Experiment design
The experiment consists of 20 plots of 625 m2 each in a 4 ha. enclosure, 5 plots per treatment: control (C), N addition (+N), water addition (+W) and addition of both nitrogen and water (+NW). The experiment started in 2013 and continues to the present. The delimitation of the plots was established semi-randomly, taking into account that those were homogeneous in terms of plant cover. Between each plot, at least 10 m buffer area was set to avoid interference among plots. The plots with water addition treatment are equipped with a semi-automatic system of sprinkler irrigation in which the amount of historical annual precipitation for the site is increased by approximately 25%, distributed in events of 5 mm each during the growing season. Freshwater extracted from a well located next to the closure was used to water addition treatment (see Fig. S1 for more details of experimental design). Water addition is canceled during the winter, which is the wettest period and to avoid also damage to the irrigation system by freezing water. Nitrogen is added twice a year by applying 100 kg/ha/yr in the form of urea and diammonium phosphate (160 kg and 154 kg, respectively). One of the additions is performed at the beginning of winter and the other at the end of winter, to favor the dissolution of the fertilizer in the soil during the wet season. The amount of N added per year was similar to that applied in manipulative experiments in other arid ecosystems (Bai et al., 2010; Ladwig et al., 2012; Reichmann et al., 2013). The addition of diammonium phosphate was performed to also add phosphorus in order to avoid any possible limitation of this nutrient that interferes with the effect of the addition of N (Luo et al., 2022).