Forage production and nitrogen dynamics in silvopastoral systems with Leucaena diversifolia in Urochloa grass-based pastures

Abstract

In tropical regions, cattle production often relies on grass monocrops (GMC), which limit productivity and land-use efficiency. Sustainable intensification of pastures is needed to meet rising demand for animal-source foods. Silvopastoral systems (SPS) with nitrogen-fixing legume trees, such as Leucaena diversifolia, offer a potential solution but may increase nitrous oxide (N₂O) emissions due to higher nitrogen (N) inputs. We conducted a one-year study comparing Urochloa GMC versus SPS with L. diversifolia on a fertile Pachic Haplustoll in Valle del Cauca, Colombia. Two grass cultivars were evaluated: Urochloa hybrid cv. Cayman and U. brizantha cv. Toledo. In SPS, tree rows created two zones: grass-only (SPS/GO) and under-tree (SPS/UT). We measured forage biomass, grass N uptake, N₂ fixation by L. diversifolia, and soil N₂O emissions under two conditions: no additional N input (+water) and after urine deposition (+urine). SPS produced 30% more forage dry matter than GMC. Leucaena diversifolia obtained 54% of its N from atmospheric fixation. The grass in SPS absorbed 8 kg N ha⁻¹ more per grazing cycle than GMC. Under no additional N inputs, soil N₂O emissions were higher in SPS/UT, likely due to greater N recycling. With urine addition, emissions were variable, with higher values in SPS/GO, but no statistical difference was found between SPS/UT and GMC, possibly due to inappropriate urine-N overdosing in GMC. Although SPS showed higher N₂O emissions under specific conditions, their greater forage yield and quality, plus potential co-benefits like lower enteric methane emissions and higher carbon capture, indicate a favorable production–environment trade-off.

Graphical Abstract

Highlights

• •Silvopastoral systems (SPS) with L. diversifolia increased 30% forage dry matter production compared to grass monocrops (GMC)
• •L. diversifolia obtained a maximum of 58% and 79% of aboveground N due to biological N fixation when integrated in SPS
• •Grass N uptake was higher in SPS compared to GMC due to complementary N uptake, instead of direct N transfer from the legume
• •The C/N ratio of plant litter was significantly lower in SPS than in GMC, favouring soil N mineralization and nitrification
• •N₂O emissions peaked in SPS: under the tree canopy when no N was added, and in SPS grass-only areas following urine patches

1. Introduction

The global demand for meat and milk is projected to increase by 138–167% by 2050 relative to 2002 levels (Rao et al., 2015, Van Eenennaam, 2024). Considering that grazing areas already represent two-thirds of the global agricultural land (FAO, 2024), to meet the future food demand, sustainable intensification of livestock systems is of special importance. Tropical America has 250 million hectares of well-drained savannas (Jank et al., 2005), and a significant share of that area is underutilized and poorly managed with extensive livestock systems.

The average cattle density in Brazil, the second largest beef producer globally, is lower than one animal stocking unit per hectare (Arantes et al., 2018, Zu Ermgassen et al., 2018). Neighboring countries, like Colombia, face similar challenges regarding livestock productivity and land use efficiency, with an average animal stocking rate of 0.7 large livestock units (equivalent to 450 kg live weight) per hectare (FEDEGAN, 2023). Improved pasture management offers a pathway to increase animal stocking rates sustainably. Nitrogen fertilization of tropical pastures has demonstrated potential to increase stocking density by an additional one to two animal units per hectare (Rincón et al., 2018, Bastidas et al., 2024), but a similar effect can be achieved with the use of N₂-fixing legumes as grass-legume pastures (Lascano and Euclides, 1996, Zu Ermgassen et al., 2018, Schultze-Kraft et al., 2018). These only require an adequate supply of essential macro- and microelements, other than N, supplied through maintenance fertilizer applications. As an added value, legumes provide other well-known environmental co-benefits (Schultze-Kraft et al., 2018).

Silvopastoral systems (SPS) are agroforestry configurations that integrate diverse forage species, including herbaceous, shrub, and tree layers, to support livestock production while delivering ecological co-benefits (Chará et al., 2019a, Sandoval et al., 2023). The legume tree Leucaena leucocephala (Lam.) de Wit has been named “the most widely used forage tree legume” (Shelton and Brewbaker, 1994), and it has indeed been widely studied for its suitability for intensive SPS in the tropics (Chará et al., 2019b). Compared to grass monocrops (GMC) of Cynodon plectostachyus, Urochloa (syn. Brachiaria) spp. and Megathyrsus (syn. Panicum) maximus, the use of L. leucocephala in SPS has been reported to, (i) improve soil fertility due to an increase in availability of N to companion grass (Banegas et al., 2019), (ii) improve forage availability, forage quality, and animal productivity, and (iii) reduce methane (CH₄) emission (i.e. per unit of dry matter intake or animal weight gain) (Rivera-Herrera et al., 2017, Rivera et al., 2024). Leucaena diversifolia (Schltdl.) Benth. is considered another promising legume tree with a broader adaptation range than L. leucocephala, in terms of soil acidity, low to medium natural soil fertility, and altitude above sea level (Peters et al., 2011). However, very few studies have assessed its forage quality and overall effect on soil N dynamics, for instance, its capacity for biological nitrogen fixation (BNF) under field conditions, or its effect on emission of nitrous oxide (N₂O) from soil.

The use of mixed pastures of grasses and legumes as a production intensification strategy instead of the use of synthetic N fertilization has a significant impact on reducing pollutant and greenhouse gas (GHG) emission sources. Some of them include reduced NH₃ volatilization, lower direct N₂O emissions after high loads of readily available mineral N, and the avoided embedded CO₂ emissions associated with fertilizer production and transportation (Schmeer et al., 2014, Schultze-Kraft et al., 2018). On the other hand, it has been reported that the use of forage legumes enhances N recycling via plant residues, and may lead to increased N₂O emissions if increased soil N supply from mineralization and nitrification exceeds the plant N demand (Gogoi et al., 2025, Boddey et al., 2020). The magnitude of these benefits remains uncertain under SPS conditions with high N inputs.

The quantification of GHG emissions from grazed pastures is of special importance since a significant share (9–19%) of the national GHG emissions in the larger economies of Latin America like Brazil, Argentina, Mexico, Colombia and Costa Rica are livestock-sourced, and in more livestock-dependent countries like Uruguay, can add to 72% of the national emissions (Arango et al., 2020). The relative contribution of N₂O to total GHG emissions varies from one country to another, for example in Colombia direct and indirect N₂O emissions in the Agriculture, Forestry and Other Land Use (AFOLU) sector accounted for 5.4% of the sector’s emissions in 2021, whereas in Uruguay it was 27% in 2020 (MA, SNRCC, 2023, IDEAM, MADS, 2024).

Considering that most of the mentioned Latin American countries have committed to ambitious GHG reduction targets, in order to effectively design strategies that mitigate N₂O emissions from pastures, it is essential to understand the mechanisms of pasture soil N dynamics and how they are influenced by grass monocropping vs SPS land use strategies.

In this study, we assessed the influence of the tree legume L. diversifolia on pasture N cycling in terms of forage productivity and N uptake, BNF, N cycling through plant litter, soil N transformations, and N₂O emissions. To that end, during 2021–2022 we continuously evaluated Urochloa spp. GMC and SPS with L. diversifolia. Our hypothesis was that integrating L. diversifolia into Urochloa-based silvopastoral systems would enhance forage productivity and N uptake compared to grass monocrops but may also lead to increased soil nitrification and higher N₂O emissions, potentially presenting a trade-off between productivity gains and greenhouse gas mitigation in tropical pastures. We expected that, despite possible increases in N₂O emissions, the overall sustainability outcome from SPS would remain positive due to substantial improvements in forage productivity and N use efficiency.

