Many regions of the globe are experiencing a simultaneous change in the dominant plant functional type and regional climatology. We explored how atmospheric temperature and precipitation control leaf- and ecosystem-scale carbon fluxes within a pair of semi-arid shrublands, one upland and one riparian, that have undergone woody plant expansion.
Through a combination of leaf-level measurements on individual bunchgrasses and mesquites shrubs and ecosystem-scale monitoring using eddy covariance techniques, we sought to quantify rates of net carbon dioxide (CO2) flux, CO2 flux temperature sensitivity and the responsiveness of these parameters to seasonal rains and periods of soil dry-down.
We found significant differences in physiological acclimation between the two plant functional types, in that the shrubs consistently conducted photosynthesis across a broader temperature range than co-occurring grasses during dry periods, yet maximum photosynthetic rates in grasses were twice that of mesquites during the wetter monsoon season. Landscape position modulated these temperature sensitivities, as the range of functional temperatures and maximum rates of photosynthesis were two to three times greater within the riparian shrubland in dry times.
Also, it was unexpected that ecosystem-scale CO2 uptake within both shrublands would become most temperature sensitive within the monsoon, when mesquites and grasses had their broadest range of function. This is probably explained by the changing contributions of component photosynthetic fluxes, in that the more temperature sensitive grasses, which had higher maximal rates of photosynthesis, became a larger component of the ecosystem flux.
Synthesis: Given projections of more variable precipitation and increased temperatures, it is important to understand differences in physiological activity between growth forms, as they are likely to drive patterns of ecosystem-scale CO2 flux. As access to stable subsurface water declines with decreased precipitation, these differential patterns of temperature sensitivity among growth forms, which are dependent on connectivity to groundwater, will only become more important in determining ecosystem carbon source/sink status.
Surface air temperatures have increased over the last 30 years across the semi-arid Southwestern United States (Cubasch et al. 2001; Backlund et al. 2008), and this region is predicted to experience warmer daytime temperatures, with more frequent warmer nights (Cubasch et al. 2001; Tebaldi et al. 2006; Christensen et al. 2007; Backlund et al. 2008; Weiss, Castro & Overpeck 2009). The US south-west is also predicted to become drier, on average, but with more extreme rain events. Western North America has experienced a 15% decrease in precipitation since 1900 (Cubasch et al. 2001), and models forecast, with fair confidence, reduced and more variable precipitation with longer interstorm rain-free periods (Overpeck & Cole 2006; Christensen et al. 2007; Backlund et al. 2008), and with slightly less confidence, more marked interannual variability in summer rains associated with the North American monsoon (Cubasch et al. 2001). In order to better understand ecosystem-scale responses to climate change, there has been substantial interest in quantitatively linking plant and ecosystem responses to temperature and water stress (Medlyn, Loustau & Delzon 2002b; Medlyn et al. 2002a; Barron-Gafford, Grieve & Murthy 2007; Lloyd & Farquhar 2008; Sage, Way & Kubien 2008; Montpied, Granier & Dreyer 2009; Kanniah, Beringer & Hutley 2010; Barron-Gafford et al. 2012; Munson et al. 2012).
Interacting with climatic change is land cover-scale shifts in the dominant plant functional type, creating an opportunity to study the nexus of physical geography, plant ecophysiology and climate science. One of the most notable transitions in vegetative cover is woody plant encroachment (WPE) into historical grasslands (Knapp et al. 2008b), which can alter interception of solar radiation (Breshears et al. 1998; Villegas et al. 2010), nutrient cycling and availability (Hibbard et al. 2001; McLain, Martens & McClaran 2008; Throop & Archer 2008), net ecosystem exchange of carbon and water (Jackson et al. 2000; Hughes et al. 2006; Scott et al. 2006b; Barron-Gafford et al. 2012), ecosystem structure and function (Jackson et al. 2002; Huxman et al. 2005; Munson et al. 2012) and controls to landscape hydrological dynamics (Seyfried et al. 2005). WPE has been documented throughout regions of varied water status (Jackson et al. 2002; Knapp et al. 2008b; Ravi et al. 2009; Van Auken 2009; Naito & Cairns 2011; Jenerette & Chatterjee 2012), including the US central plains grasslands (Briggs et al. 2005; McKinley et al. 2008), south-west desert grasslands (Buffington & Herbel 1965) and mesquite (Prosopis spp.) expansion world-wide (Harding & Bate 1991; Archer 1994; McClaran 2003).
