Dissolved and gaseous nitrogen losses in forests controlled by soil nutrient stoichiometry

The 12 catchments used in this study are located within the mountain landscapes of the Czech Republic (table 1). All but two catchments have acidic soils developed over acidic bedrock; out of 93 soil pits assessed in this study, 66% were described according to the World Reference Base soil classification as Cambisols, 13% as entic Podzols, 12% as haplic Podzols, 7% as Stagnosols and 2% as gleyic Stagnosols. Norway spruce dominates the catchment vegetation, but deciduous trees are common at two catchments, and one catchment is predominantly alpine. Catchment areas ranged from 21 to 254 ha (mean = 102 ha) and mean elevations ranged from 471 to 1301 m. Calculated runoff to precipitation ratio (runoff ratio), considered to be a measure of catchment wetness, varied from 0.08 to 0.84 (mean = 0.38). For the period 1994–2016 total inorganic N deposition (Oulehle et al 2017) averaged 12 kg N ha⁻¹ yr⁻¹ and ranged from 7 to 17 kg N ha⁻¹ yr⁻¹. The δ¹⁵N of inorganic N (both NO₃ ⁻ and NH₄) in precipitation was collated from previous studies in the territory of the Czech Republic (Buzek et al 1998, Novak et al 2016).

Soils were sampled in 2015 using quantitative soil pits up to 40 cm depth (Huntington et al 1988). Soil samples were taken during the growing season and the number of soil pits per catchment varied from 5 to 10 according to catchment size. The positions of soil sampling pits were chosen to characterize topography and forest vegetation within catchment. At each location, soil profiles were sampled in a 0.5 m² frame as follows: L (litter) layer; F + H (fermented + humus) layers combined; and mineral soil in specified depths of 0–10 cm, 10–20 cm, 20–40 cm (where 0 cm is the top of the mineral soil layer). All material from each layer was weighed and sieved in the field (1 cm), separated into stones, soil <1 cm and coarse roots. The soil samples were weighed and then returned to the laboratory where it was sieved after air-drying (mesh size of 5 mm for organic horizons and 2 mm for mineral horizons). Soil moisture was determined gravimetrically by drying at 105 °C. Air-dried soil samples were analysed for total organic C and total N by dry combustion with a CNS elemental analyser (Thermo Scientific FLASH 2000). Total P was analysed in FH horizon after digestion spectrophotometrically, available P was determined in Mehlich extract by the molybdate method. Oxalate-extractable Fe and Al were determined in 0.2 M ammonium oxalate/oxalic acid solution at pH 3. The concentrations of oxalate-extractable Fe and Al (Fe_o and Al_o) in the extracts were determined by ICP-OES (Thermo Elemental Intrepid II). Total exchangeable acidity (TEA) was determined by titration of 0.1 M BaCl₂ extracts.

Forest inventory data were measured within forest plots located in circles (500 m²) with soil pit in the centre, thus directly related to the soil chemistry. For each forest plot, three dominant Norway spruce trees were selected for chemical analysis of foliage (1st and 2nd needle year) and wood (bulk wood core). Chemical analysis consisted of total C, N and P concentrations. Total above- and belowground biomass pools were calculated according to an allometric equation derived for Norway spruce (Wirth et al 2004) using measured site DBH.

Stream N fluxes represented long-term averages of available measurements, covering 1994–2016 for NO₃ ⁻ and 2006–2016 DON losses. Losses of NH₄ were negligible and not included (Oulehle et al 2017). Nitrate (NO₃) was measured by high-performance liquid chromatography (Knauer 1000), DON was calculated as a difference between total dissolved N (chemiluminiscence detection after sample combustion; Tekmar-Dohrmann Apollo 9000) and the sum of NO₃ ⁻ and NH₄ ⁺ (indophenol blue colorimetry method).

All samples (leaves, soils, and wood) were analysed for their stable N isotope composition with an isotope ratio mass spectrometer ISOPRIME100 (Isoprime, UK) interfaced with a Vario PYRO cube Elemental Analyzer (Elementar Analysensysteme, Germany). The system was calibrated by the certified reference materials with known isotopic ratio from the International Atomic Energy Agency (IAEA, Vienna, Austria). δ¹⁵N values were referenced to caffeine (IAEA-600) and potassium nitrate (USGS32) standards. Nitrogen isotope composition was measured as δ¹⁵N (‰) following the formula δ¹⁵N = [(R_sample/R_standard) − 1] × 1000, where R_sample is the ratio of ¹⁵N:¹⁴N in the investigated sample and R_standard is the ratio of ¹⁵N:¹⁴N in atmospheric reference N₂ (purity of 99.999%).

