Highlights

  • An eddy covariance network was set up along an elevation gradient of sagebrush ecosystems.

  • Timing of precipitation and snowmelt influenced carbon balance at various elevations.

  • A relation between GEP and ET for the southwestern USA is consistent for northern sites.

  • ET and GEP may increase with climate warming above a threshold of 450 mm precipitation.

Introduction

Climate change is expected to have widespread impacts on ecohydrological processes, plant health and production, and species distribution in the semiarid Western USA, including the shrub ecosystems of the Great Basin in the Intermountain West (Chambers and others 2008, 2014; Mote and others 2014; Parmesan and Yohe 2003). Observations from the northern Great Basin indicate that increased temperatures, decreased fraction of precipitation that falls as snow, and earlier snowmelt have already occurred, and that recent warming has impacted snow hydrology (Nayak and others 2010; Klos and others 2014; Mote and others 2014). As a result, many historically snow-dominated mountain basins within the region are now rain-dominated, growing seasons are prolonged with the earlier snowmelt, and there has been an increase in the elevation of the rain–snow transition zone, that is, the elevation zone where the dominant phase of precipitation changes from rain to snow (Klos and others 2014). Assessing the impact of these ongoing changes on ecosystems relies heavily on our ability to monitor, model, and understand the ecohydrologic response to shifts in climate and meteorological conditions.

Ecosystems in the Great Basin and Intermountain West often straddle the elevation of the rain–snow transition zone. These ecosystems are typically a mix of diverse grass, shrub, aspen and conifer communities, with grass and shrub communities spanning the entire rain–snow transition and tree-dominated communities often limited to snow-dominated areas (Barbour and Billings 2000; Clark and others 2001). Across the rain–snow transition, temperatures and snow cover limit plant production in the winter/spring, and warm temperatures and water availability limit plant production in the summer (Comstock and Ehleringer 1992; Kwon and others 2008). These strong climatic limitations on plant production cause Great Basin vegetation to be sensitive to inter-annual variations in weather and suggest that they may be vulnerable to shifts in climate (Schlaepfer and others 2012; Shinneman and others 2015).

Warming and shifts in snowmelt timing interact with precipitation amount, evapotranspiration and plant-available water to produce a range of scenarios that complicate model forecasts of ecosystem processes and productivity across the rain–snow transition. Studies have shown, for example, that the relation between gross ecosystem production (GEP) and annual precipitation can be site specific and time variant (Huxman and others 2004; Gilmanov and others 2006; Jin and Goulden 2014; Biederman and others 2016), hinting that the predictive capacity of annual precipitation is limited and that a more detailed consideration of plant-available water is required. Biederman and others (2016), however, demonstrated a comparatively stronger relationship (r2 = 0.73) between productivity and evapotranspiration (ET) that is consistent across time and space, suggesting that vegetation production is not always limited by precipitation but rather by the amount of water available to the ecosystem for ET. Thus, when snowmelt and spring rains exceed the soil’s water storage capacity, ET and vegetation production are limited by soil conditions and late spring/summer precipitation (Smith and others 2011; Bowling and others 2010). Consequently, both the timing and amount of water available for plants could influence the effects of climate warming on ET and GEP.

Limited systematic observations of ET and GEP in the snow-dominated Western US shrublands have markedly limited our understanding of the effects of climate gradient and snowmelt timing, which are needed to understand how climate change may impact shrub-steppe hydrology and production. Available data have used multiple approaches and include a range of vegetation, but data from few sites were consistently collected and processed to systematically capture gradient effects. Some work has been conducted on shrub-steppe communities using chamber methods (Prater and DeLucia 2006; Mitra and others 2014; Cleary and others 2015), Bowen ratio systems (Svejcar and others 1997, 2008; Angell and others 2001; Gilmanov and others 2003, 2006; Wylie and others 2003), and eddy covariance systems (Kwon and others 2008; Reed and others 2018). Much of the available work on shrublands in the Western USA to date has been focused in the desert southwest (Goulden and others 2012; Scott and others 2006, 2015), and the similarity of southwest arid and semiarid ecosystems to the northerly snow-dominated ecosystems remains largely unknown.

Observations of ecosystem processes across gradients (elevation, precipitation, temperature, and so on) can provide invaluable information on the effects of potential environmental changes (for example, Whittaker and Niering 1975; Anderson-Teixeira and others 2011; Goulden and others 2012; Fang and others 2014), and the rate that climate and ecosystem states might transition across the gradient (Loarie and others 2009). Moreover, additional ecosystem-level measurements from the northern Great Basin provide the opportunity to compare northern Great Basin ecosystems with more intensively studied systems in the southwest USA. If the relationship between GEP and ET established for sites largely in the southwest USA (Biederman and others 2016) holds for Great Basin ecosystems, it could provide clues to the potential impacts of climate warming on these ecosystems. Thus, there is a considerable need for systematic measurements of carbon fluxes in the northern Great Basin, particularly measurements with long temporal records along an elevationally induced climate gradient.

