Assessing multiple limiting factors of seasonal biomass production and N content in a grassland with a year-round production

There is little evidence on the extent that multiple factors simultaneously limit ecosystem function of grasslands with year-round production. Here we test if multiple factors simultaneously limit (i.e., more than one factor at a time) grassland functioning in different seasons and how they interacted with N availability. In a Flooding Pampa grassland, we ran a separate factorial experiment in spring, summer, and winter with several treatments: control, mowing, shading, P addition, watering (only in summer), and warming (only in winter), each of them crossed with two nitrogen treatments: control and N addition. Grassland functioning was assessed by aboveground net primary productivity (ANPP), green and standing dead biomass, and N content at the species group level. Out of 24 potential cases (three seasons by eight response variables), 13 corresponded to just one limiting factor, 4 to multiple limiting factors, and the other 7 to no evidence of limitation. In conclusion, grassland functioning in each season was most often limited by just one factor, while multiple limiting factors were rarer. Nitrogen was the prevailing limiting factor. Our study expands our knowledge of limitations imposed by factors associated with disturbance and stress, such as mowing, shading, water availability, and warming in grasslands with year-round production.


Introduction
The limiting factors of aboveground net primary productivity (ANPP) and plant nutrient content have been studied under two main hypotheses. One of them is the Liebig's law of the minimum (Liebig 1842;van der Ploeg et al. 1999): the scarcest resource restricts plant growth and becomes the individual limiting factor. The other hypothesis proposes that multiple factors can simultaneously limit plant growth (Bloom et al. 1985;Chapin et al. 1987). Simultaneity and multiplicity (more than one at a time) are the two conditions to detect multiple limiting factors. Much of the evidence of those multiple limiting factors has come from studies conducted over complete, but short-growing seasons in northern-hemisphere ecosystems (Chapin and Shaver 1985;Chapin et al. 1995;Boelman et al. 2003;Limpens et al. 2011;White et al. 2014). However, the strong and extended dormant season and the low interannual variability of ANPP of these regions (Knapp and Smith 2001) preclude extrapolation to grasslands whose vegetation grows yearround. Thus, a more robust understanding of the multiple limiting factors of functioning in these grasslands is needed.
Some evidence has shown that plant growth is simultaneously limited by an array of environmental factors. At a local level, primary productivity may be limited by light interception, and water and nutrient availability, which, in turn, are regulated by various factors, such as temperature or precipitation (Chapin et al. 2002). Several experiments have shown responses to multiple limiting factors (Chapin and Shaver 1985;James et al. 2005;Harpole et al. 2011;Rillig et al. 2019). For example, a network of studies across five continents showed that ANPP was simultaneously limited by N, P, and K combined with micronutrients at 25 of 42 sites (Fay et al. 2015). Another example is the tallgrass prairie functioning, where patterns of ANPP are best explained by the multiple limitations of N, energy, and water than by the single limitation of N (Seastedt and Knapp 1993;Blair 1997). Often, the ecosystem response was accounted for by the behavior of some species or groups of species (Lauenroth Communicated et al. 1978;Bobbink et al. 1989;Wei et al. 2014;Van Sundert et al. 2021). Other results suggested the existence of multiple limiting factors acting simultaneously (Dukes et al. 2005;Radujković et al. 2021). There is of course more evidence that one or two factors, mostly N availability and its interactions, affect ANPP and plant nutrient content of temperate grasslands (i.e., Gao and Yan 2019;Zhao et al. 2019). Temperate Flooding Pampa grasslands extend in the eastern part of South America from 35 to 38°, where livestock production on rangelands and pastures is the main land use. These grasslands have several features that suggest that their seasonal functioning may be simultaneously limited by multiple factors. They have year-round production (no period with zero ANPP), with a maximum of 30 kg of dry matter ha −1 day −1 from late spring to the beginning of summer and a minimum of 2 kg of dry matter ha −1 day −1 in late winter (Sala et al. 1981). Thus, several environmental factors may limit ANPP and plant nutrient content in different seasons. Additionally, they include forbs and C 3 and C 4 grasses (Perelman et al. 2001(Perelman et al. , 2017. The phenology and relative dominance of C 3 and C 4 species at different times of year may also contribute to what limits grassland functioning. Some species or species groups exhibit idiosyncratic responses to stress, disturbance, and nutrient availability. For example, grazing may have a negative effect on ANPP, but this ecosystem net outcome resulted from two contrasting growth responses at species group level: the reduction of grasses and the promotion of exotic, unpalatable forbs (Rusch and Oesterheld 1997). The addition of commercial fertilizers is a known management practice in the region, but not widely used (García et al. 2002). However, there are few studies aimed at disentangling how nutrient availability limits the different species groups. In these grasslands, N addition increased the growth of grasses and P addition increased plant P content (Ginzo et al. 1982;Collantes et al. 1998;Semmartin et al. 2007;Rubio et al. 2009;Graff et al. 2020). Other effects associated with year-to-year seasonal variations in rainfall (CV = 30-50%), air temperature (CV = 4-8%), and incoming photosynthetically active radiation (CV = 3-7%) are practically unknown. Particularly, an experiment explored the extent of seasonal limitation by mowing, shading, N, and P on ANPP, below-ground plant biomass production, and nutrient dynamics in a grassland that is active throughout the year (Semmartin et al. 2007). N addition had the largest and most extended limiting factor: it tripled ANPP in spring and summer but had no effect on below-ground biomass (Semmartin et al. 2007). However, that experiment did not explore interactions between N and the other environmental factors (Semmartin et al. 2007). This lack of evidence on the interactions is not restricted to that site.