Selected stream water samples from 2018 only were analysed for their ^15/14NO₃ ⁻ ratio using a diffusion method (Goerges and Dittert 1998). Analysis of total N and ¹⁵N/¹⁴N ratio were performed on IR-MS (Delta X Plus, Finnigan, Germany) connected to the NC analyser (Elementar analyser FLASH 2000, Thermo Fisher Sci., Germany). A calculated blank correction was used (Stark and Hart 1996).

We applied an isotopic modelling approach (Houlton et al 2007, Houlton and Bai 2009, Soper et al 2018, Yu et al 2019) to estimate the proportion (f_gas) and flux of N lost via denitrification across our catchments:

$$\begin{equation}{f_{{\text{gas}}}} = { }{\delta ^{{\text{15}}}}{{\text{N}}_{{\text{TE}}}}{\text{- }}{\delta ^{{\text{15}}}}{{\text{N}}_{\text{I}}} + { }{\varepsilon _{\text{L}}}/\left( {{\varepsilon _{\text{L}}}-{ }{\varepsilon _{\text{D}}}} \right)\end{equation} \tag{ 1 }$$

in which δ¹⁵N_TE and δ¹⁵N_I represent ¹⁵N isotopic ratios of the total ecosystem (soil + vegetation) and atmospheric N inputs, respectively. For each site, differences between measured stream δ¹⁵NO₃ ⁻ and ecosystem δ¹⁵N were calculated to estimate the isotope effect () associated with dissolved N leaching (_L, assuming zero effect of δ¹⁵N-DON). Therefore _L represents the combined weighted isotope effect associated with N-NO₃ ⁻ + DON leaching (mean _L = 2.7‰, SE = 1.1‰). We used a range of assumed denitrification fractionation effects (_D) between −14 and −18‰ (Houlton and Bai 2009, Fang et al 2015, Yu et al 2019). Published data (Buzek et al 1998, 2012, Novak et al 2016) indicate that precipitation δ¹⁵N inputs (δ¹⁵N_I), including both NH₄ and NO₃ ⁻ average between −4‰ and −8‰, with a central value of −6‰ and so we used this range to assess model sensitivity to these inputs.

Total ecosystem N content in 12 intensively studied Czech catchments varied from 4.4 to 7.9 t N ha⁻¹ across catchments, with soil N comprising the largest N pool (averaging 82% of total ecosystem N, range 72%–96%) (table 1). Total mineral soil N content increased with the amount of oxalate extractable aluminium and iron (Al_o and Fe_o; p < 0.001, r = 0.92; figure 1). That is, mineral soil physicochemical properties largely determined the whole-ecosystem N stock. This relationship was independent of fine soil mass and reflected greater contributions of Spodosols (entic Podzols and haplic Podzols) at catchments with larger mineral N pools.

Zoom In Zoom Out Reset image size

Stream total dissolved N (TDN) losses were strongly negatively correlated with forest floor C:N ratio (p = 0.005, r = −0.75, figure 2(a)) and positively with forest floor N:P_available ratio (p < 0.001, r = 0.94, figure 2(b)). Nitrate (NO₃) stream fluxes dominated over DON fluxes at the majority of catchments (table 1), with an average N-NO₃/TDN ratio of 0.6. TDN losses were unrelated to foliar C:N, but increased significantly with foliar N:P ratio (p = 0.004, r = 0.76). Measured TDN runoff fluxes averaged 2.9 ± 2.3 kg N ha⁻¹ yr⁻¹ and accounted for 23% ± 17% of measured N deposition across our catchments (table 1).

Zoom In Zoom Out Reset image size

Additional examination of factors that predict forest floor stoichiometry revealed that forest floor C:N decreased (p < 0.001, r = −0.83) and N:P_available increased (p < 0.001, r = 0.88) with increasing catchment wetness (here, as runoff ratio), while the ratio of N:P_tot in forest floor was unrelated to catchment wetness (supplementary table 1 (available online at stacks.iop.org/ERL/16/064025/mmedia)). Canopy N properties correlated significantly with only one environmental variable, as foliar N concentration increased with N deposition (p = 0.025, r = 0.64). Foliar N:P ratio closely related to forest floor N:P_available ratio (p = 0.002, r = 0.80) suggesting that canopy enrichment of N over P strengthened with the increasing imbalance of forest floor N relative to available P (supplemetary figure 1).