As part of the Reynolds Creek Critical Zone Observatory (RC CZO; White and others 2015; Seyfried and others 2018), a network of core sites was established to monitor long-term carbon fluxes across an elevation/climate gradient that spans the rain–snow transition zone within Reynolds Creek Experimental Watershed (RCEW). The objectives of this paper were to: (1) present initial findings that quantify differences in water and carbon fluxes along a climate gradient within a Great Basin sagebrush ecosystem and identify controls on carbon fluxes across this gradient; (2) leverage these observations with data from previous studies within RCEW to test the relationship between GEP and ET for plant communities within the Great Basin ecosystem and compare these results to recent findings from the southwest USA; and (3) make inferences about the potential impacts of climate warming on ET and GEP for a range of precipitation and across the rain–snow transition elevation.

Methods

The site is the Reynolds Creek Experimental Watershed (RCEW) located in the Owyhee Mountains of southwest Idaho and operated by the USDA-ARS Northwest Watershed Research Center. Reynolds Creek drains north to the Snake River and ranges in elevation from 1090 to 2240 m. The diverse climate, soils and vegetation found on the watershed range from semiarid sagebrush rangeland on shallow desert soils receiving 236 mm of precipitation to forest stands (aspen, Douglas fir and subalpine fir) on deep loess soils that receive 1118 mm of precipitation, 76% of which falls as snow. Reynolds Creek has cool, wet winters and hot, dry summers. December–March precipitation typically accounts for over half of the annual accumulation.

Herein, data from the first four years of observations (water years 2015 through 2018) from the RC CZO Climate Gradient Core Network were used to quantify differences in water and carbon fluxes along a climate gradient. Cross-site differences and inter-annual variability were explored to identify differing controls on carbon fluxes across the elevation gradient. These data were then combined with data previously collected within RCEW to identify trends between in ET and GEP across the climate gradient. Finally, model simulations were used to assess potential impacts of climate change on ET and GEP.

RC CZO Climate Gradient Core Network

A primary objective of the RC CZO (Seyfried and others 2018) is to quantify and predict soil carbon processes, storage and fluxes from the pedon to the landscape scale. A network of four core observation sites was established along an elevation/climate gradient in 2014 to intensively monitor long-term carbon fluxes and soil carbon dynamics. Sites were selected to capture the range of precipitation and temperature in the RCEW while minimizing variation in other soil forming state factors (Jenny 1941), such as parent material, time, and topography to the greatest extent possible. The elevation range of selected sites spans the elevation of the current rain–snow transition within the RCEW (Figure 1). The four sites are all sagebrush dominated and located on moderate (approximately 10–15%) north-facing slopes near long-term (40 + years) meteorological stations.

Figure 1
figure 1

Location map of the research sites within the Reynolds Creek Experimental Watershed. Eddy covariance sites, designated by stars, are the Wyoming big sagebrush (WBS), low sagebrush (LoS), post-fire sagebrush (PFS) and mountain big sagebrush (MBS) sites. All are located within 800 m of a long-term climate station.

Full size image

Selected site characteristics are summarized in Table 1. Each site is identified by the dominant plant species at the site: Wyoming big sagebrush (Artemisia tridentata wyomingensis), low sagebrush (Artemisia arbuscula), post-fire mountain big sagebrush site (Artemisia tridentata vaseyana), and undisturbed mountain big sagebrush. These sites are referred to as WBS, LoS, PFS, and MBS, respectively. All sites are located on the eastern flank of the RCEW within 13 km of each other. The elevation range (1425–2111 msl) and corresponding temperature and precipitation gradients encompass most of the conditions within the RCEW and are typical of much of the Great Basin region within the USA. Precipitation phase at the lower elevation sites (WBS and LoS), is dominated by rain, whereas that at the higher elevations (PFS and MBS) is dominated by snow and accompanied by a seasonal snowpack.

Table 1 Climate, Vegetation, and Soil Properties for the Four Study Sites
Full size table

Vegetative parameters, such as plant density and leaf area index, tend to increase with elevation and precipitation (Table 1). The WBS site represents a semiarid shrub-steppe plant community with an understory of native graminoids, forbs and moss. Other shrubs present in this community include green rabbitbrush (Chrysothamnus viscidiflorus) and spineless horsebrush (Tetradymia canescens). Dominant perennial graminoids include bluebunch wheatgrass (Pseudoroegneria spicata), squirreltail (Elymus elymoides), and Sandberg bluegrass (Poa secunda). Cheatgrass (Bromus tectorum), an exotic annual grass, is present in minor quantities (~ 5% of total leaf area index, LAI). The LoS also includes native grasses, forbs and moss. Grasses include predominantly Sandberg bluegrass, squirreltail (Elymus elymoides), and Idaho fescue (Fescue idahoensis), and forbs include longleaf phlox (Phlox longifolia), pale agoseris (Agoseris glauca), and silvery lupine (Lupinus argentius). Moss (Tortula ruralis) forms the bulk of a biotic crust which can occupy some or much of the soil surface in this community. Cheatgrass is present but accounts for less than 5% of the total LAI.