In water limited systems, plant growth form influences plant and ecosystem phenology through differential growth responses to precipitation and the temporal persistence of plant activity across periods of drought or high temperature stress (Smith, Monson & Anderson 1996). In contrast to grasses, which allocate relatively little to deep roots, woody plants can develop root systems that extract shallow water resources (e.g. Ogle & Reynolds 2004), or deep, extensive rooting networks, often allowing them to act as phreatophytes in landscape positions with stable groundwater sources or areas of deep soil water recharge (Smith, Monson & Anderson 1996). This feature can effectively decouple woody plant physiological activity from incident rainfall (Williams et al. 2006). Physiological processes in grasslands tend to be more sensitive to incident precipitation than woodland or mixed vegetation ecosystems because shallowly rooted grasses have less connectivity to stable water sources at depth (Ehleringer et al. 1991; Golluscio, Sala & Lauenroth 1998; Jackson et al. 2000; Scott et al. 2000; Potts et al. 2006; Williams et al. 2006). Such differences in plant growth/life-forms traits therefore modify both ecosystem-scale hydrological patterns important to surface processes (Canadell et al. 1996; Jackson et al. 2000; Schenk & Jackson 2002; Huxman et al. 2005; Seyfried et al. 2005) and plant carbon and water fluxes that are important to biosphere–atmosphere feedbacks (Evanari et al. 1975; Hultine et al. 2004, 2006; Jenerette et al. 2009).
Leaf physiological traits are another fundamental component of biosphere–atmosphere gas exchange (Wright et al. 2004; Enquist et al. 2007). Leaf physiology is primarily associated with rates and patterns of plant carbon assimilation (A, photosynthesis), which responds to variations in vapour pressure deficit (VPD), light, [CO2] and temperature through both biochemical shifts and leaf morphological features that affect internal gas diffusion (Schulze & Caldwell 1994). The temperature response of A generally follows a dual Arrhenius function with a peaked response, such that at low temperatures A increases to an optimum beyond which rates decline depending on thermal tolerance (Farquhar, von Caemmerer & Berry 1980; Leuning 2002). C4 species tend to have higher rates of A at elevated temperatures compared with most C3 plants, but have enzyme kinetics that more strongly restrict the range of physiologically optimal temperatures (Berry & Björkman 1980). However, the expressed convexity of the temperature response across entire growing seasons depends not only on photosynthetic pathway (C3 or C4), but also the degree of physiological temperature acclimation (tendency of a species to maintain homeostasis in a changing environment), which can be affected by plant resource availability (Barron-Gafford et al. 2012). In semi-arid settings, water availability depends strongly on landscape position. In riparian areas, plant rooting attributes and access to stable groundwater may allow for temperature acclimation for deep-rooted woody C3 plants as compared to shallow rooted C4 grasses. In contrast, at upland locations, the lack of stable groundwater resources may result in both growth forms having more constrained seasonal physiological adjustment due to more prevalent water stress. The combination of these biological and resource availability features in controlling A, suggest emergent properties of ecosystem-scale biosphere–atmosphere exchange and temperature response that may be tied to landscape position.
There are clear topographical/landscape position influences on ecosystem responses to precipitation, both in terms of timing and amount of precipitation. While riparian woodlands are less sensitive to summer monsoon precipitation, whether those rains are small or large (Scott et al. 2006a; Potts et al. 2008), mesquites in upland shrublands are more responsive to larger precipitation events than smaller pulses (Fravolini et al. 2005). Knapp et al. (2008a) suggested dry-land ecosystems that become stressed under current precipitation regimes of frequent small events may become only ‘intermittently stressed’ under predicted less frequent, but larger rains because such events more completely recharge to greater soil depths. These forecasted conditions may better benefit mesquites, which appear to respond more favourably to large events, than grasses, which have evolved a rooting habit capable of quickly capturing smaller pulses. In this context, we explored the concomitant controls of temperature and precipitation on leaf- and ecosystem-scale CO2 flux in upland and riparian mesquite-encroached semi-arid grasslands.
Using these paired ecosystems, both a mix of C4 grasses and C3 mesquites, as a model for ecosystems of differential access to stable groundwater, we addressed the following questions:
(i) How does temperature differentially limit net CO2 uptake in C4 grasses and C3 mesquites in contrasting landscape positions through periods of varying precipitation and temperature extremes?; (ii) Does access to groundwater modulate when rates of ecosystem-scale CO2 uptake are more or less temperature sensitive?; and (iii) How do the component fluxes of CO2 uptake and efflux contribute to ecosystem-scale temperature sensitivity? In addressing these questions, we will better understand emergent features at the ecosystem scale of how temperature and precipitation variation differentially influence carbon dynamics within mixed vegetation ecosystems experiencing dramatic changes in available soil moisture, such as what may occur under projected climatic change scenarios.