Here, we evaluated the seasonal limiting factors of ANPP, aboveground biomass, and plant N content of a Flooding Pampa grassland with year-round production. Each season, P additions simulated the use of commercial fertilizers and were used to evaluate their limiting condition, together with mowing, which simulated grazing, and shading, which allowed to assess of the limitation by cloudiness. In summer, an irrigation treatment assessed the limitation by water availability, and in winter a warming treatment assessed temperature limitation. In addition, all these treatments were applied to control and N-fertilized plots to assess the interaction with nitrogen availability and the limitation by nitrogen itself. Our hypothesis was that grassland seasonal functioning is simultaneously limited by multiple factors (i.e., more than one at a time). We tested that hypothesis both at species group and ecosystem levels. We explored how different environmental manipulations assessed the presence of multiple limiting factors in each season.

Overall approach
We ran a set of three separate experiments in spring, summer, and winter. We evaluated simultaneously 4 factors in spring (N, mowing, shading, and P) and 5 factors in summer and winter (N, mowing, shading, P, and watering or warming). We also evaluated interactions between either mowing, shading, P, watering, or warming with N. We measured ANPP, biomass, and plant N content at the species group level (eight response variables). In total, we evaluated 24 potential cases of one or more limiting factors (three seasons by eight response variables).
We defined nutrient, water, or temperature as limitations when a response variable was increased by nutrient additions, watering, or warming, respectively (James et al. 2005). Shading reduced incident radiation, simulating a cloudy season, so we defined shading as a limitation when it decreased a response variable. Finally, mowing reduced leaf tissue available for photosynthesis, simulating herbivory, so we also defined it as a limitation when it decreased a response variable. In this context, for example, mowing and P limitations are comparable: one limiting leaf area and the other limiting P intake.
We concluded that a single factor was limiting in each season when the response variable significantly responded to only one environmental factor. In contrast, we detected multiple limiting factors in a season when the response variable simultaneously responded to two or more environmental factors, both individually (e.g., shading and warming) or jointly as explored by N interactions (e.g., N addition plus warming; Harpole et al. 2011).

Site description
Our study was conducted in a livestock ranch in the center of the Flooding Pampa (36° 16.8′ S; 58° 15.8′ W; province of Buenos Aires, Argentina), a region of nine million hectares primarily covered by natural grasslands (Baldi et al. 2006;Oyarzabal et al. 2018Oyarzabal et al. , 2020. Annual mean precipitation is 956 mm uniformly distributed among seasons. Mean monthly temperature ranges from approximately 7 °C in winter to 22 °C in summer (Soriano et al. 1992;Paruelo et al. 2007). Overall, the period of the study was wetter than usual because of the amount of annual rainfall (1406 mm), 47% higher than the mean, with a particularly humid spring, summer, and winter (respectively 70, 70, and 40% higher than the mean). Plant communities in the region have a combination of species with C 3 and C 4 photosynthetic pathways. The study was in a stand of the most widespread community in the region, a humid mesophytic meadow defined as Piptochaetium montevidense (C 3 grass), Ambrosia tenuifolia, Eclipta bellidioides, Mentha pullegium (forbs) and Chascolytrum subaristata (C 3 grass) community (Burkart et al. 1990). This community occupies flat areas in intermediate topographic landscape positions subjected to winter and spring flooding events and summer droughts almost every year, even in wet years (Soriano et al. 1992). The soil is a Typic Natraquoll with a loamy non-saline A horizon (pH 6.7), with approximately 3.5% organic carbon and 24% clay, and a thick natric B horizon (pH 7.5; Lavado and Taboada 1987). The community has a strong seasonal pattern of growth (see details in the Introduction section). ANPP is approximately 5500 kg of dry matter ha −1 year −1 (Sala et al. 1981;Semmartin et al. 2007). Three floristic groups with particular morphology and phenology and similar diversity are recognized (Semmartin et al. 2007): cool-season graminoids, warm-season graminoids, and forbs. Cool-season graminoids have a productivity peak in late winter and early spring, whereas warm-season graminoids and most of the forb species concentrate their productivity in summer and fall (Sala et al. 1981;Soriano et al. 1992). Dominant species in each group are Danthonia montevidensis, Eleocharis spp. and Lolium multiflorum (non-native) among the cool-season graminoids, Stenotaphrum secundatum, Leersia hexandra, Panicum gouinii, P. millioides, Bothriochloa laguroides, Paspalidium paludivagum and Setaria geniculata (warmseason graminoids), and Leontodon taraxacoides (nonnative), Mentha pulegium (non-native), Phyla canescens, Plantago lanceolata (non-native) and Spilanthes spp. (forbs; Semmartin et al. 2007). The site has been grazed at 0.5 cattle units ha −1 for nearly 100 years, which is considered a normal stocking rate for this region. Current grazing management involves annual resting periods during winter and spring, and the reintroduction of cattle from early summer to late autumn (Campana and Yahdjian 2020).