2. Materials and methods

2.1. Study area

The present study was conducted at the International Center for Tropical Agriculture (CIAT) in Palmira, Colombia (3°30'06.0"N, 76°20'44.7"W). The field site is located in an inter-Andean Valley at 990 m.a.s.l., with a mean daily temperature of 24°C and annual precipitation of 950 mm (annual average of 2011–2021). The soil type is classified as Pachic Haplustoll (CVC and IGAC, 2021). Mollisols were chosen for their high fertility and organic matter content, which support productive pastures and tree legumes. While Urochloa grasses tolerate acidic soils, Leucaena thrives on fertile soils; thus, a Mollisol was selected to generate experimental evidence through a proof-of-concept for system sustainability.

The parental materials of the soils of the region are mainly Holocene fluvial sediments and eroded materials originating from the adjacent Western and Central Cordilleras, 2:1 clay minerals (mainly smectite, montmorillonite, and vermiculite), with high base saturation, high pH and soluble salts (Howeler, 1986). The soil texture in the site is silty clay (40–44% of silt and clay each, 15% sand), soil bulk density ranges from 1.45 to 1.64 g cm⁻³, and soil porosity from 41 to 44% of the soil volume. Average soil pH is 7.0, organic matter 36 g kg soil⁻¹, Bray-II P 29.6 mg kg soil⁻¹, and cation exchange capacity 19 cmol kg soil⁻¹. Bray-II P was non-homogeneously distributed across the field. The average concentration in Cayman pastures was 23 mg P kg soil⁻¹, whereas it was 56 mg P kg soil⁻¹ in Toledo pastures; these values indicate an adequate supply of P (Table S1).

2.2. Experimental design

In 2013, a field experiment was established with the aim of evaluating N cycling and cattle weight gain between grazed grass monocrops and silvopastoral systems with the same grass base plus the legume tree L. diversifolia (Molina-Botero et al., 2024). Four pasture treatments (two GMC and two grass-legume SPS were evaluated): (GMC1) Urochloa hybrid cv. Cayman (CIAT BR02/1752), (GMC2) Urochloa brizantha cv. Toledo (CIAT 26110), (SPS1) cv. Cayman + L. diversifolia (ILRI 15551), and (SPS2) cv. Toledo + L. diversifolia. Two different Urochloa grass cultivars were evaluated with the aim of comparing the performance of a bred cultivar (Cayman) versus an already well-established cultivar in the market (Toledo). The major attributes of the two grass cultivars were described by Sandoval et al. (2023).

The experimental layout was factorial with three blocks, wherein the larger plots of each grass species were established with the respective level of intensification (GMC or SPS). Each plot (treatment replicate) had an area of 0.33 ha (36 m × 91 m). The grass pastures were seeded by broadcasting 8 kg of seeds per hectare. The tree legumes were established with 2000 plants per hectare in the form of double hedgerows, with 10 m distance between double hedgerows, 2 m distance between single hedgerows, and 1 m distance between plants within one row. Out of the 2000 L. diversifolia plants per hectare, 75% were maintained at 2 m height for animal browsing, and 25% were left to grow taller than 10 m to provide shade. For a schematic representation of the field experiment design, see Figure S1.

2.3. Plant biomass production and N concentration

Between July 2021 and July 2022 six evaluations of plant biomass production were performed. In each plot of the GMC treatments, two areas of 2 m × 2 m were isolated with electric fence leaving the internal 1 m² for continuous forage sampling and 0.5 m surrounding it as a buffer. In each plot of the SPS treatments four areas of 2 m × 2 m were fenced, out of which two areas were located under the L. diversifolia trees (SPS/UT), and the other two at 5 m from the L. diversifolia hedgerows, where there was grass only (SPS/GO) areas (Figure S1B). The total number of sampling areas was 36, corresponding to 12 areas in the GMC plots (2 grass species × 2 sampling areas per plot × 3 replicates), and 24 in SPS plots (2 grass species × 4 sampling areas per plot × 3 replicates). All sampling areas were excluded from animal grazing throughout the duration of the experiment to avoid possible contamination of the soil and plants’ N isotopic signatures with cattle urine- or dung-N.

Each evaluation consisted of an initial standardization, cutting the grass to a height of 20 cm, and in the SPS treatments, manually removing the available L. diversifolia foliage and non-woody branches with a diameter less than 0.5 cm at a maximum height of 2 m. After 51 days of regrowth, the grass and legume biomass samples were collected in the same manner as for the standardization. Available plant litter was also sampled, considered as dead plant parts lying on the ground or still attached to tillers. Plant and litter samples were oven-dried at 60°C for 72 h to determine dry matter (DM) content and were later milled (1 mm) for chemical analyses.

In GMC treatments, the total grass DM production per plot was calculated by averaging the grass DM harvested from the two sampling areas in each plot and extrapolating to one hectare. In SPS treatments, the grass DM production per plot was calculated by assigning a different weight to the two types of spatial conditions. The SPS/GO areas were assumed to represent 60% of the total SPS area and SPS/UT areas were considered representative of 40% of the SPS area. This ratio was determined based on measures of the photosynthetically active radiation (PAR) in the 10 m transects between two tree hedgerows. The largest effect of the tree canopy was observed in the 0–2 m and 8–10 m distance where the tree canopy reduced the PAR more than 80% compared to a fully open area, whereas in the 2–8 m segment of the transect the reduction of PAR ranged between 53% and 65%. The weighed grass DM production was then extrapolated to one hectare. The legume biomass in SPS plots was calculated by averaging the DM obtained from each shrub harvested and multiplying by the number of shrubs in the plot. The total forage production of a given SPS plot was calculated by summing the weighted grass and the legume forage DM production.

2.4. Biological N₂ fixation

In each sampling area of 4 m², soil samples (0–20 cm soil depth) were taken using a soil auger avoiding the central 1 m² used for plant sampling, then air-dried for 48 h and sieved to 0.5 mm particle size. Soil and plant samples were analyzed for total C and N concentration, and ¹³C/¹²C and ¹⁵N/¹⁴N isotopic ratios by dry combustion using an Elementar vario EL cube elemental analyzer interfaced to an Elementar VisION isotope ratio mass spectrometer (Elementar Analysensysteme GmbH, Langenselbold, Germany) at the isotope facility in the University of California Davis, USA. Total nutrient concentrations and isotopic signatures were determined from four of the six sampling events, two from the dry season (sampling 1 and 2), when accumulated precipitation between samplings was lower than 100 mm, and two from the rainy season (sampling 4 and 6), when accumulated precipitation was higher than 150 mm (Fig. 1B). Sample size per treatment for each harvest was n = 6 for isotope analysis (3 plots per treatment × 2 fenced areas per plot).

Fig. 1.

A. Forage dry matter (DM) production per sampling in Cayman and Toledo grass monocrops (GMC) and silvopastoral systems (SPS) pastures averaged over six samplings. B. Average forage (grass + legume) DM production of the four pasture treatments over six samplings of 51 days in a time span of 381 days from 9th July 2020–31st July 2021. Values represent the mean and standard deviation. Different letters on top of the standard deviation denote statistical differences between pasture treatments (A) or samplings (B) according to Tukey’s HSD test (α=0.05).

