1. Introduction

Rivers are significant sources of nitrous oxide (N₂O) emissions, with global estimates ranging from 72.8 to 291.3 Gg N y⁻¹ (Marzadri et al. 2021). Although emitted in smaller quantities, nitrous oxide (N₂O) is a potent greenhouse gas, with a global warming potential that is 273 times greater than that of carbon dioxide over a 100-year timescale (Forster et al. 2023). The riverine N₂O emissions intensity varies greatly due to their sensitivity to heterogenous habitats (Wang et al. 2022). Investigating the influencing factors of N₂O emissions in river networks is imperative for accurately estimating emissions and establishing mitigation strategies.

In lotic ecosystems, nitrification and denitrification are the main biogeochemical processes that produce N₂O (Quick et al. 2019). Nitrification is a chemoautotrophic process in which dissolved carbon dioxide is fixed by ammonium-oxidizing bacteria that ultimately convert NH₄⁺ to NO₃^-, releasing N₂O in the process (Harrison and Matson 2003). In addition, heterotrophic nitrification of organic N producing N₂O has been reported in soil with high C: N ratios (Helton et al. 2015). Nitrification rates can be controlled by NH₄⁺ availability and indirectly through competition for NH₄⁺ with heterotrophic bacteria when C availability is high (Mwanake et al. 2019). For denitrification, N₂O is produced as an intermediary by heterotrophic bacteria that reduce NO₃^- to N₂ in anaerobic conditions. Consequently, dissolved organic carbon availability, NO₃^-, and dissolved oxygen are possible controlling factors (Inwood et al. 2007).

At a catchment scale, land use changes and anthropogenic activities further regulate riverine N₂O concentrations and fluxes by altering the input of reactive N and organic carbon to rivers (Zhang et al. 2023). Rivers flowing through catchments dominated by agricultural land (cropland and pastures) had higher N₂O levels than forested catchments, likely due to excessive N inputs from fertilizers (Borges et al. 2018). Urban areas exhibited the highest diffusive N₂O fluxes, followed by sandy lands and grasslands (Hao et al. 2021). These studies highlight the influence of different degrees of human impact on N₂O fluxes, presumably caused by N and organic carbon inputs.

In African tropical rivers, very few studies have been conducted to either quantify emissions or explain the factors controlling them (Borges et al. 2015, Mwanake et al. 2019). The sub-Saharan African rivers represent ~12% of the global stream surface area (Raymond et al. 2013), positioning them as a potential critical component of the global N₂O flux. However, uncertainties remain in the amount of N₂O flux from river networks in the region. With agricultural intensification in Africa, the risk of increased N₂O emissions from streams in the region is increasing. To our knowledge, there is a lack of studies investigating the riverine N₂O emissions in Madagascar, which is the largest tropical island of the African continent (Deschamps et al. 2023). This study addresses this critical gap by quantifying N₂O fluxes and concentrations from Ikopa streams while determining the influence of land use and in situ water physicochemical properties on controlling N₂O dynamics.

2. Materials and methods

2.1 Study site

The study was conducted in the Ikopa River network located in the central highland of Madagascar’s grassland. The Ikopa River drained a watershed of 1,622 km² with an average elevation of about 1,300 m above sea level (Fig. 1). The main channel of the river spans 299.33 km and experiences a tropical climate characterized by a warm, rainy season from November to April, followed by a cooler, drier season from May to October. Annual average precipitation ranges from 1,330 to 1,906 mm. The watershed is underlain by granite bedrock covered by a thick lateritic layer. According to the literature, the sediment texture of the Ikopa River can be composed of clay (Raunet 1980) or silty clay (D’Arifat et al. 2019, Randrianantoandro et al. 2014). The sampling campaign was carried out during the first two weeks of November 2022, corresponding to the onset of the wet season. Seven sites distributed across the Ikopa River network were visited and sampled once during the daytime (10:00 AM – 4:00 PM). No rainfall events occurred during the sampling campaign, and all sites were sampled under comparable non-stormflow conditions.

Fig. 1.

a) Locations of the studied rivers and their land covers in the Ikopa River Basin. b) the proportions of different land use types in the 7 drainage buffers (circles with 2000 m diameters upstream of the sampling sites).

