Nitrogen availability links forest productivity, soil nitrous oxide and nitric oxide fluxes of a tropical montane forest in southern Ecuador

Authors


Abstract

[1] Tropical forests are important sources of the greenhouse gas nitrous oxide (N2O) and of nitric oxide (NO), a precursor of ozone. In tropical montane forests nitrogen limitation is common which affects both soil N2O and NO fluxes and forest productivity. Here we present evidence that forest productivity and N-oxide (N2O + NO) fluxes are linked through N availability along elevation and topographic gradients in tropical montane forests. We measured N-oxide fluxes, several indices of N availability, and forest productivity along an elevation gradient from 1000 m to 3000 m and along topographic gradients. Organic layer thickness of the soils increased and N availability decreased with increasing elevation and along the topographic gradient from the lower slope position to the ridges. Annual N2O fluxes ranged from −0.53 μg(N)m−2h−1 to 14.54 μg(N)m−2h−1 while NO fluxes ranged from −0.02 μg(N)m−2h−1 to 1.13 μg(N)m−2h−1. Both N-oxide fluxes and forest productivity increased with increasing N availability and showed close positive correlations with indices of N availability (C/N ratio andδ 15N signature of litterfall). We interpret the close correlations of N-oxide fluxes with total litterfall and tree basal area increment as evidence that N availability links N-oxide fluxes and forest productivity. This opens the possibility to include forest productivity as co-variable in predictions of N-oxide fluxes in nitrogen limited tropical montane forests. Especially increment of tree basal area was a promising proxy to predict soil N-oxide fluxes in these N limited ecosystems, possibly because it better reflects long-term forest productivity than total litterfall.

1. Introduction

[2] Tropical forest soils are considered the largest natural source of the greenhouse gas nitrous oxide (N2O) with a source strength of 3.0 Tg N yr−1 [Werner et al., 2007]. Additionally, they can produce considerable amounts of nitric oxide (NO), which plays a crucial role regulating troposphere ozone concentrations [Crutzen, 1979]. Estimates of the source strengths of N2O and NO are based on field studies that are still heavily biased toward tropical lowland forests. The few studies on N2O and NO fluxes carried out in tropical montane forests show that the trace gas emissions decrease with increasing elevation [Hall et al., 2004; Purbopuspito et al., 2006]. Measurements of nitrogen oxide fluxes from tropical montane forests range between 0.01 and 3.75 kg (N) ha−1yr−1 for N2O [Breuer et al., 2000; Holtgrieve et al., 2006; Ishizuka et al., 2005; Köhler et al., 2009; Purbopuspito et al., 2006] and between 0.03 and 0.4 kg (N) ha−1yr−1 for NO, an estimate that is based on two studies only [Davidson and Kingerlee, 1997]. The relative wide range of N-oxide fluxes reported may reflect the complex pattern of soil properties, nutrient and soil water availability caused by topography, drainage characteristics and disturbance history that overlay elevation gradients in montane regions. Such small-scale heterogeneity potentially affects fluxes of N2O and NO and forest productivity. Although studies in temperate regions show strong topography effects on N-oxide emissions [Corre et al., 1996; Meixner and Eugster, 1999], most of the studies done in tropical montane forests do not include information on topographic positions.

[3] Direct measurements of N2O and NO fluxes in the field are laborious and challenging, which is why several studies have been carried out to test proxies that may have the potential to predict N-oxide fluxes [Davidson et al., 2000; Ishizuka et al., 2005; Keller and Reiners, 1994]. The conceptual “hole-in-the-pipe model” (HIP) explains the main controls on N2O and NO fluxes. Rates of N cycling and N availability control the total amount of N2O + NO produced, whereas soil moisture regulates the relative importance of the two gases by influencing the main N2O and NO producing processes nitrification and denitrification through the gas diffusion rate of the soil [Davidson et al., 2000; Firestone and Davidson, 1989]. Some indices of N cycling rates and N availability (e.g., soil mineral N, net N mineralization and net N nitrification) are temporally and spatially highly variable, which limits the applicability of these indices, despite reports of positive correlations with N oxide fluxes [Veldkamp et al., 1999]. C/N ratios and δ 15N signatures of litterfall are more integrative indices of the long-term N cycle [Kahmen et al., 2008; Robinson, 2001] making them more promising predictors of N2O and NO fluxes [Davidson et al., 2000; Purbopuspito et al., 2006].

[4] In contrast to many lowland forests, productivity of tropical montane forests is limited by N, or nitrogen and phosphorous (P) together [Tanner et al., 1998]. This appears to be mainly related to the soil development where forest productivity on young soils (common in tropical montane forests) tends to be N-limited, whereas forest productivity on heavily weathered soils (common in tropical lowland forests) tends to be P-limited [Walker and Syers, 1976]. Further evidence of N limitation comes from slow net and gross N mineralization rates in montane forest soils compared to lowland forest soils [Arnold et al., 2009; Marrs et al., 1988] and lower δ 15N signatures of leaves and litter [Martinelli et al., 1999]. The microbial processes of nitrification (responsible for NO, N2O and NO3 production) and denitrification (responsible for NO, N2O and N2 production) discriminate against 15N, thus leading to 15N enrichment of soil and plant material in the ecosystem [Houlton and Bai, 2009]. An N-limited ecosystem, with small N losses through leaching and gas emissions therefore displays lowerδ15N signatures compared to ecosystems where more N is lost relative to N cycling rates [Amundson et al., 2003; Corre et al., 2010; Houlton et al., 2006]. Finally, low N concentration and high C/N ratios in leaves and litter of montane forests indicate N limitation of tree growth [Tanner et al., 1998].

[5] In summary, there is evidence that in tropical montane forests forest productivity is limited by N and that N-oxide fluxes depend on N cycling rates and N availability [Corre et al., 2010]. In the present study our goal was to analyze whether forest productivity and N-oxide fluxes are linked through N availability along elevation and topographic gradients in tropical montane forests. Our hypotheses were the following: 1) N-oxide fluxes increase with increasing N availability, 2) forest productivity increases with increasing N availability in N-limited ecosystems, and 3) forest productivity shows good correlations with N-oxide fluxes in such N-limited ecosystems. We tested our hypotheses by measuring N2O and NO fluxes, several indices of N availability, and forest productivity during one year along elevation and topographic gradients in natural tropical montane forests in southern Ecuador. Our results show that both N-oxide fluxes and forest productivity are linked through N availability and we suggest that predictions of soil N2O + NO fluxes of N-limited ecosystems can be improved using forest productivity as a co-variable.