The natural abundance of ¹⁵N (δ¹⁵N) expressed as ¹⁵N excess or depletion of the sample over the ¹⁵N/¹⁴N ratio (R) of the air (R_air = 367.6 × 10⁻⁵) (Shearer and Kohl, 1986) was determined by using the formula:

$${\mathrm{\delta }}^{15}\mathrm{N}(\mathrm{‰})=\frac{R_{\mathit{sample}}-R_{\mathit{air}}}{R_{\mathit{air}}}\times 1000$$

The δ¹³C was accordingly expressed as ¹³C excess or depletion over the ¹³C/¹²C ratio of the international Vienna Pee Dee Belemnite (VPDB) standard (R_VPDB = 1123.75 × 10⁻⁵).

The proportion of N derived from the atmosphere (%Ndfa) in the L. diversifolia tree legumes was determined by using the formula:

$$\mathrm{Ndfa}\left(\%\right)=\frac{{\mathrm{\delta }}^{15}{N}_{\mathit{ref}}-{\mathrm{\delta }}^{15}{N}_{\mathit{leg}}}{{\mathrm{\delta }}^{15}{N}_{\mathit{leg}}-B}\times 100,$$

where δ¹⁵N_ref is the δ¹⁵N signature of a non-N-fixing reference plant, δ¹⁵N_leg is the δ¹⁵N signature of the legume shoots (L. diversifolia), and B is the δ¹⁵N of legumes relying exclusively on atmospheric N₂ as the N source and accounts for any internal isotopic fractionation of legume plants (Unkovich et al., 2008). The grasses growing in the grass-legume plots in the areas under the legume trees (SPS/UT) were used as non-N-fixing reference plants. For the B value we used the minimum recorded δ¹⁵N value of the legume in each sampling as discussed by Sarabia-Salgado et al. (2020).

Nitrogen uptake by grasses was calculated by multiplying the DM production per m² by the N concentration in the DM. Crude protein (CP) concentrations of grasses and legumes were calculated by multiplying their N concentrations by the factor 6.25 (AOAC, 2009).

2.5. Soil mineral N

With every forage harvest, soil samples (0–10 cm soil depth) were also taken. Soil ammonium (NH₄⁺) and nitrate (NO₃^-) concentrations were measured for the first three soil samplings. Unfortunately, difficulties during sample storage hampered the measurement of mineral N (Nmin) in samplings 4, 5 and 6.

Air-dried soils were sieved (2 mm), and Nmin was extracted by shaking 30 mL 1 M KCl with 3 g of soil (Li et al., 2012) for 30 min and then filtered using Whatman paper. The concentration of NH₄⁺ was determined in the KCl extracts from the formation of the indophenol complex and measuring the absorbance at 667 nm, and of nitrate (NO₃^-) from alkalinization with sodium salicylate and measuring the absorbance at 410 nm (Forster, 1995).

2.6. N₂O emissions from urine patches

To determine differences in N₂O emissions from soil in the different pasture treatments in the rainy season of 2021 (October to December) we experimented to measure both the background soil N₂O emissions of the pastures under no additional N inputs (by adding water to the soil), and also at high N input by simulating urine deposition in patches and measuring the N₂O emissions over 52 days using the manual static chamber method (Davidson et al., 2002).

Each static chamber consisted of a PVC collar (10 cm height, 27 cm diameter) and a cap (25 cm height, 27 cm diameter). The measurement footprint area was 0.057 m² and the chamber volume was 17.17 L. The caps were equipped with two ports, each fitted with a rubber septum, one for gas sampling and the other to insert a thermometer. On one side of the cap, a venting tube was installed to equalize the pressure conditions inside and outside the chamber (Davidson et al., 2002). Static chambers were set up in each of 36 sampling areas, and 1 L of urine was deposited (+urine) in the collar and surrounding area in a ring extending 15 cm from the collar. In half of the sampling areas (n = 18), additional control chambers were set up and received 1 L of deionized water (+water) to maintain similar levels of soil moisture between treated and control areas. The experiment therefore included measurements from a total of 54 chambers.

Cattle urine was collected from a farm near the CIAT campus (16 km distance) by manual vulval stimulation of four dairy cows at around 6:00 am. To obtain enough urine volume, the urine collection was repeated three consecutive days at the same hour. The cows were consuming a mixed diet with the grasses Cynodon nlemfuensis and Megathyrsus maximus, with a limited inclusion rate of the legumes Leucaena sp. and Gliricidia sepium. The urine obtained every day (∼13 L) was stored in a cold room (<10°C) after subsampling of 50 mL preserved at −20°C with 10% v/v H₂SO₄ for analysis of N concentration. After the third day, the total of 40 L of collected urine was mixed, and subsamples of 50 mL were analyzed to determine possible changes in N concentration during storage time.

The chamber collars were buried 5 cm into the soil the day before taking baseline measurements of N₂O. Between 09:00 and 10:00 am on each measurement day, the static chambers were closed and gas samples of 24 mL were taken from the chamber headspace at intervals of 20 min (0, 20, 40, 60 min) using a syringe. The gas samples were then deposited in vacuum-empty glass vials for analysis of N₂O concentration using a Shimadzu GC-2014 gas chromatograph in the laboratory of greenhouse gases of CIAT. Immediately after the baseline measurement in each sampling point, either 1 L of the mixed urine or water was homogeneously deposited in the collar and surrounding area in a ring extending 15 cm from the collar. The N concentration of the urine used was 6.1 g N L⁻¹, equivalent to an N load of 24.4 g N m⁻². The emissions of N₂O were measured 1, 2, 3, 4, 8, 9, 11, 16, 18, 22, 24, 28, 32, 36, 43, and 52 days after urine deposition as previously described. Accumulated N₂O emissions over the period of 52 days were estimated by linear interpolation for the unsampled time periods, assuming constant emission rates between sampling points. The equations used to determine the gas flows and accumulated emissions were the same as described in Costa Jr et al. (2022).

Every measurement day, temperature in the chamber headspace, soil temperature and soil moisture were recorded. Likewise, 10 g of topsoil (0–10 cm) were sampled from the area surrounding the collar, air-dried and Nmin was extracted to determine changes in the ammonium (NH₄⁺) and nitrate (NO₃^-) pools over time (in situ net N transformation rates) following the same procedure as described earlier.

2.7. Data analysis

Data on forage DM productivity, nutrient concentrations and isotopic signatures were analyzed with a repeated measures ANOVA, considering the grass species and pasture treatment (GMC, SPS/GO, SPS/UT) as fixed factors, and each unique sampling area was considered as a random factor nested within the factor “sampling time”. For the analysis of forage DM productivity, the soil available P (Bray-II) was included as a covariate in the model. For N₂O emissions, only the accumulated emissions were subjected to statistical tests. Multiple comparisons were tested using the Tukey’s HSD test with a level of significance of 0.05. All statistical analyses were performed in R v 4.2.1. using the packages “nlme”, “car”, “lsmeans”, and “multcomp”.

3. Results

3.1. Forage biomass production and N concentration

The production of total available forage (including both grass shoots and legume leaves/branches) was significantly affected by both the grass species and the pasture type. The highest DM was produced by Toledo-SPS with 1197 kg DM ha⁻¹ sampling⁻¹, followed by Toledo-GMC with 965 kg DM ha⁻¹ sampling⁻¹, Cayman-SPS with 931 kg DM ha⁻¹ sampling⁻¹, and the lowest in Cayman-GMC with 668 kg DM ha⁻¹ sampling⁻¹ (Fig. 1A). On a yearly basis, total DM production was equivalent to 8.59, 6.90, 6.66, and 4.78 Mg DM ha⁻¹ yr⁻¹. Legume biomass represented 21% of total DM production in Cayman SPS and 11% in Toledo SPS. On average, SPS showed 30% higher total forage production than GMC, resulting in an additional yield of 1.77 Mg DM ha⁻¹ yr⁻¹ on average in SPS compared to GMC systems, and Toledo pastures showed 35% higher forage production than Cayman pastures for both GMC and SPS.