Materials and methods
Study sites and species
This study used two former south-east Arizona USA semi-arid grasslands that have been converted to shrubland following prolonged woody plant encroachment. The riparian site (31.566°N, 110.133°W) is located on an old alluvial terrace at an elevation of 1237 m, bordering the San Pedro River. Soils consist mainly of gravelly sandy loam layers interspersed with clay and gravel lenses (Scott et al. 2006b). The eddy covariance tower footprint at this site is dominated by velvet mesquite (Prosopis velutina), with a canopy cover of ca. 51%. Understorey canopy is principally the C4 bunchgrass, big sacaton (Sporobolus wrightii; ca 27% total cover), and an additional 32% of understorey intercanopy soil space fills in with annual herbaceous species, most commonly Viguiera dentata (Cav.) Spreng. Mesquite height averaged 3.7 m, mean bunchgrass canopy height was 1.25 m, with a mean depth to groundwater of 6.5 m (Scott et al. 2006b; Potts et al. 2008). The upland site is located in the Santa Rita Experimental Range (31.8214° N, 110.8661° W, elevation: 1116 m) SSE of Tucson, AZ, USA. Soils here are a deep sandy loam (Scott et al. 2009). Mesquite cover at this site is approximately 35%, with vegetation of the intercanopy space dominated by a mosaic of perennial C4 bunchgrasses (Eragrostis lehmanniana Nees, Digitaria californica Benth and Bouteloua eriopoda) and seasonally bare soil. Intercanopy plant cover of perennial grasses, forbs and sub-shrubs is approximately 22% (Scott et al. 2009). Mesquite averaged 2.5 m in height, and the mean depth to groundwater at the upland site probably exceeds 100 m, as nearby wells measured depths to groundwater of 100 and 154 m. Mean annual precipitation (1971–2000) from stations around the riparian shrubland ranges from 313 to 386 mm (Scott et al. 2006b) and is (1937–2007) 377 mm at the upland site (Scott et al. 2009), with about 50% falling between July–September during the North American Monsoon.
Eddy covariance measurements
Ecosystem-scale carbon dioxide, water vapour and energy fluxes were monitored at the riparian site from 2003 to 2008 and at the upland since 2004 till present (Scott et al. 2006b, 2009). Detailed descriptions of instrumentation, sensor heights and orientations, and data processing procedures are given elsewhere (Scott et al. 2004, 2006b, 2009; Potts et al. 2008). Briefly, at both sites, instrumentation on 8 m towers measures all variables needed to quantify 30 min averages of net ecosystem exchange of CO2 (NEE), air temperature (Tair), VPD, air pressure, photosynthetic photon flux density (PPFD), shortwave and net radiation, and precipitation. We filtered CO2 flux data over poor turbulent mixing periods by removing data associated with a friction velocity (u*) below 0.15 m−2 s−1 (Scott et al. 2006b). Ecosystem-scale gross ecosystem productivity (GEP) was calculated as:
where REco is ecosystem respiration, derived from the night-time NEE data processed using procedures in Reichstein et al. (2005). We converted NEE to net ecosystem productivity (NEP = −NEE), so that ecosystem-level data follows the sign convention of leaf-level ecophysiological data.
Eddy covariance data from 2007 was divided into 40-day blocks representing the pre-monsoon (DOY 140–180), monsoon (190–230) and post-monsoon (280–320) periods for each site. These periods were selected because they represent typical climatic conditions of their respective seasons and encompass periods when canopies were present and stable. Each seasonal batch of NEP data was separated into 16 2.5 °C temperature bins from 0 °C to 40 °C. Within each bin, NEP was regressed against PPFD to assess T-effects at different light levels across the day and season (Huxman et al. 2003; Barron-Gafford et al. 2012). Light-saturated NEP rate for each 2.5 °C bin of Tair was modelled using a rectangular hyperbola (nonlinear least-squares regression; SigmaPlot 11.0, spss, Chicago, IL, USA):
where αe' is the apparent ecosystem quantum yield, NEPsat is light-saturated net CO2 exchange, and Re is respiratory CO2 exchange at PPFD = 0 (Ruimy et al. 1995). The NEPsat at each of 16 temperature bins were plotted for each site for each seasonal period to assess a T-response function for each ecosystem. The Q10 of REco was calculated as:
where Tair was referenced to a common temperature of 25 °C. Rates of REco at this common temperature (REco25) were also used for between-ecosystem comparisons.
