Detailed regional predictions of N 2 O and NO emissions fro a tropical highl nd rainforest

Introduction Conclusions References


Introduction
Soils are both a source and a sink of gases such as carbon dioxide (CO 2 ), methane (CH 4 ) and nitrous oxide (N 2 O).All these gases are potent greenhouse gases (GHG).In particular, N 2 O is 298 times as potent as CO 2 and ranks third after CO 2 and CH 4 as a global warming agent (Meehl et al., 2007).In addition, NO emitted from soils acts as Figures

Back Close
Full a precursor of tropospheric ozone, which contributes to the greenhouse effect as well (Delmas et al., 1997).
The concentrations of the atmospheric GHGs have accelerated during the past century, indicating an urgent need for process-based understanding of the factors influencing the exchange of these gases between the soil and atmosphere at a range of scales, as a route to developing effective mitigation strategies.Natural sources of atmospheric N 2 O and NO are significant in the respective global budgets and include oceans and tropical forest soils.The importance of tropical forests as sink and source of carbon GHGs is relatively well known, but N 2 O and NO fluxes from tropical forest soils are far less well characterized.On its own, tropical forest soils are identified as a major natural source for atmospheric N 2 O (Matson et al., 1990;Bouwman et al., 1993;Mosier et al., 1998;Breuer et al., 2000;Kiese et al., 2002Kiese et al., , 2003;;Butterbach-Bahl et al., 2004;Werner et al., 2007b) and are responsible for an approximately 1.3 ± 0.3 Tg nitrogen (N) yr −1 (Werner et al., 2007a).Similarly, based on the few existing data, NO from tropical forest soils represent a significant source with estimates ranging from 1.1 to 3.0 Tg N yr −1 within the global atmospheric budget (Davidson and Kingerlee, 1997;Gut et al., 2002;Butterbach-Bahl et al., 2004).Further simulation results, using an N isotope coupled mechanistic biogeochemical model, showed that total gaseous losses, including N 2 , from tropical rain forest soils in Hawaii contributed for ∼ 26-48 % of total N losses (Bai and Houlton, 2009).
Production of N 2 O and NO in tropical forest soils is closely linked to the microbial processes of nitrification when ammonia is oxidized to nitrate in the presence of oxygen, and denitrification when nitrate used as an electron acceptor instead of oxygen, as well as chemical process of chemo-denitrification when nitrite react with metallic cations such as Fe (II) (e.g.Davidson et al., 2000;Serca et al., 1994;Butterbach-Bahl et al., 2004).

Conclusions References
Tables Figures

Back Close
Full and LandscapeDNDC (Haas et al., 2012), simulate soil climate, litter decomposition and plant growth, in order to determine soil C and N turnover and predict the soilatmosphere exchange of GHGs.Irrespective of extensive research on microbial nitrification and denitrification and their contribution to N 2 O and NO fluxes (e.g.Matson and Vitousek, 1990;Davidson et al., 2000;Butterbach-Bahl et al., 2004;Lisboa et al., 2011;Rowlings et al., 2011) the magnitude of gaseous nitrogen (N) emissions for global sources are still highly uncertain at all scales.Amongst others a high spatio-temporal variability in environmental factors is a reason for the high uncertainty of current regional and global estimates of GHG emissions from soils, in general (Kesik et al., 2005;Werner at al., 2007a).Moreover, for most regions little information exists on the spatial heterogeneity and/or average values of physicochemical soil properties, driving biogeochemical models.Currently, global estimations of N 2 O emissions for tropical forest rely mostly on the International Soil Reference and Information Center (ISRIC-WISE, with focus on tropical and subtropical regions), which does not distinguish between agriculture and natural land cover.
The estimation of N 2 O emission from tropical forest in such a global study is based on soil properties, like OC, which are predominantly collected for agriculture soils that can be an obvious source of bias.Furthermore, global studies do not explicitly differentiate highland from lowland tropical forests, which strongly affects climate and soil properties.Given the uncertainty associated with current estimations of N 2 O and NO emissions from global tropical forests, which rely on relatively few soil inventories from tropical forests, no reliable spatial explicit predictions for N-trace gases are available for the tropical forests.
Given the importance of N-trace gases in the GHG balance of tropical forests there is an urgent need to quantify the impact of the regional distribution of soil properties and vegetation characteristics on the model estimates of these gases.To do so, emphasis should be put on using regionally available legacy soil data to predict spatial N-trace gas emissions.Introduction