Even though we recorded a decrease in photosynthetically active radiation (PAR) of 30% in the SPS areas under the trees (SPS/UT) due to shade, compared to SPS areas with grass-only (SPS/GO), the decrease in grass DM production between these two areas was not statistically significant (89 g DM m⁻² sampling⁻¹ vs. 101 g DM m⁻² sampling⁻¹).

Over the experimental period of 381 days, forage DM production showed moderate variation, except for the sampling during the last rainy period, where all treatments showed a decrease of 72% in DM production compared to the previous sampling (Fig. 1B).

The N concentration of legume foliage was similar over the year and did not differ according to the associated grass (average 41.4 mg N g DM⁻¹, Table 1). On the contrary, the N concentration of grasses in the final sampling of the rainy season was higher than the others, likely because of less dilution of N uptake in plant DM since significantly less biomass was produced from that sampling.

Table 1.

N concentration of legume foliage (leaves + young stems/branches) and grass shoots measured in four samplings (two per season).

Plant type	Pasture		N concentration (mg N g DM−1)				Average
			Dry season		Rainy season
			Sampling 1	Sampling 2	Sampling 4	Sampling 6
Legume leaves + young stems	Cayman	SPS/UT	43.9 (5.6)	40 (2.6)	40.1 (1.4)	41.2 (2.6)	41.3 A
	Toledo	SPS/UT	40.6 (5.9)	40.9 (4.4)	42.7 (7.1)	41.7 (2.6)	41.5 A
Grass shoots	Cayman	GMC	12.3 (3.2)	11.4 (2.4)	9.4 (1.1)	18.5 (2.6)	12.9 a
		SPS/GO	19.3 (2.9)	17.5 (2)	18.8 (3.4)	23.5 (1.8)	19.8 bc
		SPS/UT	25.5 (5.7)	20.3 (2.5)	19.7 (3)	26.9 (1.6)	23.1 c
	Toledo	GMC	14.5 (2.5)	13.4 (3.9)	11.8 (1.2)	19.4 (2.3)	14.8 a
		SPS/GO	14.6 (2.7)	13 (2.5)	12.6 (2.8)	20.9 (2.3)	15.3 a
		SPS/UT	17.4 (3)	17.4 (3.9)	17.4 (2.2)	23.9 (2.9)	19.0 b

Values are means and standard deviations in brackets. Different uppercase letters indicate statistical differences of legumes’ N concentration with different accompanying grass according to Tukey’s HSD test (α=0.05). Lowercase letters indicate statistically significant differences in grass N concentration under different pasture types according to the same test. GMC: grass monocrop; SPS/GO: silvopastoral system grass-only area; SPS/UT: silvopastoral system under tree area.

The average N concentration of grass shoots was higher in both SPS treatments compared to GMC, regardless of the grass species, especially higher in the SPS/UT areas, with an increase from 12.9 to 23.1 mg N g DM⁻¹ in Cayman pastures, and from 14.8 to 19.0 mg N g DM⁻¹ in Toledo pastures (Table 1).

Under the assumption that grazing cattle would have an intake of the same legume proportions as the offered forage, which was of 21% in Cayman-SPS and 11% in Toledo-SPS (Fig. 1), the weighted CP of diets based on these pastures would be 8.0% for Cayman-GMC, 14.9% for Cayman-SPS, 9.25% for Toledo-GMC, and 12.1% for Toledo-SPS.

3.2. Biological N₂ fixation (BNF)

The δ¹⁵N signature of grass shoots, legume foliage, and soil was not statistically different between Cayman and Toledo pastures, except for the grasses in GMC (Fig. 2). The average δ¹⁵N of the soil in all pastures was around +5.7‰. In SPS, grasses had a δ¹⁵N between +2.78 and +3.02‰ for Cayman and Toledo, respectively, and legumes had a δ¹⁵N around +0.72‰ when associated with Toledo grass, and +1.67‰ when associated with Cayman grass. The largest difference to the δ¹⁵N signature of the legumes was observed in the grasses in GMC pastures, being +2.90‰ in Toledo-GMC, and +1.16‰ in Cayman-GMC.

Fig. 2.

δ¹⁵N (‰) signature of grass shoots, legume foliage, and soil of different pastures. Average of four samplings for grass shoots and legume pods (sampling 1, 2, 4 and 6), and three for soil (sampling 1, 2 and 4). GMC: grass monocrop; SPS/GO: silvopastoral system grass-only area; SPS/UT: silvopastoral system under tree area.

The average Ndfa of the legume L. diversifolia ranged between 30% and 58% in Cayman-SPS and between 52% and 79% in Toledo-SPS. In the last sampling of the rainy season, a notable increase in the N fixation rate was detected (Table 2).

Table 2.

δ¹⁵N (‰) signature of grass shoots and legume leaves in SPS and Ndfa (%) of legumes in SPS of two grass cultivars (Cayman, Toledo).

Sampling #	Cayman-SPS/UT			Toledo-SPS/UT
	δ15N (‰)		Ndfa (%)	δ15N (‰)		Ndfa (%)
	Grass	Legume	Legume	Grass	Legume	Legume
1	+4.40	+2.92	49%	+4.04	+1.69	57%
2	+3.78	+2.91	30%	+3.04	+1.24	52%
4	+1.87	+0.89	41%	+2.11	-0.21	68%
6	+2.04	+0.27	58%	+2.30	0.00	79%

SPS/UT: silvopastoral system under tree area. Each value represents the average of six replications (2 SPS/UT sampling areas per plot) per treatment per sampling.

N uptake by the grasses both in Cayman and Toledo pastures was higher in SPS than GMC, by 3–22 kg N ha⁻¹ sampling⁻¹ in Cayman-SPS, and up to 8 kg N ha⁻¹ sampling⁻¹ in Toledo-SPS (Table 3). Although N fixation of L. diversifolia was higher when associated with Toledo, the grass in Cayman-SPS pasture perceived a greater N benefit, possibly due to the greater legume proportion in Cayman-SPS (21% of total forage DM) compared to Toledo-SPS pastures (10% of total forage DM) (Fig. 1A).

Table 3.

N uptake of grasses in different pasture treatments over six samplings.

Pasture		Grass N uptake (kg ha−1)				‘Extra’ N uptake (kg ha−1) with respect to GMC				Weighted ‘extra’ N uptake with respect to GMC
						Sampling #
		1	2	4	6	1	2	4	6	1	2	4	6
Cayman	GMC	7	8	9	4	-	-	-	-	-	-	-	-
	SPS/GO	19	30	19	7	13	22	10	3	15	22	7	3
	SPS/UT	24	32	12	6	17	23	3	2
Toledo	GMC	13	19	10	5	-	-	-	-	-	-	-	-
	SPS/GO	18	20	15	4	5	0	5	-1	8	4	7	-1
	SPS/UT	24	29	18	4	11	10	9	-1

Each value represents the average of six replications per treatment per sampling. GMC: grass monocrop; SPS/GO: silvopastoral system grass-only area; SPS/UT: silvopastoral system under tree area.

3.3. C and N concentration of soil and plant litter

The plant litter in the SPS pastures contained two to three times more N than that of GMC pastures, significantly reducing its C/N ratio from 47–55 in GMC to 19–22 in SPS (Table 4). At the same time, the soil N concentration was slightly higher in the SPS than in the GMC pastures. The δ¹³C signature of the soil was significantly lower in the SPS than in the GMC pastures.

