Over 1 year we followed the seasonal variations of N2O and CH4 fluxes at a tropical rain forest site in Australia. In addition, meteorological parameters, litter fall and decomposition, plant species composition, and concentrations of NH4+/NO3− in the soil and N2O and CH4 in the soil atmosphere were measured. N2O emissions showed a pronounced seasonal pattern with highest rates in the wet season (108.6 μg N m−2 h−1) and significantly lower rates during dry season (mostly <10 μg N2O-N m−2 h−1). N2O emissions were positively correlated to N2O-concentrations in the soil profile and to moisture, but not to concentrations of NH4+ and NO3−. The annual emission of N2O (N = 6015) was 0.97 kg N ha−1 yr−1, and, thus, approximately 7 times lower than a previous estimate for the year 2000. The marked differences in N2O emissions between different years indicate that the interannual variability of N2O emissions from rain forest soils cannot be neglected. With regard to CH4 the soil functioned throughout the entire year as a significant sink. Rates of CH4 uptake during the dry period (35–68 μg CH4 m−2 h−1) were higher as compared to the wet period (4–45 μg CH4 m−2 h−1). A close linear correlation between soil moisture and magnitude of CH4 uptake was found. The calculated annual CH4 uptake (N = 6090) is 3.2 kg CH4 ha−1 yr−1. This implies that tropical rain forest soils function as significant sinks for atmospheric CH4 on a global scale.
 The magnitude of N2O emissions, as well as of CH4 uptake, is highly variable and strongly influenced by environmental conditions, like soil texture, substrate availability and, in most published data sets, also highly dependent upon changes in soil moisture [e.g., Davidson, 1993; Kiese and Butterbach, 2002; Breuer et al., 2000; Steudler et al.,1996]. However, due to the harsh environmental conditions and the problems with infrastructure, most of the estimates of N2O emissions or uptake of atmospheric CH4 by tropical rain forest soils are based on sporadic measurements [e.g., Matson et al., 1990; Verchot et al., 1999], which do not fully reflect the spatial and temporal dynamics of emission/deposition of these gases as was recently demonstrated by Kiese and Butterbach-Bahl . The limited availability of data and particularly the absence of measurements covering entire years are the main reason why estimates of the source and the sink strength for N2O and CH4, especially for tropical rain forest soils, are still highly uncertain [Breuer et al., 2000; Kroetze and Mosier, 2002]. To improve these estimates a dual approach needs to be followed. At first more measurements of rates of trace gas exchange between different tropical ecosystems and the atmosphere are required, fulfilling both representativeness and long-term coverage of measurements in order to understand seasonal variability and to identify environmental drivers which control the magnitude of trace gas emission/deposition. On the basis of the results of these measurements it will be necessary to develop new tools for upscaling emissions from a site to a regional scale. At present the best prospects for this task are biogeochemical models, which are able to simulate all processes and mechanism involved in N- and C-trace gas emissions from soils. However, even these models will need detailed data of a high temporal resolution, which should cover at least entire seasons or years, for further development and validation. On the basis of this background, we conducted a study that aimed to acquire a unique, long-term data set on N2O emissions and CH4 uptake by a tropical rain forest soil and to report also on other environmental parameters which may affect the magnitude of trace gas fluxes.
2. Material and Methods
2.1. Study Site
 The study was conducted at a typical tropical rain forest site (145°54′E, 17°16′S, 80 m a.s.l.) in the Coastal Lowlands of the “Wet Tropics,” Queensland, Australia, close to the village of Bellenden Ker, approximately 70 km south of Cairns. Mean annual precipitation at the site is 4360 mm and mean annual temperature is 24.3°C. The precipitation regime is characterized by a pronounced seasonality with 70% of annual precipitation falling during the wet season, which normally lasts from November to April, and only 30% during the dry season lasting from May to September. The soil type is a Ustochrept derived from granitic parent material characterized by a high sand fraction (60%) and medium silt (20%) and clay (20%) contents. The soil pH in the uppermost layers of the mineral soil is 4.1 indicating acidic conditions due to intensive weathering and leaching processes. The organic carbon content is around 3.1% the nitrogen content is 0.26% resulting in a C/N ratio of 12:1 [Kiese and Butterbach-Bahl, 2002]. The complex mesophyll vine forest [Tracey, 1982] at our site has a basal area of almost 50 m2 ha−1 and a average canopy height of 20 m. The forest structure is comparable to many of the lowland rain forests in South East Asia. Plant biodiversity is very high with over 130 plant species including 63 different kind of trees occurring in a defined plot of 20 m by 50 m. (Table 1). All experiments described below were conducted within this area. For further information about site properties, see Kiese and Butterbach-Bahl .