Experimental design
We carried out three factorial experiments, one per season, with three replicates per treatment. In spring, summer, and winter, the basic design was two levels of N addition (0 and 25 g m −2 ) interacting with other factors (Table A.1): control, mowing, shading, or P addition (factorial 2 × 4 = 8 treatments; n = 3; total 24 plots in spring). Additionally, watering was included in summer (factorial 2 × 5 = 10 treatments; n = 3; total 30 plots in summer), and warming was included in winter (factorial 2 × 5 = 10 treatments; n = 3; total 30 plots in winter). Thus, the total number of plots was 84 (24 + 30 + 30). In autumn, flooding precluded the addition of N and P, so the experiment was not included. In spite of this, the spring, summer and winter experiments covered most of the grassland seasonality because these seasons explain approximately 90% of annual ANPP (Sala et al. 1981).
Each plot corresponded to an experimental unit (i.e., the area in which one treatment was randomly applied): it consisted of a 2 × 2 m moveable cage (height 0.7 m) that excluded domestic cattle, the main grazers in the region, through a 15 × 15 cm iron mesh on each side and the roof. Although long-term grazing exclusion of these grasslands drastically changes species composition and aboveground and below-ground biomass (Chaneton 2006;Graff et al. 2020;López-Mársico et al. in press), the three-month grazing exclusion of our treatments did not have effects that could mask experiment results. Each season, the plots were randomly scattered within an area of approximately 1 ha, a stand of the most widespread community in these grasslands (see above). Moveable cages were being placed in the same general area, though never overlapping. The random assignation of each treatment to each moveable cage was made in each season. Therefore, the seasonal experiments were separated, and the season was not analyzed as a factor. Some manipulations were made as a single event at the beginning of each season (mowing, N and P additions) and the others were either maintained (shading, warming) or repeated (watering) throughout the season.
Experimental manipulations were as follows. N and P addition consisted of single high dose of 25 g.m −2 at the beginning of each seasonal experiment. N was added as ammonium nitrate and P as calcium triple super phosphate. This dose approximates the highest nutrient addition annual rate used in the region and in other grasslands (0-25 gm −2 for N and 0-21 gm −2 for P; Hooper andJohnson 1999, Borer et al. 2014). Mowing consisted of a single biomass removal event at the beginning of each seasonal experiment, made with a lawn mower approximately at 3-5 cm height. Mown biomass was taken out of the plots. Shading consisted of fully covering the plots with 50% optically neutral black shade cloth for the duration of the season. Shade cloth 1 3 covered the four sides and roof of each moveable cage. Air warming during winter was achieved by placing open-top chambers (1.8 × 1.8 m), with nine transparent, acrylic sloping bands on the top (band width = 18 cm, spacing = 0 cm, angle 30º). Sloping bands allowed gas exchange and did not block rainfall or radiation. The summer watering treatment consisted of adding water approximately when soil water content fell below field capacity (38%). Soil water content was measured in the top 10 cm before water addition (TRASE SYSTEM I ® , Soil Moisture Equipment Corp., California, USA), which coincides with ~ 7-10 days after the previous watering. Therefore, watering was not carried out at regular intervals, but when it was required based on the water content of the soil. Total summer watering was 140 mm, 45% of total rainfall during the season, distributed in 6 events of 23 mm on average.