Leaf-level physiological measurements of gas and water exchange
Measurements of leaf-level night-time leaf respiration (RLeaf) and daytime net photosynthesis (A) were conducted across a 25 + °C range of Tair on mesquites and grasses at each site throughout 2007. Following the methods described by Barron-Gafford et al. (2012), we used a LI-6400 portable photosynthesis system (LI-COR, Lincoln, NE, USA) with a red–blue light source (LI-6400-02b) providing constant PPFD of 1500 μmol m−2 s−1 during daytime A estimations. Cuvette CO2 was held to 375 p.p.m. by mixing outside air with CO2 from a cylinder injection assembly attached to the instrument. A small, white reflective cap was placed on top of the cuvette to minimize heat gain by the chamber, while still allowing for thermal dissipation by the chamber's cooling fans. An initial measurement flux was taken at ambient T to serve as an indicator of RLeaf or A under ambient conditions and as a metric against which to compare the T-response curve. Finding a similar reading prior to the initialization of the response curve and at that same T during the subsequent response curve confirmed, we did not induce any perturbation by our protocol. After this initial measurement, the temperature of a Pelletier-exchange cooled temperature block was set to 5 °C to lower chamber T to its minimum reachable level. Once chamber T had stabilized, the two infrared gas analysers within the instrument were matched, and gas exchange data were logged. After this, the block temperature was increased in 3–5 °C increments, and the leaf was given a minimum of 2 min to stabilize in response to cuvette T. This was repeated until the maximum potential chamber temperature was reached. Leaf temperature was continuously measured with a fine-wire type-T thermocouple pressed to the underside of the leaf. We occasionally added a small amount of water to the instrument's CO2 soda lime scrubber to avoid excessive VPD at higher temperatures, with care taken to avoid any condensation in the instrumentation tubing. Five individual mesquite and five individuals of the dominant grass (S. wrightii at the riparian E. lehmanniana and at the upland) were sampled at each site, with measurements taken on intact leaves midway up the south side of the canopy. Estimates of gross photosynthesis (AGross) were made by summing A and RLeaf. All gas exchange leaf samples were harvested after measurement and stored in paper envelopes. Sample leaf area was determined (CI-202, CID Bio-Science, Camas, WA, USA), and then samples were air-dried.
These protocols were repeated during three periods throughout the growing season: the pre-monsoon drought (DOY 171–173), monsoon peak (DOY 223–225) and post-monsoon dry-down (DOY 284–286) in 2007. We chose these periods to identify the influence of landscape position on (i) patterns of photosynthetic upregulation in response to warm-season rains, (ii) changes in T-sensitivity among plant functional types in response to prolonged soil moisture availability and (iii) acclimation potential as the system returned to a dry state.
Determination of temperature optima and the degree of temperature-limitations to CO2 uptake and statistical analysis
Amax and NEPmax and optimum temperature (Topt) for each were estimated from the single peak of a temperature response curve fit to the leaf- and ecosystem-level data. A custom model was developed in MATLAB 2009b (MathWorks, Natick, MA, USA) to fit the data, based on the energy of activation and deactivation model presented by Leuning (2002) for estimating the temperature sensitivity of maximum catalytic rate of the carbon-fixing enzyme Rubisco and the maximum electron transport rate (Barron-Gafford et al. 2012; Richardson, Chatterjee & Jenerette 2012). The model estimates a peak in the temperature response function and allows for asymmetry in the sub- and supraoptimal portions of the response curve. From there, a metric of the convexity of the temperature response function was derived by quantifying the range of temperatures over which a leaf or ecosystem was assimilating 50% (Ω50) and 75% (Ω75) of Amax and NEPmax. Ω50 and Ω75 illustrate the difference between the upper and lower temperatures at which Amax and NEPmax declined by 75% and 50%, respectively. Ω50 illustrates variation in the temperature sensitivity at the edge of the plant's or ecosystem's functional range, while Ω75 illustrates a plant or ecosystem's ability to assimilate carbon in the range of temperatures most immediate to Topt.
A split-plot, repeated-measures analysis of variance (rm-anova; Statistix v. 8.0, Analytical Software, Tallahassee, FL, USA) was used to test for differences in leaf-level Amax, Topt, Ω50 and Ω75, and RLeaf between the two sites (landscape position), three sampling periods and two plant functional types. The between-treatment, whole-plot effect was growth form (mesquite versus grass), using the growth-form-by-replicate interaction as the whole-plot error term and an α of 0.05. The within-treatment, subplot effects were seasonal periods (pre-monsoon, monsoon, post-monsoon) and the growth-form-by-season interaction, using the growth-form-by-season-by-replicate interaction as the subplot error term. An rm-anova was used to test for differences in ecosystem-level NEPmax, Topt, Ω50 and Ω75, and REco between the two sites (landscape position) and the same three sampling periods (pre-monsoon, monsoon, post-monsoon).
Total 2007 precipitation at the riparian and upland site was 263 mm and 330 mm, respectively (Fig. 1a,b). Average weekly air temperature (T) peaked during the pre-monsoon at ~40 °C (DOY 140–180) and was consistently warmer at the upland site. Average T differed as little as 1 °C between the sites during the monsoon (DOY 190–230) and by 8.4 °C in dry periods (DOY 280–320; Fig. 2c). Average daily temperature fluctuation (daily Tmax− Tmin) at the riparian site (23.4 °C) was greater than at the upland site (14.2 °C) due to lower Tmin at the riparian site, not differences in Tmax. With the exception of late winter, VPD was higher at the upland site (Fig. 1c,d, Fig. 2d). Pre-monsoon weekly average diel VPD were 1.1 kPa higher at the upland site and 0.5 kPa through monsoon and post-monsoon periods. Average 5, 10 and 50 cm soil moisture were 4.4, 4.3 and 7.0% in upland site soils vs. 6.8, 8.8 and 9.0%, respectively, at the riparian site (Fig. 1e–h). Rainfall increased weekly average 5 and 10 cm soil moisture more at the riparian site than at the upland site. Upland site soils were consistently drier at 50 cm than at the riparian site (Fig. 2e).