Table 1. List of Tree and Vine Species and Frequencies of Diameter at Breast Height (DBH) > 10 cm Found at the Bellenden Ker Measuring Site Within the 50 m by 20 m Area in Which All Measurements Were Carried Outa
Stem Area, cm2
Biomass Index, m3
This floristic composition is typical of well-drained soils along the foothills of the Bellenden Ker Range.
Canthium sp. (Whitfield Range BH1020 FRK)
Myristica globosa subsp. muelleri
2.2. Measurement of Soil-Atmosphere N2O and CH4 Exchange
 N2O emissions and CH4 uptake were monitored continuously over a one year period from November 1, 2001, to October 31, 2002, thus, covering all seasons, i.e., wet, dry and intermediate conditions. For the measurements a fully automated recording system was used, which allowed simultaneous determination of N2O emissions and CH4 uptake with a 5-hour time resolution. The system consisted of a gas chromatograph (Texas Instruments, SRI 8610C) equipped with a 63Ni electron capture detector (ECD, Vichy Valco, Switzerland) for N2O analyses and a flame ionization detector (FID) for CH4 analyses, an automated sampling unit for gas sampling and five measuring chambers. Details of the measuring system and modes of calibration have already been described by Breuer et al.  and Kiese and Butterbach-Bahl .
 For the flux measurements, two sampling positions were randomly selected for each chamber within the 20 m by 50 m study area. Combinations of two or three chambers were moved weekly between the different sampling positions to increase representativeness and to reduce the bias of the chambers to the measuring site, for example, by reducing rainfall or by excluding part of the litter fall. A full measuring cycle for the determination of N2O emissions and CH4 uptake lasted 300 min, during which the chambers were automatically closed for the first 100 min and were kept open for the last 200 min. On the basis of four gas concentration measurements in the headspace of the closed chambers, N2O and CH4 fluxes were calculated by linear regression and were corrected for temperature and air pressure. The system was automatically calibrated by standard calibration gases after each measuring cycle.
2.3. Measurements of N2O- and CH4-Concentration at Different Soil Depths
 Weekly measurements of N2O- and CH4-concentration at different soil depths (5 cm, 10 cm, 15 cm, 30 cm) were carried out during the investigation period from November 2001 to October 2002 at three different sampling positions within the defined 20 m by 50 m area. For that, sampling tubes, which consist of a hydrophobic gas permeable membrane (Akzo Nobel Faser AG, Germany), were fixed to 1/8″ stainless steel tubes. To allow gas sampling, the stainless steel tubes were closed with a rubber septum and probes were inserted horizontally into the soil at 5 cm, 10 cm, 15 cm and 30 cm soil depths. For further details, see Kiese and Butterbach-Bahl . Soil air samples where taken with 10-mL gas-tight syringes (Baton Rouge, Louisiana, USA) and analyzed for N2O- and CH4-concentrations by gas chromatography (described above) within a few minutes after sampling.
2.4. Measurement of Litter Mass, Litter Fall and Litter Decomposition Rates
 The litter mass on the soil was determined monthly from five replicates by randomly collecting all litter from quadrates of 0.125 m2. Measurements of litter fall were carried out weekly by sampling five square litter traps each with a collecting area of 1 m2. The litter traps consisted of a fiber-glass gauze with a mesh size of 2 mm which was fixed 50 cm above the ground level. All litter sampling was done within the 20 m by 50 m study area. For litter sampling we included twigs (<2 cm diameter), leaves, flower buds, flowers, bark, fruits and seeds. All sampling material was oven dried at 70°C for 24 hours before weighted.
 Litter decomposition rates were determined using wire mesh cages (1.0 cm mesh on metal frames) to exclude falling litter. Five cages, each 1 m by 1 m by 0.5 m high, were placed over five sampling positions. Litter from all of these positions was collected, bulked, mixed and then subdivided into five portions of equal mass. Each portion was then placed back on the ground beneath a cage. Any litter fall on the tops of the cages was removed weekly to maintain typical micro-climatic conditions. The decomposition of litter, determined as mass loss, was followed by taking 0.125 m2 subsamples at establishment and at subsequent monthly intervals. The experiment was carried out for wet (November 2001 to April 2002) and dry season conditions (June to November 2002). To allow comparison with other published data sets we calculated the decomposition constants (k-values) based on the single exponential decay model by Olson .