The duration of each seasonal experiment was established according to vegetation phenology and logistic possibilities: spring accounted for most cool-season species growth and fructification (27 September to 19 December; 83 days); summer was the season of maximum warm-season species growth and dispersion (19 December to 22 March; 93 days); and winter was the period of minimum cool-season species growth (30 June to 20 September; 83 days). The response variables were assessed at the end of each season.

Aboveground biomass and aboveground net primary productivity (ANPP)
Initial and final standing total aboveground biomass was harvested in one sub-plot of 0.5 × 0.5 m per plot. Initial biomass was harvested at the beginning of each season, in sub-plots located in untreated areas (for the mowing treatment, it was estimated after mowing in ad hoc mowed areas next to the corresponding plot, the number of replicates was 3). Final biomass was sampled in each plot at the end of each season. Biomass was separated into green and standing dead biomass. The green fraction was in turn separated into three species groups: cool-season graminoids, warm-season graminoids and forbs. Samples were oven dried (70 °C) for 48 h and weighed.
ANPP was estimated by a method that accounts for the simultaneity of production and senescence (Sala et al. 1981;Sala and Austin 2000): new leaves emerge while others senesce or fall to the ground.
where P i was the productivity of individual species groups " i " in the time interval " t ", estimated as the difference between final and initial green biomass of group " i ". Because productivity cannot be negative, only positive P i values were considered. S c was a correction factor that accounted for the simultaneous nature of production and senescence.
where the first term of Eq. 2 represents the increment of standing dead biomass (SD), and the second term represents the sum of decrements of green biomass of individual species groups (G i ). To be included in the calculation of Eq. 1, Sc must be greater than 0, i.e., when the accumulation of standing dead biomass is greater than the senescence detected by the decrease of green biomass (Sala and Austin 2000). The majority of standing dead biomass from a previous season fall to soil and it is incorporated as litter in a few months (Chaneton et al. 1996). Consequently, the ANPP estimation was not affected from senescence of previous seasons.

Plant N and P content
Total N and P content in aboveground biomass were measured on milled subsamples of the four fractions of final harvested biomass (cool-season graminoids, warm-season graminoids, forbs, and pooled standing dead biomass). For N content, the semi-micro Kjeldahl method was used (Ma and Zuazaga 1942). For P content, 100 mg were digested with a solution of nitric-perchloric 3:1 (Johnson and Ulrich 1959). P content was measured by inductively coupled plasma atomic emission spectroscopy (ICP-AES, model 1000-III, Shimadzu Corp ® , Tokyo). While N content was measured in all treatments, P content was only measured under treatments of N and P additions. A previous study on this site had shown that mowing and shading practically did not affect P content (Semmartin et al. 2007). Due to reduced data on plant P content, we only show and discuss extensively plant N content.

Effects of experimental manipulations
We evaluated to what extent the experimental manipulations induced the expected environmental changes: fertilization increased soil nutrient availability, shading reduced incident radiation, watering increased water availability, mowing reduced the amount of leaf tissue available for photosynthesis, and warming increased air temperature. The N and P additions were evaluated in terms of soil inorganic N and P in the top 10 cm of soil approximately 45 days after nutrient additions. Three subsamples were collected in each plot. Fresh soil was sieved through a 2 mm mesh within 48 h of sampling. For nitrogen, a subsample of ~ 10 g of soil was extracted for determination of inorganic N (NH4 + and NO3-) in 50 ml 2N KCl and then filtered. We corrected soil N content for soil water content determined by placing a weighed subsample in a drying oven at 105 °C for 48 h (Robertson et al. 1999). Soil extracts were analyzed using a colorimetric analysis of inorganic nitrogen in liquid extracts (Alpkem ® autoanalyzer, O-I Corporation, College Station, Texas, USA). The detection limit from the manufacturer is 0.09 ppm for NO3-and 0.15 ppm for NH4 + , although due to some adjustments in the specifications of the autoanalyzer, the working limits of detection in our laboratory was 0.03 ppm for NO3-and 0.12 ppm for NH4 + (Tagliazucchi pers. comm.). For phosphorus, fresh sieved soil was air dried. Extractable P content was measured according to Bray-Kurtz P1 protocol (Bray and Kurtz 1945). Shading was evaluated by measuring incident radiation above and below the shade cloth with a handheld radiometer during a sunny midday in spring (angle of view = 21°; SKYE ® SKR100, Skye Instruments