2. Material and Methods

2.1. Study Area

[6] The study area is located in the Cordillera del Consuelo, which forms part of the eastern chain of the Andes in southern Ecuador. Three study sites were selected along an elevation gradient (Table 1): 990–1200 m a.s.l. (Bombuscaro, 04° 06′S, 78° 58′W, hereafter called the ‘1000 m site’), 1800–2100 m a.s.l. (San Francisco, 03° 58′S, 79° 04′W, hereafter called the ‘2000 m site’) and 2800–3000 m a.s.l. (Cajanuma, 04° 06′S, 79° 11′W, hereafter called the ‘3000 m site’). The 1000 m site is located close to the city of Zamora, Zamora-Chinchipe province within the Podocarpus National Park (PNP). Natural vegetation consists of “premontane rain forest” [Homeier et al., 2008, p. 89] and most important tree families are Moraceae, Rubiaceae and Melastomataceae. The 2000 m site is situated between the cities of Loja and Zamora within the Reserva San Francisco (RSF), a private reserve that borders the PNP. Vegetation can be classified as “lower montane rain forest” [Homeier et al., 2008]. Most important tree families of the 2000 m site are Lauraceae, Euphorbiaceae, Melastomataceae and Rubiaceae. The 3000 m site is located south of the city of Loja within the PNP. Forest is classified as “upper montane rain forest” [Homeier et al., 2008]. Cunoniaceae, Clusiaceae and Melastomataceae are the most important tree families of this site. Stand height decreased with elevation and from the lower slope positions to the ridge positions within every site. The annual mean temperature decreased from 19.4°C at the 1000 m site to 9.4°C at the 3000 m site, annual precipitation increased from 2230 mm at 1000 m, and 1950 mm at 2000 m, to 4500 mm at 3000 m (Table 1) [Moser et al., 2007]. Rainfall showed little seasonality [Emck, 2007]. For most of the study area, Paleozoic metamorphic schists and sandstones with some quartz veins form the parent material for soil development. Only at the 1000 m site soil parent material of half of the plots consists of deeply weathered granitic rock of the Jurassic Zamora granitoide formation [Litherland et al., 1994].

Table 1. Site Characteristics of the Topographic Positions Across the Elevation Gradient
 1000 m2000 m3000 m
Lower SlopeMidslopeRidgeLower SlopeMidslopeRidgeLower SlopeMidslopeRidge
  • a

    Moser et al. [2007].

  • b

    Means (±SE, n = 6) within rows followed by different letter indicate significant differences among topographic positions at each elevation (lowercase letters) and differences among elevations at each topographic position (capital letters) (one-way ANOVA with Fisher's LSD test atP ≤ 0.05).

Height a.s.l. (m)990–12001800–21002800–3000
Air temperature (°C)a19.415.79.4
Precipitation (mm yr−1)a223019504500
Gravimetric moisture (g g−1)b1.0 (0.2)aA0.9 (0.2)aA1.8 (0.2)bA1.7 (0.3)aB2.1 (0.3)aB4.3 (0.1)bB2.1 (0.2)aB2.9 (0.3)bC4.5 (0.2)cB
Organic layer thickness (cm)b2.5 (0.7)2.6 (1.0)A6.5 (1.9)A4.0 (0.9)a12.2 (3.1)aB24.0 (2.9)cB6.6 (0.7)a14.2 (1.6)bB22.2 (3.4)cB
Organic layer         
pH (H2O)b4.9 (0.5)a4.8 (0.6)a3.7 (0.1)b4.6 (0.2)a4.8 (0.5)a3.4 (0.1)b4.0 (0.3)3.7 (0.1)3.6 (0.1)
C/Nb18.5 (0.7)A18.6 (1.5)A20.5 (0.7)A15.7 (0.8)aB15.6 (0.6)aA21.1 (0.7)bA20.5 (0.8)aA23.9 (1.4)bB26.7 (1.3)cB
δ 15N (‰)b3.7 (0.5)A2.1 (0.7)A1.9 (0.5)4.0 (0.3)aA3.0 (0.2)aA1.0 (0.4)b1.4 (0.2)aB0.2 (0.3)bB0.6 (0.2)ab
Total P (mg (P) g−1)b1.1 (0.1)1.2 (0.1)A0.6 (0.02)A1.2 (0.1)a1.1 (0.2)aA0.5 (0.0)bAB0.8 (0.1)aA0.6 (0.1)abB0.4 (0.02)bB
Mineral soil layer (Ah)         
pH (H2O)b5.0 (0.3)A5.1 (0.4)A4.2 (0.1)4.6 (0.2)aA4.7 (0.4)aA3.7 (0.1)3.7 (0.1)B3.7 (0.0)B3.7 (0.1)
Base Saturation (%)b65.9 (19.6)aA59.9 (18.2)aA2.2 (0.4)b46.3 (14.9)A42.5 (17.3)A3.3 (0.7)8.2 (0.5)B6.9 (1.3)B8.3 (1.5)
C/Nb16.3 (0.6)aA15.0 (0.7)a21.3 (1.2)bB13.2 (1.1)B14.1 (1.5)17.0 (0.8)A14.2 (0.2)aAB18.0 (1.1)b22.3 (1.0)cB
δ 15N (‰)b5.8 (0.1)A4.7(0.3)A5.2 (0.1)5.9 (0.3)A5.1 (0.2)A5.2 (0.2)4.1 (0.3)B3.4 (0.3)B4.8 (0.6)
Total P (mg (P) g−1) b0.7 (0.2)0.8 (0.2)A0.6 (0.1)A0.5 (0.1)a0.7 (0.1)aA0.2 (0.0)bB0.3 (0.0)0.2 (0.0)B0.1 (0.0)B
Sand (%)42.4 (8.6)A46.9 (8.9)33.0 (7.3)26.7 (2.1)AB25.8 (5.3)23.9 (2.9)21.8 (3.9)B32.1 (6.7)36.0 (4.8)
Silt (%)b30.6 (7.5)A35.8 (8.2)35.6 (5.8)A55.3 (3.0)B53.1 (3.8)60.9 (1.8)B57.5 (0.3)B49.7 (2.5)50.0 (2.0)AB
Clay (%)b26.9 (2.6)a17.3 (2.3)b31.5 (2.1)aA18.1 (2.1)21.1 (2.1)15.2 (2.1)B20.7 (3.6)18.2 (4.3)14.1 (3.3)B

2.2. Experimental Design

[7] We installed 54 plots (20 × 20 m) at three elevation levels (1000 m, 2000 m, and 3000 m) forming an elevation transect from 1000 m to 3000 m. Eighteen plots per elevation were distributed over the topographic positions lower slope, midslope and ridge, with six replicates each. Vertical deviation of plots from the pre-determined levels was never more than 200 m. All plots were covered by mature, closed-canopy forest that is representative for the respective topographic position. Plot size was small enough to ensure homogeneity of environmental factors and forest structure. Each plot contained six litter traps and four chamber bases, which we placed in a random design along two orthogonal transects. For the calculations, we corrected plot size for inclination.