Whole-ecosystem δ¹⁵N signature averaged 0‰ (range from −1.7 to +2.9 ‰), and increased significantly with stream TDN losses (p = 0.001, r = 0.81; supplementary figure 2(A)) and consequently with catchment wetness (p = 0.002, r = 0.78; supplementary figure 2(B)). On average, TDN leaching was isotopically heavier than N remaining in the ecosystem (fractionation effect of leaching, _L, of +2.7‰), thus ecosystem ¹⁵N enrichment likely arose as consequence not from leaching but from a highly fractionating process such as denitrification. Isotopic estimates of denitrification fluxes across our catchments were sensitive to the values assigned to denitrification fractionation effect (_D) and precipitation δ¹⁵N (supplementary figure 3). When the _D value was set to −16‰ and the precipitation input δ¹⁵N was set at −6‰ (see section 2), average gaseous N loss was estimated to be 47% of total N loss (range = 38%–62%). Alternatively, using the D of −48‰ calculated for soils incubated under laboratory conditions (Wang et al 2018) estimated gaseous N loss amounted to only 17% of total loss. Precisely measured hydrologic losses of N for the study sites (averaging 2.9 ± 2.3 kg N ha⁻¹ yr⁻¹), allow translation of fractional N losses to an average calculated denitrification flux of 2.5 ± 2.2 kg N ha⁻¹ yr⁻¹ (table 1), amounting to 20% ± 14% of measured N deposition. Calculated gaseous N fluxes thus increased with increasing TDN fluxes (p < 0.001, R² = 0.88), and the magnitudes of both N losses were associated with forest floor C:N and N:P_available ratios (figure 2), both of which provide metrics of ecosystem N saturation.

Measured average N deposition across our catchments over the period 1994–2016 was 12 kg N ha⁻¹ yr⁻¹, and ranged from 7.3 to 17 kg N ha⁻¹ yr⁻¹ (table 1). Based on our mass balances, catchment N retention (i.e. N deposition—stream TDN—gaseous N loss) was relatively high (mean = 6.3 kg N ha⁻¹ yr⁻¹ or 57% of N deposition) but with striking variation from 0% to 97%. In contrast to dissolved and gaseous losses, N retention was significantly correlated only with forest floor C:P_available (p = 0.019, r = −0.66) and N:P_available (p = 0.014, r = −0.68) but not with forest floor C:N ratio, suggesting a dominant role of soil available P in determining ecosystem N retention (supplementary figure 4).

We sought to test whether the tight relationship between measured stream TDN losses and estimated gaseous N losses was consistent across forested biomes worldwide. We identified 11 additional forested catchments where stream inorganic N export was measured alongside N gaseous estimates based on the soil ¹⁵N isotopic balance method, spanning a broad ecological and climatic range (four tropical (Fang et al 2015, Soper et al 2018), five subtropical (Yu et al 2019), and two temperate (Fang et al 2015)). Thus our dataset more than doubles the total number of catchments globally for which a full isotopic mass balance is possible. Based on the combined dataset, we observed a remarkably consistent relationship between stream NO₃ ⁻ export and denitrification flux, spanning environmentally diverse forest ecosystems (with annual air temperature range from 3 °C to 26 °C and precipitation range from 615 mm to 3220 mm) and variations in N flux of 2–3 orders of magnitude (figure 3). Based on this apparently general relationship, gaseous N flux (in kg N ha⁻¹ yr⁻¹) is predictable from stream NO₃ ⁻ flux using a power function ($f\left( x \right) = a{x^b}$, where a = 1.74 and b= 0.62; figure 3). The relationship also shows that gaseous N loss dominates over dissolved inorganic N loss when N losses are low, and the relative importance of gaseous loss decreases with increasing NO₃ ⁻ export and with overall N loss.