LTD, UK). Warming during winter was assessed as the difference between air temperature inside transparent acrylic shelters (n = 2) and control plots (n = 2). The air temperature was measured every 30 min for 21 days with a temperature sensor located 15 cm above the ground and connected with an automatic data logger (HOBO H8 ® , Onset Computer Corporation, Massachusetts, USA). Watering in summer was evaluated by measuring volumetric soil water content in the top 10 cm before water addition (TRASE SYSTEM I ® , Soil Moisture Equipment Corp., California, USA), which coincides with ~ 7-10 days after the previous watering. The experimental manipulations effectively achieved the intended environmental changes in most cases. The N and P additions increased availability during summer and winter (n = 3; P < 0.04), but not in spring. Soil N for control and fertilized plots was, respectively, 17 and 21 ppm in spring, 9 and 154 ppm in summer, and 9 and 43 ppm in winter. Soil P for control and fertilized plots was, respectively, 14 and 18 ppm in spring, 13 and 108 ppm in summer and 10 and 57 ppm in winter. The lack of effect of N and P additions on soil availability in spring could be related to rapid mineralization and/or run-off or lixiviation produced by rain or an eventual short flooding event. Regardless, N and P were more absorbed in plots with N and P additions than in the non-fertilized ones (see Results and Discussion sections), so N and P addition in spring were effective. The proportion of aerial biomass removed by mowing was 91-96%. Shading reduced ~ 50% of incident photosynthetically active radiation (n = 3; P < 0.001) without affecting the red:far-red ratio (P = 0.60). The air temperature difference was on average 2 °C (n = 2; P < 0.001). Watering significantly increased soil water content by 15-82% in three of the four dates in which it was measured (n = 3; P < 0.05).

Statistical analyses
Data of each experiment in spring, summer and winter were analyzed with N addition (2 levels; -N and + N), other factors (4 or 5 levels depending on the season, see Table A.1; control, mowing, shading, P addition, warming and watering) and interaction among N level and other factors (2 × 4 or 2 × 5) as fixed effects, and plots as random effects. We performed mixed-effect models (lme function in nlme package, Pinheiro et al 2017) with R (R version 3.6.3, R Core Team 2020), followed by ad hoc contrasts (P < 0.05), namely untreated vs treated, to avoid inflating the likelihood of type I error. The rationale and purpose of analyses were to detect differences between the treated and untreated plots. For example, we were interested in knowing if there was in spring a different ANPP between N 0 -Control and N 25 -Control plots, or between N 25 -Control and N 25 -Shaded ones (Tables A.1 and A.2). Therefore, we were not interested in all differences among levels. The mixed-effect models allow modeling the variance. We modeled variance heterogeneity (varIdent) if significantly improved model fit (AIC; Zuur et al. 2009). The normality assumption was fit.

Results
Overall (see more details in the next paragraphs), ANPP (response variable 1) was limited by three factors in spring (N, P and shading) and by one factor in summer (N) and winter (warming). Green biomass of warm-season graminoids (2) was limited by three factors in summer (mowing, N and P), and one factor in winter (mowing), whereas green biomass of cool-season graminoids (3) was limited by one factor in winter (mowing). Green biomass of forbs (4) was only limited by one factor in winter (P). Total standing dead biomass (5) was limited by four factors in spring (N, P, mowing, and shading), and one in summer and winter (mowing). The plant N content of warm-season graminoids (6) was limited only by one factor in all seasons (N). The plant N content of cool-season graminoids (7) was limited by one factor in summer (N), and three in winter (N, mowing, and shading). The N content in total standing dead biomass (8) was only limited by one factor in summer and winter (N). Summarizing, out of 24 potential cases (three seasons by eight response variables), 13 corresponded to just one limiting factor, 4 to multiple limiting factors, and the other 7 to no evidence of limitation. Grassland functioning in each season was most often limited by just one factor, while multiple limiting factors were rarer. Nitrogen was the prevailing limiting factor.

Aboveground net primary productivity (ANPP) and plant biomass
ANPP was differentially affected by the experimental manipulations across seasons. In spring, N addition increased ANPP by 66%. When P has added alone, ANPP did not change, but when added to N-fertilized plots ANPP increased by 39%. Whereas when shading was done in N-fertilized plots, ANPP was reduced by 43% (Fig. 1A). In summer, N addition increased ANPP by 64% (Fig. 1B). Finally, in winter, warming increased ANPP by 215% ( Fig. 1C and Tables 1 and A.2). Thus, ANPP was limited by three factors in spring (N, P, and shading) and one in summer (N) and winter (warming).