2.2 Water sampling and analyses

The water temperature, pH level, and conductivity were measured in situ with portable field probes (HI 991300, Hanna Instrument, Romania). Moreover, the surface water dissolved oxygen and turbidity were determined with a portable DO meter (HI 9147, Hanna Instrument, Romania) and a turbidity digital meter (TES Digital turbidity meter 1386, Taipei, Taiwan), respectively. Hanna multiparameter probes were calibrated according to the manufacturer’s specifications.

The surface water for total phosphorous (TP) and total nitrogen (TN) contents was sampled with 30 mL acid-washed High-Density Polyethylene (HDPE) plastic bottles at 10 cm below the surface water, while water samples used for dissolved organic carbon (DOC), ammonium (NH₄⁺), and nitrate (NO₃^-) were filtered using 0.45 µm polyethersulfone (PES) syringe filters, and collected respectively into 20 mL glass bottles and 80 mL acid-washed HDPE plastic bottles. Concentrated sulfuric acid (2 mL) was added for sample preservation (Zhang et al. 2020). The bottles were stored inside a portable cooler (-4°C) filled with ice packs and then frozen (-20°C) until further analysis in the laboratory. A HUMAS water kit analyzer made in South Korea using the colorimetrical method and a UV-spectrophotometer (UV- visible spectrophotometer T60, UK) was used to quantify the TP (HS-TP-L, range 0.01 to 3 mg. L^-1), TN (HS-TN-U, range 0.02 to 5 mg. L^-1), NH₄⁺ (HS-NH₃(N)-U, range 0.03 to 1 mg. L^-1), and NO₃^- (HS-NO₃(N)-L, range 0.2 to 5 mg. L^-1) contents in water, and the DOC concentration were estimated with the Total organic carbon analyzer (TOC-L Shimadzu, Kyoto, Japan) by Non-Purgeable Organic Carbon.

2.3 Greenhouse gas collection and measurement

Dissolved N₂O concentrations were measured using the headspace equilibration method. In the field, a 60 mL syringe fitted with a three-way stopcock was used to equilibrate 40 mL of stream water sampled at wrist depth with 20 mL of ambient air (Tang et al. 2021, Weiss and Price 1980). After the syringe was vigorously shaken underwater for 1 minute, 20 mL of the headspace was immediately transferred into vacuum glass tubes (Labco Exetainer®, 12 mL, Buckinghamshire, UK) (Billett and Moore 2008). Ambient air was also collected using a 20 mL syringe, then it was injected into vacuum glass tubes, and triplicate samples were collected and analyzed at each site. Before analyzing the gas samples with a gas chromatography equipped with an electron capture detector (uECD, ChroZen GC System, Youngin Chromass, South Korea), the apparatus was calibrated with standard N₂O gases ranging from 1 to 10 ppm. Water temperature and atmospheric pressure were recorded via a 1450 micro station data logger (Spectrum Technology 3684 WDI, Illinois, USA), a temperature sensor (Spectrum Technology 3684 WDI, Illinois, USA), and a barometric pressure sensor (Spectrum Technology 3671, Illinois, USA).

The headspace equilibrium method was used to quantify the dissolved N₂O concentrations in the stream water, while the floating chamber method was used to assess diffusive N₂O fluxes as well as the CH₄ ebullition from streams (Zhang et al. 2020). At each site, three identical floating chambers were made of inverted 30 L buckets, and they were covered with aluminum foil to reflect sunlight and minimize internal heating (Zhang et al. 2020). After mixing the gas contents of the chambers three times with a 20 mL syringe, 20 mL of gas samples were extracted and injected into vacuum tubes at 0, 5, 10, 20, 40, and 60-minute time points (Crawford et al. 2013). By analyzing each collected gas concentration with gas chromatography, the N₂O concentration variations over time are recorded (Wang et al. 2021). The diffusive fluxes were calculated according to the following equation:

where F is fluxes (mmol m⁻² d⁻¹) across the water-air interface, C_t and C₀ are the N₂O concentrations in the chamber headspace at time t and 0, respectively (ppmv), P_atm is the atmospheric pressure (Pa), V is the volume of the chamber headspace (m³), A is the chamber area in contact with water (m²), time is the time of the linear gas concentration increase (min), R is the universal gas constant (8.314 m 3 Pa mol⁻¹ K⁻¹), and T is the air temperature (K) during the measurement.