2.5. Measurements of Leaf Area Index (LAI)
 LAI data at the Bellenden Ker site were obtained using the standard methodology for the Licor LAI 2000 developed at the CSIRO Tropical Forest Research Centre, Atherton. Observation were made facing eastward either on clear days with the afternoon sun behind the western ranges, or on overcast days in the late afternoon. Readings were taken at 10 pegged locations along a 90-m transect spanning the study area. The sensor head was held at a standard height of 0.3 m above the ground, and at least 0.5 m from the nearest understorey foliage. No trunks of >30 cm DBH were located within 3 m of the sensor head. Large gaps were not encountered.
2.6. Measurements of Climate Parameters
 An automatic climate station (Campbell 21×) was installed at the top of a 25-m high tower 100 m from the measuring area. This meteorological station acquired hourly data on the following parameters: global radiation, air pressure, air temperature, relative air humidity, precipitation, wind direction and wind speed.
2.7. Measurements of Ammonia and Nitrate Concentrations in the Soil
 For ammonia and nitrate analyses and soil moisture content soil samples from the uppermost 5 cm of the mineral soil were taken weekly in three replicates. Each replicate soil sample was sieved to pass through a 2-mm mesh and divided into two subsamples. The first subsample was utilized for determining gravimetric soil moisture, being dried to constant weight at 105°C. Soil moisture content was expressed as percentage of soil dry weight. For NH4+ and NO3− determinations, 15 grams of the second subsample were extracted in 60-mL 2N KCl, shaken for 45 min, and allowed to stand for 15 min before being separated using a Whatman #42 filter paper prewashed with 2N KCl. For each subsample, 13 mL were frozen for future analysis in the Laboratory of CSIRO, Brisbane, where NH4+ and NO3− were determined colorimetrically [Henzell et al., 1968].
 All statistical analyses were performed with SPSS 8.0 (SPSS Inc., United States) and Microcal Origin 6.1. Tests of significance of differences were either performed by using the multiple range test (LSD) by ANOVA or by using a parametric t-test or nonparametric U-test (Mann-Whitney) (SPSS 8.0, SPSS Inc.).
3.1. Environmental Conditions
 In the annual measuring period from November 1, 2001, to October 31, 2002, precipitation at the rain forest site at Bellenden Ker was in total 2678.2 mm (Table 2, Figure 1). This value is only 61% of the mean annual average precipitation for the 20 year period 1980–1999 at this site (4360 mm, Bureau of Meteorology, Brisbane, Australia). The 2001–2002 wet season was the driest observed in the Wet Tropics of Queensland for decades. Rainfall during the 2002 dry season (May/June 2002 to October 2002) also was much lower compared to previous years. However, the pronounced seasonal pattern which can be observed in this part of Australia was still evident in 2001/2002; that is, 80.7% of total rainfall was observed during the wet season (November 2001 to April 2002), whereas only 19.3% of total rainfall occurred during the dry season (May to October 2002). The seasonal pattern of soil moisture closely followed the pattern of rainfall; that is, highest values of soil moisture were recorded in the period from February to April 2002 (approximately 30%), whereas significantly lower values were measured during the period from June to October 2002 (16.5–20%) (Figure 1). Mean annual temperature for the observation period was 23.8°C. The lowest mean daily air temperature was recorded in July 2002 (17.3°C), whereas the maximum occurred in December 2002 (31.4°C) (Table 2). Daily sums of global radiation were highest with values of up to 8050 W m−2 day−1 from October to December 2001. Lower values were observed from May–September 2002, when daily sums of global radiation were <5000 W m−2 day−1 (Figure 1).
Table 2. Main Characteristics of the Meteorological Parameters Temperature, Precipitation, Radiation and Air Humidity at the Rain Forest Site at Bellenden Ker, Australia, During the Measuring Period November 1, 2001, to October 31, 2002
Measurements are missing for the following periods of time: 01/01/02–01/03/02, 01/06/02–02/23/02, 05/27/02–06/02/02, and 07/29/02–08/04/02.
Data from Bureau of Meteorology, Brisbane, Australia.