Ecosystem-scale flux and temperature responses
By DOY 85, weekly average net ecosystem exchange of CO2 (NEE) at the riparian site were negative, indicating net CO2 uptake, while the upland site did not attain negative weekly average NEE until DOY 203 (Fig. 2a). For the pre-monsoon, the riparian site was a net CO2 sink (ΣNEE=−24.5 gC m−2), while the upland site was a net source (+ 13 gC m−2; Fig. 2b). At the start of the monsoon, ΣNEE was −73.6 and + 27.2 gC m−2 at the riparian and upland sites, respectively (Fig. 2a,b). By the end of the monsoon, the riparian site received 193 mm rain with a ΣNEE of −123 gC m−2, the upland site received 222 mm and a ΣNEE + 10.3 gC m−2. During the post-monsoon period, NEE continued to show that both sites were a net sink for CO2.
The temperature response of ecosystem-level net ecosystem productivity (NEP; illustrated as positive values sensu ecosystem photosynthesis) and ecosystem respiration (REco) differed significantly between upland and riparian shrubland sites and varied between seasonal periods (Fig. 3a–c). Pooled across sites, maximum rates of NEP (NEPmax) were greatest within the monsoon period (F2, 16 = 37689; P < 0.0001). Lowest NEPmax was attained at the upland site during the pre-monsoon period and over the post-monsoon at the riparian site. There was a significant two-way site-by-season interaction (F2,16 = 5449; P <0.0001), due to Topt at the upland site varying between seasonal periods (19.7 °C, 28.1 °C and 22.1 °C for pre-monsoon, monsoon and post-monsoon, respectively), but not at the riparian site (27.1, 27.7 and 26.3 °C). NEPmax also had a significant site-by-season interaction (F2,16 = 4852; P <0.0001), because NEPmax increased 420% from pre-monsoon to the monsoon in the upland, but only 45% at the riparian site. Post-monsoon NEPmax declined 50% from monsoon period levels at both sites.
Ω50 and Ω75 for the riparian site NEP were ca. twice that at the upland, averaging 52 + 4% higher across the seasonal periods (F1, 4 = 65737 and 151190, respectively; P <0.0001). Both shrublands reduced Ω50 and Ω75 in from pre-monsoon to the monsoon, indicating greater ecosystem T-sensitivity during the period of greatest water availability. The reduction in NEP Ω50 was more pronounced for the riparian (38%) than the upland (7%) site. Post-monsoon Ω50 and Ω75 increased 30% at the upland site, while the riparian site Ω50 of increased 39% and 29% in Ω75, indicating a broader ecosystem T-response than during the monsoon (Fig. 3).
REco temperature responses also differed between seasonal periods for both sites, although the riparian site consistently had more negative REco25 (Fig. 3d–f; F1,4 = 2114.55; P <0.0001). During the pre-monsoon, there was little REco in the upland site, and the Q10 of REco was 0.98, indicating limited REcoT-response, while REco25 at the riparian site was −2.8 μmol m−2 s−1 and Q10 = 1.5 (Fig. 3d). REco25 was most negative over the monsoon (F2,16 = 684.60; P <0.0001), with a curvilinear T-response; the Q10 of the riparian site was 2.2 for the 15–25 °C range, but only 1.0 for the 25–35 °C range. There was a significant site-by-season interaction (F2, 16 = 181.14; P <0.0001) in that the monsoon increased REco more markedly at the upland site (844%) than at the riparian (57%), with REco25 of −3.4 and −4.4 μmol m−2 s−1, respectively. REco25 declined 56% at the upland and 45% at the riparian site over the post-monsoon, but was 4 times higher than pre-monsoon levels (Fig. 3d,f).