A prescribed fire was conducted at the PFS site in the fall of 2007. Vegetation at this site is recovering following the fire disturbance. Mountain snowberry (Symphoricarpos oreophilus), a resprouting shrub, is currently the dominant shrub species, but green rabbitbrush and juvenile mountain big sagebrush are also present. Dominant perennial graminoids include mountain brome (Bromus marginatus), Idaho fescue, and Hood’s sedge (Carex hoodii). Studies revealed that LAI, ET, and GEP at the site recovered to pre-burn levels within about 3–4 years post-fire (Flerchinger and others 2016; Fellows and others 2018). At the MBS site, mountain snowberry is also a common shrub species. Dominant perennial graminoids here include mountain brome, Idaho fescue, and squirreltail. Silvery lupine, parsnipflower buckwheat, bigflower agoseris (Agoseris grandiflora) and Nuttall’s violet (Viola nuttallii) are dominant forbs.

Soils at the sites are underlain by volcanic material (basalt/welded tuff) and have received variable additions of aeolian material. Soils were previously described following Schoenberger and others (2002) and quantified for soil carbon and physical properties (see Table 1). In brief, all soils are Mollisols, with the lower elevation sites, WBS and LoS, classified as Typic Argixerolls and the two upper sites classified as Pachic Haploxerolls. The relatively high silt contents, especially at MBS and PFS, reflect aeolian inputs. Strong horizonation in the form of an argillic horizon (and hence higher clay contents) is evident at the WBS and LoS sites. Soil organic carbon contents increase strongly with elevation (5.4–17.9 kgC m−2 in the top 120 cm), whereas measurable amounts of inorganic carbon were found only at WBS and LoS sites, and were prominent only at the WBS site (13.1 kgC m−2).

Meteorological and Flux Measurements

Short and long wave radiation, air temperature and humidity were collected at all sites every 30 min using a four-component net radiometer (CNR-1, Kipp & Zonen, Delft, The Netherlands), and a temperature/humidity probe (HMP155C, Vaisala, Helsinki, Finland). Ground heat flux was measured with six heat flux sensors (HFT3, REBS, Seattle, WA) installed 0.08-m deep within the soil, and three sets of self-averaging thermocouples installed at 0.02 and 0.06-m deep. Measured soil moisture was used to compute volumetric heat capacity of the soil above the heat flux plates. Soil moisture was measured hourly near the PFS site using two sets of time domain reflectometry probes installed at 0.10 m and every 30 min at the remaining sites using Hydra-probe II soil moisture sensors (Stevens Water Monitoring System, Inc., Portland, OR) installed at 0.05 m.

Climate stations operated by the USDA-ARS with 40 + years of observations are located within 70–800 m of each site (Figure 1). Ascending elevationally, these are the Nancy Gulch, Lower Sheep, Upper Sheep and Reynolds Mountain climate stations. Observations at these sites include air temperature, humidity, wind speed and direction, and solar radiation, currently recorded at 15-minute intervals. Dual-gauge systems especially designed for the windy and snow-dominated conditions prevalent in the area were used to measure precipitation (Hanson and others 2004), which was processed hourly. Data from these sites were used to gap-fill missing records at the eddy covariance sites.

Latent heat, sensible heat, water, and carbon dioxide fluxes were measured at all sites using eddy covariance (EC) systems. EC systems consisted of a three-dimensional sonic anemometer (Model CSAT3, Campbell Scientific, Inc., Logan UT) and an open path infrared gas analyzer (IRGA; Model LI-7500a, LI-COR, Inc., Lincoln, NE) sampled at 10 Hz. Systems were mounted to towers between 1.7 and 2.5 m above the plant canopy; heights were 2.05, 2.09, 3.5, and 2.5 m above the ground surface for the WBS, LoS, PFS, and MBS site, respectively.

The 30-minute fluxes of carbon and water were summed daily and by water year to obtain daily and annual ET, net ecosystem production (NEP), Reco and GEP. Details of filtering, processing, gap-filling, and quantifying uncertainty are given in the supplementary online material. In brief, EC data were processed using EddyPro® software (Lincoln, Nebraska USA; https://www.licor.com), carbon fluxes were gap-filled and partitioned using REddyProc software available online (Reichstein and others 2005), ET fluxes were gap-filled according to Flerchinger and Seyfried (2014), and uncertainty analysis included uncertainty due to measurement errors (Richardson and others 2006), gap-filling, and friction velocity threshold (Wutzler and others 2018). Friction velocity thresholds quantified by REddyProc varied seasonally from 0.05 to 0.08 m s−1 at WBS to typically 0.10–0.25 m s−1 at the remaining sites, but was as high as 0.36 m s−1 at PFS during the fall of 2017. Missing observations, filtered data, and non-turbulent conditions resulted in data gaps ranging from around 15% at WBS to 40% at MBS for most years. However, power supply problems during 2015 caused total gaps as high as 52% at the LoS site, resulting in greater uncertainty in the 2015 annual NEP estimates. Energy balance closure over the four years was 0.82, 0.91, 0.80, and 0.79 for the WBS, LoS, PFS, and MBS sites, respectively. Fluxes were not corrected for energy balance closure so that consistent comparisons could be made with those published by Biederman and others (2016).