Conclusions References
Tables Figures

Back Close
Full Legacy soil data is data collected during historical soil surveys e.g. for the purpose of soil mapping are generally collected without any statistical criteria.Such legacy soil data can, however, be valuable for biogeochemical models as it is the best and the only available data.This also links to the wider discussion about uncertainty in biogeochemical models (Werner et al., 2007a), which have recently highlighted the potential effort for improving spatial information on soil and vegetation data from tropical forests.
In this study we incorporate historical soil data to predict regional N 2 O and NO emissions from the mountainous tropical rainforest in Nyungwe national park, Rwanda, using the ForestDNDC-tropica biogeochemical model (Werner et al., 2007a).Using these predictions we aim to answer key questions such as: ( 1) what is the N 2 O and NO source strength of a mountainous tropical forest soil? ( 2) To what extend do these values differ from current estimates?(3) And how sensitive is the ForestDNDC-tropica model to changes in driving (historical) input variables.

Site description
Nyungwe forest in southwestern Rwanda ( 2• 17 S -2 • 49 S/29 • 03 E -29 • 29 E, 1485-2950 m a.s.l., Fig. 1a) is one of the largest and best-preserved mountainous rainforests in Africa.The forest occupies an area of about 100 000 ha dividing the Nile and the Congo river catchments.Overall temperatures vary between 8 and 29 degrees and yearly rainfall ranges from 1308 to 2071 mm yr −1 , with the rainy season lasting from September until May.The forest contains various ecosystems ranging from dense forest, bamboo groves to marshes, and contains approximately 1105 plant species, as well as high biodiversity of fauna.(Graham et al., 1995;Masozera and Alavalapati, 2004;Plumptre et al., 2007) The vegetation is characterized by Entandrophragma excelsum, Parinari excelsa, Prunus Africana and Octotea usambarensis (Van Ranst et al., 1997).Introduction

Conclusions References
Tables Figures

Back Close
Full

Model description
Model description DeNitrification-DeComposition (DNDC) models (Li et al., 1992) have been developed for simulating ecosystem C and N-cycling and associated emission of trace gases.In particular, the ForestDNDC-tropica model includes sub-models that simulate soil climate, litter decomposition, plant growth and biogeochemical processes contributing to trace gases emissions in order to determine soil C and N turnover and predict the soil-atmosphere exchange of GHG from tropical forest soil.
In the soil climate sub-model, daily climate data is used to calculate soil temperature, moisture and oxygen profiles.This is done by considering soil properties (texture), and plant and microbial turnover processes of C and N. Forest growth is calculated depending on temperature, water and nitrogen availability.Litter production, water and nitrogen demand of plants and root respiration is linked with the soil and the decomposition sub-model.N-trace gas production, consumption and emission are calculated within the sub-models nitrification and denitrification.
In general, these simulations are carried out by considering soil properties such as texture, OC and N content, mineral soil pH and bulk density (BD), vegetation parameters (aboveground biomass, wood mass, leaf mass and floor mass) and climate data (daily precipitation, minimum and maximum temperature).A detailed description of the above mentioned input parameters is given in subsequent sections.In the present study, no modifications of the model were performed and the model set-up remained

Conclusions References
Tables Figures

Back Close
Full covering the Rwandan territory (Imerzoukene and Van Ranst, 2002;Verdoodt and Van Ranst, 2006a;Verdoodt and Van Ranst, 2006b).Topsoil (0-30 cm) data of sand, silt, clay, gravel, organic carbon (OC), total nitrogen (TN), mineral soil pH, and litter layer pH for our study area was extracted from the database.We selected 147 soil profiles located within the Nyungwe forest.Summary statistics for all selected soil properties are shown in Table 1.Missing values for TN (30 % of the selected profiles) were estimated by regression equation which was based on a statistical analysis of the complete datasets: TN = 0.0994 × OC 0.7178 (R 2 = 0.74, p < 0.05).Since soil bulk density values are largely missing in the original survey, bulk density was derived using specific pedo-transfer functions (PTF) for tropical highland forest soils (Gharahi Ghehi et al., 2012b).