2.3. Soil Characterization

[8] Twelve to sixteen soil profiles per elevation were dug until the underlying parent material was reached (0.5–1.5 m depth). We opened the profiles outside the boundaries of the plots under closed canopy and undisturbed, representative vegetation. Bulk density was determined for mineral soil using steel cores (250 cm3) and for the organic layer by measuring the thickness and dry weight of an area of 400 cm2 [Blake and Hartge, 1986; Schlichting et al., 1995]. We expressed moisture content as gravimetric moisture content and as water-filled pore space (WFPS), assuming a particle density of 2.65 g cm−3 for mineral soil [Linn and Doran, 1984] and of 1.4 g cm−3 for the organic layer [Breuer et al., 2002]. Disturbed soil samples for chemical analysis were taken from the organic layer (as a pooled sample covering the whole depth of the organic layer) and from each mineral soil horizon. Oven-dried samples (45°C for three days) were sieved and stored in plastic bags until air transport to Germany. All samples were analyzed for total carbon, nitrogen, phosphorous, total base cations and pH (H2O). In addition, effective cation exchange capacity (ECEC) was measured for the mineral soil horizons. Total C and N were determined using ground samples and a CNS Elemental Analyzer (Elementar Vario EL, Hanau, Germany). Total P and total base cations were measured using Inductively Coupled Plasma-Atomic Emission Spectrometer (ICP-AES; Spectro Analytical Instruments, Kleve, Germany) after pressure digestion in 65% HNO3 [Heinrichs et al., 1986]. ECEC was determined from sieved samples by measuring cations in percolates of soils with unbuffered 1 M NH4Cl using ICP-AES [König and Fortmann, 1996]. Base saturation was calculated as the percentage of exchangeable base cations (Na, K, Ca and, Mg) of all exchangeable cations. Soil pH was measured in a suspension of soil in H2O with a 1:1 ratio for sieved mineral soil and a 1:10 ratio for material from the organic layer. Texture was determined by sedimentary fractionation of the clay fraction (25°C, 21 h, 30 cm fall height) following the Atterberg method after wet sieving of the sand and coarse silt fraction (630 μm, 200 μm, 63 μm and, 20 μm) and after destruction of organic matter with 30% H2O2and Fe oxides with 4% Na-Dithionite-Citrate solution [Schlichting et al., 1995].

[9] We measured δ15N signatures on composite soil samples consisting of four subsamples taken with an auger within 1 m distance of the chamber bases. Depending on the depth of the organic layer, we took samples from organic layer (0–10 cm depth and if present also 10–20 cm depth) and three depths of the mineral soil (0–10 cm, 10–20 cm, 20–30 cm depth). Samples were dried at 45°C for three days, sieved and stored in plastic bags until analysis in Germany. Ground samples were analyzed for δ15N signature using an elemental C and N analyzer (NA1500, CE - Instruments, Rodano, Milano, Italy) and an isotope ratio mass spectrometer (Delta V Plus, Thermofisher, Bremen, Germany). The isotope signatures were expressed asδ 15N (‰):

display math

where Rsample = 15N/14N is the ratio of the sample and Rstd = 15N/14N is the ratio of the reference (atmospheric N2) [Martinelli et al., 1999].

2.4. Forest Structure and Productivity

[10] We approximated forest productivity by annual tree basal area increment and annual total litterfall. In the study plots we made an inventory of all trees with stems ≥0.1 m diameter at breast height (dbh) and their basal area was calculated. After approximately one year we measured all trees again and we calculated the increment of plot basal area as the sum of the individual tree increments per plot over one year (diameter difference between first and second measurement was calculated for one year).

[11] Total litterfall was collected monthly (bi-weekly at 1000 m) from six litter traps (60 × 60 cm) per plot. Total litterfall, including leaves, twigs, epiphytes, flowers and fruits, was dried at 60°C and weighed. Total litterfall of December 2008 was collected, ground and sent to Germany for analysis. Litter samples of the six litter traps were proportionally pooled for each plot and analyzed for total C and N, total P, total base cations andδ 15N signatures as described above for soil samples.

2.5. N2O and NO Flux Measurements

[12] We measured N2O fluxes from static closed chambers with a base area of 0.044 m2. Six months prior to the first measurements, we installed chamber bases, inserting them approximately 0.02 m into the soil. The chamber bases were not moved until the end of the sampling period. Chambers had a mean volume of 13 L; lids were equipped with two Luer lock valves for sampling and pressure compensation during sampling. Gas samples were taken at 2, 14, 26, and 38 min following chamber closure and stored in pre-evacuated glass containers (60 ml). To assure that tubing and glass containers were evacuated and containers were filled to atmospheric pressure, pressure was checked with a manometer during sampling. Gas samples were analyzed using a gas chromatograph (Shimadzu GC-14B, Duisburg, Germany) equipped with an electron capture detector (ECD) combined with an autosampler [Loftfield et al., 1997], which was located in a laboratory near our field sites in Ecuador. We analyzed gas samples typically within 48 h following sampling. Gas concentrations were calculated by comparing integrated peak areas of samples with four standard gases (320, 501, 1001, and 3003 ppb N2O; Deuste Steiniger GmbH, Mühlhausen, Germany). N2O measurements were conducted bi-monthly over a period of one year (May 2008–May 2009).