3.2. Seasonal Variability of NH4+ and NO3− Concentrations in the Soil, Leaf Area Index, Litter Fall, Forest Floor Litter Mass and Litter Decomposition
 Over the entire observation period from November 1, 2001, to October 31, 2002, measurements of leaf area index (LAI) and forest floor litter mass were carried out at monthly intervals and litter fall and NH4+ and NO3− concentrations in the soil were measured at weekly intervals (Figure 2). NH4+ and NO3− concentrations in the soil varied in a range of 0.05–3.4 μg NH4-N g−1 soil dry weight (SDW) or 0.9–5.7 μg NO3-N g−1 SDW, respectively. Highest concentrations of NH4+ and NO3− in the soil were found at the beginning of the wet season. A second peak of high concentrations of NO3− in the soil (>5 μg N g−1 SDW) was observed at the start of the dry season, i.e., from mid-May 2002 to June 2002, whereas NH4+ concentrations in the soil decreased with time during the wet season but increased with time during dry season conditions (Figure 2). Mean average NH4+ and NO3− concentrations in the soil over the entire observation period were 1.4 ± 0.1 μg NH4-N g−1 SDW and 2.6 ± 0.2 μg NO3-N g−1 SDW.
 During the study period the LAI only varied in a very narrow range of 5.2–6.3 m2 m−2 (mean: 5.5 ± 0.1 m2 m−2). Generally, higher values of LAI (mostly >5.7 m2 m−2) were recorded during wet season conditions, whereas the LAI in the dry season was slightly lower (<5.5 m2 m−2). The lowest value of LAI (5.2 ± 0.1 m2 m−2) was recorded at the end of the observation period in October 2002. The latter finding fits well with the observed maximum in litter fall (Figure 2). However, compared to seasonal variations in the LAI litter fall was much more variable with time. A first peak of litter fall was observed in January 2002 with 0.51 ± 0.06 t ha−1 week−1, during a period of time with heavy storms and rainfall events. During dry season conditions litter fall was mostly lower only occasionally exceeding 0.2 t ha−1 week−1. Except at the end of the observation period, after 3–4 months of dry season conditions, rates of litter fall increased and reached an absolute maximum of 0.81 ± 0.07 t ha−1 week−1 (Figure 2). During the entire observation period, litter fall was 9.0 ± 1.0 t ha−1 yr−1. The average N-content of the leaf litter was 1.63% and annual total litter fall nitrogen was approximately 150 kg N ha−1 yr−1. The amount of forest floor litter mass was highest in the early stages of the wet season in January 2002 (6.0 ± 0.6 t ha−1). Thereafter, forest floor litter mass declined steadily and reached a minimum in September 2002 with 2.6 ± 0.3 t ha−1. On average the litter mass of the forest floor was 4.1 ± 0.3 t ha−1.
 With regard to the dynamics of litter decomposition significant differences were found between the wet and dry seasons (Figure 3). Litter decomposition was approximately two-fold higher during wet season conditions as compared to dry season conditions. The half-life of litter during wet season was 42 days (90% decay: 159 days) while the half-life of litter during the dry season was 83 days (90% decay: 276 days). The decomposition constants (k-values) were 6.1 yr−1 for the wet season and 3.0 yr−1 for the dry season, respectively.
3.3. Seasonal Variability of N2O Emissions
 By the use of the fully automatic measuring system it was possible for the first time to measure in at least daily resolution the rates of exchange of N2O and CH4 between the soil and the atmosphere in a tropical rain forest ecosystem continuously over an entire year (Figure 4, Table 3). During the 1-year observation period, more than 6000 valid N2O emission rates were obtained, spanning 348 of 365 days. Mean N2O emissions from the different positions varied by approximately a factor of 4 in a range of 5.6–20.1 μg N2O-N m−2 h−1. The maximum observed N2O emission rate was 108.6 μg N2O-N m−2 h−1, whereas minimum N2O emission rates at all positions were lower than the detection limit of the system (<1.0 μg N2O-N m−2 h−1) (Table 3). Highest N2O emissions were observed during the wet season (Figure 4) and during the transition period from wet to dry season (May 2002). During this period of time, the mean N2O emission from all positions only occasionally dropped below 10 μg N2O-N m−2 h−1 (mean for November 1, 2001, to May 15, 2002: 16.3 ± 0.3 μg N2O-N m−2 h−1). Figure 4 clearly shows that rainfall events during this humid period, especially if they last for some days, directly lead to increased N2O emissions. During such a period, after 2 weeks of continuous rainfall (sum of precipitation: 410 mm) in May 2002, when soil moisture reached a value of 34%, the highest mean (average of 5 chambers) N2O emission of 62.0 μg N2O-N m−2 h−1 was also observed.