Weather data
No long-term meteorological data from the forest were available and therefore daily weather data for the years 1981-1993 from the nearest three climatological weather stations at an altitude of ∼ 2000 m and an average distance to the forest of < 5 km were used (Minagri andCTB/BTS, 1993a, 1993b).In addition to past thirteen years climate data, we used more recent years (2007)(2008)  Daily weather data were spatially interpolated for the forest using the Thiessen polygon approach in GIS (geographic information system) (ESRI ArcMap 9.3 software).The mean annual temperature and precipitation (1981)(1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993) for the climate polygons were 16.8 • C and 1458 mm, respectively.For the northern region of the Nyungwe forest, decreased amounts in annual precipitation and higher values of mean annual temperature have been observed.The average annual mean temperature and precipitation (2007)(2008) at Uwinka climate station is 14.5 • C and 1824.7 mm, respectively.The rainfall during the recent years was slightly higher than the average rainfall observed Introduction

Conclusions References
Tables Figures

Back Close
Full

Vegetation data
We used the Lund-Potsdam-Jena General Ecosystem Simulator (LPJ-GUESS, Smith et al., 2001) to derive a regional distribution of aboveground biomass estimates for Nyungwe forest.The LPJ-GUESS is an object-oriented, modular framework for modeling the dynamics of ecosystem structure and functioning at scales from plot to global scales, and at varying levels of process detail.The LPJ-GUESS model simulates vegetation dynamics and structure based on the tropical evergreen plant function type (PFT), which represent individual physiological, morphological and bioclimatic (Smith et al., 2001;Sitch et al., 2003).Taking time series of climate data (as mentioned above) and, given constraints of latitude, topography, and Nyungwe soil characteristics (as mentioned above), as input, the LPJ-GUESS model dynamically computed spatially explicit transient vegetation composition in terms of plant functional types, and their associated carbon and water budgets.Aboveground biomass (wood, leaf and floor mass) was extracted from the output of the LPJ-GUESS model and used as input drivers for ForestDNDC-tropica.The fine root mass was estimated as 0.7 × leaf mass (Kiese et al., 2005;Werner et al., 2007a).The spatial distribution of the simulated aboveground biomass is shown in Furthermore, our simulated aboveground biomass up to 400 t C ha −1 were of similar magnitude as the reported aboveground biomass of ∼ 500 t C ha −1 for the northern part of the Nyungwe forest (Nsabimana, 2009).Introduction

Conclusions References
Tables Figures

Back Close
Full

Model runs
In order to predict N 2 O and NO emission from Nyungwe forest, soil, climate and vegetation data were integrated into a GIS database covering Nyungwe forest, with a spatial resolution of 0.025 • × 0.025 • .Many tropical forests are faced with a lack of soil data, thus requiring the generation of parameter maps based on rather limited datasets.The Inverse Distance Weighting (IDW) technique (Myers, 1993), a weighted average interpolator, was used for preparing the spatial distribution of model input parameters for grid cells where no observational data were available.We used a conservative linear weights function of 1 to interpolate the values.An IDW approach was preferred over kriging techniques because little spatial auto-correlation was observed for the selected properties.An example of the spatial distribution of the aboveground biomass and selected soil characteristics is shown in Fig. 1.Since many existing soil datasets for natural forested ecosystems comprise generally less than 30 % of the size compared to our dataset (ISRIC-WISE global soil profile data set; Nsabimana et al., 2008;Kiese et al., 2008;Werner at al., 2007a), we are argue that the spatial map of model input parameters (Fig. 1) is the best possible estimate of the spatial patterns of the topsoil property in tropical forested ecosystems at such a scale today.All input data was formatted in ForestDNDC-tropica compatible formats.Finally, ForestDNDC-tropica was run in a daily time-step for the years 1981-1993 and 2007-2008 on this final vegetation and soil dataset.The model was run for a two years simulation period, using the first year as spin-up period in order to allow model internal C and N pools to equilibrate.The estimated N 2 O emissions are converted to CO 2 equivalents for comparison with other GHG emissions.We used N 2 O 100-yr global warming potential (GWP = 310).