[13] Nitric oxide (NO) was measured in the field using open dynamic chambers, which were closed for six minutes using the same chamber bases as for N2O sampling. Analysis of NO was conducted using a Drummond LMA-3D NO2 Analyzer with chemiluminescence detector (Drummond Technology, Ontario, Canada). NO was analyzed following oxidation to NO2 using a CrO3 catalyst and reaction with a Luminol solution. As the CrO3catalyst is sensitive to relative humidity the gas stream was monitored with a portable relative humidity meter (GFTH 95, Greisinger electronic GmbH, Regenstauf, Germany) and the relative humidity was held below a value of 40% by mixing the air from the chamber with make-up air that was dried using silica gel. Flow rate through the chamber was approximately 1.5 L min−1. The detector was calibrated in the laboratory before field measurement using a standard gas (3000 ppb NO; Deuste Steiniger GmbH, Mühlhausen, Germany) which was diluted using make-up air to the range of concentrations measured in the field. Detector signal and time were recorded every five seconds with a CR800 Series data logger (Campbell Scientific, Utah, USA). NO measurements were carried out four times within the same year as the N2O measurements. All gas fluxes were calculated from the linear increase of gas concentration of the chamber air over time, multiplied by the density of air and the ratio of chamber volume to soil surface area [Köhler et al., 2009]. Dilution of the NO concentration in the chamber by outside airflow through the chamber is negligible during the initial linear part of the concentration increase [Bakwin et al., 1990]. Zero fluxes were included. Annual gas fluxes were calculated from the average of the monthly measurements expressed on a yearly basis.

2.6. Soil Mineral N, Net N-Mineralization and Net Nitrification

[14] Soil mineral N was determined on a composite soil sample, which consisted of four subsamples per plot. We collected these samples from the top 0.05 m of the surface (organic layer or mineral soil), within one meter distance of the chamber bases concomitantly to every gas sampling. In the field, part of the composite sample was added to a polyethylene bottle containing 150 ml of 0.5 molar potassium sulfate (K2SO4) solution. The remaining sample was stored in plastic bags for determination of gravimetric moisture content in the laboratory after drying for 24 h at 105°C. Extraction of soil mineral N was finished within 12 h after field sampling. Extracts were filtered through filter papers (4 μm nominal pore size) after one hour of agitation, and frozen immediately after a drop of chloroform was added. Samples remained frozen during transport by air to Germany, where analysis was conducted. Ammonium (NH4+) and nitrate (NO3) concentrations of the extracts were analyzed using a continuous flow injection colorimeter (Cenco/Skalar Instruments, Breda, Netherlands) [Arnold et al., 2009].

[15] We determined net rates of soil N mineralization and nitrification once a year. To determine initial NH4+ and NO3 levels (T0) four soil samples were taken per plot. A subsample of the composite plot sample was extracted in the field, as described above. The remaining subsample was put into a plastic bag, and inserted into the soil for a ten day incubation period (“buried bag method,” [Hart et al., 1994]). After incubation, soil mineral N was extracted as described to obtain T1 − NH4+ and NO3 concentrations. Net N mineralization and nitrification rates were calculated from the difference in mineral N concentrations between T1 − and T0 [Hart et al., 1994].

2.7. Statistical Analyses

[16] Statistical analysis was carried out on the plot mean N2O and NO fluxes (average of four chambers). Using the N2O flux data, we first tested the spatial independence of the six plots for each topographic position, and these plots were spatially independent based on the rank version of von Neumann's ratio test [Bartels, 1982]. For testing the effect of elevation and topography and of continuous variables in time series data (i.e., bi-monthly measurements of N2O, NO fluxes and soil mineral N), we applied linear mixed effects models (lme) with elevation and topographic position, or the continuous variable as fixed effects, and replicate plots nested in time as random effect. We used lme for time series data because they account for temporal correlation among observations on the same experimental unit [Piepho et al., 2004]. Models describing time series data were tested for autocorrelation effects by including a first order autoregressive structure and for heteroscedasticity of residual variance by including variance functions [Bliese and Ployhart, 2002]. For testing the elevation and topography effects on data that were measured only once (i.e., soil characteristics, vegetation parameters and net rates of soil N transformation), we used two-way analysis of variance (ANOVA) with Fisher LSD test. For testing the influence of soil N transformation rates on N-oxide fluxes, we used the average N-oxide fluxes across a year for each plot. The average N2O and N2O + NO fluxes were first log-transformed to correct for the non-normal distribution of the data set. The relationships among net N mineralization, net nitrification, N2O, NO and N2O + NO fluxes were assessed by lme, which includes different error variances of the different spatial scales of the hierarchical design [Crawley, 2009] and hence showed better fit (P < 0.05) than the general least square model (gls). Additionally, lme models are unaffected by randomly missing data [Piepho et al., 2004]. The relationships among soil characteristics, vegetation parameters and annual N-oxide fluxes were assessed by linear regression on the means of six plots for each topographic position (n = 9; 3 topographic positions of 3 elevations). Linear regression was used because lme did not improve the fit of the gls model on the 9-point database (P < 0.60). Means with ±1 standard error are given in the text. Effects were accepted as statistically significant ifP ≤ 0.05. Analyses were conducted using R version 2.10.1 (R Development Core Team, 2010, http://www.R-project.org/) and Statistica8.

3. Results

3.1. Soil Characteristics

[17] Organic layer thickness increased with elevation and from the lower slope position to the ridge positions (P < 0.001 for elevation and topography effect). Spatial variation in organic layer thickness was high within short distances ranging from 0.00 m to 0.18 m at the 1000 m site, and from 0.08 m to 0.55 m on the ridges of the 2000 m and 3000 m sites. Soils at lower slope positions, at midslope positions and on well drained ridge positions of the 1000 m and 2000 m sites were Eutric or Dystric Cambisols [IUSS Working Group WRB, 2006] (Typic Eutropepts and Typic Dystropepts [Soil Survey Staff, 1975]). Soils at the ridge position of the 1000 m site holding thicker organic layers were Humic Cambisols (Typic Humitropept). On most ridge positions of the 2000 m and 3000 m sites, water logging lead to the development of Histic Planosols and Stagnic Histosols (Typic Albaquults). At the 3000 m site, Stagnic Cambisols and Planosols (Aquic Dystropept, Typic Albaquults) dominated at the lower and midslope positions. Table 1 summarizes soil characteristics of the different elevations and topographic positions.

3.2. Forest Structure and Productivity

[18] Both elevation and topography influenced dbh (P = 0.03 for elevation effect and P < 0.001 for topography effect) and stem density (P = 0.01 for elevation effect, P = 0.001 for topography effect and the interaction), whereas tree basal area only differed with topographic position (P = 0.02). Mean dbh decreased and stem density increased with increasing elevation. Along the topographic gradient, mean dbh decreased from the lower slope position to the ridges whereas stem density increased slightly (Table 2).

Table 2. Forest Structure, Productivity, and Total Litterfall Characteristics of the Topographic Positions Across the Elevation Gradient
 1000 m2000 m3000 m
Lower SlopeMidslopeRidgeLower SlopeMidslopeRidgeLower SlopeMidslopeRidge
  • a

    Means (±SE, n = 6) within rows followed by different letter indicate significant differences among topographic positions at each elevation (lowercase letters) and differences among elevations at each topographic position (capital letters) (one-way ANOVA with Fisher's LSD test atP ≤ 0.05).