Zoom In Zoom Out Reset image size

Extrapolating this cross-biome relationship globally, we calculated the global forest denitrification flux based on published NO₃ stream exports from 100 catchments covering tropical (n = 28), subtropical (n = 9), temperate (n = 31) and boreal (n = 32) forest biomes (supplementary table 2). Weighted by biome area, our calculated current stream TDN flux from global forests is 13.5 Tg N yr⁻¹ (table 2). Calculated area-weighted N gaseous losses increased in order boreal (0.7 kg N ha⁻¹ yr⁻¹) < temperate (2.7 kg N ha⁻¹ yr⁻¹) < tropical/subtropical (3.3 kg N ha⁻¹ yr⁻¹) forests, resulting in total annual flux of 8.9 Tg N yr⁻¹ (table 2). This analysis predicts that total dissolved and gaseous N losses (22.4 Tg N yr⁻¹) nearly balance N deposition to global forests (23.1 Tg N yr⁻¹), that do also receive uncertain amounts of N input from N fixation. Gaseous losses comprised 40% of the total N loss.

Soils comprised the largest N store within all of our study catchments, as is typical of most forests (Vesterdal et al 2008, Cremer et al 2016). Storage capacity for N in the mineral soil appears to be controlled by soil physicochemical properties, particularly aluminium- and iron-oxyhydroxides, which also have been shown to predict soil carbon stocks across the U.S. and to immobilise dissolved organic matter within mineral soil horizons (Kalbitz et al 2000, Bingham and Cotrufo 2016). Our results (figure 1) suggest that organic matter was preferentially protected in spodic horizons by interaction with poorly crystalline minerals represented by the oxalate soluble Al and Fe fraction (Kleber et al 2005, Spielvogel et al 2008). The occurrence of Spodosols generally increases with elevation across our catchments (Chuman et al 2021), and so do Al_o + Fe_o stocks. Thus, catchment soils with high rates of mineral weathering, e.g. those with more water availability and dissolved organic matter leaching, are predisposed to accumulate relatively large quantities of organic N due to the development of soils with high sorption capacities in spodic horizons.

Catchment wetness was also the strongest correlate among environmental variables influencing forest floor C:N:P stoichiometry, which in turn governed N losses. Across our environmental gradient, wetter forests are associated with progressively thicker forest floor (Oulehle et al 2008) (moder/mor like organic horizons) and more acidic conditions (table 1). Increasing exchangeable acidity (Al and H⁺) promotes stabilization of organic P in the forest floor, thus negatively affecting P availability (Tahovská et al 2018). Reduced P availability associated with increased catchment wetness combined with increasing N deposition to result in surpluses of forest floor N over C and available P. These responses are consistent with the N saturation hypothesis (Aber et al 1989, 1998) that predicts eventual increases in nitrification and NO₃ ⁻ loss to surface waters as a result of chronic N deposition, with new information on soil P limitation and N gas losses. Our data suggest that dissolved N losses in runoff are tightly linked to forest floor nutrient stoichiometry, with observed increases in N leaching as forest floor C:N ratio decreases (Gundersen et al 1998a, Lovett et al 2002, Dise et al 2009) and N:P_available ratio increases (figure 2). The strikingly tight relationship between dissolved NO₃ ⁻ losses and forest floor N:P_available further suggests that chronic N enrichment leads to progressive P limitation. This observation is consistent with forest growth responses from Switzerland (Braun et al 2010) and canopy chemistry from the Adirondack region in the US (Crowley et al 2012) suggesting shifts towards P limitation as consequence of increasing N deposition.

Compared across catchments, total ecosystem ¹⁵N enrichment increased with catchment wetness, which pointed to the importance of gaseous losses. This ecosystem ¹⁵N enrichment could not be produced by N deposition, which is isotopically lighter than ecosystem N, or by leaching, which is isotopically heavier. By contrast, gaseous losses usually have large fractionation factors (Mariotti et al 1981, Denk et al 2017), which enrich the N that remains in the ecosystem. Denitrification can occur even in well-drained soils, in low-oxygen microsites or in temporarily or locally developed episodes (McClain et al 2003, Wexler et al 2014). Denitrification estimates calculated by isotopic model approaches (Houlton and Bai 2009, Fang et al 2015) are challenged by uncertainties associated with the assumed fractionation effects of the N loss processes. Across our catchments, the isotopic effect of dissolved N leaching varied between −1.6 and 10.7‰, with an average of +2.7‰. The large isotope effect associated with denitrification used in our study (range of −14 to −18‰) and elsewhere (Houlton and Bai 2009, Wexler et al 2014, Weintraub et al 2016, Soper et al 2018, Yu et al 2019) resulted in calculated average gaseous losses of 2.5 ± 2.2 kg N ha⁻¹ yr⁻¹ across our catchments. This assigned _D between −14 and −18‰ is less fractionating than the denitrification isotope effect calculated for forest soils under laboratory conditions with an abundant supply of NO₃ ⁻ (−48‰ on average; Wang et al 2018). Underexpression of the _D under field conditions compared to laboratory conditions is expected to arise from more effective nitrate consumption (e.g. Högberg 1997, Houlton et al 2006). Nevertheless, a correct understanding of isotopic fractionation and its variability during denitrification poses a critical challenge for further application of the ¹⁵N natural abundance method.