Model validation with measured N-trace gases fluxes
To (1 where EI is extrapolated incubation N 2 O or NO data, x meas is the laboratory measured N 2 O or NO value, day mod is sum of day of year of simulated WFPS values at 40-60 %, 60-80 % and 80-100 % by ForestDNDC-tropica. To further evaluate the model, the Friend beta sensitivity index (β) (Friend et al., 1993) was computed to assess the influence of individual model parameters on model output on a regional scale (e.g., Butterbach-Bahl et al., 2004).Different model parameters were varied one-at-a-time over their plausible range while keeping the others at their nominal value.We used the climate data of the year 1993 and calculated variation of simulated average N 2 O emissions in response to changes in plus-minus 1 for pH, and plus-minus 25 % for BD and in plus-minus 50 % for other input parameters.
Sensitivity index is than defined as: where P 1 and P 0 are the individual input parameter that increased and decreased, respectively; and the β index is calculated from the resulting change in N 2 O emissions Introduction

Conclusions References
Tables Figures

Back Close
Full ).The deviation of the β value from zero would imply a proportional to the sensitivity of a given parameter and the sign of β indicates the direction of correlation (a negative or positive correlation) (Kiese et al., 2005;Werner et al., 2007a).We reported the average β values for all 147 data locations in the historical data set.

Change in historical soil data
The soil properties have been obtained at least three decades ago and may have for 1981-1993 periods, with an average of 2.0 ± 0.87 kg N ha −1 yr −1 .The emission rates for the period 2007-2008 range from 2.5 to 5.0 kg NO-N ha −1 yr −1 , with an average value of 3.8 ± 0.42 kg NO-N ha −1 yr −1 .Significantly lower NO emissions (0.8-2.0 ha −1 yr −1 ) are seen in the center and towards the south of the forest, regions with high soil pH (pH > 4) or a relatively high OC ( > 4 %) content (Fig. 3).

Model validation with measured N-trace gases fluxes
For the 31 sites simulated mean N 2 O and NO emissions were generally < 15 g N 2 O-N ha −1 d −1 .Although most of the laboratory N 2 O and NO agreed well with simulated data, a few sites showed measurements that were three to four times higher (Fig. 5).The high measured N 2 O and NO emission values of > 15 g N ha −1 d −1 (Fig. 5) in laboratory study of Gharahi Ghehi et al. (2012a) were somewhat surprising.They pointed out that chemo-denitrification might be an important production pathway due to high free iron concentration in these acidic soils.
Overall, the simulated N 2 O and NO data showed an underestimation of the emission by ForestDNDC-tropica.Correlation analysis between measured and simulated N 2 O Introduction