Structure and productivity
Mean dbh (≥10 cm)a22.6 (1.4)22.6 (1.9)A18.7 (1.4)A23.8 (2.6)a19.0 (0.6)bAB16.8 (0.6)bAB24.0 (1.8)a15.9 (0.3)bB14.7 (0.7)bB
Stem density (400 m2)a32.8 (3.5)36.0 (3.0)22.6 (7.4)A30.4 (3.6)a45.4 (2.8)b45.4 (3.6)bAB34.4 (6.8)49.3 (7.3)56.7 (4.8)B
Basal area (m2 ha−1)a44.8 (5.0)57.2 (15.0)20.3 (7.2)48.9 (8.6)40.4 (4.3)29.2 (3.0)44.8 (4.6)28.1 (4.2)28.6 (4.2)
Total litter production (Mg ha−1yr−1)a7.0 (0.3)aA8.6 (0.5)bA6.4 (0.5)aA9.5 (0.5)aB9.8 (0.6)aB7.1 (0.3)bA4.7 (0.2)C4.8 (0.5)C4.5 (0.2)B
Basal area increment (m2 ha−1yr−1)a0.6 (0.1)AB0.8 (0.2)A0.7 (0.2)A0.9 (0.2)aB0.8 (0.1)aA0.4 (0.1)bAB0.4 (0.03)A0.3 (0.1)B0.3 (0.1)B
 
Total Litter
N (mg g−1)a16.5 (0.4)A15.9 (0.6)A15.3 (1.2)A21.1 (1.7)aB19.3 (1.2)aB13.8 (0.7)bA12.0 (0.4)aC9.9 (0.8)aC7.1 (0.3)bB
N (g m−2)a11.6 (0.8)A13.7 (0.8)A10.1 (1.5)A19.7 (1.4)aB19.2 (2.2)aB9.7 (0.6)bA5.7 (0.3)C4.7 (0.5)C3.1 (0.2)B
C/Na31.7 (1.0)A32.6 (1.4)A35.8 (2.9)A24.8 (2.4)aA27.0 (2.2)aA39.6 (2.2)bA43.8 (1.8)aB55.5 (5.0)bB77.9 (3.1)cB
δ 15N (‰)a1.6 (0.2)A1.1 (0.4)A1.2 (0.4)A3.4 (0.4)aB1.8 (0.3)bA−1.3 (0.4)cB0.3(0.2)aC−1.4 (0.7)bB−2.6 (0.5)bC
P (mg (P) g−1)a0.8 (0.1)A0.8 (0.1)A0.7 (0.1)A1.3 (0.1)aB1.2 (0.2)aB0.6 (0.0)bAB0.6 (0.0)A0.4 (0.0)C0.3 (0.1)B
P (g m−2)a0.6 (0.1)aA0.7 (0.1)aA0.5 (0.1)bA1.2 (0.1)aB1.2 (0.3)aB0.4 (0.04)bA0.3 (0.02)aC0.2 (0.02)bC0.2 (0.03)bB

[19] Both elevation and topography affected annual total litter production and annual tree basal area increment (P < 0.001 for annual litter production, P < 0.001 for tree basal area increment and elevation and P < 0.05 for tree basal area increment and topography). Litter production and tree basal area increment were highest at the lower slope position of the 2000 m site and lowest at all positions at 3000 m (Table 2). We found strong positive correlations between annual litter production and N concentrations in litterfall (P < 0.001, R2 = 0.82) and tree basal area increment and N concentrations in litterfall (P < 0.001, R2 = 0.83) (Figures 1a and 1b). Furthermore, several indices of N availability, e.g., C/N ratio and δ 15N signatures of litterfall displayed strong correlations with litter production (P < 0.01, R2 = 0.62 for C/N ratio (Figure 2a) and P = 0.02, R2 = 0.54 for δ 15N signatures of litterfall) and tree basal area increment (P < 0.01, R2 = 0.70 for C/N ratio (Figure 2b) and P < 0.001, R2 = 0.81 for δ 15N signatures of litterfall).

Figure 1.

Linear regression of litterfall N and (a) total litter production (Y = 0.77 + 0.42 *X, P < 0.001, R2 = 0.82), and (b) tree basal area increment (Y = −0.14 + 0.05 * X, P < 0.001, R2= 0.83) of the topographic positions across the elevation gradient. Symbols used for data are point-down triangles, lower slope; circles, midslope; triangles, ridge; black, 1000 m; gray, 2000 m; open, 3000 m.

Figure 2.

Linear regression of C/N ratio of litterfall and (a) total litter production (Y = 11.0–0.09 * X, P < 0.01, R2 = 0.62), and (b) tree basal area increment (Y = 1.06 − 0.01 * X, P < 0.01, R2= 0.70) of the topographic positions across the elevation gradient. Symbols used for data are point-down triangles, lower slope; circles, midslope; triangles, ridge; black, 1000 m; gray, 2000 m; open, 3000 m.

3.3. N-Oxide Fluxes and Soil Moisture

[20] Nitrous oxide fluxes did not display clear seasonality at any of the sites, probably because of the lack of a pronounced dry season (Figure S1 in the auxiliary material). Elevation as well as topography influenced N2O fluxes (P < 0.001 for elevation and topography effect) and NO fluxes (P < 0.001 for elevation and P = 0.02 for topography). N2O and NO fluxes were highest at the lower slope positions of the 2000 m site and lowest at the 3000 m site (Table 3). Topography affected N2O fluxes at 2000 m (P < 0.001) and NO at 3000 m (P = 0.02). In both cases gas fluxes decreased toward the ridge positions. Ridge positions at the 3000 m site showed negative annual N2O fluxes and zero NO fluxes, while soils at the lower slope and midslope positions emitted N2O and NO.

Table 3. Average N-Oxide Fluxes (±SE) (μg (N) m−2 h−1) of the Topographic Positions Across the Elevation Gradient
 1000 m2000 m3000 m
Lower SlopeMidslopeRidgeLower SlopeMidslopeRidgeLower SlopeMidslopeRidge
  • a

    Means (±SE, n = 6) within rows followed by different letter indicate significant differences among topographic positions at each elevation (lowercase letters) and differences among elevations at each topographic position (capital letters). (Linear mixed effects models with Fisher's LSD test at P ≤ 0.05.)