Table 3. Means (± SE), Minimum and Maximum of N2O Emissions and CH4 Uptakes for the 10 Positions Investigated at the Rain Forest Site at Bellenden Ker, Australiaa
N2O Emission, μg N m−2 h−1
CH4 uptake, μg CH4 m−2 h−1
Different superscript (A–G) indicate significant differences between individual positions (P < 0.05). cV: coefficient of variation; N: number of valid rates of N2O emissions/CH4 uptake.
5.6 ± 0.2A
40.8 ± 0.4A
7.3 ± 0.2B
46.9 ± 0.4B
11.1 ± 0.5C
28.5 ± 0.3C
17.2 ± 0.7D
31.2 ± 0.4D
6.3 ± 0.2A,B
42.7 ± 0.3E
11.7 ± 0.4C
36.5 ± 0.4F
20.1 ± 0.7E
30.7 ± 0.5D
16.2 ± 0.5D
31.0 ± 0.4D
14.0 ± 0.6F
38.8 ± 0.3G
11.9 ± 0.4C
36.6 ± 0.3G
11.0 ± 0.2
36.7 ± 0.1
 Compared to the relatively high and temporally variable N2O emissions in the humid season, N2O emissions during the dry season were almost constant and on a significantly lower level (mostly <10 μg N2O-N m−2 h−1) (Figure 4). The mean N2O emission rate for the dry period (here May 15, 2002 to October 31, 2002) was 4.0 ± 0.1 μg N2O-N m−2 h−1 (Figure 4). Even short-term increases of soil moisture due to sporadic rainfall events hardly led to increased N2O emissions during the dry period, which was exceptionally long lasting and having unusual low rainfall. Only during a short rainy period in mid August 2002, when approximately 90 mm of rainfall was observed within 2 weeks and when soil moisture increased from approximately 15% to well above 20%, did N2O emissions increase to the dry season maximum of 13.0 μg N2O-N m−2 h−1 (Figure 4). The average N2O emission for the entire observation period, i.e., November 1, 2001, to October 31, 2002, was 11.0 ± 0.2 μg N m−2 h−1, which equates to an annual N2O release of 0.97 ± 0.02 kg N ha−1 yr−1.
 The data presented show that the magnitudes of N2O emissions obviously correlate with changes in soil moisture. This close positive relationship can be best described by an exponential equation (Figure 5). A correlation between N2O emissions and NO3− or NH4+ concentrations in the soil could not be demonstrated, irrespective of which data transformations (e.g., log transformation of N2O emission, mineral N concentrations, NO3− to NH4+ ratio or sum of inorganic N in the soil) were applied. Furthermore, also the amount of litter fall nitrogen was not correlated to the magnitude of N2O emissions.
 The seasonal dynamic of N2O emissions was highly correlated with changes in N2O-concentrations in the soil air (Figures 4 and 6). Maximum mean N2O-concentrations in all investigated soil depths were observed during rainy periods, for example, mid-February or beginning of May, when N2O-concentrations at the 5 cm soil depth increased to values of up to 700 ppbv N2O and at 30 cm soil depth of up to 1450 ppbv N2O (Table 4). During the dry period from mid-May to end of October, N2O-concentrations in all soil layers were relatively constant and close to the ambient atmospheric N2O-concentration of approximately 315 ppbv (soil air 5 cm: 316–365 ppbv N2O; soil air 30 cm: 339–390 ppbv N2O). Only during the rainy period at the end of August 2002 did N2O-concentrations in the soil air increase significantly from the very low levels observed before. During this period, N2O-concentrations of up to 393 ppbv N2O at 5 cm soil depth and 429 ppbv N2O at 30 cm soil depth were observed.
Table 4. Means (± SE), Minimum and Maximum of N2O- and CH4-Concentrations (ppbv) in the Soil Air at Different Soil Depths at the Rain Forest Site at Bellenden Ker, Australiaa
Different superscripts indicate significant differences (P < 0.05) in N2O- or CH4-concentrations between different soil depths (A–C) and periods of time (1,2). CV: coefficient of variation; N: number of measurements.