Table 4.

Concentrations and ratios of C and N and isotopic signatures of plant litter and soil in different pasture treatments.

Sample	Pasture		C concentration	N concentration	C/N	δ13C	δ15N
			mg g−1			‰
Plant litter	Cayman	GMC	337 (55) ab	9 (5) ab	47 (24) bc	-18 (5) b	-0.3 (0.8) a
		SPS/GO	340 (58) ab	18 (3) c	19 (4) a	-24 (3) a	+ 0.9 (1.2) b
		SPS/UT	371 (36) b	21 (4) c	19 (6) a	-25 (2) a	+ 0.5 (0.5) abc
	Toledo	GMC	311 (67) a	6 (2) a	55 (26) c	-14 (1) b	+ 0.4 (1.4) bc
		SPS/GO	329 (59) ab	13 (6) b	34 (26) ab	-21 (4) b	-0.1 (0.6) abc
		SPS/UT	354 (48) ab	18 (5) c	22 (13) a	-24 (3) a	+ 0.3 (0.5) abc
Soil	Cayman	GMC	15.0 (1.8)	1.33 (0.1) A	11.3 (1)	-15.0 (0.5) D	+ 5.9 (0.9)
		SPS/GO	15.2 (1.8)	1.50 (0.1) AB	10.2 (0.9)	-16.4 (0.6) BC	+ 5.8 (1.0)
		SPS/UT	16.8 (5.9)	1.54 (0.3) B	10.8 (2.0)	-18.5 (1.7) A	+ 5.2 (1.3)
	Toledo	GMC	16.2 (2.2)	1.44 (0.1) AB	11.2 (1.2)	-15.3 (0.8) D	+ 5.8 (0.8)
		SPS/GO	15.3 (2.0)	1.47 (0.2) AB	10.5 (1.3)	-15.9 (1.2) CD	+ 5.8 (0.5)
		SPS/UT	16.5 (2.2)	1.54 (0.2) B	10.7 (0.7)	-17.0 (1.0) B	+ 5.5 (0.6)

Values indicate the average and standard deviation in brackets of 6 replicates per sampling, with four samplings for plant litter and three samplings for soil (i.e. n = 18 for soil and n = 24 for plant litter). Different lowercase letters indicate statistical differences of plant litter for the respective variable between pasture treatments according to Tukey’s HSD test (α=0.05). Different uppercase letters indicate statistical differences of soil for the respective variable between pasture treatments. GMC: grass monocrop; SPS/GO: silvopastoral system grass-only area; SPS/UT: silvopastoral system under tree area.

3.4. Soil mineral N

The average soil N-NH₄⁺ concentration across all pastures for three samplings ranged from 5.0 to 8.8 mg kg soil⁻¹, whereas the N-NO₃^- concentration ranged from 7.3 to 16.9 mg kg soil⁻¹ (Table 5). The total soil mineral N (Nmin) concentration was lower in GMC (16.1 and 13.4 mg N kg soil⁻¹) than in SPS pastures (from 18.0 to 22.1 mg N kg soil⁻¹). Because there were no statistical differences in the N-NH₄⁺ concentration between pasture types, the differences in total Nmin were primarily attributed to variations in N-NO₃^-.

Table 5.

Soil (0–10 cm) mineral N concentration (NH₄⁺ and NO₃^-) in different pasture types in the first three soil samplings and the average of the three.

Pasture		N-NH4+(mg kg soil−1)				N-NO3-(mg kg soil−1)
		Sampling #
		1	2	3	Average	1	2	3	Average
Cayman	GMC	9.3 (6.9)	4.1 (2)	13.1 (12.3)	8.8 (8.6)	8.8 (2.3)	4.9 (2.3)	8.2 (2.9)	7.3 (3.0) a
	SPS/GO	4.9 (1.8)	4.2 (1.0)	7.5 (2.2)	5.5 (2.2)	20.8 (8.1)	10.6 (10.1)	12.5 (3.1)	14.6 (8.5) bc
	SPS/UT	6.2 (2.3)	3.8 (1.2)	7.5 (2.0)	5.9 (2.4)	25.6 (11.8)	12.8 (6.5)	12.1 (1.9)	16.9 (9.8) c
Toledo	GMC	4.6 (1.1)	4.9 (3.4)	5.5 (0.5)	5.0 (2.0)	13.9 (4.8)	4.9 (2.3)	7.3 (1.2)	8.7 (4.9) ab
	SPS/GO	5.8 (2.1)	6.8 (5.9)	6.1 (1.5)	6.3 (3.3)	20.7 (13.3)	11.5 (9.8)	8.4 (3.1)	13.5 (10.6) abc
	SPS/UT	9.0 (9.3)	4.0 (1.4)	6.2 (1.7)	6.2 (5.3)	20.2 (3.2)	6.9 (2.1)	13.5 (5.1)	13.5 (6.5) abc

Values are the average and standard deviation in brackets of 6 replicates per pasture type. Different lowercase letters indicate statistical differences of NO₃^- availability between pasture treatments according to Tukey’s HSD test (α=0.05). GMC: grass monocrop; SPS/GO: silvopastoral system grass-only area; SPS/UT: silvopastoral system under tree area.

In most pasture soils, the dominant Nmin pool was NO₃^-, especially in SPS soils, and the NO₃^-/NH₄⁺ ratio was close to 2 in Toledo and close to 3 in Cayman pastures.

3.5. Soil N₂O emissions

Over the N₂O measurement period, we observed three periods of significant rainfall. The first occurred from days 0–5 after water/urine application, with accumulated precipitation of 25 mm. The second occurred from days 28–40 with accumulated precipitation of 97 mm. The third occurred on the day 48 with a precipitation of 41 mm. The total precipitation during the 52 days of N₂O measurement was 168 mm (Fig S2).

The precipitation periods had a significant effect on the soil water-filled pore space (WFPS), reaching values as high as 70% after the rainfall events, and as low as 40% between rainfall periods. Nevertheless, the WFPS varied little among pasture treatments, with a daily standard deviation of 4–7% WFPS between treatments. Differences were not detected in WFPS between the areas that received water or urine (Fig S2).

Daily N₂O emissions from soil followed a pattern similar to that of soil WFPS, both in the areas that received water (+water) or urine (+urine) as initial input. Two major emission peaks, an initial pulse within the first 5 days and a second between days 28 and 40, coincided with rainfalls (Fig. 3 C, D and Fig. 4 C, D). Daily emissions from +water soils ranged between 0 and 7 mg N-N₂O m⁻² day⁻¹ throughout the 52 days. On the contrary, the +urine soils showed a first emission peak of about 10 times the order of magnitude observed in +water soils. For example, at day 4, the N₂O emissions of +urine Cayman pasture soils were 49, 87 and 54 mg N-N₂O m⁻² day⁻¹ for GMC, SPS/GO and SPS/UT, respectively, and of Toledo pasture soils 34, 73, 37 mg N-N₂O m⁻² day⁻¹ for GMC, SPS/GO and SPS/UT, respectively.

Fig. 3.

Accumulated (A) and daily (B) nitrous oxide emissions, and nitrate (E, F) and ammonium (G, H) concentration of soil in Cayman and Toledo pastures over 52 days of measurements after application of 1 L of water in the chamber area and surroundings. Each value corresponds to the mean of three chambers. Error bars in panels A and B correspond to the standard deviation of the mean. GMC: grass monocrop; SPS/GO: silvopastoral system grass-only area; SPS/UT: silvopastoral system under tree area.