N2Oa1.90 (1.38)A3.53 (1.13)4.14 (1.56)14.54 (2.13)aB3.68 (1.46)b1.15 (1.48)b1.63 (1.14)A1.34 (1.43)−0.53 (0.94)
NOa0.75(0.10)1.11 (0.32)A0.72 (0.31)A1.11 (0.36)1.13 (0.33)A0.14 (0.08)B0.19 (0.06)a0.06 (0.04)abB−0.02 (0.03)bB
N2O + NOa2.66 (1.32)A4.63 (1.36)4.85 (1.77)A15.66 (2.27)aB4.82 (1.50)b1.28 (1.51)cB1.82 (1.09)C1.41 (1.46)−0.54 (0.92)C

[21] At all elevations N2O fluxes exceeded NO fluxes. We did not calculate a N2O/NO ratio because negative N2O fluxes occurred and NO fluxes were close to zero. Gravimetric moisture content increased with elevation and from the lower slope positions to the ridges (P < 0.001 for the influence of elevation and topography effect). Throughout the sites gravimetric moisture content did not correlate with N2O + NO fluxes (P = 0.06, R2 = 0.33) or the ratio of N2O/ N oxide emissions (P = 0.08, R2 = 0.29). Excluding the elevation effect, gravimetric moisture content did not influence N2O + NO fluxes within either of the sites. We did not use WFPS as index for soil moisture content, because a comparison of calculated WFPS of organic layers and mineral soil was not possible.

3.4. Indices of N Availability and Their Control on N-Oxide Fluxes

[22] Elevation affected mineral N content of the topsoil (P < 0.001 for NH4+ and NO3). We found declining mineral N concentrations with increasing elevation. Topography did not have significant influence on mineral N content of the soil, with the exception of the 2000 m site (P < 0.001). At all sites, extractable NH4+ was the dominant form of mineral N and in 60% of the extracts NO3 was below detection limit. At the 3000 m site, NO3 was below detection limit in 90% of the extracts (Figure S2 in the auxiliary material). Elevation had stronger influence on net N mineralization rate (P < 0.001) and net nitrification rate (P = 0.01) than topography (P = 0.05 for net N mineralization and net nitrification). Both net N mineralization and net nitrification rates were declining with increasing elevation, whereas there was no clear trend along the topographic gradient of either elevation (Table 4). Throughout all sites and within each site, bi-monthly measured mineral N contents did not significantly influence N2O and NO fluxes. Net N mineralization and net nitrification rates did not affect annual means of the N2O, NO or N2O + NO fluxes.

Table 4. Average Soil Mineral N, Net N Mineralization and Net Nitrification Rates of the Topographic Positions Across the Elevation Gradient
 Unit1000 m2000 m3000 m
Lower SlopeMidslopeRidgeLower SlopeMidslopeRidgeLower SlopeMidslopeRidge
  • a

    Means (±SE, n = 6) within rows followed by different letter indicate significant differences among topographic positions at each elevation (lowercase letters) and differences among elevations at each topographic position (capital letters). (Linear mixed effects models with Fisher's LSD test at P < 0.05.)

  • b

    Means (±SE, n = 6) within rows followed by different letter indicate significant differences among topographic positions at each elevation (lowercase letters) and differences among elevations at each topographic position (capital letters) (one-way ANOVA with Fisher's LSD test atP ≤ 0.05.)

Extractable NH4+akg (N) ha−12.2 (0.4)A1.8 (0.5)A1.5 (0.5)0.8 (0.2)B0.9 (0.3)AB0.9 (0.2)0.6 (0.1)B0.7 (0.1)B0.8 (0.1)
Extractable NO3akg (N) ha−10.4 (0.1)A0.4 (0.1)A0.6 (0.3)A0.4 (0.1)aA0.3(0.1)aA0.04 (0.02)bB0.0B0.02 (0.01)B0.03 (0.01)B
Net N mineralisationbkg (N) ha−110d−13.6 (1.2)A3.1 (0.9)0.8 (0.6)AB1.7 (0.4)B1.4 (0.7)1.5 (0.3)AB0.2 (0.2)B0.6 (0.4)−0.4 (0.3)B
Net nitrificationbkg (N) ha−110d−13.9 (1.5)A1.7 (1.3)0.4 (0.6)0.8 (0.5)B1.8 (0.6)0.02 (0.3)0.0B0.04 (0.04)0.0

[23] Nitrogen cycling indices such as C/N ratio and δ 15N signatures of litterfall, organic layer and upper mineral soil were affected by elevation as well as topography (P < 0.001 for elevation and topography effect on C/N ratio and δ 15N signature of all three substrates). C/N ratios of litterfall, organic layer and upper mineral soil increased and δ 15N signatures of litterfall and organic layer decreased along the topographic gradient form the lower slope position to the ridges at all sites (Tables 1 and 2). Differences in these N cycling indices between topographic positions and between elevations were less pronounced in the mineral soil, indicating that δ 15N enrichment and N cycling were restricted to the organic layer. The lower slope positions at the 2000 m site showed lowest C/N ratios and highest δ 15N signatures of litterfall and organic layer. Negative δ 15N signatures of litterfall occurred at the ridges of the 2000 m site and at midslope and ridge positions of the 3000 m site. Throughout the elevation gradient, we found strong correlations of N2O + NO fluxes with C/N ratio and δ 15N signatures of litterfall (P = 0.02, R2 = 0.54 for C/N ratio, and P < 0.001, R2 = 0.83 for δ 15N signatures) (Figures 3a and 3b) and of the organic layer (P = 0.01, R2 = 0.57 for C/N ratio, and P = 0.01, R2 = 0.58 for δ 15N signatures). Ratios of C/N and δ 15N signatures of the mineral soil did not correlate with N2O + NO fluxes. Finally, we found strong correlations of N2O + NO flux with proxies of forest productivity like annual litter production (P = 0.02, R2 = 0.53) and tree basal area increment (P < 0.01, R2 = 0.72) (Figures 3c and 3d). Both proxies also showed strong correlations with NO fluxes (P < 0.001, R2 = 0.77 for annual litter production, and P = 0.01, R2 = 0.65 for tree basal area increment) and with N2O fluxes (P < 0.05, R2 = 0.44 for annual litter production and P = 0.05, R2 = 0.38 for tree basal area increment).

Figure 3.