Conclusions References
Tables Figures

Back Close
Full and NO fluxes yielded non-significant correlations of r = 0.24, p > 0.05 and r = 0.16, p > 0.05, respectively.However, when excluding high emission values (emissions from the laboratory data sets, we observed a significant correlation between measured and simulated values of N 2 O (r = 0.42, p < 0.05) and NO (r = 0.51, p < 0.05) fluxes (Fig. 5). Figure 6 illustrates the parameters included in the sensitivity analysis by varying the individual parameters by ±1 for pH, ±25 % for BD and ±50 % for other input parameters.This simple model sensitivity test was previously used in several modeling studies, for example Butterbach-Bahl et al. ( 2004 soil data inside the tropical forest (Rwanda database).Specifically with regard to soil OC contents our database comes up with higher values as compared to the Werner et al. ( 2007a) study, as in the latter study soil OC contents of the most prominent soil classes in Africa was 1.38 % (our database: average of 5.5 %).Second, there are also differences in simulated years and origins of climate data.
The average model estimates in our study for N 2 O (3.8 kg N ha −1 yr −1 ) and NO (2.0 kg N ha −1 yr −1 ) for the Nyungwe forest are also supported by field measurements from other studies.On average, the mean annual N 2 O emission from tropical forest in Australia and southwest China was 1-3 kg N ha −1 yr −1 (Kiese et al., 2005;Werner et al., 2007a).Higher N 2 O emissions (2.6 kg N ha −1 yr −1 ) for forests in Kenya were reported by Werner et al. (2007b).In general, measurement of N 2 O fluxes from tropical mountain forests range between 0.01 and 3.75 kg N ha −1 yr −1 (Breuer et al., 2000;Ishizuka et al., 2005;Holtgrieve et al., 2006;Purbopuspito et al., 2006;K öhler et al., 2009) and between 0.03 and 0.4 kg ha −1 yr −1 for NO (based on only one study of Davidson and Kingerlee, 1997).However, NO emissions up to 3 kg ha −1 (for a period of 3 months and not for one year) were reported for rain forests in Queensland, Australia at the onset of the rainy season (Butterbach-Bahl et al., 2004).Compared to other published NO emission data for other tropical rain forest ecosystems, the simulated NO emissions of up to 5 kg ha −1 yr −1 at the Nyungwe forest were high.This can be the effect of low soil pH values as ForestDNDC-tropica considers a chemo-denitrification algorithm for pH values lower than 4 as co-process of NO production only (pH < 4 cover ∼ 92 % of the total forest area).The different rainfall pattern between 1981-1993 and 2007-2008 periods within the study area might be related to shifts in weather patterns due to adjacent land use/land cover change (LULCC) as an important climate forcing (Mahmood et al., 2010) or altitude difference of almost 500 m between the old stations and the Uwinka station which explains at least the differences in temperature.There is also a known but not welldocumented gradual decrease in rainfall from west to east, and since two of the old stations are located east of the forest they are likely underestimating the precipitation in the forest.

Model and model parameter uncertainty
The high NO emissions simulated during wetter years could be a consequence of the rapid mineralization of litter accumulated during the dry period.Specifically at the onset of the rainy season (six months) the effect of the rainfall events on surface litter decomposition is direct and more pronounced as compared to decomposition dynamics of root litter and organic matter in the mineral soil.Our high NO emissions simulated during wetter years most likely are the result of higher NO production in the surface litter layer.Though NO production was also simulated in these years in the mineral soil, this contributed only little to total emissions, since NO consumption in the mineral layer was increasing at the time.We emphasize that the simulated responses of high NO emissions to the rainfall events are in agreement with observation by Butterbach-Bahl et al. (2004).They observed that wetting events at the onset of the rainy season resulted in peak NO emissions cumulating to several kg of NO-N within a few weeks.
For our N 2 O emissions such a clear effect between wetter and dryer years is not visible, since N 2 O production in the model occurs mainly in the mineral layer.Again, increased soil moisture will increase anaerobiosis and simulate further conversion of N 2 O to N 2 , so that total emissions are less affected.
The variation in simulated N 2 O and NO emissions in the Nyungwe forest also reflected the texture and soil physico-chemical properties.For instance, N 2 O emissions exceeding 4 kg ha −1 yr −1 were found in the northwestern part of the forest, which

BGD Introduction Conclusions References
Tables Figures

Back Close
Full is characterized by high clay and OC contents and low pH (< 4).Several authors (N ömmik, 1956, Weier andGillam, 1986;Granli and Bockman, 1994) have shown that low pH decreases the activity of the N 2 O-reductase, thereby increasing production of N 2 O, rather than N 2 from denitrification.For nitrification, it has also been demonstrated that low pH < or = 4 values favor N 2 O production (Sitaula and Bakken, 1993;Martikainen and De Boer, 1993;Kesik et al., 2006).All this is in agreement with field observations of N 2 O emissions for different tropical lowland forest sites in Australia by Kiese and Butterbach-Bahl (2002) (Bremner, 1997).Under acidic conditions chemo-denitrification may be a significant source of NO, N 2 O, N 2 when NO − 2 is reduced to NO, N 2 O or N 2 with Fe (II) as an electron donor (Van Cleemput and Baert, 1984;Cooper et al., 2003).
We have evidence of chemical pathways for NO and N 2 O production in the Nyungwe forest soils (see Supplement for detailed description).Our laboratory experiment argues that abiotic denitrification can be an important source of N 2 O and NO under certain circumstances, i.e. low pH and present of reduced inorganic compounds (e.g.