Combining measured stream TDN fluxes with isotope-model derived denitrification estimates (figure 2) suggested non steady state N ecosystem balance, as the sum of these losses were less than N deposition inputs, yielding an N retention term by difference. The long-term accumulation of N in the soil derived from N deposition with a distinct δ¹⁵N signature (usually negative δ¹⁵N) (Bragazza et al 2005, Solga et al 2006) could impact the isotopic model calculations, leading to underestimates of the proportion of N losses occurring as N gases. Nonetheless, calculated N retention was strongly related to the forest floor concentration of available P and its ratios to total C and N, indicating an important role of P limitation in the capacity of forest ecosystem to retain N from deposition.

Across forest biomes, denitrification fluxes derived from ecosystem δ¹⁵N values increased with ecosystem NO₃ ⁻ export (figure 3). The same overall relationship between N gas and NO₃ ⁻ leaching loss held across tropical, subtropical and temperate forests, indicating a generalized response to increased NO₃ ⁻ availability within ecosystems. Moreover, increasing N losses associated with progressive P scarcity (widening of N/P ratios in both forest floor and foliage) across our temperate forest gradient are generally consistent with observed increases of foliar N:P from high to low latitudes (McGroddy et al 2004) and increasing N losses towards tropical regions (Hedin et al 2009). The largest N loss fluxes to dissolved N and denitrification were calculated for the tropical/subtropical forest biome, followed by temperate and boreal biomes. The relative importance of dissolved N over gaseous N losses tended to increase with increasing NO₃ ⁻ export, indicating that denitrification dominates when N losses are low but that denitrification capacity may be exceeded in N saturated ecosystems (figure 3).

Systematic evidence of increased N losses with increasing N availability in our regional analysis of temperate forest catchments match broader ecological observations of decreased N use efficiency in high-N forests (Vitousek 1982). Contemporary global average TDN export from catchments across different forest biomes (supplementary table 2), ranges from 5.3 kg N ha⁻¹ yr⁻¹ in tropical/subtropical forests to 1.6 kg N ha⁻¹ yr⁻¹ in boreal forests, with a global average of 4.2 kg N ha⁻¹ yr⁻¹, i.e. 13.5 Tg N yr⁻¹ (table 2).

N gas losses have long been challenging to constrain, as they include the greenhouse gas N₂O produced by both nitrification and denitrifiaiton, as well as production of elemental N₂. Given the chain of possible N transformations as mineral N availability increases, including the coupled processes of nitrification and denitrification (both associated with formation of the greenhouse gas N₂O) it is possible to constrain estimates of gaseous N losses based on hydrological NO₃ ⁻ losses, which are routinely measured as part of catchment monitoring programs. Extrapolating our empirical relationship between N gas and NO₃ ⁻ fluxes to forests worldwide yields an estimated values of 8.9 Tg N yr⁻¹ (2.8 kg N ha⁻¹yr⁻¹). This estimate of N gas loss is much lower than past N-budget-based estimates of denitrification N losses from non-agricultural soils worldwide (58 Tg N yr⁻¹) based on N mass balances that include highly uncertain and widely varying N fixation fluxes (Van Drecht et al 2003, Seitzinger et al 2006). Bouwman et al (2013) further provided denitrification estimate of 28 Tg yr⁻¹ for non-agricultural soils applying revised (and much lower) estimates for BNF (Vitousek et al 2013) as N input to their mass balance calculations. This revised rate of denitrification recalculated to forest land is then similar to our independent denitrification estimate for forested land, thereby supporting the evidence of lower BNF in natural ecosystems (Vitousek et al 2013, Sullivan et al 2014) compared to earlier assessments (Cleveland et al 1999, Galloway et al 2004).