Fig. 4.

Accumulated (A) and daily (B) nitrous oxide emissions, and nitrate (E, F) and ammonium (G, H) concentration of soil in Cayman and Toledo pastures measured over the period of 52 days after the application of 1 L of cattle urine in the chamber area and surroundings. Each value corresponds to the mean of six chambers. Error bars in panels A and B correspond to the standard deviation of the mean. GMC: grass monocrop; SPS/GO: silvopastoral system grass-only area; SPS/UT: silvopastoral system under tree area.

It is worth highlighting that in +water soils the second rainfall period generated a peak of emissions of similar magnitude to the first peak of emissions, whereas in the +urine soils the peak generated by the second rainfall period was approximately 80% lower than the first emission peak of emissions, despite the soil WFPS being similar for +water and +urine soils.

Despite the difference in the scale of N₂O emissions generated by the initial urine-N input, the pasture composition showed a significant effect on the N₂O emissions for both input types. In +water soils, the accumulated N₂O emissions were 4, 35, and 69 mg N-N₂O m⁻² for Cayman GMC, SPS/GO, and SPS/UT, respectively, and 5, 10, and 44 mg N-N₂O m⁻² for Toledo GMC, SPS/GO, and SPS/UT respectively (Fig. 3A and B).

Over the 52 days, soil NH₄⁺ and NO₃^- fluctuated in +water soils but the difference between pasture treatments was relatively constant. Nitrate concentration was usually higher in SPS than GMC soils whilst for NH₄⁺ it was rather the other way round: Excluding days of atypically high values, soil NO₃^- in Cayman was on average 63% higher in SPS pastures than GMC (16 vs 10 mg N-NO₃^- kg soil⁻¹), whereas soil NH₄⁺ in GMC (22 mg N-NH₄⁺ kg soil⁻¹) was 21% higher than SPS/GO (19 mg N-NH₄⁺ kg soil⁻¹) and 48% higher than SPS/UT (15 mg N-NH₄⁺ kg soil⁻¹). In Toledo pastures soil NO₃^- was 80% higher in SPS/GO (20 mg N-NO₃^- kg soil⁻¹) and 19% higher in SPS/UT (13 mg N-NO₃^- kg soil⁻¹) compared to GMC (11 mg N-NO₃^- kg soil⁻¹), whereas soil NH₄⁺ in GMC (21 mg N-NH₄⁺ kg soil⁻¹) was 10% higher than SPS/GO (19 mg N-NH₄⁺ kg soil⁻¹) and 43% higher than SPS/UT (15 mg N-NH₄⁺ kg soil⁻¹). N₂O emissions from +water soils were highest in SPS/UT pastures for both Cayman and Toledo, which were the pastures with the highest NO₃^- concentration and lowest NH₄⁺ concentration in soil.

Accumulated N₂O emissions from +urine soils were 352, 613, and 404 mg N-N₂O m⁻² for Cayman GMC, SPS/GO, and SPS/UT, respectively, and 453, 685, and 540 mg N-N₂O m⁻² for Toledo GMC, SPS/GO, and SPS/UT, respectively (Fig. 4A and B). Although NO₃^- production was higher in all SPS pastures compared to the GMC (Fig. 4E and F), higher N₂O emissions were only observed in the SPS/GO areas but not in the SPS/UT areas (Fig. 4C and D).

The higher N₂O emissions from +urine soils compared to +water soils were likely related to in-situ nitrification of N added with urine. On the first day after the application of urine, the soil NH₄⁺ concentration in Cayman pastures was enriched in 196, 138, and 148 mg N-NH₄⁺ kg soil⁻¹ in GMC, SPS/GO, and SPS/UT, respectively, compared to the NH₄⁺ concentration of the soils that only received water. Toledo pasture soils were enriched by 87, 78, and 123 mg N-NH₄⁺ kg soil⁻¹ in GMC, SPS/GO, and SPS/UT, respectively, in the +urine compared to +water areas. The N load applied was equivalent to 166 mg N kg soil⁻¹.

By the 11th day after the application of urine when most of the NH₄⁺ was consumed, most pastures showed the peak of NO₃^- production, with 56–109 mg N-NO₃^- kg soil⁻¹ of additional NO₃^- compared to +water soils on the same day for Cayman and about 89 mg N-NO₃^- kg soil⁻¹ of additional NO₃^- compared to +water soils on the same day for Toledo pastures.

The accumulated N₂O emissions in both +water and +urine conditions differed statistically between pasture types but not between grass species (P = 0.08) (Table 6). Under both +water and +urine conditions, N₂O emissions were higher in SPS than GMC, but statistically significant in only one of the two SPS zones. In the case of +water soils, only the difference in emissions of SPS/UT and GMC (57.1 vs 4.8 mg N-N₂O m⁻²) was statistically significant. In +urine soils, only the increase of emissions in SPS/GO with respect to GMC (649 vs 402 mg N-N₂O m⁻²) was statistically significant.

Table 6.

Accumulated nitrous oxide emissions over 52 days of measurements after application of 1 L of water or urine in manual static chambers.

Pasture type		Accumulated emissions in 52 days (mg N-N2O m−2)
		+water (n = 3)	+urine (n = 6)
GMC		4.8 A	402 a
SPS/GO		22.9 A	649 b
SPS/UT		57.1 B	472 ab
Standard error		8.1	52
P-value	Pasture type	0.00	0.00
	Grass species	0.08	0.08
	Grass × Pasture type	0.43	0.91

Each value corresponds to the Least-Squares Means and are averaged by pasture type regardless of the grass species, which was not statistically significant. Different letters next to the treatment’s “lsmeans” indicate significant differences between pasture types according to Sidak test (α=0.05). GMC: grass monocrop; SPS/GO: silvopastoral system grass-only area; SPS/UT: silvopastoral system under tree area.

4. Discussion

4.1. Forage productivity and N sources used in GMC and SPS pastures

Toledo pastures were on average 35% more productive than Cayman pastures. One likely explanation for this difference lies in the contrasting water-use strategies of the two grasses. Cayman is classified as a “water spender”, a strategy associated with high transpiration rates, deep rooting, and limited tolerance to prolonged dry seasons. Its optimal ecological niche corresponds to 900–1200 mm of annual precipitation. In contrast, Toledo is considered a “water saver”, capable of tolerating both short and long dry seasons and adapted to lower rainfall conditions of 700–800 mm yr⁻¹ (Cardoso and Rao, 2019).

Given that total precipitation during the experimental period (August 2021–July 2022) was 855 mm, it is plausible that Cayman experienced moderate drought stress and was therefore unable to reach its productive potential. These differences were likely amplified by the physical condition of the soil at the site. The soil contained 40–44% of both clay and silt, and 15% sand, with bulk density values ranging from 1.45 to 1.64 g cm⁻³ in the topsoil (0–5 cm), and 1.60–1.69 g cm⁻³ in deeper soil layers (20–60 cm, Table S3). For soils of this texture, such bulk densities indicate possible compaction, creating substantial mechanical impedance to root growth. Under these conditions, root morphology becomes a major determinant of productivity, especially when water availability is already at the lower end of the optimal range for a species’ water-use strategy. Although Cayman is a decumbent grass, we observed lower soil cover compared to Toledo at our site; this reduced cover likely contributed to higher topsoil bulk density in Cayman pastures (Table S3) due to animal trampling during the wet season.

