiration, another important determinant of soil CO2 flux beside heterotrophic soil respiration. For example, Knohl et al.  and Högberg et al.  demonstrated that autotrophic soil (root) respiration is driven by the photosynthetic activity of the plants/trees, which is directly linked to air temperature and photosynthetically active radiation.
4.4. Environmental Controls on Trace Gas Fluxes
 Although many environmental controls (e.g., soil temperature, soil moisture, C:N ratio, pH, soil N content) were considered as influencing factors for trace gas fluxes from temperate forest soils [Schindlbacher et al., 2004; Kitzler et al., 2006; Pilegaard et al., 2006], previous studies at Höglwald and other temperate forests showed that soil temperature and soil moisture were the most important factors [Gasche and Papen, 1999; Papen and Butterbach-Bahl, 1999; Borken et al., 2002; Rosenkranz et al., 2006a]. Therefore, correlation analyses were performed exclusively with soil temperature and WFPS as the dominating regulators in our study. We generally found significant positive relationships between soil temperature and trace gas fluxes, which were in good agreement with results from other studies [Borken et al., 2002; Schindlbacher et al., 2004; Pilegaard et al., 2006]. However, high soil water content during freezing and thawing periods clearly overrode the effect of soil temperature on N2O emissions. The mean Q10 values (±SE) calculated for the temperature dependency (5–15°C) of N2O, NO and CO2 fluxes were 2.76 ± 0.71, 2.50 ± 0.82 and 2.12 ± 0.17, respectively. The larger range of Q10 values for N2O and NO suggests that the potential temperature effect on N trace gas fluxes from Höglwald was subject to higher seasonal and interannual variability than that on soil CO2 efflux. The strong temperature dependencies of N2O and NO emissions in our study are in excellent agreement with Q10 values reported from previous studies [Sitaula and Bakken, 1993; Gasche and Papen, 1999; Papen and Butterbach-Bahl, 1999]. The temperature sensitivity of soil respiration at Höglwald was also in line with the median value of 2.4 from various soils [Raich and Schlesinger, 1992], but was lower than some earlier results [Buchmann, 2000; Borken et al., 2002]. The large range of Q10 values reported in the literature may be due to different regions, land uses, tree species and soil depths from which temperature dependencies were deduced.
 Several studies have demonstrated that WFPS values around 60% act as an important threshold for nitrification and denitrification activities, e.g., below 60% WFPS nitrification activities increases with increasing soil moisture and decreases dramatically when WFPS exceeded 60%, while denitrification activity showed an entirely antithetic relationship [Davidson, 1991; Kusa et al., 2006; Mu et al., 2008]. In our study, soil WFPS values were seldom above 60%, indicating that nitrification might be the primary pathway of N-trace gas production at the Höglwald Forest site, which is consistent with previous studies at our site [Butterbach-Bahl et al., 2002]. This is confirmed by the observation that NO fluxes dominated N2O emissions, resulting in high NO:N2O emission ratios during most of the entire observation period, since NO production in soils is mainly originating from nitrification rather than from denitrification [Kiese and Butterbach-Bahl 2002; Butterbach-Bahl et al., 2004b]. Relatively high NO flux rates were found around 45–50% WFPS (Figure 7), which was in accordance with previous studies [Gasche and Papen, 1999; Li et al., 2007]. However, in a study of six European forest soils, Schindlbacher et al.  found large differences in the optimal moisture level (range from 15% to 65% WFPS) for NO emission, suggesting that each soil will have a specific soil moisture optimum for NO production according to its specific soil characteristics, e.g., nitrogen content and availability, pH, soil texture and bulk density, but also to its distinct microbial community. In addition, increased NO emissions were observed following soil rewetting after longer dry periods (i.e., summer 2004 and 2006 as well as April 2007), which is accordance with some previous studies [Davidson et al., 1991; Smart et al., 1999; Stark et al., 2002]. However, in contrast to N2O during freeze and thaw events, we have never observed a stimulation of NO fluxes after dry periods to levels higher than “normal” flux rates. This might be due to the fact that the soil never really dried out, as can be seen from the WFPS data (Figure 2b), i.e., WFPS was never below 30%.
 In our study, CO2 emission rates were negatively correlated with increasing WFPS, and the optimum soil moisture level for soil respiration calculated from the correlation function was around 35% WFPS. A gradual decline in soil respiration due to increasing soil moisture was also reported by other observations [e.g., Kiese and Butterbach-Bahl, 2002; Brümmer et al., 2009]. Kiese and Butterbach-Bahl  further pointed out that the effect of soil moisture on soil respiration is characterized by a strong seasonality; that is CO2 emission rates were positively correlated with WFPS during the dry season or the transition period between dry and wet season, but negatively correlated with soil moisture during the wet season. In addition, Kitzler et al.  reported that CO2 release was reduced during periods of heavy rain, when the soil water content was between 50 and 65% WFPS. One possible explanation for this could be oxygen deficiency in soil due to diffusion restrictions during periods with high WFPS values, thus limiting the activity of organisms only capable of aerobic respiration and favoring denitrification [Kiese and Butterbach-Bahl, 2002].
 To our knowledge, the 5 year data set of soil-atmosphere exchange of N2O, NO and CO2 at the temperate, nitrogen-saturated Norway spruce forest site Höglwald reported here is a worldwide unique data set for terrestrial ecosystems. On the basis of these long-term continuous measurements, we could constrain the annual budgets of trace gases with a high degree of certainty, and demonstrate substantial seasonal and interannual variation of gas flux rates at Höglwald. Seasonal patterns of soil N2O fluxes were characterized by event emissions, generally occurring during thawing of soil frozen for several weeks. In contrast to N2O emissions, the seasonal patterns of NO and CO2 soil-atmosphere exchange were relatively stable and followed the annual course of temperature throughout the investigation period. However, we were also able to demonstrate a substantial increase in CO2 emissions during freeze and thaw periods. Soil temperature and soil moisture were found to have significant effects on soil-atmosphere exchange of the three trace gases at our study site. Our results show that correlation of trace gas fluxes with soil temperature was stronger than that with soil moisture. However, soil moisture could become the crucial regulator during some extreme emission events (like freeze and thaw periods). In addition, our results demonstrate the importance of field measurements at high temporal resolution to improve our understanding of the biogeochemical N and C turnover processes in temperate forest soils as well as the processes involved in trace gas production, consumption and emission. Furthermore, the comprehensive data set in our study can be very useful for future model testing and validation to reduce the uncertainties associated with regional and national trace gas emission estimates.
 This research was supported by the Helmholtz Association of German Research Centers in the framework of the program-oriented funding (POF), period 2004–2008, and by the Integrated Project NitroEurope IP, funded by the European Commission. One of the authors, Xing Wu, would like to thank the Helmholtz Association of German Research Centers and the China Scholarship Council (CSC) for providing financial support within the Junior Scientists Exchange Program.