N2O-Concentration in the Soil Air
463 ± 17A,1
494 ± 23A,1
632 ± 36B,1
740 ± 50C,1
333 ± 4A,2
342 ± 5A,2
375 ± 7B,2
388 ± 9B,2
CH4-Concentration in the Soil Air
1213 ± 25A,1
870 ± 22B,1
403 ± 18C,1
274 ± 27D,1
1254 ± 10A,1
928 ± 11B,2
450 ± 7C,2
253 ± 11D,1
3.4. Seasonal Variability of CH4 Uptake
 During the entire observation period the soil at the rain forest site at Bellenden Ker was a sink for atmospheric CH4 (Figure 7, Table 3); that is, rates of CH4-oxidation were always higher than possibly occurring CH4-production. Lowest rates of CH4 uptake were observed at late April/early May, i.e., at the end of the wet season 2002, when uptake rates at individual measuring positions were as low as 4.1 μg CH4 m−2 h−1 (Table 3). With the start of the dry season and with declining values of soil moisture, CH4 uptake rates increased rapidly and reached a maximum in July 2002 with mean values in the range of 40–50 μg CH4 m−2 h−1 (Figure 7). The coefficients of variation show that the seasonal variability of CH4 uptake by the soil was much lower as compared to the seasonal variability of N2O emissions (Table 3). For most of the individual positions the coefficient of variation of CH4 uptake was <30% and, thus, approximately 3 times lower than the coefficient of variation for N2O emissions. However, as was also shown for N2O emission, the magnitude of CH4 uptake was strongly linked to observed soil moisture values. Figure 5 shows that the relationship between soil moisture and rates of CH4 uptake could be best described with a simple linear regression and that rates of CH4 uptake decrease with increasing values of soil moisture. The mean rate of CH4 uptake over the entire observation period, which is based on more than 6000 valid uptake rates, was 36.7 ± 0.1 μg CH4 m−2 h−1 or 3.21 ± 0.01 kg CH4 ha−1 yr−1.
 A weak but significant relationship was found between CH4 uptake rates and concentrations of NH4+ in the soil (r2 = 0.15, p < 0.05, f(x) = 29.1 + 5.3 x), whereas a correlation between NO3− concentrations in the soil and CH4 uptake could not be demonstrated. The function given illustrates that CH4 uptake rates increased with increasing NH4+ concentrations in the soil.
 The comparable low seasonal variability of CH4 uptake at our rain forest site was also mirrored by only small seasonal changes in CH4-concentrations in the soil air at different depths (Table 4 as well as Figure 7). The lowest CH4-concentrations in the soil air with values of 144–386 ppbv CH4 were observed for the 30-cm layer, whereas CH4-concentrations in 5-cm soil depth were significantly higher (1038–1446 ppbv CH4). In contrast to the results obtained for N2O, no significant relationship between soil air CH4-concentrations and rates of atmospheric CH4 uptake could be demonstrated (Figure 8).
 The measured values for leaf area index of 5.2–6.3 m2 m−2 at our rain forest site at Bellenden Ker, Australia, fall within the range of values reported for tropical forests elsewhere in the world [Grace et al., 1995; Vourlitis et al., 2001; Kitayama and Aiba, 2002]. Also, the annual rate of litter fall at our site (9.0 ± 1.0 t ha−1) is in very good agreement with reported values for other tropical rain forest sites in Australia (7.3–11.0 t ha−1) [Spain, 1984; Stocker et al., 1995] or on other continents [Erickson et al., 2002] (and see, for example, compilation of data by Vitousek ). Furthermore, the mass of litter on the forest floor (2.6–6.0 t ha−1) agrees well with other reports from tropical rain forests in Australia [Spain et al., 1984], Brazil [Klinge et al., 1975] or Nigeria [Hopkins, 1966]. The calculated decomposition rates with k-values of 6.1 yr−1 (wet season) and 3.1 yr−1 (dry season) are in the upper range of published values for tropical regions [e.g., Kumar and Deepu, 1992; Muoghalu et al., 1994; Vitousek et al., 1994; Aerts, 1997], but in very good agreement with reports by Spain  for tropical rain forests in Australia. This may be due to the quite narrow C/N ratio of leaf litter at our site, which is in the range of 20–30 (unpublished data).