The kN₂O estimation method used in this study is widely adopted in studies when both headspace equilibration and floating chamber methods are used together to assess gas velocity transfer of the gas emissions (Wu et al. 2021). FN₂O was adopted to estimate the diffusive gas exchange velocity according to Fick’s law with the following equation:

where kN₂O is the gas exchange velocity of N₂O (m. d⁻¹), FN₂O is the N₂O flux calculated from the floating chamber, K_H is the Henry constant adjusted for the water temperature and salinity, dN₂O water and dN₂O air are the dissolved N₂O concentration in the surface water and atmosphere, respectively.

2.4 Geographical analysis and Statistical analysis

Land use characteristics were identified using watershed delineation and data with a 10 m resolution provided by the European Space Agency (Zanaga et al. 2022). Circular upstream buffers with diameters of 100 m, 500 m, 1000 m, and 2000 m were extracted by ArcGIS Pro 10.4 software (ESRI Co., USA). Land cover was classified as water, forest, shrubland, grassland, and cropland (Fig. 1). The area of each land use within the buffers was retrieved and analyzed. Field observations revealed two contrasting landscape settings among the study sites. One group consisted of open, grass-dominated catchments with limited tree cover, whereas the other comprised catchments characterized by mixed vegetation, including forest, shrubland, and cropland. To facilitate comparison between these contrasting landscape types, sites were categorized as grassland areas when grassland represented more than 45% of the land cover, while non-grassland sites had lower grassland coverage (4 to 43%). This classification was used as an exploratory landscape grouping rather than a predefined ecological threshold. To determine spatial changes in the environmental factors, nutrients, and greenhouse gases, a Mann-Whitney U test was used. Spearman multiple correlation analyses were carried out between environmental factors and N₂O concentrations and fluxes, since most of the variables did not present a normal distribution (Shapiro-Wilk normality test, p < 0.05). Principal component analysis (PCA) was conducted on key physiological variables (DO, NO₃^-, NH₄⁺, DOC, water temperature, flow velocity, TP, and depth) to explore environmental gradients across the study sites.

3. Results

3.1 Patterns of riverine dissolved N₂O concentration and fluxes

The average N₂O concentration ranged from 7.76 to 47.6 nmol L^-1 (mean was 31.3 ± 15.7 nmol L^-1). The atmospheric N₂O concentration was 328 ppb, equivalent to 13.4 nmol L^-1. A moderate variability was observed with a coefficient of variation equal to 50%, but no clear spatial pattern was evident. N₂O saturation ranged from 117 to 925%, with an average of 544 ± 252%, indicating intense oversaturation of N₂O in the rivers and subsequent potential to emit to the atmosphere. The N₂O flux spanned two orders of magnitude (24.53 to 638.94 µmol m^-2 d^-1) with an average of 177.02 µmol m^-2 d^-1.

3.2 Variation of physicochemical properties

Physiochemical parameters were relatively consistent across the study sites, with several variables showing narrow ranges and moderate dispersion. Water temperature averaged 25.2 ± 2.5°C, while DO was 51.2 ± 14.2%. Electrical conductivity (EC) averaged 81.4 ± 22.6 µS cm⁻¹, and TP concentrations were low, with a mean of 0.048 ± 0.012 mg L⁻¹. Nitrogen species were generally low, NO₃^- averaging 0.27 ± 0.31 mg L^-1 (Fig. 2 (a)) and NH₄⁺ 0.15 ± 0.06 mg L^-1. TN concentrations were also low (0.17 ± 0.16 mg L^-1). DOC averaged 3.53 ± 4.03 mg L^-1, while the DOC: NO₃^- ratio was 15.2 ± 13.3 (Figs. 2 (b) and (c)). The estimated gas transfer velocity (k) calculated from velocity and slope measurements ranged from 0.06 to 4.53 m d^-1 with a mean value of 2.01 ± 1.29 m d^-1.

Fig. 2.

Variations of the key parameters across river sampling sites with different land use categories: a) nitrate concentration, b) DOC concentrations, c) DOC: NO₃^- ratios. Different letters indicate a significant difference among groups.