Vegetation Composition and Structure

Total plant cover, green leaf area index (LAI) and vegetation biomass sampling at the WBS, LoS, and MBS sites were conducted as part of a vegetation monitoring program initiated on the RCEW in 2015. Vegetation cover was measured using the point-intercept method (also referred to as point quadrat method; Clark and Seyfried 2001). At each point, total plant “hits” were recorded, and each plant hit was identified as a transpiring or non-transpiring plant part to obtain total plant and green leaf area indices. Protocols were refined each year. For 2015, thirteen one-meter frames were sampled on each of the three 50-m transects, for a total of 39 frames at each site with 20 points per frame. For subsequent years, randomly located one-meter frames increased from 90 frames with 20 points per frame in 2016 to 150 frames with 100 points per frame in 2018. Vegetation cover at the PFS site was sampled in 2017 at 12 randomly selected 1-m2 plots each with five frames, as described by Flerchinger and Seyfried (2014) and Flerchinger and others (2016) for years 2005 through 2012. Results for all sites are reported in Table 1.

Previous Data Collected at RCEW Sites

Data from the RC CZO Climate Gradient Network were supplemented using data previously collected within the RCEW. These data include: pre- and post-fire years (2006 through 2012) from the Upper Sheep site (that is, the PFS site; Flerchinger and others 2016; Fellows and others 2018); and four years of data (2004–2007) from the Reynolds Mountain meteorological station (Flerchinger and others 2010; Reba and others 2009; Fellows and others 2019). Pre-fire vegetation at the Upper Sheep site consisted primarily of mountain big sagebrush, snowberry and an understory of native grasses and forbs. The Reynolds Mountain site is a windswept mixed sagebrush (low and mountain big sagebrush) site near the MBS site but receives less effective precipitation due to snow being blown from the windswept ridge. Ten-hertz data from both studies were systematically re-processed and filled using similar procedures as the core network sites to provide a consistent long-term dataset.

Ecosystem Metrics

Evaporative fraction, that is, the fraction of available energy allocated to evapotranspiration, is an effective indicator of soil moisture and water stress (Nutini and others 2014). Daily averages of evaporative fraction (EF) throughout the growing season were used as a measure of ecosystem water stress and computed from

$$ {\text{EF}} = {\raise0.7ex\hbox{${\text{LE}}$} \!\mathord{\left/ {\vphantom {{\text{LE}} {\left( {{\text{H}} + {\text{LE}}} \right)}}}\right.\kern-0pt} \!\lower0.7ex\hbox{${\left( {{\text{H}} + {\text{LE}}} \right)}$}} $$
(1)

where LE and H are the latent and sensible heat fluxes. EF values close to 1.0 indicate no water stress while values closer to zero indicate increasing water stress. The advantage of EF over other measures, such as the ratio of ET to potential ET, is that EF can be computed directly from the turbulent flux measurements, and unlike the Bowen Ratio (H/LE), it remains numerically stable as LE approaches zero.

The main growing season (MGS) was identified for each water year to examine seasonal differences in the fluxes. The start and end of the MGS were identified by the date on which four consecutive days of GEP rose above and fell below, respectively, the median GEP for the year (Schmidt and others 2012).

Model Simulations

Potential effects of climate warming on snow accumulation, ET and GEP were assessed using simulations from the Simultaneous Heat and Water (SHAW) Model. The SHAW model has been tested and applied extensively over a range of vegetation types in semiarid and arid environments (Flerchinger and others 1996, 2012; Flerchinger and Seyfried 2014; Chauvin and others 2011; and so on). The SHAW model is a numerical representation of a vertical, one-dimensional system composed of a multi-species plant canopy, snow cover if present, plant residue, and the soil profile. This study uses a version of the model described by Yu and Flerchinger (2008) that uses a Ball–Berry model to simulate photosynthesis and stomatal conductance and was updated to include within-canopy routines described by Flerchinger and others (2015). The model simulates the surface energy balance, evaporation, photosynthesis, transpiration, soil water, snow accumulation/melt and heat and water fluxes using a detailed, physically based solution to the energy and mass balance equations. The detailed physical representation incorporated in the model makes it suitable for capturing the effects of climate warming on snow accumulation and melt, soil water recharge, and water availability on ET and GEP.

Simulations of ET and GEP were validated using observations from the 2015 and 2016 water years as described in the supplementary online material. The Nash–Sutcliffe R2 (otherwise known as Model Efficiency) for simulated daily ET over both water years was 0.59, 0.65, 0.84, and 82 for WBS, LoS, PFS, and MBS, respectively. Model efficiency for daily GEP was 0.61, 0.79, 0.80 and 0.79.

After validation, simulations were conducted for observed and potential climate change scenarios using observations from water years 2006 through 2016. A potential + 2 °C warming scenario, which could be experienced within the next 50 years (Chambers and others 2008; Mote and others 2014), was explored by adding 2 °C to each hourly weather observation from the nearby meteorological station for each site; wind speed, relative humidity, and solar radiation were assumed unchanged. Climate predictions are less certain with respect to precipitation, ranging from a 10% decrease to an 18% increase (Chambers and others 2008; Mote and others 2014). So, in addition to the 2 °C warming, scenarios were explored for precipitation changes ranging from − 50% to + 200% by adjusting each hourly precipitation observation accordingly. Although excessive, these scenarios provide sufficient overlap in precipitation between the driest and wettest sites to be informative.