Components fluxes ~ Leaf-scale temperature responses
Amax and Topt differed between seasonal periods (F2,40 = 732.33 and 8.64; P <0.05 for Amax and Topt, respectively), with significant three-way site-by-species-by-season interactions in both variables (F2,40 = 7.02 and 11.52; P <0.05 for Amax and Topt, respectively; Fig. 4). At both sites, higher Amax were attained by bunchgrasses during the monsoon (48.8 + 3.1) compared with mesquite (19.8 + 1.7), while mesquite Amax were higher in the pre- (14.9 + 1.3 and 10.1 + 1.1 for mesquite and grasses, respectively) and post-monsoon (17.8 + 1.8 and 9.90 + 1.2 for mesquite and grasses, respectively). At the upland site, Amax increased 124% and 1041% between pre-monsoon and monsoon periods, in mesquite and grasses, respectively (Fig. 4). Post-monsoon Amax in mesquites were 21% lower than monsoon peaks, and 86% lower in grasses, which remained higher than pre-monsoon levels. Ω50 and Ω75 of A for both growth forms peaked in the monsoon and were greater in mesquites across the pre-monsoon, monsoon and post-monsoon periods, respectively (Fig. 4). Monsoon Ω50 were 64% and 81% higher than pre-monsoon Ω50 in the upland site mesquite and grasses, respectively, and post-monsoon Ω50 and Ω75 reductions were more substantial in grasses (Fig. 4).
Although Amax and Topt for mesquites and grasses differed between seasonal periods of differing water availability, they were less variable in riparian site plants (Fig. 4). Riparian site mesquite Amax were seasonally invariant and averaged 22.6 + 0.4 μmol m−2 s−1. Amax of riparian site grasses increased 234% from pre-monsoon to monsoon periods and dropped to 75% of pre-monsoon levels in the post-monsoon (Fig. 4d–f). Amax was 7.3 μmol m−2 s−1 higher in mesquites than grasses in pre- and post-monsoon periods, but mesquite Amax were 32.2 μmol m−2 s−1 lower than in grasses during the monsoon. Ω50 and Ω75 for riparian mesquites were greatest in the pre-monsoon and were lower and similar over the monsoon and post-monsoon. Riparian grass Ω50 and Ω75 in the monsoon were 11% and 26% higher than pre-monsoon ranges. Changes in Ω75 were greater than Ω50 for mesquites, while the opposite was true for grasses, indicating a greater expansion and contraction in area around Amax in mesquites, and more change at the edges of the T-response for riparian site grasses (Fig. 4d–f). Topt did not change in riparian mesquites (27 °C), while grass Topt was 33 °C in pre-monsoon and monsoon and 29 °C in post-monsoon periods.
Pooled across the plant functional types, seasonal rates of Amax were greater at the riparian than the upland site, although differences were smaller during the monsoon. Pre-monsoon, rates of Amax were three times greater in riparian mesquites than in upland site plants (22.4 + 1.9 vs. 7.4 + 0.7, respectively), and more than four times that of grasses (16.6 + 1.6 vs. 3.7 + 0.7; Fig. 4a,d). Post-monsoon, Amax was nearly twice as great at the riparian site for both mesquite (Fig. 4c, mesquite: 22.5 + 2.0 vs. 12.6 + 1.6) and grasses (Fig. 4f; 13.8 + 1.3 vs. 6.0 + 1.2). Pre- and post-monsoon Ω50 for both growth forms were nearly twice as great in riparian plants than in upland plants, although these landscape differences became muted within the monsoon (Fig. 4d–f).
RLeaf showed a near-linear T-response in all periods, regardless of landscape position or growth form (Fig. 5). Upland RLeaf25 became progressively more negative through time for both growth forms, with greatest rates in the post-monsoon (F2,40 = 57.71; P <0.0001; Fig. 5). RLeaf25 was higher in grasses than in mesquite during the monsoon, with the opposite in the pre- and post-monsoon. RLeaf Q10 in upland mesquites was greater in the monsoon (1.89) than pre- (1.50) or post-monsoon (1.58; Fig. 5a–c). Seasonally pooled RLeaf25 at the riparian site were twice that of upland plants (F1,40 = 2244.09; P <0.0001; Fig. 5). Leaf-level RLeaf25:AGross were more T-sensitive during pre-monsoon than monsoon periods, indicated by curvilinear responses in upland site plants and steeply sloped responses in riparian site plants (Fig. 6a,c). During the monsoon, riparian plant RLeaf had slightly negative response (Q10 of 1.6 + 0.1), while upland plants had a Q10 indicating stronger T-response (1.9 + 0.1). At the ecosystem level, REco was rarely more than 45% of GEE (Fig. 6b,d), except across temperatures where NEP was nearly negative (dashed lines in Fig. 6). At both sites, pre-monsoon REco:GEE peaked at lower T than in the monsoon, and these correspond with respective Topt for NEP.
How components of ecosystems combine to collectively exchange materials and energy with the atmosphere and the associated hydrological network is an important challenge. The complexities of growth form, photosynthetic pathway and the potential for acclimation and/or genetic adaptation all suggest emergent properties of how environmental variables are related to ecosystem exchanges at larger spatial scales. This study provides comprehensive quantification of the temperature sensitivity of both the dominant vegetative components within semi-arid shrublands and the entire ecosystems themselves. By repeating measurement campaigns across seasonal periods and landscape positions, we were able to compute changes in thermal sensitivity due to periods of varying precipitation and estimate the role of component fluxes in driving ecosystem-scale responses.