PCA was conducted on key physicochemical variables; the first two principal components explained 71.3% of total variance, with the first principal component and second principal component accounting for 40.3% and 31.5%, respectively (Fig. 3). The first principal component represented a gradient contrasting sites characterized by higher ammonium concentrations and depth with those with higher flow velocities. The second principal component was primarily associated with NO₃^-, DOC, and DO. The PCA biplot revealed that NO₃^- and DOC were positively correlated, as indicated by their similar vector orientations, whereas NH₄⁺ and TP were negatively associated with these variables. Flow velocity was largely independent of nutrient-related variables. Sampling sites were widely dispersed in ordination space and did not form distinct clusters according to land cover classification, indicating that local environmental conditions varied independently of broad land-use categories.

Fig. 3.

Principal component analysis of the key parameters across river sampling sites.

3.3 Influence of land cover and physicochemical controls on N₂O

Grassland-draining river presented a numerically higher dissolved N₂O concentration (35.96 nmol L^-1) than non-grassland-draining rivers (25.17 nmol L^-1), but the difference was not statistically significant (Fig. 4 (a)). Diffusive N₂O flux did not differ significantly between grassland and non-grassland rivers (Fig. 4 (b)). Among physicochemical properties, grassland and non-grassland rivers exhibited a significant difference in EC, and DOC was statistically higher in non- grassland rivers. Although the mean NO₃^- presented a difference, it was not statistically significant. Despite these variabilities, no significant differences in remaining physicochemical properties were detected between land cover types.

Fig. 4.

Variability in a) dissolved N₂O concentrations and b) fluxes across river sampling sites with different land use categories.

No significant correlations were found between N₂O concentrations and the key parameters. In particular, N₂O showed no relationship with NO₃^-, and weak, non-significant relationships with NH₄⁺ (r = 0.29, p = 0.56) and DOC (r = −0.39, p = 0.40).

4. Discussion

4.1 Patterns of riverine dissolved N₂O concentration and fluxes

The dissolved N₂O concentration in this study falls within the range reported for global river networks spanning from 2.86 to 1,335.71 nmol L^-1 (mean value of 54.64 nmol L^-1) (Wang et al. 2022), but exceeded the average concentration observed in tropical rivers (23.2 nmol L^-1). Furthermore, the diffusive N₂O fluxes were comparable to the global headwater stream emissions (109.06 ± 10.91 μmol m^–2 d^–1), but were a higher than reported for other tropical headwater streams (43.62 ± 16.36 μmol m^–2 d^–1) (Li et al. 2021). Diffusive N₂O fluxes exceeded the values reported for rivers draining natural (10.96 μmol m^–2 d^–1) and agricultural (15.6 μmol m^–2 d^–1) land cover (Ho et al. 2020). These elevated fluxes suggest Ikopa tropical streams may represent an important yet poorly quantified source of N₂O, highlighting the potential contribution of tropical African riverine systems to the regional greenhouse gas emissions

Despite relatively limited anthropogenic disturbance in some sites, the elevated diffusive N₂O fluxes indicate active nitrogen transformation processes and efficient gas transfer dynamics. No significant difference in dissolved N₂O concentrations and fluxes was detected between grassland and non-grassland rivers. This likely reflects the mixed land-use mosaic surrounding most catchments, where grassland-dominated sites often still contained a large proportion of cropland coverage and non-grassland sites were similarly heterogeneous (Fig. 1). Consequently, land cover categorizations may not adequately capture the hydrological, biogeochemical, and microbial controls governing N₂O dynamics. Future studies incorporating finer-scale land cover characterization (e.g. proportion of land cover within the riparian zones) may help better resolve land cover influences on riverine N₂O dynamics

4.2 Weak relationship between N₂O and physiochemical variables

Riverine N₂O production is governed by multiple interacting microbial pathways, including nitrification and denitrification, which may respond differently to oxygen variability, nutrient concentrations, and carbon inputs (Borges et al. 2018). In the present study, dissolved N₂O concentration and fluxes were not significantly correlated with the measured dissolved nitrogen species, DOC, DO, or water temperature. Rather than indicating the absence of environmental control, these results indicate that N₂O dynamics were influenced by multiple interacting processes operating simultaneously across sites. The limited sample size and strong spatial heterogeneity may also have reduced the ability to detect simple linear relationships.