Fe (II)).
In the present version of ForestDNDC-tropica NO and N 2 O is mainly simulated via microbial nitrification and dentrification pathways.Although chemo-denitrification has been included in ForestDNDC-tropica (soil pH < 4) our study suggests that rates of

Conclusions References
Tables Figures

Back Close
Full chemo-denitrification for NO production in tropical soils may still be underestimated and that it is likely to be necessary to introduce such a mechanism for N 2 O as well.
In addition to pH, bulk density has a significant influence on N 2 O and NO emissions.Increasing bulk density increases the anaerobic zone and increasing the probability for saturated moisture conditions following rainfall.This triggers denitrification and N 2 O emissions and decreases NO emissions (Stange et al., 2000;Kiese et al., 2005).Therefore, in order to further improve N 2 O and NO emissions from tropical forest, improved spatial maps of soil pH and bulk density are needed.In general, soil BD measurements have rarely been made in soil surveys.However, it has been shown that organic content can be used as a sole predictor of topsoil BD in tropical forest soils (Gharahi Ghehi et al., 2012b).

N 2 O and NO emissions: an integrated approach
Although uncertainties remain in our model results we are confident that, given the available driving data, these results give us a good insight into the spatial and temporal variability in N 2 O and NO emissions.A comparative analysis of current with past TN, We acknowledge that these changes could affect our estimates of N 2 O and NO.However, soil structural properties free of anthropogenic disturbance are in general more stable and change little over time.On a regional to global scale, N 2 O emission from e.g.grassland in West/East Africa regions is estimated to be 15.0 Mt CO 2 equivalents (Stehfest and Bouwman, 2006).
Considering our small study area (1113 km 2 ) our estimate with a range of 0.19-0.25 Mt CO 2 equivalents (373-505 t N 2 O-N = 0.19-0.25 Mt CO 2 ) for 1981-2008 confirms the potential importance of tropical forests as a major source of atmospheric N 2 O on a regional, national and even global scale.

Conclusions
Although, there is still a considerable uncertainty associated with our emission estimates, our results are providing a first spatial explicit predictions of N 2 O and NO for tropical forest at this scale.The results are in line with other studies, confirming substantial N 2 O and NO emissions from tropical forests, thereby suggesting that N 2 O and NO losses play a major role in the N cycle of tropical forests.Though there are still several limitations (e.g.clustering of legacy data; chemo-denitrification only considered for NO but not for N 2 O) our results show that state of the art process-based models in combination with site-specific (historical) input data can be used to improve tropical N 2 O and NO emission estimates.
Temporal variability has shown to be sensitive to known processes such as changes in precipitation.Our study further corroborates the notion that soil BD content and pH are the most influential factors driving spatial variations in soil NO and N 2 O emissions for tropical forest soils and should therefore receive priority is monitoring campaigns.However, some abiotic processes that contribute to, but no completely integrate, into our N trace gas modelling effort were found.In particular, chemo-denitrification processes on acidic soils seem to be under represented in the current ForestDNDCtropica model.Introduction