Evidence from studies on Urochloa genotypes under drought indicates that stress resistance depends largely on the ability of the root system to capture water from deeper soil layers (Cardoso and Rao, 2019). Compacted soils, however, restrict this ability in genotypes with thinner nodal roots. Recent research shows that nodal root diameter in tropical forage grasses is strongly correlated with the ability to penetrate hard soil layers (Huot et al., 2025). Toledo exhibits larger nodal root diameters (∼ 2.0 mm, Cardoso et al., 2013, De la Cruz Jiménez et al., 2021). In contrast, Cayman has finer nodal roots (∼1.5 mm, Cardoso, Mayorga Cobos, 2024). Since deep vertical root growth is known to contribute to greater root penetration capacity in compacted soils, it is highly likely that Cayman’s productivity was constrained by its inability to access deeper moisture reserves caused by restriction of root penetration. On the other hand, Toledo, with its thicker roots, could probably be less restricted to explore soil profiles and maintain superior productivity over the course of the experiment. Thus, the finer root system of Cayman might have prevented it from penetrating the compacted soil layers to access moisture, effectively causing a "functional drought," whereas the thicker-rooted Toledo might be capable to penetrate compacted soil layers and maintain productivity.

The combined grass plus legume forage DM production over the six samplings indicated 30% higher forage availability in SPS compared to GMC pastures (Fig. 1A). Under similar pedo-climatic conditions as our study, Calle et al. (2013) reported 27% higher biomass production in SPS with L. leucocephala, compared to GMC pastures, and 64% increase in CP of the SPS forage diet. Our CP measurements suggest a similar increase in SPS, with comparable CP concentrations to those compiled by Chará et al. (2019b) for GMC (10.8% CP) and SPS with L. leucocephala (13.9–15.7% CP) in diverse agroecological conditions of Colombia. In both studies, the density of L. leucocephala shrubs per unit area was higher (4000–40,000 ha⁻¹), and all had access to browsing by cattle.

Higher forage productivity in the SPS than GMC could be explained by the higher N uptake in SPS pastures (Trumpp et al., 2025). The legume L. diversifolia showed moderate rates of N₂ fixation, with an average of 45% Ndfa in Cayman-SPS and 64% in Toledo-SPS, which were comparable to previous reports for the same species using the ¹⁵N pool dilution method [42–54% Ndfa, (Rao and Giller, 1993)]. Although N fixation rates were higher in Toledo-SPS pastures, possibly due to either higher soil P availability or lower N availability, the increase in N uptake observed in Cayman-SPS pastures with respect to the GMC pasture was higher than in Toledo (Table 3), possibly due to the higher proportion of legume in Cayman-SPS compared with the Toledo-SPS. The highest legume N fixation rates in both SPS were detected in the last sampling of the rainy season (Table 2), possibly as a mechanism to stimulate plant growth when biomass production was decreasing (Fig. 1B).

Few studies have quantified N₂-fixation by Leucaena species in grazed pastures and published research has focused almost exclusively on L. leucocephala. Under similar soil conditions as our study, Conrad et al. (2018) and Sarabia-Salgado et al. (2020) reported in neutral to alkaline Vertisols, that 73% and 95% of L. leucocephala-N was derived from atmospheric fixation in grazed pastures. In the isotope system that we evaluated (SPS/UT), the average δ¹⁵N signature of soil was +5.36‰, whereas that of grasses was +2.95‰, and the value of legumes was +1.22‰. As the non-N₂-fixing reference plant, we used the grass growing near the L. diversifolia trees to ensure that both the reference and legume would have access to the same soil N source. Oberson et al. (2013) recommended sampling different reference plants to reduce uncertainty in the determination of N₂ fixation with the ¹⁵N natural abundance method. In the study of Sarabia-Salgado et al. (2020), the authors used the average δ¹⁵N signatures of grasses and forbs sampled in the plot as a reference plant. Because the δ¹⁵N signature of the forbs sampled was as high as +7.5‰, the resulting reference value was about 1.0‰ to 4.4‰ higher than that of the grasses sampled. In our study system, we also sampled forb plants, but no single species was present in most of the plots to be considered as a suitable reference plant. Because the δ¹⁵N signature of our grasses was substantially lower than that of the soil N (by 2.40‰), it is possible that by using them as non-fixing reference plants we would underestimate the N₂-fixation capacity of L. diversifolia to some degree. If a reference plant with a δ¹⁵N signature only 1‰ higher had been used, the Ndfa calculation would increase by at least an additional 10%.

The foliar N concentration of grasses was higher in SPS than GMC pastures (Table 1). However, because the δ¹⁵N signatures of grasses growing near legumes in SPS systems were not lower than the δ¹⁵N signatures of grasses growing alone (GMC), the extra N uptake observed by grasses in SPS cannot be interpreted as direct N transfer from the legume to the grass, but rather as an indirect benefit through an enhanced N availability from soil mostly in the form of nitrate-N (Table 4). Gubsch et al. (2011) have proposed that increased leaf N concentration of plants growing near legumes without reduced δ¹⁵N indicates uptake of additional mineral N from the soil that is not consumed by legumes. This process has been referred to in other studies as ‘complementary N uptake’ (Bessler et al., 2012) or ‘N sparing’ (Chalk, 1998).

Cayman grass in GMC system had a lower δ¹⁵N signature (+1.1‰) than all the other grasses, regardless of whether associated with L. diversifolia or not (Fig. 2), and the δ¹⁵N signature was comparable to that of the legumes (average of +1.7‰ and +0.7‰ for L. diversifolia in Cayman-SPS/UT and Toledo-SPS/UT, respectively). Such an exceptionally low δ¹⁵N signature might indicate associative N₂ fixation by the grass when grown as GMC. The ability of Urochloa grasses to obtain N by associative fixation has been reported in previous studies (Boddey and Victoria, 1986, Reis et al., 2001). Moreover, Villegas et al. (2020) suggested that Urochloa grasses might also swap N sources when grown in association with legumes.

4.2. Effect of L. diversifolia on soil N cycling and N₂O emissions

Nitrous oxide emissions from soils with no additional N input (+water treatment) were higher in SPS pasture soils, especially in SPS/UT areas compared to GMC areas (Table 6, Fig. 3A and B). The higher N₂O emissions in SPS pasture soils could be explained by the effect of the legume on the N cycling through plant litter (Cruz et al., 2026). The plant litter in the SPS pastures, consisting of grass residues with higher N concentration and also legume residues (e.g. leaves, seeds and pods) was richer in N than the plant litter of GMC (Table 4). Indeed, this was reflected by the observed value of the C/N ratio of the plant litter, which was significantly lower for SPS (average 23) than GMC (average 50). Robertson and Groffman (2015) proposed that organic materials with a C/N ratio lower than 25 stimulate N mineralization, whereas materials with C/N ratio higher than 25 stimulate N immobilization. Higher N mineralization in SPS pasture soils might explain the higher soil Nmin (NH₄⁺ + NO₃^-) observed in SPS than in GMC soils (Table 5). Since denitrification is a process carried out mostly by heterotrophs (Zhu et al., 2025), such microorganisms likely benefit as well from the potentially greater availability of C mineralized from plant litter of SPS than that of GMC (Table 4).

The effect of the legume in the soil N cycling can be also evidenced by the incorporation of legume biomass C in the soil organic matter. The soils in SPS pastures have not only more total topsoil N, but also a more depleted δ¹³C signature, characteristic of the legume biomass C (−29‰ for legumes and −13‰ for grasses, regardless of the pasture type, Table S2). In the Cayman-SPS/UT area, the soil’s δ¹³C signature was depleted by −3.5‰ compared to the GMC, and in the Toledo-SPS/UT area, the soil’s δ¹³C signature was depleted by −1.7‰ compared to the GMC (Table 4). In similar SPS, higher N stocks in subsoil (20–40 cm) have been reported a few years after the introduction of the tree legumes (Pessoa et al., 2025).