Linear regression of log-transformed annual N2O + NO fluxes and (a) C/N ratio of total litterfall (Y = 0.22 3 − 0.01 * X, P = 0.02, R2 = 0.54), (b) δ 15N of litterfall (Y = −0.18 + 0.09 * X, P < 0.001, R2 = 0.83), (c) total litter production (Y = −0.62 + 0.07 *X, P = 0.02, R2 = 0.53), and (d) tree basal area increment (Y = −0.53 + 0.69 * X, P < 0.01, R2= 0.72) of the topographic positions across the elevation gradient. Symbols used for data are point-down triangles, lower slope; circles, midslope; triangles, ridge; black, 1000 m; gray, 2000 m; open, 3000 m.

4. Discussion

4.1. Organic Layers and Nitrogen Availability Along Elevation and Topographic Gradients

[24] The increasing thickness of organic layers with increasing elevation and from lower slopes to ridge positions have been reported for other tropical montane forests [Arnold et al., 2009; Grubb, 1977; Purbopuspito et al., 2006]. Thick organic layers imply that considerable amounts of nitrogen are stored and not actively cycling in the ecosystem. Grubb [1977]suggested that this was one of the reasons for reduced N-availability in tropical montane forests. Thick organic layers develop if decomposition is slower than organic matter (above- and belowground litter) input. While the lower temperatures (causing slower decomposition,Table 1) together with an increase in root/aboveground biomass ratio [Leuschner et al., 2007] probably explains the increasing thickness of organic layers with elevation, water logging has been suggested to contribute to the slow decomposition of organic matter on the relatively flat ridge positions at 2000 m and 3000 m [Leuschner et al., 2007; Wilcke et al., 2002]. Additionally, the feedback of tree vegetation to the environmental conditions at the exposed ridges in combination with the low nutrient availability probably affected species composition [Homeier et al., 2010] and reduced litter quality at the ridge positions (as shown by the high C/N ratios, Table 2) further reduced decomposition rates. In summary, we found evidence that a combination of climatic and hydrological conditions and the resulting plant feedbacks contributed to the observed large differences in organic layer thickness of these tropical montane forest soils.

[25] Several lines of evidence suggest that parallel to the increasing thickness of organic layers, N availability decreased with elevation and from lower slope to ridge positions. First, net N-mineralization and net nitrification rates decreased with increasing elevation. Similar findings were reported in studies comparing tropical forests at different elevations [Arnold et al., 2009; Grubb, 1977; Köhler et al., 2009; Marrs et al., 1988; Silver et al., 1994]. Second, the C/N ratios of litterfall (Table 2) and of soil organic layers (Table 1) suggest decreasing N availability with elevation and from the lower slope to the ridge position. In ecosystems with low N availability nutrient use efficiency is high [Vitousek, 1982] leading to high C/N ratios in plant tissue. In our study, this was evident at the ridge positions at the 2000 m site and all positions at the 3000 m site. Third, the decreasing δ 15N signatures of litterfall and of the organic layer with increasing elevation and from lower slope positions to the ridges imply decreasing N availability. Signatures of δ 15N reflect the long-term behavior of the soil N cycle of an ecosystem. Forest ecosystems with high N availability have highδ 15N signatures because isotopic light N is preferentially lost from the ecosystem owing to fractionation during nitrification and denitrification, leaving 15N enriched N behind [Robinson, 2001]. Sotta et al. [2008] showed this in the δ 15N signatures of litterfall and soil, which reflected the magnitude of soil N cycling rates of a lowland forest in the Brazilian Amazon. The δ 15N signatures of litterfall also correlated with nitrogen losses from tropical forests in Indonesia [Purbopuspito et al., 2006] and Panama [Corre et al., 2010]. On volcanic ash soils in Hawaii, however, δ 15N signatures in mineral soil and leaves depended on soil development stage rather than elevation [Vitousek et al., 1989]. At our study sites, slope processes rejuvenate soils at all elevations, leading to relatively young soils throughout the study sites. Our elevation gradient thus does not reflect a chronosequence in terms of soil age. Consequently, nitrogen availability was not controlled by soil development. Finally, at the 1000 m site most of the mineralized N was also nitrified, whereas at 3000 m net nitrification was only 3% of net N mineralization. This suggests that microbial N immobilization increased along the elevation gradient implying increasing competition of nitrogen with increasing elevation [Arnold et al., 2009].

[26] We can summarize that in these tropical montane forests organic layer thickness increased and N availability decreased with increasing elevation and from lower slope to ridge positions. At the 2000 m site, the lower slope and midslope positions displayed a higher N availability than the 1000 m site, which indicates that at this elevation the ‘topography effect’ was more pronounced than the ‘elevation effect’. These results suggest that in these tropical montane forests thickness of organic layers and N availability can be predicted based on elevation and topographic position [Pennock and Corre, 2001].

4.2. N Availability and Forest Productivity

[27] Annual litter production at our study sites was within the range reported for tropical montane forest sites [Röderstein et al., 2005; Vitousek, 1984; Vitousek et al., 1995], whereas values for the lower and midslope positions of the 2000 m site were on the high end compared to other studies [Hall et al., 2004; Purbopuspito et al., 2006; Röderstein et al., 2005; Veneklaas, 1991]. Our finding that forest productivity (expressed as annual litter production and tree basal area increment) decreased with increasing elevation and from the lower slope positions to the ridge positions reflects the results from earlier studies carried out in tropical montane forests [Grubb, 1977; Homeier et al., 2010; Moser et al., 2007; Takyu et al., 2002; Tanner et al., 1998] and mirrors the systematic change in N availability. However, litterfall does not always change systematically with elevation in tropical mountains as shown by Clark et al. [2001] and Röderstein et al. [2005]. Stem growth and basal area increment of tropical forest trees is determined by site conditions and the succession stage of the forest stand and it has been shown to vary strongly across tropical forest landscapes [Clark et al., 2001; Malhi et al., 2004]. In contrast to findings from tropical forests in Brazil, where forest productivity decreased with soil development [Anderson et al., 2009; Malhi et al., 2004], soil age did not control productivity along our elevation gradient because we did not have systematic differences in soil development. Since we studied only old-growth forests, our plots contain primarily late-successional, slow-growing tree species and our basal area increment rates between 0.3 and 0.9 m2 ha−1 yr−1 are probably below the forest stands' average because we omitted early successional stands which contain many pioneer species. However, our values are in the range of basal area increment rates reported from neotropical forests (0.3–1.1 m2 ha−1 yr−1 [Malhi et al., 2004]), with basal area growth rates of the lower and midslope positions of the 2000 m site being in the upper range, indicating the high productivity of these sites.