These responses of ANPP integrated the individual response of biomass fractions. In spring, N addition increased total green biomass by 23% ( Fig. 2A) and increased total standing dead biomass by 60%. When P was added to N-fertilized plots, total standing dead biomass increased by 30%. Whereas when shading or mowing was done in N-fertilized plots, total standing dead biomass was reduced by 50-70% (Tables 2, A.3, A.4 and A.5). Thus, the positive effect of N plus P addition on ANPP in spring was mainly accounted by an increase of total standing dead biomass, which indicates that N plus P addition equally increased growth and senescence. The negative effect of N plus shading on ANPP in spring was accounted for by a decrease in total standing dead biomass, showing that N addition plus shading reduced growth and senescence. Differently, in summer, the positive effect of N addition on ANPP was accounted for by an interactive effect of N plus P addition on total green biomass (Fig. 2B, and Tables A.3, A.4, A.5), which in turn was accounted for a positive effect of N addition on green biomass of warm-season graminoids (Tables 2 and A.6). In winter, the positive effect of warming on ANPP was accounted for a positive effect of warming on total green biomass ( Fig. 2C and Table A.4).
Some treatments, particularly mowing, affected final biomass at the end of the season without affecting ANPP. Mowing without N addition reduced total green biomass by 78% in summer and mowing reduced total green biomass by 65% in winter (  (control, mowing, shading, P addition, watering, and warming). Asterisks indicate significant contrasts between the treatment and its corresponding control (Tables 1 and A.2). In spring (A), N addition and P plus N addition increased ANPP, and N addition and shading interactively reduced ANPP. In summer (B), N addition increased ANPP. In winter (C), warming increased ANPP. Bars are ± 1 standard error. Panels show all treatments when interaction was significant (spring) or only the main effects when not (summer and winter). Note different scales on y-axes A.5). These reductions of biomass by mowing are consistent with the lack of effect on ANPP (i.e., biomass was removed at the beginning of each season when mowing was done, see Methods), and due to the lack of a positive response of ANPP to mowing (Fig. 1) the difference translated to the end of the season. In contrast, in spring, mowing did not reduce significantly total green biomass ( Fig. 2A), probably because there was a marginal increase of ANPP compensating for the initial biomass removal (Fig. 1). Thus, total green biomass was limited by one factor in spring (N), three factors in summer (N, P and mowing), and two factors in winter (mowing and warming). Total standing dead biomass was limited by four factors in spring (N, P, shading and mowing), and one in summer and winter (mowing). At the species group level, mowing reduced green biomass of warm-season graminoids by 43% in summer and by 81% in winter, and mowing reduced green biomass of cool-season graminoids by 60% in winter (Tables 2, A.6 and A.7). N addition and P addition in summer increased green biomass of warm-season graminoids by 36-54%. P addition in winter increased the green biomass of forbs by 167% (Tables 2, A.6 and A.8). Thus, the green biomass of warmseason graminoids was limited by three factors in summer (N, P, and mowing) and one in winter (mowing), while green biomass of cool-season graminoids was limited by one factor in winter (mowing). The green biomass of forbs was limited by one factor in winter (P).

Plant N content
N content in total green biomass did not respond to the interaction between N addition and other factors, but to some individual factors in summer and winter ( Fig. 3 and Table 2). N addition increased N content in total green biomass by 33-41% in summer and winter, a response mainly accounted for by the two groups of graminoids ( Fig. 3B-C, and Tables A.9, A.10 and A.11). Mowing increased N content in total green biomass by 66% in winter due to the impact on coolseason graminoids (Fig. 3C and Tables A.3 and A.12). N addition increased N content in warm-season graminoids by 24% in spring while shading increased N content in coolseason graminoids by 44% in winter, both without affecting N content in green biomass (Tables A.9 and A.11). N content of total standing dead biomass increased by 30-47% after the addition of N in summer and winter (Tables 2 and  A.12). Thus, N content in warm-season graminoids was limited only by one factor in all seasons (N), while N content of cool-season graminoids was limited by one factor in summer (N), and three in winter (N, mowing and shading). N content of total standing dead biomass was limited by one factor in summer and winter (N).