Our results show several important conclusions, which are described in detail below. First, there was a difference in the physiological acclimation among the two plant functional types. Mesquites were able to maintain net carbon assimilation across a range of temperatures 40–75% greater than co-occurring grasses during dry periods, but grasses outperformed mesquites during the monsoon. Secondly, landscape position played an important role in modulating these temperature sensitivities, as the range of functional temperatures and maximum rates of A were two to three times greater within the riparian shrubland in dry times. Thirdly, there was a disconnect between ecosystem- and leaf-scale temperature sensitivities of carbon uptake. Ecosystem carbon fluxes were most temperature sensitive within the monsoon for both shrubland locations, when mesquites and grasses displayed their largest ranges of photosynthetic temperature tolerance. This shows that ecosystem-scale functional restrictions were not due to constraints in plant photosynthetic capacity within either growth form, but rather were tied to the shifting relative contributions of component fluxes. Finally, the relative contribution of respiration to gross CO2 uptake at both leaf- and ecosystem scales were more temperature sensitive in the upland ecosystem than the riparian shrubland. Soil- and leaf respirations from the nitrogen-rich mesquite leaves were both major contributors to ecosystem CO2 fluxes. Collectively, our results highlight both a differential physiological acclimation to temperature among these plant functional types and that the relative contribution of these component fluxes directed ecosystem-scale temperature sensitivity of the shrublands depending on access to soil moisture (as summarized in Fig. 7).
Differential temperature sensitivity of grasses and woody plants and the role of landscape position in modulating plant activity
Plant functional types differed significantly in their rates of Amax, the range of temperatures across which they were able to conduct net A, and their responsiveness to the onset of precipitation. Importantly, we detected a greater physiological capacity in mesquites than grasses within dry periods, such that Amax and Ω50 in mesquites were nearly twice those of bunchgrasses. Collectively, these findings illustrate that ecosystems undergoing vegetative change are likely to see enhanced performance in the woody plants relative to native grasses, particularly under predicted precipitation patterns of longer interstorm periods of drought (Cubasch et al. 2001; Overpeck & Cole 2006; Backlund et al. 2008). Our classic understanding of historical woody plant expansion, developed through extensive observation and experimentation, suggests that high levels of herbivory by livestock yields (i) spread of mesquite seed and (ii) less above-ground grass biomass and fine fuel, which greatly reduces or eliminates grassland fires (see Archer, Schimel & Holland 1995; Van Auken 2000 and references within). Here, we have identified a potential mechanism for continued expansion despite much better range management practices that keep livestock grazing in check and efforts to restore native semi-arid grasslands. Rises in atmospheric CO2 concentrations may exacerbate this landscape transition given its potential to overly favour C3 relative to C4 species. Previous work has suggested that elevation has not been a primary cause of woody expansion in the past (Madany & West 1983; Conley, Conley & Karl 1992; Archer, Schimel & Holland 1995), although this may need to be revisited in light of climate change projections. One might predict that as more ecosystems lose subsurface waters (Seager et al. 2007), native grasses will become more reliant on incident precipitation to mitigate restrictions imposed by atmospheric temperature on plant and ecosystem function. Hence, greater access to subsurface water and increased atmospheric CO2 concentrations would both benefit woody plant productivity and expansion (Fig. 7; climate change analogue).
Disconnect between ecosystem- and leaf-scale temperature sensitivities
Despite a considerable increase in NEPmax, both shrublands experienced a significant increase in temperature sensitivity from the pre-monsoon to monsoon conditions. This reduced range of functional temperatures, in spite of increased peak performance, was contrary to our expectations, given that both ecosystems would have more available soil water during the monsoon than any other time of year. We found that the increased temperature sensitivity of the upland shrubland was not due to a restriction in the leaf-level plant physiological capacity of either growth form, as Ω50 and Ω75 actually increased an average of 68% and 83% in the mesquites and grasses from the pre-monsoon to the monsoon (Fig. 4a,b). Rather, the ecosystem-scale reduction in Ω50 and Ω75 was tied to (i) the relative contribution of component fluxes and the transition of dominant contribution from a species with a wider range of temperature function (mesquites) to one more constrained by temperature (grasses) and (ii) the significant increase in green leaf area of grasses in response to summer rains (personal observation). During the pre-monsoon, grasses had minimal rates of A and were contributing little to total ecosystem-scale flux, but in the transition to the monsoon, average grass Amax was nearly three times (157%) that of the mesquites and therefore probably the dominant contributor to total ecosystem flux. Bunchgrasses in this region typically have very low functional leaf area index (LAI) in the pre-monsoon, but LAI increases ~620% in periods of peak performance (Hamerlynck et al. 2010), while mesquite LAI remains relatively constant at ~1.6 (Scott et al. 2004). Ω50 and Ω75 of NEP increased to their highest levels among all seasonal periods for both shrublands in the post-monsoon, and we hypothesize that this is due to the concomitant effects of a 45% and 72% greater Ω50 and Ω75 in the mesquites than the grasses and a greater capacity for assimilation in mesquites than grasses in the post-monsoon, as quantified by an average of 89% higher Amax.
