The study was realized at the Rio Mayo Experimental Station of the Instituto Nacional de Tecnología Agropecuaria (INTA, 45º24′11″S; 70º17′37″W), located at the southwest of Chubut Province, Argentina. The site is characterized by a shrub-gramineous steppe from the southern of the Western District of the Patagonian Province (Oyarzabal et al. 2018). Soils are of coarse texture (sandy) with a cemented-calcareous layer at a depth of about 45 cm and have very low organic matter content (0.40%; Paruelo et al., 1988). Mean annual precipitation in the study site is close to 150 mm, and occurs as small events (88% are lower to 5 mm, Cavallaro et al. (2020)). Mean monthly temperature varies from 1°C in July to 15°C in January.
The Patagonian steppe is characterized by vegetation with low species diversity and low plant density, with a plant cover of between 30 and 40% (Pereyra et al. 2017). The vegetation is distributed in patches composed of shrubs and grasses on a bare soil matrix (Soriano et al. 1994). The dominant shrub species are Azorella prolifera (Cav.) G.M. Plunkett & A.N. Nicolas (ex Mulinum spinosum), Adesmia volckmannii Phil. and Senecio filaginoides DC. The dominant grass species are Pappostipa speciosa (Trin. & Rupr.) Romasch., Pappostipa humilis (Cav.) Romasch. and Poa ligularis Nees ex Steud. While shrubs have dimorphic root systems with a taproot and lateral roots (Bucci et al., 2009), grasses have superficial fibrous root systems that are horizontally distributed in the upper 20 cm of the soil profile (Soriano et al. 1987).
Until 2013, the livestock management practice in the study site consisted in a low stocking rate of 0.14 sheep per hectare only from May to October. During the rest of the year the animals were moved to sites with abundant forage. A 4 ha enclosure was delimited in May 2013, and fertilization and irrigation treatments were established. The experimental design consisted of 20 plots of 625 m2 each, separated from each other by 15 m, which were randomly assigned to one of the four treatments (control, fertilization (F), irrigation (I) and fertilization-irrigation (I + F)). The plots with the irrigation treatment were equipped with a semi-automatic sprinkler irrigation system. Irrigation water was extracted from a local well of 6 m depth and stored in 5 containers with a total capacity of 13000l. Due to the proximity to the Andean mountains, electric conductivity of groundwater, determined using a conductivity meter (HI 98311, Hanna instruments, Woonsocket, USA) was low (0.16 dS/m), similar to other wells in the research area. The annual precipitation in irrigation plots was increased by approximately 20–25%, and in each irrigation event ~ 5 mm of water were applied, being 6–8 irrigation events per year. The frequency of irrigation events depended on weather conditions (no wind and no natural rain events close to the date of irrigation treatments). Irrigation was carried out during spring and summer, and was canceled during fall and winter, which is the wet period and also to avoid damage to the irrigation system due to freezing of water near the rooting layers. Nutrient addition was done twice a year (once in autumn and once in early spring) applying urea and diammonium phosphate in an amount of 100 kg/ha/year of nitrogen and 75 kg/ha/year of phosphorus, since N and P are the two most limiting nutrients. Another reason for the addition of both nutrients is that nitrogen deposition could further reduce the availability of phosphorus in the soil (Zhang et al. 2013), but the addition of phosphorus could mitigate this limitation, as has been reported for other steppes (Huang et al. 2018).
Soil physico-chemical properties
In spring, 6 cores from the upper soil layer (5 cm diameter, 5 cm depth) were obtained randomly (3 cores close to vegetation and 3 in bare soil) and mixed well to produce one composite sample (n = 3, three of the five plots of each treatment). All soil samples were analyzed for organic matter, total nitrogen, available phosphorus, pH and electrical conductivity. Soil organic matter, total nitrogen and available phosphorous were analyzed in the Soil Laboratory (INTA Chubut, Argentina). Soil organic matter was determined using the wet oxidation method (Walkley and Black 1934). Soil total nitrogen was determined with the Kjeldahl method (Bremner 1996). Available phosphorus was measured using the Olsen method (Olsen et al. 1954). Soil pH was determined in a solution 1:2.5 of distilled and deionized H2O. Electrical conductivity was determined in a saturated extract and corrected for temperature, and values were reported in dS m− 1 at 25ºC.