Discussion
Much of the evidence for the multiple factors hypothesis has come from studies of seasonal ecosystems, during short, complete growing seasons (Chapin et al. 1995;White et al. 2014). In grasslands, the multiple factors hypothesis was tested broadly only with nutrient manipulations: often, ANPP is simultaneously limited by multiple nutrients (Fay et al. 2015). Our study expands our knowledge of limitations imposed by factors associated with disturbance and stress, such as mowing, shading, water availability, and warming in grasslands with year-round production (Semmartin et al. (Tables A.3 and A.4). In spring (A), N addition increased biomass. In summer (B), mowing without N addition interactively reduced biomass, whereas P plus N addition interactively increased biomass. In winter (C), mowing reduced biomass, while warming increased biomass. Bars are ± 1 standard error. Panels show all treatments when interaction was significant (summer) or only the main effects when not (spring and winter). Note different scales on y-axes 2007; Campana and Yahdjian 2020). Out of 24 potential cases (three seasons by eight response variables), 13 corresponded to just one limiting factor, 4 to multiple limiting factors, and the other 7 to no evidence of limitation. In contrast to previous findings in seasonal ecosystems with short-growing seasons, grassland functioning in each season was most often limited by just one factor, while multiple limiting factors were rarer. The identity of those limiting factors varied among seasons. In addition, the limiting factors acted on certain species groups and differed among response variables. Nitrogen was the prevailing limiting factor. The limiting factors of grassland ANPP shifted among seasons. In spring N, P and shading limited ANPP, in summer only N, whereas in winter temperature was the limiting factor. Some of these effects were traceable to green biomass of particular species groups. Similarly, the limiting factors of N content in green biomass were mainly one or two, with variations among seasons. In summer, N addition increased N content, whereas in winter N content was increased by N addition or mowing. In addition to all these significant responses it is equally remarkable that ANPP and plant N content of this grassland were unresponsive in some seasons to drastic manipulations such as adding maximum fertilizer doses, removing nearly 100% of aerial biomass, reducing incoming radiation by 50% or increasing summer precipitation by 45%.
Nutrient additions in spring affected ANPP but not N content. P plus N addition increased ANPP but did not affect N content (Figs. 1 and 3). In contrast, P plus N addition increased plant P content in spring, and P addition increased plant P content in summer and winter . These results indicate that some N added was indeed absorbed by plants and used to grow, a response called sufficient supply (growth increases but nutrient concentration remains constant; James et al. 2005). Contrarily, P added was indeed absorbed by plants and accumulated in aboveground tissue, a response called luxury consumption (i.e. accumulation of nutrients above levels that promote growth; James et al. 2005). P luxury consumption as a response to P addition was previously reported for this grassland (Ginzo et al. 1982;Semmartin et al. 2007;Graff et al. 2020) but had not been observed after P plus N addition. P luxury consumption was revealed in this study both at the ecosystem level and graminoids species group level (Tables A13-A17), suggesting that P availability did not play a key role as a limiting factor of primary production. At the individual level, species that show luxury consumption during nutrient flushes may use these reserves to support growth after soil reserves are exhausted (Chapin 1980) and/or to buffer the impact of adverse conditions such as defoliation events (Oyarzabal and Oesterheld 2009). At the ecosystem level, the effect of P luxury consumption in temperate grasslands is less known (Graff et al. 2020).
In contrast, nutrient additions in summer and winter differently affected ANPP and N content. In summer, N addition increased both ANPP and plant N content (Figs. 1B and 3B), indicating a response called limitation (both growth and nutrient content increase following nutrient addition; Table 2 Biomass fractions (g m −2 ) in different seasons under treatment of N addition (N 0 and N 25 ) in interaction with other factors (control, mowing, shading, P addition, watering, and warming) Mean values ± 1 standard error. Data were pooled to show the main effects because interactions between N addition and other factors were rare (Table A.3). Numbers in bold indicate that the manipulation significantly differed from the N 0 or Control (P < 0.05). For example, in summer, green biomass of warm-season graminoids was increased by N addition from 199 to 308 g m −2 (in bold), and it was decreased by mowing from 251 to 144 g m −2 (in bold; see specific contrast in Table A.6). There were only two significant interactions affecting total standing dead biomass in spring (not shown): P addition increased dead biomass only in N-fertilized plots, while shading reduced dead biomass only in N-fertilized plots (Tables A.3 (James et al. 2005). To our knowledge, this is the first evidence of N luxury consumption in the Río de la Plata grasslands (Soriano et al. 1992). In the tallgrass prairie, dominant grasses draw on internal N reserves accumulated in the absence of fire, as suggested by the relatively high N tissue content observed in the unburned plots, and the decline in tissue N content with repeated burning (Blair 1997). Warming in winter increased ANPP by 200% but did not affect plant N content. This warming-induced increase in ANPP may be a direct effect of the increased rate of growth, which could be particularly important for the cool-season graminoids during winter (Alward et al. 1999;Mueller et al. 2016), or an indirect effect of increased N availability, resulting from increased rate of N mineralization (Liu et al. 2016). The latter mechanism is unlikely in our case because ANPP did not respond to N addition in winter ( Fig. 1C and Table 1). N addition did increase plant N content, though, which suggests that low-temperature limited growth more than N absorption (Fig. 3C). Individual grassland sites showed both neutral and positive responses of ANPP to 1-5 years of experimental warming (Rustad et al. 2001). Our results show a very positive response of ANPP to 3 months of experimental warming.