More specifically, 16 simulations were run for each site to include every combination of temperature changes (+ 0 °C and + 2°) and precipitation changes (− 50%, − 20%, + 0%, + 20%, + 50%, + 100% + 150% and + 200%). In all simulations, the first two years of the 11-year simulation (water years 2006–2016) were discarded from the analyses to allow simulated temperature and moisture conditions to adjust to the changed climate forcing. Feedback of perturbed temperature and precipitation on maximum LAI was based on the ratio of simulated GEP to the base simulation (i.e. no perturbations), and temporal variation in LAI was adjusted for the timing of complete snowmelt as described by Flerchinger and others (2016).

Results

Precipitation for water years 2015 through 2018 deviated from the measured 30-year mean by − 9% to + 33% at the sites (Table 1), but deviation changed with elevation for the different years. Precipitation for water year 2015 was + 16% from the mean at the lowest elevation and − 7% at the highest elevation. Conversely, water year 2016 precipitation was − 9% of the mean at the lowest elevation and +18% at the highest elevation (see Table 1). The winter of 2017 experienced heavy snowfall, resulting in precipitation being around + 40% of the mean at all sites. Water year 2018 was slightly drier than normal at all sites, but within 10% of the mean. Cold-season (December–March) precipitation at the four sites deviated from the 30-year average by − 7 to − 18% in water year 2015, +3 to +44% in 2016, +65 to +114% in 2017 and −18 to −32% in 2018.

Cross-Site Contrasts and Seasonal Trends

Maximum snow depth at the sites varied from less the 30 cm at WBS to around 130 cm (2017) at MBS. Complete snowmelt varied between mid-January and mid-March (day of year, DOY, 15–69; Table 2) for all years at the two lower elevation sites; however, WBS was free of snow for nearly all of water year 2015. By contrast, complete snowmelt at the higher elevation sites occurred 1½ to 2 months later. Response of ET, NEP, and GEP is consistent with this difference in snowmelt date, with fluxes picking up at least a month later at the higher elevation sites (Figure 2).

Table 2 Annual Carbon and Water Fluxes, Snowmelt Date, and Percentage of Fluxes Attributed to the Main Growing Season for Sagebrush Ecosystems Across an Elevation Gradient
Full size table
Figure 2
figure 2

Average daily A evapotranspiration (ET), B net ecosystem production (NEP), and C gross ecosystem production (GEP) for water years 2015 through 2018. (Negative values of NEP indicate net flux of CO2 to the ecosystem.) Data are plotted for the Wyoming big sagebrush (WBS), low sagebrush (LoS), post-fire sagebrush (PFS) and mountain big sagebrush (MBS) sites.

Full size image

On average, ET and GEP began to increase at the two lower elevations around February–March (Figure 2), coincident with the start of the MGS (Table 2). However, with limited precipitation, ET and GEP decreased sharply, first at WBS and then at LoS (early May to June; Figure 2A, C), and the end of the MGS occurred in late July (Table 2). By contrast, daily ET rates did not decline at PFS and MBS until late June, and the MGS extended through to the end of the water year at MBS. Water stress, inferred indirectly from evaporative fraction (EF) plotted in Figure 3, is consistent with available precipitation at the different sites (Table 1). Based on Figure 3, the two higher elevation sites do not typically experience appreciable water stress (taken as EF < 0.5) until August, whereas WBS experienced some degree of water stress for nearly the entire spring and summer season. EF declined rapidly at both lower elevation sites through June.

Figure 3
figure 3

Average daily A precipitation and B evaporative fraction (i.e. fraction of total turbulent heat flux, H + LE, attributed to ET) over the growing season for years 2015 through 2018 for the Wyoming big sagebrush (WBS), low sagebrush (LoS), post-fire sagebrush (PFS) and mountain big sagebrush (MBS) sites. Precipitation is the daily average for all for sites; Figures 5 and S2 present precipitation for each individual site.

Full size image

Onset and magnitude of ecosystem productivity is strikingly similar at the two lower elevations as well as the two higher elevation sites (Figure 2C) until peak productivity is reached. Subsequent to peak productivity, LoS and MBS maintained their productively longer compared to WBS and PFS, respectively. Although ecosystem productivity typically began earlier (Feb/Mar) at the two lower elevation sites (Figure 2C; Table 2), the higher elevation sites quickly outpaced their productivity upon arrival of spring after snowmelt. Average MGS length was consistent across sites, ranging from 155 to 159 days, except for LoS where MGS length averaged 133 days. MGS productivity accounted for roughly 70% of the GEP at the lower elevation sites compared to nearly 90% at the higher elevations; thus, the lower elevation sites were better able to take advantage of off-season productivity, not being as influenced by snow cover. MGS respiration averaged 56–66% of the annual Reco for all sites.