 This paper reports for the first time a complete data set of N2O emissions from a tropical rain forest site with at least daily resolution. This huge dataset allowed us (1) to identify short-term changes in the magnitude of N2O emissions, (2) to demonstrate seasonal variability of N2O emissions with a hitherto unknown time resolution and (3) to calculate precise annual loss rates of N2O from our site. The short-term variability of N2O emissions from tropical rain forest soils and its dependency on rainfall events have already been highlighted by Breuer et al.  and Kiese and Butterbach-Bahl  for different rain forest sites in Australia and by Vitousek et al.  during a rewetting experiment of a tropical forest soil in Chamela, Mexico. Our observations show, that it is essential to have continuous measurements of N2O emissions, at least during wet season conditions, to reliably estimate annual loss rates. This statement can be further supported by a simple statistical analysis. Using the entire data set, we calculated mean annual N2O emissions (CH4 uptake) based on randomly selected data sets assuming daily, weekly or monthly measuring intervals. Such an analysis shows that the reduction of the measuring frequency from 4–5 flux rates per day to daily, weekly or monthly intervals would reduce the accuracy of the calculated annual N2O emission (CH4 uptake) by approximately 10% (5%), 20% (10%), and 40% (20%), respectively. From these figures it can be concluded that weekly measurement of trace gas fluxes provide a reliable basis for estimating annual budgets of N2O- and CH4-exchange at our site. However, it is necessary to point out that for further improvement and validation of biogeochemical models which simulate the biosphere-atmosphere trace gas exchange it is of utmost importance to have detailed data available such as daily or subdaily trace gas fluxes in order to be capable to proof that the processes involved in trace gas production, consumption and emission are reliably reproduced.
 The seasonal variability of N2O emissions found in this study, with significantly higher N2O emissions during wet season as compared to the dry season, is in general agreement with the seasonal changes in decomposition rates at our site and with previous reports about N2O emissions from soils of tropical rain forests worldwide [e.g., Garcia-Méndez et al., 1991; Steudler et al., 1991; Verchot et al., 1999; Breuer et al., 2000; Melillo et al., 2001]. The pronounced seasonal variations in N2O-emissions are most likely associated with significant changes in the composition of the microbial community involved in N-cycling and N2O-production, since results from Kiese et al.  indicate that the abundance of denitrifiers at our site is approximately 5 times higher than at a montane rain forest site with a colder and dryer climate. However, at present data on seasonal changes in microbial community composition are not available to further support this hypothesis. The higher N2O emissions in the wet season as compared to the dry season are also mirrored by higher concentrations of N2O in the soil atmosphere. N2O-concentrations in the soil air were highest with up to 1440 ppbv N2O in 30 cm soil depth and lowest close to the surface [<702 ppbv N2O]. However, these data do not allow to conclude that N2O-production does mainly occur in the deeper soil horizons, since soil N2O-concentrations do not only depend on N2O-production but also on rates of N2O-consumption and diffusion. Breuer et al.  have shown for different montane rain forest soils that the N2O-production in such soils is mainly concentrated in the uppermost centimeters of the mineral soil or even in the litter layer. The measured N2O-concentrations in the soil air (340–1440 ppbv N2O) are lower than reported by Verchot et al. , who found N2O-concentrations of up to 8200 ppv N2O in a primary rain forest in eastern Amazonia, but in the same range of N2O-concentrations as observed by McSwiney et al.  in tropical rain forest soils in Puerto Rico. However, these values are difficult to compare since sampling depth and soil properties, and thus the permeability of the soil for gases, will be different between the sites. Nevertheless, the observed pronounced seasonal dependency of N2O-concentrations in the soil atmosphere is in good agreement with results reported by Verchot et al. . NH4+ as well as NO3− concentrations in the uppermost 5 cm of the soil did not significantly correlate with the magnitude of N2O emissions at all. This finding is in agreement with results from other studies for tropical forest soils and temperate soils, in which correlations between seasonal changes in N2O emissions and soil mineral N concentrations could not be presented [e.g., Keller and Reiners, 1994; Verchot et al., 1999; Breuer et al., 2000; Papen and Butterbach-Bahl, 1999]. Only if mineral N concentrations and N2O emissions across different sites are compared, do significant correlations between N2O emissions or N2O + NO emissions and NO3− and/or NH4− concentrations in the soil become obvious [Keller and Reiners, 1994; Verchot et al., 1999; Davidson and Verchot, 2000; Davidson et al., 2000]. From our point of view, the latter correlation only indicates differences in the nutrient status between sites. However, such a correlation does not allow the conclusion that seasonal changes in N2O emissions at a given site are reflected by changes in soil NO3− and NH4+ concentrations. In this context, one should also be aware of the dynamics of microbial N turnover processes such as nitrification in temperate and tropical forest soils, which have been observed to be in the range of 20- >300 mg NH4-N m−2 d−1 (see compilation of results from different groups by Breuer et al. ). The observed rapid cycling of NH4+ and NO3− leads to short turnover times of the inorganic N pools in the soil. For example, Verchot et al.  showed for different temperate forest sites in the United States that the mean turnover time of the NH4+ and NO3− pools in the soil was only 1.7 ± 0.4 d−1 and 4.4 ± 2.5 d−1, respectively.
