Mechanisms of plants vs. streams competition
The unusual temporal match between the vegetation change (typically faster) and stream expansion processes (typically slower) at the El Morro catchment combined with its aridity offer a unique opportunity to assess the five mechanisms by which plants and streams may compete for water. Unsaturated water uptake (mechanism 1, Figure 1), has been likely the dominant source of transpired water under the deep groundwater table conditions before the 70s (Figure 4). Diverse sources of evidence point to the absolute prevalence of mechanism 1 and its capacity to fully inhibit stream flow in dry sedimentary plains in Argentina and elsewhere, including geoelectrical profiles and deep sediment cores showing dry vadose zones with an uninterrupted buildup of atmospheric chloride accumulation under native vegetation (Santoni et al., 2010; Jayawickreme, Santoni, Kim,Jobbágy, & Jackson, 2011; Amdan et al., 2011 in our study region; Scanlon et al., 2006 for a global synthesis).
At the El Morro catchment, like in many semiarid sedimentary regions subject to intense cultivation, mechanism 1 has been relaxed thanks to the reduced capacity of annual crops to exhaustively use of vadose moisture (Peck & Williamson, 1987; George et al., 1997; Santoni et al., 2010; Giménez et al., 2016). As a result, water table levels have raised, likely following pulses of high rainfall (Giménez et al., 2016), creating the opportunity for a more widespread contact between plant roots and groundwater. While in first place, such hydrological shifts may favor localized groundwater use in lowland areas or incipient streams by waterlogging-tolerant species (mechanism 2), contact zones can be widespread enough to favor a more distributed use of groundwater in the whole catchment (mechanism 3). At our study catchment, the progressive appearance of wetland communities displaying an intense groundwater consumption was manifested in water table depth and greenness patterns (Figures 4, 6 and 9), providing qualitative evidence for mechanism 2 (Figure 1). Complementary, base streamflow provided more quantitative evidence, with stream segments having shallow incisions and greater proportions of their shores flanked by wetlands showing stronger summer flow drops than the deeply incised upper Río Nuevo segment that was seasonally invariant (Table 1). Assuming that summer baseflow drops reflect the effect of mechanism 2, it would have accounted for 48 and 20% water yield reduction in summer and the whole year, respectively, at the closure of the catchment (Table 1). Mechanism 2 may not only capture groundwater feeding streams but streamflow directly as well, yet, disentangling these two sources has proven difficult (Dawson & Ehrlinger 1991; Bowling, Schulze, & Hall, 2017).
Distributed use of groundwater by vegetation (mechanism 3 , Figure 1) was confirmed in the El Morro catchment. With water table depths >7 m caldén forests used groundwater (Figure 7), as documented earlier for forest relicts north of our site (Gimenez et al., 2016). More generally, across the semiarid and subhumid loessic plains of Argentina, Hungary and China tree stands within herbaceous matrices have often shown intense groundwater use (Jobbágy & Jackson, 2004; Nosetto,  et al., 2013; Yasuda et al., 2013; Toth et al., 2014). Mechanism 3 could have massive effects on streamflow if a large fraction of the El Morro catchment gets reforested. A simple comparison of the trajectories of water table depths at site C (Figure 5) suggest that the mean drop of 93 cm during three years in the forest (69 cm in the croplands) would have get magnified to 196 cm if winter level recoveries would not have taken place (something expectable in full forest coverage scenario). The more precise estimate of groundwater depletion rates derived from diurnal water table depth fluctuations suggests potential depression of 4-5 meters per year (5 active at 3.1 cm  d-1, Figure 7). These rough estimates suggest that the water table raises that expanded wetlands and created new streams could revert in less than a decade with full reforestation. Paired catchment experiments have shown how streams in dry regions can fully vanish following massive afforestation, yet the relative weight of mechanism 1, 2 and 3 explaining these ecohydrological transitions was not disentangled (Farley, Jobbágy, & Jackson 2005; Jobbágy et al., 2013). Mechanism 3 opens the possibility for opposing feedbacks, as hinted by tree ring observations for the same Prosopis species at another site (Bogino & Jobbágy, 2011), where raising water tables boosted tree growth first (negative feedback), but caused their massive waterlogging die-off afterwards (positive feedback). Croplands consumed groundwater too at our catchment, but their shallow rooting depth limited the extent of this effect (Figure 6, Supplementary material)(Nosetto et al., 2009). Rather than assuming rigid and additive effects of vegetation types in the catchment, the fact that mechanism 3 can compensate the relaxation of mechanism 1 creates more opportunities for plants to take away water from streams. The notable match between rooting depths and water table depths globally (Fan et al., 2017), suggests that the dynamic coexistence of these two mechanisms may be widespread.