Annual GEP for the sites averaged 385, 549, 684, and 818 gC m−2, respectively, from lowest to highest elevation for the four water years (Table 2). WUEeco (taken as the ratio of GEP to ET) averaged 1.26, 1.26, 1.35 and 1.61 gC mm−2 H2O from lowest to highest elevation. Figure 2B indicates carbon uptake (negative NEP) at all sites early in the spring, beginning first with the low elevation sites. With low precipitation input, the low elevation WBS site typically transitions to a carbon source in June, coincident with the drop in ET and GEP at the site. Despite differences in precipitation input, seasonal magnitude and timing of NEP responded similarly at the two higher elevation sites, with net carbon uptake not occurring until May and persisting through mid-September. All sites were typically a carbon sink, ranging from a 100 to 200 gC m−2 net flux to the ecosystem (Table 2), with the higher elevations tending to be greater carbon sinks. Notable exceptions to this are WBS for water year 2015 and MBS for water year 2017, which will be discussed later.

Inter-Annual Variability and Trends Across Sites

Inter-annual variability in precipitation was similar across sites, with coefficient of variation (CV) ranging only from 17 to 23% (Table 2). ET was most variable at WBS, both in terms of standard deviation (Stdev) and CV (34 mm and 11%, respectively). In contrast, Stdev and CV for annual ET at PFS was only 11 mm, or 2%. GEP varied the least at the two higher elevations in terms of CV. NEP was most variable at WBS and MBS, with CV’s of 68 and 47%, respectively. Observed year-to-year trends shown in Figure 4 and provided in Table 2 are consistent across sites, with a few exceptions. NEP at WBS for 2015 and at MBS (and to a lesser extent at PFS) for 2018 stands out as outliers from the general trends in fluxes across the elevation gradient.

Figure 4
figure 4

Inter-annual variability in A precipitation, B evapotranspiration (ET), C gross ecosystem production (GEP), D ecosystem respiration (Reco), and E net ecosystem production (NEP) across an elevation gradient of sagebrush ecosystems. Error bars are the combined measurement, gap-filling, and friction velocity threshold uncertainties (σm, σg, and σu*) added in quadrature. (Negative values of NEP indicate net flux of carbon to the ecosystem.).

Full size image

Annual ET at WBS was strongly controlled by precipitation, the two being within 55 mm of each other every year. Years 2016 and 2018 were very similar at WBS in terms of precipitation, ET, GEP and Reco. Cumulative GEP for 2015 fell short of the other years beginning in April (Figure 5D). Despite precipitation and ET being similar for 2015 and 2017, sums for GEP, Reco and NEP differed by 154, 50, and 104 gC m−2, respectively. This can be partly attributed to the timing of precipitation. July rains bringing 69 mm of water at WBS resulted in a noticeable increase in NEP, associated with increased respiration and ET at the two lower elevation sites (Figures 5A–C and S2a, b, c). The pulse of respiration between July 4 and August 14 (DOY 185 and 226) at WBS accounted for a total of 57 gC m−2, or an estimated 27 gC m−2 above the trend immediately before and after the pulse. Prior to this pulse, cumulative Reco for 2015 was very similar to 2016 and 2018, but it ended the year 24–43 gC m−2 higher. Grasses and forbs at WBS, which constitute approximately 45–55% of the LAI, had already senesced (as observed during site visits and substantiated by the end of the MGS in Table 2) and could not take advantage of the water. Despite 2015 GEP being low for WBS, 2015 ET was highest out of the four years, resulting in WUEeco falling to 0.92 gC mm−1 for the year, compared to the site average of 1.26 gC mm−1. Off-season Reco was particularly high for 2015 (Table 2); the combination of reduced GEP and additional respiration after the July rains resulted in WBS being carbon neutral for 2015.

Figure 5
figure 5

Cumulative precipitation (PPT), evapotranspiration (ET), net ecosystem production (NEP), and gross ecosystem production (GEP) for water years 2015 through 2018 for the Wyoming big sagebrush (A–D) and mountain big sagebrush (E–H) sites. (Negative values of NEP indicate net flux of carbon to the ecosystem.).

Full size image

A pulse in ET and Reco was also observed at LoS in response to the 2015 July rains (Figure S2). However, there was also an increase in GEP after the July rain as the MGS had not ended at this site (Table 2). As a result, there was only a slight increase in cumulative NEP at the site (approximately 18 gC m−2 at LoS compared to 36 gC m−2 at WBS; Figure S2c and 5c). Additionally, GEP for 2015 was not lower than the other years at this site.

Observations at MBS suggest that GEP is inversely related to precipitation (r2 = 0.78). Higher precipitation at the higher elevation often translates to a delay in snowmelt (Table 2), with GEP also being negatively correlated to the date of complete snowmelt (r2 = 0.68). Complete snowmelt for years 2016 and 2017 was 8–12 days later than in 2015. The influence of this snowmelt delay can be seen in Figure 5H as GEP picks up in early April during 2015 while GEP during 2016 and 2017 does not begin to rise until May. MGS was also delayed compared to 2015 (Table 2). As a result, annual GEP was much less during 2016 and 2017. This, combined with larger respiration fluxes through the heavy snow year of 2017 (Figure 5G), resulted in the MBS site being substantially less of a carbon sink during 2017.