 It is well known that well aerated forest soils are significant sinks for atmospheric CH4 [e.g., Crill, 1991; Castro et al., 1995]. Though there are many publications which have investigated the sink strength of temperate forest soils for atmospheric CH4 [e.g., Steudler et al., 1989; Sitaula et al., 1995; Butterbach-Bahl and Papen, 2002], published data sets on rates of CH4 uptake by tropical rain forest soils are still scarce. The few reports available show tropical rain forest soils can function as sinks as well as sources for atmospheric CH4 and that fluxes are in a range of −300- + 550 μg CH4 m−2 h−1 [Keller et al., 1983, 1986; Tathy et al., 1992; Delmas et al., 1992; Keller and Reiners, 1994; Steudler et al., 1996; Verchot et al., 2000]. Our measurements on CH4 uptake at the Bellenden Ker site are the first which report on a full annual cycle of CH4 uptake by a tropical rain forest soil, thereby demonstrating that a pronounced seasonal variability exists, with significant higher rates of CH4 uptake in the dry period as compared to the wet period. The range of CH4 uptake observed within this study (4.1–68.4 μg CH4 m−2 h−1) is well within the span given above. However, it is remarkable that even in periods with heavy rainfall, as, for example, at the beginning of May 2002, the soil at our site still function as a weak, but still significant sink for atmospheric CH4 and that during these periods of time the CH4-concentrations in the soil air remained significantly below atmospheric concentrations at all soil depths monitored. This indicates the high drainage capacity of the soil at our site and the high gas permeability of the soil at least down to approximately 30 cm. However, soil moisture and, thus the gas permeability of the soil for atmospheric CH4 and O2, remains the main controller of the magnitude of CH4 uptake at our site. This observation is in good agreement with many other reports on controls of CH4 uptake by forest soils [e.g., Crill, 1991; Adamsen and King, 1993]. The negative correlation between N2O emissions and rates of CH4 uptake (f(x) = 41.7 − 0.55 × N2O emission; r2 = 0.340; p < 0.01) further support the finding, that CH4 uptake rates at our site mainly depend on CH4 and O2 diffusion from the atmosphere to the soil, whereas N2O-emissions are higher if O2 diffusion is reduced due to, for example, increased soil moisture. The finding that higher rates of CH4 uptake were found at times with higher concentrations of NH4+ in the soil is surprising, since results of other studies indicate that increased availability of inorganic N in forest soils inhibit CH4-oxidation [e.g., Adamsen and King, 1993; Schnell and King, 1994]. However, since this correlation was very weak and NH4+-concentrations in the soil were at the lower end of observed inorganic N-concentrations in tropical and temperate forests [e.g., Verchot et al., 1999; Papen and Butterbach-Bahl, 1999] and moreover highly correlated to soil moisture; that is, higher NH4+-concentrations were found at lower soil moisture values, we doubt that NH4+-concentrations in tropical forest soils are a good predictor for CH4 uptake rates. The calculated annual CH4 uptake at our site of 3.21 kg CH4 ha−1 yr−1 is in the range of observed annual rates of CH4 uptake for temperate forest soils [e.g., Steinkamp et al., 2000; Butterbach-Bahl and Papen, 2002; Borken and Brumme, 1997]. In accordance with the work of Verchot et al.  on CH4 uptake in Amazonian tropical forest soils our results demonstrate that tropical rain forest soils are indeed significant sinks for atmospheric CH4 on a global scale.
 The authors thank Hans Papen for many valuable discussions which helped to improve the design and evaluation of measurements and Niels Johannsen for his excellent assistance during the establishment of the automatic measuring system and the first weeks of measurements. We are in particular indebted to Spiro Buhagiar and Jex Healey from the Bellenden Ker Cablecar Inc. (BTS), Queensland, Australia, for their generous hospitality during our stays at the field sites and for their support in running the equipment successfully over an entire year. Furthermore the authors want to express their gratitude to Andrew Ford, Tropical Forest Research Centre, Atherton, Australia, for providing an extensive plant species list for the site. This research was funded by the Deutsche Forschungsgemeinschaft (DFG), Germany, under contract BU 1173-3.