While the previous discussion focused on the mechanisms by which plants can take water away from streams, our study illustrated how streams had the reciprocal effect on plants (Figure 1). By deepening their incisions (mechanism 4 , Figure 1), streams not only created the hydraulic gradients needed to capture groundwater (comparison of positions 1 to 3 in site C in Figure 5), but hampered mechanism 2 (wetland at site B in Figure 5, drying wetland in Figure 9) and, likely, mechanism 3 (cropland at the same site, Figure 5), favoring their own baseflow (positive feedback). While in the flat areas of the sedimentary plains of Argentina raising groundwater recharge and levels lead to flooding (Kuppel et al., 2015, Gimenez et al., 2020), the more tilted and shallow sedimentary system of the El Morro has likely favored the observed abrupt sapping process that tend to decoupled water table levels from the surface, as they were in the 70s (Figure 4). An approaching saturation of stream density is hinted by their regular spacing and the observed episode of rapid “stream piracy” (Calvache & Viseras, 1997) where they may be too close (effect of the upper Río Nuevo and Uke tributaries on the stream that they flanked, Figure 2b and 8). Theoretical approaches explaining the switch of groundwater surfaces from topography- to recharge-dominated show that increases in recharge can couple water tables to the surface (linear effect) while stream length and depth growth can decouple them (quadratic and linear effect, respectively) (Haitjema & Mitchell-Bruker, 2005; Gleeson, Marklund, Smith, & Manning, 2011). A simple exercise with one of these models suggests that the stream network is reestablishing a recharge-controlled equilibrium in the La Guardia and Río Nuevo sub catchments, but is still far in the Quebrachal sub catchment (Table 2, supplementary material).
The final mechanism by which streams can take away water from plants involves the burial of riparian and wetland areas with fluvial sediments (mechanism 5 , Figure 1). In the El Morro catchment this mechanism accounted for a relatively small fraction of the area (Table 2) but had a very strong effect on vegetation activity that lasted at least 4 years (Figure 9). While likely negligible in terms of its contribution to baseflow yields, this mechanism can create ideal conditions for repeated sapping erosion events as it tends to sustain saturated conditions along the present and future stream bed for longer periods, increasing the chances of successive high precipitation years accumulating their effects. Dry sedimentary regions experiencing deep recharge and water flushing for the first time in geological scales may have salinization as an additional mechanism that favors the partition of water to streams. Increasing salinity in soils and groundwater has the capacity to reduce or even eliminate completely plant transpiration fluxes, as found already in the El Morro catchment (Jobbágy et al 2020) and in deforested areas of the plains of Argentina and Australia (Marchesini, Fernández, & Jobbágy, 2013; Marchesini et al. 2017). While the “winner” flux under salinization is direct evaporation, the restriction of transpiration would hamper mechanisms 2 and 3 favoring water table level raises and, as a result, sapping or surface processes due to the reduced unsaturated thickness of the catchment.  To some extent natural ecosystems adjust to increasing salinity when salt-tolerant species like Tamarix sp . start dominating wetlands (Rios, 2020) or where native forest relicts consume groundwater with >20 dS/m of electric conductivity (Gimenez et al., 2016, at site C in this study). Currently cultivated species in our study system (soybean, maize, alfalfa) however do not show this adjustment capacity (Ashraf & Wu, 2011, Jobbágy et al., 2020).