Relationships between ANPP and precipitation (ambient + experimental)
The interannual variability in water inputs ranged from 92 to 233 mm in the plots without experimental water addition and from 122 to 263 in plots with water addition. A positive relationship with the same adjusted slope (0.21) was found between annual water input and shrub ANPP in all treatments (Fig. 7A). The intercept of the ANPP-water input relationship was lower in +NW than in C (p =0.004), while in +N and +W there was no difference compared to C (p >0.05) (Fig. 7A). Grass ANPP also was positively and linearly correlated to water input (Fig. 7B), although the nitrogen addition (i.e., +N and +NW treatments) had a higher positive effect (p≤0.004) (Fig. 7B). The +W treatment did not affect the response of grass ANPP to water input compared to C (p >0.5) (Fig. 7B). The relationship between total ANPP and water input was significantly modified by the treatments. The +N and +NW treatments enhanced the intercept of the function fitted to the ANPP-water input relationship (p<0.001), while the +W treatment decreased it (p =0.03) (Fig. 7C).

Discussion

Seven-years monitoring of a large-scale field nitrogen and water addition experiment in a Patagonian steppe showed that plant productivity, cover and mortality responded differentially to soil nitrogen addition and water addition. This response also differed between and within functional groups (grasses and shrubs). Although water availability is considered the most important environmental factor limiting productivity in drought-prone ecosystems, our results indicate that addition of 25% more water than historical annual precipitation resulted in a total ANPP similar to the ANPP in the control plots, while the nitrogen addition had a positive effect on ANPP.
The total ANPP in control plots reported in our study was similar to that estimated for other semiarid steppes (e.g., Bai et al., 2010; Gao et al., 2011). The range of ANPP values obtained in our study in the control plots (41-98 g m-2 yr-1) is consistent with that reported by Jobbágy & Sala (2000) for a similar Patagonian steppe (21-75 g m-2yr-1). We determined the ANPP in early December, while that in previous studies developed in the Patagonian steppes (e.g., Gherardi & Sala, 2015; Jobbágy & Sala, 2000; Yahdjian & Sala, 2006), ANPP has been estimated in January. Despite the similar range of ANPP between our study and the others, we consider late November or early December as the most accurate measurement period to estimate ANPP since the maximum NDVI and MSAVI2 were observed in this period. Therefore, measurements done in January instead at the end of November or early December could result in relatively low ANPP estimates in the Patagonian steppe.