[28] The correlations of forest productivity with proxies of N availability (C/N ratio and δ 15N signature of litterfall) (Figures 2a and 2b) suggest that forest productivity depended on N availability and that the forest sites we studied were (co-)limited by nitrogen. This is in line with the results from other studies, that found a positive correlation between tree growth and N availability [Cavelier et al., 2000; Homeier et al., 2010; Tanner et al., 1992] in neotropical montane forests. Also, nitrogen addition experiments in tropical montane forests have shown significant increases in forest productivity following nitrogen additions, which indicates N limitation [e.g., Adamek et al., 2009]. This was also the case in a nutrient addition experiment conducted close to our sites (Homeier et al., unpublished data, 2010).

4.3. N-Oxide Fluxes and N Availability

[29] The N2O fluxes in these tropical montane forests were lower than most studies published from tropical lowland forests [Köhler et al., 2009; Werner et al., 2007]. They fit within the range of values reported for tropical montane forests, with a study from Hawaii reporting lower fluxes [Matson and Vitousek, 1987; Riley and Vitousek, 1995] and studies from Panama [Köhler et al., 2009], Indonesia [Purbopuspito et al., 2006], Australia [Breuer et al., 2000] and Hawaii [Holtgrieve et al., 2006] reporting slightly higher N2O fluxes. The lower slope position at the 2000 m site was the only site that displayed much higher N2O fluxes than expected. Only few studies measured NO fluxes from tropical montane forests. Their results are in the same range as our measurements at 1000 m and 2000 m elevation [Köhler et al., 2009; Purbopuspito et al., 2006]; but higher than our fluxes measured above 2000 m [Purbopuspito et al., 2006].

[30] The N2O + NO fluxes largely followed the gradients that we detected for N availability. This was clear for the declining N2O + NO fluxes with elevation, but was not significant for topographic position at all elevations, due to the high spatial variability between sites (Table 3). Other studies have also reported lower N2O emissions and even N2O uptake at ridge positions compared to lower slope positions [Bowden et al., 1992; Corre et al., 1996; Reiners et al., 1998]. For the different proxies for N availability that we tested, we did not find good correlations of N-oxide fluxes with mineral N, net N mineralization and net nitrification rates as have been reported in a range of studies carried out in tropical ecosystems [Erickson et al., 2001; Holtgrieve et al., 2006; Keller and Reiners, 1994; Matson and Vitousek, 1987; Riley and Vitousek, 1995]. The reason is probably the high temporal and spatial variability of mineral N pools and net N mineralization and net nitrification rates [Veldkamp et al., 1999]. We did however find strong correlations with more integrative indices of the long-term N cycle: C/N ratio andδ 15N signature of litterfall (Figures 3a and 3b) which has also been shown in other studies conducted in tropical ecosystems [Davidson et al., 2000; Erickson et al., 2001; Purbopuspito et al., 2006]. The combined results of very low N-Oxide emissions, low N turnover rates and negativeδ15N signatures of plant litter at high elevations and on the ridges illustrate that slow N mineralization combined with N uptake by the vegetation probably resulted in marginal nitrification rates (often not detectable at our sites) and consequently minimal N-Oxide fluxes [Handley et al., 1999]. Furthermore, at these wet sites any soil nitrate that may have been produced was probably completely consumed by denitrification, thus reducing the expression of the isotope effect at the ecosystem scale [Houlton et al., 2006]. In contrast, at the slightly drier and warmer sites, higher δ15N signatures illustrate the strong discriminating effect of nitrification and denitrification caused by the high N-oxide emissions [Houlton and Bai, 2009]. A test of C/N ratios and δ15N signatures of organic layers and mineral soil as proxies for N2O + NO fluxes, resulted in weaker correlations and supported similar findings by Purbopuspito et al. [2006] in montane forests of Indonesia.

4.4. Nitrogen Availability as a Link Between Forest Productivity and N-Oxide Fluxes

[31] The close correlation of indices of N availability with forest productivity and N-oxide fluxes we interpret as evidence that nitrogen availability controls both forest productivity and N-oxide fluxes in the studied tropical montane forests. The good correlations of indices of forest productivity with N oxides fluxes in our study (Figures 3c and 3d) suggest that data on forest productivity have the potential to be used as predictors of N2O + NO fluxes. However, it is important to note that we can only expect close correlations between indices of forest productivity and N oxide fluxes in ecosystems that are (co)-limited in nitrogen. If nitrogen is not the limiting nutrient, forest productivity does not increase with N availability while N-oxide fluxes do. Thus, in ecosystems that are not N-limited (e.g., many tropical lowland forests), we cannot expect that forest productivity correlates with N availability and N-oxide fluxes.

[32] Information on ecosystem productivity, especially on litter production rates is available for a wide range of ecosystems, including tropical montane forests which opens the possibility to include these data as predictors or as co-variables to improve prediction of N-oxides fluxes. However, data on litter production can be quite variable and in some studies extremely high litter production rates with large standard errors have been reported (possibly caused by animals, strong winds, etc). As a result, the close correlation of N2O + NO flux with litter production that we found in our study could not be confirmed after including data from other studies carried out in tropical montane forests [Holtgrieve et al., 2006; Purbopuspito et al., 2006; Schuur and Matson, 2001] Tree basal area increment however showed a closer correlation to N2O + NO fluxes compared with litter production and is probably a better long-term measure of forest productivity compared to litter fall. In addition, tree basal increment is estimated by measuring the whole plots and not by collecting a random sample, as is the case with litterfall. Presently the scientific community is doing considerable efforts to improve estimates of aboveground biomass and changes in aboveground biomass of tropical forests. The results of our study illustrate that in N-limited forests these improved biomass estimates may be used to improve N-oxide flux estimates as well.

Acknowledgments

[33] We thank Patricio Salas, Angel Macas, Nixon Cumbicus, Fabian Cuenca, Jaime Pena, Vicente Samaniego and Marco Silva for excellent field and laboratory assistance; Christoph Scherber and Marife Corre for the statistical and professional support; the laboratory staff of the Buesgen Institute, University of Goettingen, for their assistance in laboratory analysis; the Ministerio del Ambiente for research permits, the Nature and Culture International (NCI) in Loja for providing the study area and the research station; the Universidad Técnica Particular de Loja for cooperation; the Deutsche Forschungsgemeinschaft for funding this project as subprojects A2.4 and A2.2 (Ve219/8-1 A2.2 and Le762/10-1) of the research unit “Biodiversity and sustainable management of a mega diverse mountain ecosystem in southern Ecuador” (FOR 816).

Ancillary