Conclusions References
Tables Figures

Back Close
Full  Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | unchanged as compared to the Werner et al. (2007a) study.A more detailed description for the ForestDNDC-tropica model is given in Werner et al. (2007a).
Discussion Paper | Discussion Paper | Discussion Paper | data measured at the Uwinka climate station (2 • 28 43.3 S, 29 • 12 00 E, 2465 m a.s.l.) located within the Nyungwe forest which has been established on 2007.The climatological data used, consisted of daily precipitation and daily minimum and maximum temperatures.
Discussion Paper | Discussion Paper | Discussion Paper | in most of the years during the period 1981-1993.In the rainy seasons of the years 2007-2008, monthly rainfall was approximately more than 150 mm and higher than during the years 1981-1993 (Fig. 2).
Fig. 1b.The simulated aboveground biomass for the Nyungwe forest (ranging from 74 to 400 t C ha −1 ) were in similar magnitude to those reported by Werner et al. (2007a) (100-225 t C ha −1 for tropical rainforest soils for eastern Africa).
Discussion Paper | Discussion Paper | Discussion Paper | validate the ForestDNDC-tropica model output, N 2 O and NO emission data for soils of the Nyungwe tropical forest by Gharahi Ghehi et al. (2012a) were used.The latter Introduction Discussion Paper | Discussion Paper | Discussion Paper |study carried out a medium-term incubation experiment with soils from 31 locations within the Nyungwe forest.Since the incubation data covered three water filled pore space levels (50, 70 and 90 % WFPS), a weighted mean WFPS average was used here for comparing soil incubation results and model simulations.On the basis of the laboratory simulated WFPS we calculated a weighted mean average of N 2 O and NO emissions for each individual location (percent time at 40-60 %, 60-80 % and 80-100 % WFPS) using simulated WFPS values by ForestDNDC-tropica to extrapolate the incubation data.The extrapolated incubation N 2 O and NO data were calculated using the following formula: EI = [(x meas ) 50 %WFPS × day mod40−60 %WFPS ] + [(x meas ) 70 %WFPS × day mod60−80 %WFPS ] + [(x meas ) 70 %WFPS × day mod80−100 %WFPS ] day mod40−100WFPS Discussion Paper | Discussion Paper | Discussion Paper | changed in time due to logging activities; so changes of historical soil data over time can affect the overall uncertainty of GHG inventories.The possibility of changes of historical soil data over time was shown by selecting 29 locations out of the 147 sites in the Nyungwe forest, for which a soil sampling was performed by Gharahi Ghehi et al. (2012a) during September 2009.We carried out a comparative analysis of simulated values N 2 O and NO using current with past input soil properties to test whether changes of historical soil data would significantly affect model outcome.The simulated N 2 O emission for Nyungwe forest for the period 1981-1993 range from 2.8 to 5.5 kg N ha −1 yr −1 , with an average of 3.8 ± 0.52 kg N ha −1 yr −1 .The emission rates for the period 2007-2008 vary between 2.2 and 4.7 kg N 2 O-N ha −1 yr −1 , with an average value of 3.7 ± 0.40 kg N 2 O-N ha −1 yr −1 .The highest N 2 O emissions (4-5.5 kg N ha −1 yr −1 ) are simulated in regions with a relatively high clay (> 30 %) or OC (> 4 %) content, namely north -northeast and south of the Nyungwe forest (Fig. 3).Lower emissions are estimated for the eastern part of the forest, an area with predominantly sandy soils.Total emissions from the entire Nyungwe forest are estimated Discussion Paper | Discussion Paper | Discussion Paper | at 373-504 t N 2 O-N yr −1 (mean = 439 ± 50 t NO-N yr −1 ) for the period 1981-1993 and 384 t N 2 O-N yr −1 and 489 t N 2 O-N yr −1 for 2007 and 2008, respectively (Fig. 4).The magnitude of simulated NO emission ranges from 0.8 to 5.1 kg N ha −1 yr −1 Discussion Paper | Discussion Paper | Discussion Paper | ),Kiese et al. (2005), andWerner et     al. (2007a).Overall, soil BD and pH were the most sensitive parameters for both N 2 O and NO emissions.The observed sensitivity of N 2 O emission rates for the Nyungwe forest was comparable to other published N 2 O simulation data for tropical rain forest ecosystems using the same model.Kiese et al. (2005) and Werner et al. (2007a) reported a pronounced sensitivity of N 2 O emission rate to changes in pH and BD in tropical forest ecosystems in Australia and global tropical forests, respectively.The sensitivity of N 2 O and NO emissions on leaf mass, temperature, rainfall, OC and clay content variability were lower as compared to changes in soil pH and BD.Overall, the calculated β-values in this study were lower as compared to those in the Kiese et al. (2005) and Werner et al. (2007a), e.g. a β-value of 2.5 for pH in Werner et al. (2007a) compared to 0.67 in our study.These lower β-values probably come from the smaller range of possibly values due to the restricted