The principal component analysis suggests contrasting hydrological and nutrient processing environments. The first principal component indicates a gradient between nutrient-retentive and hydrological dynamics. Increased turbulence and flow velocity enhance gas transfer efficiency, potentially promoting elevated N₂O emission independently of dissolved N₂O concentrations (Mwanake et al. 2019). Our results suggest that variability in N₂O emissions was more influenced by turbulence-driven gas transfer. This may explain why high fluxes were observed despite the weak relationships with measured environmental variables and N₂O concentrations.

Nitrification was likely not strongly nitrogen- limited, as indicated by the absence of correlation between N₂O and NH₄⁺. In this study, NH₄⁺ concentrations exceeded those previously reported in agricultural tropical streams (NH₄⁺= 0.04 mg L^-1) (Masese et al. 2017). In contrast, mean NO₃^- concentrations in this study were lower than those stated in tropical rivers (1.44 mg L^-1) (Mwanake et al. 2019), suggesting that elevated N₂O concentrations in the studied rivers are unlikely to be explained solely by high inorganic nitrogen loading. These findings indicate that internal nitrogen cycling processes may play a more important role than external nutrient enrichment in regulating N₂O production in these Ikopa rivers.

One plausible explanation is the elevated DOC: NO₃^- ratio across sites, which may favor heterotrophic microbial activity. When DOC: NO₃^- ratios are high, microbial communities may experience relative nitrogen limitation, potentially enhancing the rapid assimilation of NO₃^- and promoting denitrification-related pathways under localized redox conditions gradients (Taylor and Townsend 2010). Although direct denitrification process measurements were not conducted, the combination of low NO₃^- concentration, elevated DOC: NO₃^- ratios, and N₂O supersaturation suggests that internal microbial nitrogen cycling, rather than external NO₃^- enrichment, may contribute to N₂O production. This interpretation is further supported by the second principal component, which was associated with NO₃^-, DOC, and DO, indicating a gradient linked to oxidized nitrogen and organic matter dynamics.

4.3 Implications and limitations

A limitation of the present study is that each site was sampled only once, representing a spatial snapshot of riverine N₂O dynamics at the onset of the wet season. The dataset does not capture temporal variability associated with seasonal changes, rainfall events, or fluctuations in river discharge, all of which may influence N₂O production and emissions. Although the present study was based on a limited number of sampling sites, it highlights the substantial spatial heterogeneity of N₂O dynamics in tropical rivers. The absence of a strong statistical relationship between N₂O dynamics and observed environmental variables can be induced by the relatively small number of sampling sited and emphasizes the need for future studies incorporating more extensive spatial coverage, seasonal sampling, and direct measurement of microbial nitrogen transformation processes. The statistical analyses should be interpreted as explanatory and intended to identify potential environmental controls. Nevertheless, by quantifying N₂O fluxes from streams in Madagascar, this study provides the critical baseline data that helps reduce uncertainties in African N₂O budgets and suggests that localized environmental conditions may outweigh broad land-cover effects in regulating tropical river greenhouse gas emissions.

5. Conclusion

Ikopa tropical rivers were consistently supersaturated with a source of N₂O, with concentrations and fluxes exceeding those typically reported for tropical rivers highlighting their potential contribution to regional greenhouse gas emissions. Despite the absence of significant relationships between N₂O dynamics and physiochemical variables, multivariate analyses revealed contrasting hydrological and nutrient-processing environments among sites. The elevated N₂O concentrations and emissions observed in this study are unlikely to be explained solely by inorganic nitrogen enrichment. The combined influence of hydrological exchange, turbulence-driven gas transfer, and localized microbial pathways likely regulates N₂O production and emissions in these tropical rivers. This study contributes to reducing uncertainties on riverine N₂O emissions in Madagascar by providing first baseline measurements of riverine N₂O concentrations and fluxes. Future studies integrating seasonal monitoring and direct process measurements, and finer- scale land cover characterization will be essential to better constrain the mechanisms controlling N₂O production and emissions in tropical African rivers.