Soil respiration, temperature and water content
Soil respiration was determined in all treatments, twice per season throughout a year. In each plot, one measurement was made near vegetation and the other on bare soil (at least 50 cm away from vegetation), and then the measures were recalculated for the percentage of coverage of each type of patches and then were added together. A portable closed chamber (6400-09, LI-COR) connected to a gas exchange system (LI-6400, LI-COR) was used to measure soil respiration. Soil PVC collars (10 cm in diameter and 4.4 cm in height) were inserted into the soil at 2.2 cm depth one day before each measurement date to allow stabilization of the CO2 efflux before the measurements. At each measurement date new collars were installed in other sites of each plot. Each measurement took between 5 to 20 minutes, depending on the time of year (in winter the measurements took longer because the rates were relatively low), and included three averaged consecutive cycles. To avoid the effects of strong diurnal fluctuations in air temperature on soil CO2 efflux, measurements were made between 11:00 and 15:00 h on consecutive days with similar climatic conditions in each study period. In a previous study in this site, Silletta et al. (2019) observed a transient CO2 realease in vegetated patches immediately after a rain pulse of only 3 mm, called “Birch effect”. Therefore, soil respiration determinations in this study on irrigation and I + F plots were performed at least five days after irrigation events.
At the time of each soil respiration measurement, soil volumetric water content was measured with ECH2O probes (Decagon Devices, Inc.) at 0–10 cm depth near the soil collars. The probes have an accuracy of 0.03 m/m and a maximum temperature sensitivity of 0.003 m3/m3 per 1°C. The probes were buried on the date previous to the measurements (one probe per collar). The probes were calibrated in the laboratory with soil samples obtained from the study site (Pereyra et al. 2017). Soil temperature was measured at 0–10 cm depth with a soil temperature probe from the LI-6400.
Soil root density
At the end of each soil respiration measurement, soil cores of 10 cm diameter and 10 cm depth inside each soil collar were collected and were transported to the laboratory. Root density was determined in the upper 10 cm of the soil because the highest root density is found in the upper soil layer (0–10 cm) at the study site and decreases exponentially with increasing depth (Pereyra et al. 2017). Fresh soil was carefully separated from the fine roots by sieving the soil through a mesh of 0.5 mm. Fine roots were washed and oven-dried at 70°C until constant weight. Root density was determined as root dry mass per unit volume of soil.
Soil nitrogen dynamics
Soil ammonification and nitrification rates was estimated in situ during each incubation period as the difference between initial and final content of ammonium and nitrate, respectively, in tubes that prevented plant uptake (Raison et al. 1987). PVC tubes of 5 cm diameter and 10 cm height were buried into the soil at 7.5 cm depth, in places close to the vegetation and on bare soil that were later integrated into a single data in three of the five plots of each treatment. The tubes were placed inside the soil in three seasons (spring, summer and autumn) and they were collected after 1–2 months. At the same time as the placement of the tubes and close to each tube, a soil sample of 5 cm in diameter and 5 cm depth was extracted for the determination of the initial inorganic N content. The collected soil of each tube was transferred into a plastic bag and then transported to the laboratory. Fresh soil samples were sieved with a 0.5 mm mesh, and once frozen they were sent to the Soil Laboratory (INTA Balcarce, Argentina) to analyze the inorganic nitrogen (NH4-N and NO3-N) by a distillation method (Bremner and Keeney 1965). The ammonification and nitrification rates were determined from the difference between the initial and final content of ammonium and nitrate in the soil, respectively, during each incubation period, divided by the number of days. Soil inorganic nitrogen and mineralization rates data were converted to aerial basis using bulk density measurements. The NH4-N/NO3-N ratio of the soil was calculated for each sample with the initial content values in each of the three seasons.
Statistical analyzes were performed using the software R version 4.0.1 (R Development Core Team, 2021). We used linear models (LMs) with F-test to evaluate the effect of treatment on soil physico-chemical properties. We used linear mixed effects models (LMEs) with F-test to evaluate the effect of the interaction between season and treatment on soil water content, root density and soil respiration, with month as random effect. A multiple regression model was tested, using LME, to evaluate the effect of the interaction between soil temperature, soil water content and root density on soil respiration, with treatment as a covariate and month as a random effect. For this analysis, continuous variables were centered and scaled, and the quadratic factor of soil temperature and soil water content were included in the model. Linear models with F-test were used to evaluate the effect of the interaction between treatment and season on initial NH4-N and NO3-N content, NH4-N/NO3-N ratio and ammonification and nitrification rates. Linear regressions were tested between the initial NH4-N and NO3-N content with the ammonification and nitrification rates, respectively, using the previously estimated means per season and treatment. The LMEs were carried out using the function "lme" of the R package "nlme" version 3.1–148 (Pinheiro et al. 2020). The Cox and Snell`s R2 was calculated with the function "nagelkerke" of the R package "rcompanion" version 2.3.25 (Mangiafico 2020). When necessary, models were adjusted using variance models, and the selection of the best model was based on Akaike`s information criterion (AIC) (Burnham and Anderson 2002). The simplification of the fixed effects of all models to reach adequate minimum model was carried out by hypothesis test (F-test). Tukey’s post-hoc analysis was used for multiple comparisons in all models when the F-test was significant, using the function"glht" of the R package "multcomp" version 1.4–13 (Hothorn et al. 2008).