simulation region.For example Werner et al. (2007a) selected 1000 simulated grid cells out of a global dataset to assess model sensitivity, whereas we determined the sensitivity of modeled N 2 O and NO emissions on variations in input parameters by selecting all 147 locations sites in the Discussion Paper | Discussion Paper | Discussion Paper | 3.3 Change in historical soil data The soil properties have been obtained at least three decades ago and may have changed in time due to logging activities; so changes of historical soil data over time can affect the overall uncertainty of GHG inventories.The possibility of changes of historical soil data over time was shown by selecting 29 locations out of the 147 sites in the Nyungwe forest, for which a soil sampling was performed by Gharahi Ghehi et al. (2012a) during September 2009.We carried out a comparative analysis of simulated values N 2 O and NO using current with past input soil properties to test whether changes of historical soil data would significantly affect model outcome.To the best of our knowledge, simulated and N 2 O and NO emission from the Nyungwe forest represents the first explicit regional N 2 O and NO emission inventory for a tropical forest on the African continent.The estimated annual average N 2 O emission from Nyungwe forest (∼ 3.8 kg N ha −1 yr −1 ) is in line with simulated annual N 2 O emissions from tropical rainforest soils for eastern Africa (∼ 3.5 kg N ha −1 yr −1 ) by Werner et al. (2007a).On average the entire Nyungwe forest emits 439 t of N 2 O-N and 244 t of NO-N per year (1981-1993) and ∼ 384-489 t N 2 O -N and ∼ 489-384 t NO-N for period 2007-2008.Werner et al. (2007a) estimated the N 2 O source strength of tropical forest soils of Africa (assuming a total rainforest area of 3.055 × 10 6 km 2 ) to be 344 000 t N 2 O yr −1 , which is equivalent to 113 t for our area.The latter number is significantly lower than our mean estimate of 439 t N 2 O-N yr −1 (1981-1993).This discrepancy may be due to two principal factors.First, Werner et al. (2007a) used the soil profile data from the ISRIC-WISE soil database and linked these to the global FAO soil map, whereas we used 147 Discussion Paper | Discussion Paper | Discussion Paper | Werner et al. (2007a)  andButterbach-Bahl et al. (2004) pointed out that spatial variations in regional N 2 O and NO emissions were mainly resulting from differences in the spatial distribution of soil and climate characteristics.For example, the two-fold increase of total annual NO emission for 2007-2008 years (384-489 t NO-N yr −1 Discussion Paper | Discussion Paper | Discussion Paper | compared to 1981-1993 years (244 ± 16 t NO-N yr −1 ) indicates that large discrepancy in climate data play a major role in the N dynamics.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | pH, clay and OC content values for a select subset of 29 locations (see Gharahi Ghehi et al., 2012a) displayed significant differences for all soil driving information but the clay values (p > 0.05).This data shows that soil properties have changed over the last three decades.The probable reasons for these changes in the Nuyngwe forest based on personal communication (G.Wallin, August 2012) and several field observations (P.Boeckx, 2005-2012) could be the following: (i) forest fires in eastern part of the forest in 1999, (ii) traditionally more disturbances occurs in the eastern part, although locally some mining and logging activities have been done in the west as well and (iii) disturbance caused by people normally have occurred where the forest have been accessible, thus in the same areas where the most extensive soil sampling has been carried out.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | and Zhao, Z. C.: Global climate projections, in: Climate Change 2007, The Physical Basis, edited by: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K. B., Tignor, M., and Miller, H. L., Contribution of Working Group I to Fourth Assessment Report of IPCC on Climate Change, Cambridge University Press, Cambridge, UK/NY, USA, 2007.Minagri et CTB/BTS.: Interpr étation des donn ées climatiques du Rwanda d'apr ès Pa-Discussion Paper | Discussion Paper | Discussion Paper |
which revealed that low pH was a crucial factor driving high N 2 O emissions.The high NO emissions were found in the center and towards the north and northeastern of the forest, which are growing on soil characterized by low soil pH and relatively high OC or high clay contents.Moreover, all in the study of Serca et al. (1994) the combined influence of low pH and high OC was used to estimate the NO production potential of forest soils in the Mayombe forest.Furthermore, Gharahi Ghehi et al. (2012a) suggest that high N 2 O and NO emissions for some sites in the Nyungwe forest are possibly due to chemo-denitrification processes.Chemo-denitrification is thought to occur when nitrite (NO