At high N input (+urine treatment), N₂O emissions were several times higher than those from no additional N input, likely due to nitrification of added N. Accumulated emissions were higher in the SPS than in the GMC pastures. In the SPS/GO areas, the increase in the N₂O emissions was of 61% compared to the GMC and statistically significant (P = 0.00), whereas in the SPS/UT, the increase was of 17% but not statistically significant. In our study, spatial variability was much more relevant at high N input than at no additional N input. Even though in the +urine treatment, we sampled 12 gas chambers per pasture type (GMC, SPS/GO, and SPS/UT) across both Cayman and Toledo, the observed intra-group variability was substantial, complicating the detection of consistent emission patterns.

One limitation of our experiment with urine patches lies in the use of the same N load for all pastures. While in +water soils, differences in the background Nmin (34–37% higher in SPS than GMC) might have been the main driver of N₂O emissions, the addition of the same amount of high load of urine-N to all pasture treatments might have masked existing differences in the native Nmin pool. Moreover, the N₂O emissions from urine patches in GMC pastures were likely overestimated. Durango et al. (2021) evaluated animal N excretion of Cayman GMC and SPS with L. leucocephala under the same pedo-climatic conditions as our study, and reported 55% higher urine-N excretion of cattle grazing SPS (6.1 g N L⁻¹) than cattle grazing GMC (3.9 g N L⁻¹). In our study, we applied urine with an N concentration of 6.1 g N L⁻¹, corresponding to urine-N excretion levels of animals grazing SPS. In future studies, the use of different N loads may be more appropriate for every pasture treatment according to the expected animal N excretion.

4.3. SPS as a climate change mitigation strategy in tropical livestock systems

Our results indicate that SPS emit more N₂O than GMC pastures. However, in order to provide a comprehensive diagnosis of the GHG balance of the production system, it is imperative to consider potential trade-offs with other sources of emission and the production of commodities (i.e. emission intensity) (Moraes et al., 2025). For example, Molina-Botero et al. (2024) analyzed the weight gain of Brahman x Angus beef cattle grazing the same pastures as the present study for 12 months. They found 13% higher (P = 0.025) animal liveweight gain (LWG) and 15% lower net methane emissions (P = 0.007) in the cattle grazing the SPS compared to those grazing the GMC pastures, due to better forage diet quality, and the presence of anti-methanogenic compounds contained especially in the legumes.

In the Caribbean region of Colombia, Rivera et al. (2024) used a life-cycle assessment approach to calculate GHG balances of SPS with L. leucocephala compared to GMC pastures, considering cattle enteric CH₄ emissions, N₂O emissions from urine and feces (but no background N₂O emissions), and sinks due to C sequestration. They found that the inclusion of L. leucocephala increased animal production (LWG) up to four times compared to the baseline scenario, and mitigated GHG emissions per hectare up to 86% mostly due to C sequestration in soil. Thus, despite the risk of increasing N₂O emissions, considering both soil processes and animal-level impacts, SPS not only improves forage yield and quality but also enhances system sustainability by improving feed efficiency, reducing methane emissions intensity, and fostering C sequestration, possibly leading to a positive trade-off for environmental and productivity outcomes.

Our study underscores the significance of considering spatial heterogeneity and measuring background N₂O emissions in grazed pastures to accurately assess overall N₂O fluxes. While previous research has often focused on N₂O emissions from urine patches and the background emissions are used just as a control for the calculation of N₂O emission factors, our findings demonstrate that these background emissions can differ significantly depending on the pasture composition, and though generally low compared to the emissions of urine patches, they could be substantial when scaled up to larger areas or over a longer term. For example, the accumulated emissions of Cayman-SPS +water (averaging the GO and UT area) in our experiment over 52 days were 48 mg N-N₂O m⁻². Assuming that emissions were constant over the year and considering a global-warming potential (GWP) of N₂O of 273 CO₂eq (AR6, IPCC), the yearly CO₂eq emissions would be 1478 kg CO₂eq ha⁻¹ yr⁻¹ for Cayman-SPS pasture, compared to 121 kg CO₂eq ha⁻¹ yr⁻¹ for Cayman-GMC pastures. These emission dynamics suggest that nitrification-driven N₂O production in SPS soils may be strongly influenced by moisture pulses following N-rich inputs, particularly under tree canopies.

On the other hand, assuming that we had captured all the N-N₂O emissions from the urine patch during the 52 days of experimental period, the total CO₂eq emissions would be equivalent to 0.03–0.07 kg CO₂eq L urine⁻¹. Using the urine volume excreted per animal per day for Cayman-GMC and Cayman-SPS as reported by Durango et al. (2021), the annual CO₂eq emissions of urine patches would range between 112 and 245 kg CO₂eq animal⁻¹ yr⁻¹. We acknowledge that these calculations oversimplify the reality observed in the field, however, although with significant assumptions, they illustrate the importance of not ignoring the soil-borne N₂O emissions, especially in SPS towards determining an appropriate GHG balance of the production system.

5. Conclusion

This study partially confirmed our initial hypothesis. Integrating L. diversifolia into Urochloa-based silvopastoral systems significantly increased forage productivity and nitrogen uptake compared to grass monocrops, demonstrating the potential of SPS to improve pasture performance. However, higher soil nitrification and elevated N₂O emissions observed in SPS highlight possible environmental trade-offs that require careful management.

To maximize benefits, farmers should regulate legume density and implement rotational grazing to minimize urine deposition hotspots. These practices, combined with the carbon sequestration potential of trees, can help offset localized N₂O emissions. At a broader scale, incorporating SPS with L. diversifolia offers an opportunity to reduce greenhouse gas emission intensity in livestock systems by increasing productivity and lowering enteric methane emissions. This approach aligns with national climate pledges aiming to reduce agricultural emissions while improving food security. Future GHG inventories should account for both emission sources and sinks, including soil and biomass carbon stocks, to accurately reflect the mitigation potential of silvopastoral systems.

CRediT authorship contribution statement

Daniel M. Villegas: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Catalina Trujillo: Writing – review & editing, Methodology, Data curation. Idupulapati M. Rao: Writing – review & editing, Methodology, Conceptualization. Enrique Murgueitio: Writing – review & editing, Methodology, Funding acquisition, Conceptualization. Julian Chará: Writing – review & editing, Methodology, Funding acquisition, Conceptualization. Julian E. Rivera: Writing – review & editing, Methodology, Conceptualization. Jon Moorby: Writing – review & editing, Project administration, Methodology, Funding acquisition, Conceptualization. Mauricio Sotelo: Writing – review & editing, Methodology. Juan A. Cardoso: Writing – review & editing, Methodology, Funding acquisition. Astrid Oberson: Writing – review & editing, Validation, Supervision, Methodology, Conceptualization. Jacobo Arango: Writing – review & editing, Supervision, Project administration, Methodology, Funding acquisition, Conceptualization.

Declaration of Generative AI and AI-assisted technologies in the writing process

During the preparation of this work the authors used Microsoft Copilot in order to improve the readability of specific sentences and code debugging during statistical analysis. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Funding

This study was funded by the UK’s Biotechnology and Biological Sciences Research Council (BBSRC) grant BB/S01893X/1 “Advancing sustainable forage-based livestock production systems in Colombia” (CoForLife). Funding was also received from the CGIAR Science Programs on Sustainable Animal and Aquatic Food Systems (SAAF), Climate Action (CA), Multifunctional landscapes (MFL), and Scaling for Impact (S4I).

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary material

Supplementary material