We found that a greater connectivity to groundwater both relaxed dependence of ecosystem performance on precipitation and enhanced potential for ecosystem-scale temperature acclimation. Sap flux-based measures of whole-tree transpiration in upland mesquite have illustrated a significantly reduced total daily water use between conditions of lower available soil moisture and higher VPD, relative to an environment characterized by abundant soil moisture and relatively lower evaporative demand (Dugas, Heuer & Mayeux 1992). Reduced rates of NEP at higher temperatures and lower Ω50 and Ω75 within our upland relative to the riparian shrubland were not due to higher RSoil efflux countering plant uptake at higher temperatures in the upland (Fig. 6b), but rather may be linked to reduced rates of A under the combined atmospheric conditions of high temperatures and high VPD at the upland site (Fig. 2c,d).
The relative contribution of respiration to gross CO2 flux at the leaf- and ecosystem-scales
There was a strong temperature response of ecosystem-scale respiration (REco; Q10 > 1.5) within the riparian shrubland during all seasons and in the upland during the monsoon and post-monsoon (Fig. 3d–f). The near-zero efflux rates and lack of a temperature response in REco within the dry upland during the pre-monsoon were probably tied to the lack of soil moisture and minimal rates of leaf-level A within both growth forms that would yield few root exudates to feed microbial respiration. REco25 was significantly greater in the riparian site than the upland site during all seasonal periods, but the greatest differences between the shrublands came during this pre-monsoon period when efflux rates were an order of magnitude larger (Fig. 3d). Relatively high REco rates have been documented in mixed and mesquite-dominated riparian ecosystems during the dry pre-monsoon because of mesquite access to groundwater (Scott et al. 2006b). By also measuring component fluxes of mesquite and grass RLeaf, we could estimate their relative contributions to ecosystem efflux. We found that a significant portion of REco came from foliar respiration, as RLeaf25 within riparian shrubland was −6.0 and −1.9 μmol m−2 s−1 in the mesquites and grasses. The relative proportion of REco to gross ecosystem exchange of carbon (GEE) remained between 20 and 40% for both shrublands, except at temperature extremes when NEP itself was near-zero, resulting in virtually all CO2 exchange being an efflux from the ecosystem (Fig. 6b,d). The shape of this REco:GEE temperature response function was much more convex in the upland shrubland, suggesting that the relative contribution of respiration to gross flux CO2 flux was more sensitive to temperature than in the riparian shrubland. The temperature at which this REco:GEE function and the temperature at which NEPmax peaked corresponded almost exactly with one another within shrublands during each growing season, underscoring the fact that these ecosystems were functioning best at specific temperatures, beyond which their carbon balance became less of a net sink for CO2.
Summary and conclusions
Leaf- and ecosystem-scale carbon fluxes were monitored throughout multiple seasonal periods to provide a quantitative linkage among atmospheric temperature, water availability and ecosystem processes across multiple landscape positions and temporal scales. We found connectivity to stable groundwater sources decoupled leaf- and ecosystem-scale temperature sensitivity relative to comparable sites lacking such access (Fig. 7). Groundwater access not only resulted in a near-doubling of the range of temperatures across which the ecosystem could assimilate CO2 at near-peak rates, but also actual rates of net ecosystem productivity being 1.5 times greater when precipitation was relatively abundant and 5 times greater when it was not. Given projections of more variable regional precipitation and increased temperatures, differences in physiological capacity among growth forms are likely to drive patterns of ecosystem-scale carbon flux, depending on the degree of woody versus grass cover within shrublands. As access to stable subsurface water declines with decreased precipitation input, these differential patterns of temperature sensitivity among growth forms dependent on connectivity to groundwater will only become more important in determining ecosystem carbon source/sink status.
This work was supported by Philecology Foundation of Fort Worth, Texas and NSF-DEB 04189134 and 0414977 to T.E.H. Additional support was provided by (1) the University of Arizona Water, Environmental and Energy Solutions program through the Technology and Research Initiative Fund, (2) the USDA-ARS, and (3) the Sustainability of semi-Arid Hydrology and Riparian Areas (SAHRA) Center under the STC Program of the National Science Foundation, Agreement No. EAR-9876800. The authors thank J.L. Bronstein, D.L. Venable and R.L. Minor for providing insightful comments on the manuscript. We also thank R. Bryant for assistance with equipment maintenance, M. McClaran and M. Heitlinger who oversee research carried out within the Santa Rita Experimental Range, and the many undergraduate and graduate students who have collectively worked on these sites over the last decade.