Mowing did not affect ANPP, which indicates that the grassland was fully compensated in terms of ANPP. Grazing promotes a diversity of responses, which can range from damage to benefit depending on the changes in the relative growth rate suffered by grazed plants in response to defoliation (Oesterheld and McNaughton 1991). Negative responses are more frequent than positive, both at the individual and the community level (Oesterheld et al. 1999;Ferraro and Oesterheld 2002;Carlyle et al. 2014). In these grasslands mowing and mowing plus N addition, promotes compensatory responses that mitigate or even overcompensate for the removal of biomass Oesterheld 1996, 2001;Semmartin et al. 2007). Our results show that ANPP was similar between mowed and un-mowed control plots over the three seasons and independently on N addition. Therefore, mowing promoted a full compensatory response (Oesterheld and McNaughton 1991), even when defoliation was intense (91-96% of biomass removal). Our experimental approach was to simulate grazing using a mechanical defoliation or mowing. It had at least three limitations to mimicking actual grazing. First, there was one-time biomass removal instead of season-round. Second, all plants were cut and not just the more preferred ones. Third, trampling and urine and fecal depositions were absent. Thus, we caution our results are limited to the effects of an event of defoliation at the beginning of each season.
The lack of effect of shading on ANPP in summer and winter suggests an increase of radiation use efficiency. According to the so-called radiation use efficiency logic, based on Monteith ecophysiological model (Monteith 1972), ANPP is determined by the amount of photosynthetically active radiation absorbed by the canopy (APAR), and the efficiency with which that energy is transformed in aboveground dry matter. APAR is the product of incoming photosynthetically active radiation (PAR) and the fraction absorbed by the canopy (fPAR; Hill et al. 2004;Piñeiro et al. 2006). Our experimental shading reduced PAR by half. Visual observations during the experiment suggested that aerial cover of green biomass, a major determinant of fPAR, was similar between shaded and control plots. Consequently, shaded plots had lower APAR. However, the shading did not affect ANPP in summer and winter, similar to . Asterisks indicate significant contrasts between the treatment and its corresponding control (Tables A.3 and A.9). In summer and winter B and C, N addition increased N content, while in winter C mowing increased N content. Bars are ± 1 standard error. Data were pooled to show the main effects because interactions were not significant (Table A.3). Note different scales on y-axes a previous study in this grassland (Semmartin et al. 2007). Therefore, we interpret that shading increased the radiation use efficiency. Previous studies revealed that radiation use efficiency could be significantly enhanced by an increase in diffuse radiation (Huang et al. 2014;Druille et al. 2019). Under cloudy or aerosol-laden skies, PAR was more diffuse and more uniformly distributed in the canopy (He et al. 2013). We hypothesize that a more diffused PAR under experimental shading increased radiation use efficiency and thus maintained similar ANPP as controls. The increase in the radiation use efficiency would have occurred at low (winter), intermediate (spring) or high (summer) PAR level.
In contrast, at intermediate PAR level (spring), shading in N-fertilized plots reduced PAR and radiation use efficiency did not compensate that reduction. Thus, ANPP was reduced by shading plus N in spring (Fig. 1). The Flooding Pampa is one of the less transformed grassland areas in the world (Baldi et al. 2006;de Abelleyra et al. 2019;Oyarzabal et al. 2020). Mild winters and evenly distributed monthly precipitation, together with a mix of C 3 and C 4 grasses with peak growth at different times of the year resulted in a year-round growing season (Paruelo et al. 1995;Perelman et al. 2001Perelman et al. , 2017. Grasslands with similar characteristics extend in the eastern part of South America from approximately 31 to 38° Latitude S and from 57 to 66.5° Longitude W (Oyarzabal et al. 2020). We do not expect our results be directly extrapolated to the entire region, but we suggest that similar shifting of limiting factors throughout the year likely takes place. Thus, there is the great challenge of identifying limiting factors and potential management decisions in specific subregions or landscape areas (López-Mársico et al. in press). In addition, we have not explored all the potential interactions among limiting factors, such as shading due to cloudiness and air temperature, invertebrate herbivory and nutrient availability, or phosphorus availability and several environmental factors (McNaugthon et al. 1989;La Pierre and Smith 2016). We also did not explore sequential limiting factors. Basically, a resource limiting in any season could indirectly limit biomass production next season and there could be seasonal interactions (López-Mársico et al. in press). For example, N addition in spring might increase P or water demand in summer. This evaluation could be incorporated in future experiments. Its factorial design could be extended through seasons, for a fuller appreciation of limiting factors in ecosystems with year-round growing season.