Similar to the MBS site, GEP at the PFS site was delayed during 2016 and 2017 compared to the 2015 (Table 2). Also, GEP was lower for 2016 and 2017 and highest for 2018. However, the response of GEP at PFS did not exhibit the high correlation with snowmelt date. Water year 2018 resulted in the highest GEP and Reco fluxes at both MBS and PFS. Similar to MBS, a combination of below average GEP and above average Reco at PFS resulted in 2017 being an abnormally weak carbon sink compared to other years at this site.

Observed fluxes at the four core network sites are consistent with fluxes measured previously at nearby sites within RCEW. Annual GEP for MBS averaging 880 gC m−2 is slightly higher than the range of values measured from 2004 to 2007 at the drier, windswept Reynolds Mountain site (418 to 632 gC m−2). However, average NEP of −142 gC m−2 for the MBS is in the middle of range of that at the Reynolds Mountain site (−75 to −185 gC m−2). Annual GEP and NEP for PFS are well within the range of values measured from 2006 to 2012 at the site (532 to 813 gC m−2 for GEP and −60 to −226 gC m−2).

The relationship between GEP and ET presented in Figure 6 for the four core network sites and previous sites at RCEW shows a relatively strong correlation (r2 = 0.61) and is consistent with that for 21 sites covering a diverse range of semiarid biomes in the Western USA given by Biederman and others (2016; r2 = 0.73). An F-test indicated that the Biederman relationship significantly accounted for the variance in observed GEP for the RCEW sites (p < 0.05).

Figure 6
figure 6

Plot of measured annual gross ecosystem production (GEP) versus annual evapotranspiration (ET) for several sites within the Reynolds Creek Experimental Watershed (RCEW) along with the best fit line (GEP = 1.7ET −138 gC m−2; r2 = 0.61) for the Reynolds Creek sites. Also plotted is the relationship derived by Biederman and others (2016) for a diverse range of semiarid biomes. Previous sites within RCEW were reported by Fellows and others (2019; Reynolds Mtn site) and Fellows and others (2018; Upper Sheep site).

Full size image

Simulated Impacts of Climate Changes

Simulated warmer temperatures resulted in advances in timing of ET and GEP at all sites but not overall cumulative ET and GEP at the lower sites, whereas ET and GEP increased at the higher elevation sites. Figure 7 presents the difference in cumulative average ET and GEP for the nine water years (2008 through 2016) between observed conditions and a potential climate warming of +2 °C. Despite the warmer temperatures, cumulative ET and GEP at the two lower elevation sites were nearly unchanged, although the timing was advanced. For example, cumulative ET was 23 mm and GEP was 44 gC m−2 higher by May at LoS for the warmer climate, but it was within 0.2 mm and 1.5 gC m−2 by the end of the water year. Timing of ET and GEP was also advanced for the two higher elevation sites under the +2 °C scenario, but they had adequate moisture to sustain productivity longer, resulting in an additional 40 mm of ET and 103 gC m−2 at MBS. Thus, the influence of warmer temperatures on annual ET and GEP was more pronounced with higher precipitation, as will be explored further.

Figure 7
figure 7

Difference between A average simulated cumulative ET (mm) and B average simulated cumulative GEP (gC m−2) for current climate and +2 °C scenario (that is, ΣET+2C − ΣET and ΣGEP+2C − ΣGEP).

Full size image

The fraction of observed Dec-Mar sub-freezing precipitation under the current climate that falls at temperatures above −2 °C ranges from 30% at MBS to 48% at WBS. Thus, the influence of the simulated warmer climate on snow cover was profound at all elevations, as shown in Figure 8. Both the amount and duration of snow cover at the two higher elevation sites were affected significantly, resulting in the sites being free of snow approximately one month earlier under the warmer climate. Indeed, simulated average snow conditions at MBS under the warmer climate were very similar to the average snowpack at PFS under the current climate. The simulated average snowpack at the two lower elevation sites was much more transient under the warmer climate (Figure 8A), allowing for periods of plant growth and transpiration throughout the winter.

Figure 8
figure 8

Average simulated snow water equivalent for water years 2008–2016 for measured weather conditions and under a climate warming scenario of +2 °C for the four sagebrush sites along an elevation transect. Lower elevation sites are plotted in panel (A) and upper elevation sites are plotted in panel (B).

