1. Grasslands store a third of global terrestrial carbon but are vulnerable to carbon loss due to inappropriate livestock grazing. Grasslands management can be improved with a mechanistic understanding of biogeochemical processes that determine carbon storage, such as plant litter decomposition.

2. Herbicides, such as glyphosate, are used to improve the quantity and quality of the forage. In the Flooding Pampa, the most extensive cattle-grazed natural grassland and one of the few remnant temperate grasslands in South America, glyphosate is applied to promote Lolium multiflorum, a forage grass associated with a fungal endophyte non-toxic for cattle.

3. We studied five mechanistic pathways in which the application of glyphosate can alter litter decomposition. We grouped them into single application pathways, through effects on living plants (1), leaf litter (2) and bare soil (3), and repeated annual application pathways, through legacies on ecosystem properties (4) and through the growth of an annual forage grass with a fungal endophyte (5). Single application pathways were tested in a greenhouse experiment using leaf litter of L. multiflorum and of a native dominant grass. Repeated annual application pathways were tested through a field experiment with 3-year annual glyphosate application using leaf and root litter of L. multiflorum with and without endophyte association.

4. Glyphosate application on living plants produced leaf litter with 70% higher nitrogen content and 140% higher decomposition constant than naturally senesced litter. In contrast, glyphosate application on naturally senesced leaf litter reduced decomposition constant by 20%. Glyphosate application on the soil did not affect the decomposition of naturally senesced leaf litter but accelerated the decomposition of the glyphosate-killed plants even more.

5. Legacies of repeated annual application of glyphosate resulted in a notable reduction in plant cover (45%) and potential soil respiration (57%), with a consistent acceleration of leaf (53%) and root (18%) litter decomposition. Furthermore, the association of endophytes in L. multiflorum plants reduced leaf litter decomposition by 22%. On the contrary, the association of endophytes did not alter root litter decomposition.

6. Glyphosate application on living plants and legacies of repeated application on the ecosystem stimulate litter decomposition, which can result in a net carbon loss from grasslands. In other ecosystems, the net result on decomposition would depend on the relative cover of vegetation, aboveground litter, and bare soil. This study highlights that glyphosate application should be considered when evaluating sustainable management to preserve and enhance soil carbon storage in grasslands.

Study site and application of glyphosate

The study site was a humid mesophytic grassland in the Flooding Pampa, a vast region of around 9 million hectares in the province of Buenos Aires, Argentina. The mean annual temperature is around 15°C and the mean annual rainfall is 885 mm (Soriano and Paruelo 1992). The landscape has a treeless physiognomy and an extremely flat topography with periodic flooding during autumn–spring in lowland, except in ridge areas with well-drained sandy soils (Burkart et al. 1998). The field experiment was carried out in a commercial livestock farm (35° 01´S, 57° 50´ W). The plant community is dominated by C₃ and C₄ grass species (see details in Druille et al. 2015). Soil is classified as a typical Natracuol (US Soil Taxonomy), characterized by having a nonsaline acid A1 horizon, and a highly alkaline saline B2 horizon (Lavado and Taboada 1988).

Glyphosate was not applied at the study site before, even though glyphosate application in the late summer is a common practice in the region with a 3 l/ha dose (Rodriguez and Jacobo 2010). We applied this dose (1440 g of acid equivalent/ha) of a commercial glyphosate formulation Glacoxan® in field and greenhouse experiments with a 20 l backpack sprayer with a constant pressure of 3 bars.

Pathways of single-glyphosate application

To evaluate pathways of single application of glyphosate effects through living plants (1), leaf litter (2), and soil (3), we set up a litter decomposition experiment in a greenhouse. For these pathways, we used plant material that is naturally found in the field at the end of the summer when glyphosate is applied in the Flooding Pampa. At that time of the year, Paspalum dilatatum, the native dominant perennial C₄ grass, can be found as a living plant and as plant litter. In turn, Lolium multiflorum, which is an introduced annual winter C3 forage grass, is only found dead as plant litter. Considering that in this grassland the vast majority of L. multiflorum plants are endophyte-infected with Epichloë occultans (Gundel et al. 2009), in this experiment we used L. multiflorum plants associated with the endophyte. Together, for the living plant pathway we used P. dilatatum (1) and for the leaf litter pathway (2) we used litter produced by P. dilatatum and L. multiflorum with endophyte plants.

Paspalum dilatatum and Lolium multiflorum plants were grown in 1 m x 1 m monoculture plots in the experimental campus of the School of Agronomy at the University of Buenos Aires. L. multiflorum plants grew from seeds with naturally high level (82%) of endophyte association (E+), and from seeds without endophyte (E-) obtained experimentally following Omacini et al. (2004). We collected fresh senesced plant litter of both species and, in the lab, sorted leaf litter from other plant organs. Then, in the P. dilatatum plots, we removed all dead plant material to applied glyphosate on living plants. After 15 days, we collected P. dilatatum plants killed by glyphosate and separated the leaf litter from other organs. We determined the total carbon (%C) and nitrogen (%N) content of all types of litter by Dumas combustion with a TruSpec elemental analyzer (LECO, St. Joseph, MI, USA) at the University of Buenos Aires.

We prepared litterbags containing leaf litter from P. dilatatum plants killed by glyphosate (Plant Gx) and from naturally senesced P. dilatatum and L. multiflorum E+ plants. Litterbags were made of fiberglass mesh, which is the most common used material for litter decomposition studies (Harmon et al. 1999, Bradford et al. 2002), and that we have successfully used before (Omacini et al. 2004, Vivanco and Austin 2006, 2019).  We used 0.5 g of each litter type in 11 cm x 9 cm litterbags with a 3 mm opening on the upper face and a 2 mm opening on the lower face. We prepared plastic containers with 1.2 kg of soil from the study site, which had not received prior glyphosate treatment. Half of the litterbags containing naturally senesced leaves were sprayed with glyphosate (Litter G+) and the other half was sprayed with water (Litter G-). Half of the soil containers were sprayed with glyphosate (Soil G+) and the other half was sprayed with water (Soil G-). We assigned litterbags (Plant Gx, Litter G+, Litter G-) to soil containers (Soil G+, Soil G-) in a factorial design and kept them moistened with regular watering (n=5).

We assessed litter decomposition as litter mass loss over time. We collected litterbags after 140 and 270 days of incubation. Litterbags were dried for 48 h at 65°C; soil and debris were removed from litter and were oven-dried again for determination of dry mass. We estimated the decomposition constant k using a single exponential decay model by regressing the log of the fraction of mass remaining against time. The decomposition constant integrates the dynamics of litter mass loss over time and it is a useful parameter to compare between litter types and treatments (Wieder and Lang 1982). We used ln (M_t/M_o) = –kt, where M_o is the initial dry mass, M^t is the dry mass at time t, and k is the decomposition constant (Swift et al. 1979). Linear regressions were performed by setting the intercept to zero. In the few cases when samples did not fit a significant regression, values were considered outliers and were replaced by the mean of the treatment, following the missing value procedure (Steel and Torrie 1980, Vivanco and Austin 2008).

Pathways of repeated annual application of glyphosate in natural grasslands

We evaluated pathways of repeated annual application of glyphosate through legacies in ecosystem properties (4) and through the enhancement of endophytic grass (5) (Fig. 1) on decomposition of leaf litter and roots in a field experiment in the Flooding Pampa. In this field experiment, we previously studied the impacts of glyphosate application on beneficial soil microorganisms (Druille et al. 2013, 2015, 2016). We established 10 plots (1.5m x 1.5 m) in an area of similar floristic composition and randomly assigned them to control (Ecosystem G-) or glyphosate application (Ecosystem G+) treatments. Every April (late summer in the southern hemisphere) for three consecutive years, we applied 3 l / ha of water to Ecosystem G- plots and 3 l / ha (1440 g acid equivalent / ha) of commercial glyphosate formulation (Glacoxan®) to Ecosystem G+ plots. We applied these treatments using a 20 l backpack sprayer with a constant pressure of 3 bars. Cattle grazing was avoided during the experiment by keeping an electric wire around the experimental area. To avoid biomass accumulation and the consequent aging of grasslands, we made a harvest of plant biomass using a lawn mower set to leave 10 cm stubble every year before application of the treatment.

To evaluate pathways of repeated annual application of glyphosate, we used litter produced by plants of L. multiflorum with (E+) and without (E-) endophyte that was accumulated above and below ground (leaf and root litter). We prepared 14 cm x 14 cm litterbags made of 2 mm fiberglass mesh. We placed leaf litterbags on the ground and root litterbags buried 5 cm belowground. Considering that the place where the litter was deposited (above and belowground) can interact with the type of litter (leaf and root litter), we placed a common substrate (stem litter) litterbags on the ground and buried at 5 cm to assess the effects of the above and belowground environment. The experiment started 15 days after the third year of application of glyphosate (n = 4) and we collected litterbags at 30, 140 and 260 days. We assessed ash-free dry mass (500°C oven for 4 h) to estimate the decomposition constant k as described in Section 2.3. Together, this experiment evaluated the relative importance of pathways 4 and 5 and provides information about the effect of an aerial symbiosis on root decomposition of the host, which has not been evaluated previously.

We assessed above and belowground ecosystem properties in Ecosystem G- and Ecosystem G+ plots. We measured plant cover in December (when the last litterbag pickup occurred) in 10 plots of each level of glyphosate application. For estimation of plant cover, we used the line intercept method proposed by Canfield (1941). We determined potential water evaporation at ground level by measuring the water loss of wet filter papers. We used preweighted oven-dried filter papers and wet them in the field to full water-holding capacity. Filter papers were weighed immediately before and after incubation on the ground for 1 hour at midday in May to calculate water loss. We measured two filter papers per plot in 5 replicates for each level of glyphosate application.

We determined soil gravimetric water content from 10 cm depth soil cores taken in August and December (second and third litterbag harvest dates, respectively). We also determined soil organic matter content and soil potential respiration from soil cores taken in May, approximately one year after the decomposition experiment was installed in the field. Soil organic matter content was determined by total combustion in an oven at 500°C for 4 hours. We determined soil potential respiration by incubating a 15-g sample at 25°C, in a 200-ml vial with gastight septum caps. The soil was pre-incubated at water field capacity for 48 h without seedlings or any plants. CO₂ production was measured 2, 4 and 7 days after a 24-h incubation period with an infrared gas analyzer (PP Systems EGM-4, Amesbury, Massachusetts, USA). We used five replicates per level of glyphosate application for soil measurements.