Full size image

A plot of simulated ET versus precipitation for a range of precipitation scenarios under observed 2008-2016 temperatures follows a 1:1 trend until about 400 or 500 mm (Figure 9A). This trend held for all sites regardless of elevation, soil or vegetation, suggesting that nearly all precipitation is used for ET until approximately 400 mm. This threshold agrees well with Hibbert (1983) who found that there is little potential for increasing streamflow by reducing ET through rangeland vegetation management unless precipitation exceeds 450 mm. GEP in Figure 9B shows a similar trend with GEP increasing linearly until a precipitation threshold around 400–500 mm. The plot of the ratio between ET for the +2 °C scenario to ET for the current temperature regime (ET+2C/ET) shown in Figure 9C for all precipitation scenarios shows very little change in ET under a warmer climate below 400 to 500 mm of precipitation, but the influence of warmer temperatures caused as much as a 10% increase in ET at 800 mm of precipitation. Figure 9D, however, shows an interesting result, with GEP decreasing under the +2 °C scenario below 400 mm precipitation and increasing by as much as 15% at higher precipitation. The decrease in GEP can be attributed to an additional 2–3% of the ET being consumed by evaporation rather than transpiration under the warmer climate and low precipitation.

Figure 9
figure 9

Simulated annual A ET and B GEP versus precipitation for the water years 2008–2016 using observed temperature and a range of precipitation scenarios (−20% to +200%); and C, D the ratio of average annual ET and GEP for the +2 °C warming scenario to average annual ET and GEP for +0 °C (ET+2C/ET and GEP+2C/GEP, respectively) versus average precipitation for a range of precipitation scenarios. Sites include the Wyoming big sagebrush (WBS), low sagebrush (LoS), post-fire sagebrush (PFS) and mountain big sagebrush (MBS) sites.

Full size image

Climate change scenarios presented herein assume the current atmospheric CO2 concentration of approximately 400 ppm. Each climate scenario was run a second time using a CO2 concentration of 475 ppm (https://www.co2.earth/2100-projections?), which is an 18% increase. Although this increase had essentially no impact on ET, the simulated increase in GEP was around 18% at all sites.

Discussion and Conclusions

Results from our study show strong elevational dependence in the timing, onset, and magnitude of ET, GEP, and NEP that is consistent with the local aridity gradient (Figure 4). Specifically, results indicate that the low elevation WBS site was carbon–neutral for 2015 (partly due to a post-rain summer respiration pulse), whereas the cooler and wetter sites at higher elevation were carbon sinks. All sites were found to be carbon sinks for the remaining years; however, the high elevation MBS site exhibited a sharp decline in net carbon uptake during the 2017 heavy snow year. In addition to finding strong precipitation controls on ET and GEP at the lower drier sites, findings suggest a combination of temperature and soil moisture limitation on GEP in cold shrub-steppe ecosystems at the higher elevation sites. Timing of complete snowmelt exhibited a strong control on annual GEP at the highest elevation; this is consistent with findings of Fellows and others (2019) for a nearby (~ 800 m separation) aspen site, but they found no correlation for the nearby windswept, mixed sagebrush Reynolds Mountain site. Collectively, our data and the warming simulations indicate that climate changes will have differential influence on carbon uptake along this elevation gradient through changes in temperature and timing and amount of water.

The regression between GEP and ET presented by Biederman and others (2016) for sites largely in the southwestern USA was substantiated herein (Figure 6) with 27 site-years of data, suggesting commonality in the underlying mechanisms and vegetation efficiencies across the Western USA despite stark differences in climate and precipitation timing. This commonality can enable comparisons between cold shrublands and other more heavily studied ecosystems and provide clues to the potential impacts of climate warming. For example, annual WUEeco herein ranged from 0.96 to 1.87 and tended to be lowest at the dry, lower elevation sites. This is consistent with the Biederman regression which suggests zero GEP, and therefore zero WUE, at about 80 mm of ET (although truly extrapolating the relation to zero GEP is not warranted; Figure 6).

Results indicate that ET and GEP may increase with climate warming at higher, wetter elevations but decrease at lower elevations, and that sensitivity is tightly linked to early snowmelt, water availability, and prolonged growing season. The two lower elevation sites use all the available water in the current climate. Measured ET was nearly equal to annual precipitation, and simulated ET did not increase with a warmer climate at these lower elevation sites, suggesting that ET and GEP are limited by total water supply. Indeed, results suggest that ET may increase with climate warming only where precipitation is above a threshold of about 450 mm (Figure 9). For example, the warming impact was greatest at the MBS site where modeled ET increased by about 40 mm and GEP increased by about 100 gC m−2 y−1 (Figure 7). Nearly all of the increase in simulated ET and GEP at the MBS site under the +2 °C scenario occurred before June, the date of average snow ablation under the current climate (Figure 8). This is consistent with the fact that GEP under the current climate at MBS is correlated with date of snowmelt, suggesting that GEP at the MBS site is largely limited by temperature and growing season. The Upper Sheep PFS site +2 °C scenario also showed a large increase in cumulative ET and GEP by June, however it was unable to sustain ET throughout the growing season, resulting in an average increase of 30 mm of ET and about 55 gC m−2 y−1 of GEP. The modest increase at the Upper Sheep PFS site is consistent with results from Flerchinger and others (2016) indicating that the PFS site uses nearly all precipitation input available to it during most years. Thus, ET and GEP at the PFS site is likely on the cusp being water limited.

Observations of energy and mass fluxes along an elevation/climate gradient can provide clues as to what further climate warming might impose on the ecosystems and to their potential response if the rain–snow transition elevation continues to rise. The four sites located along a climate gradient within RCEW have annual average temperatures of 9.1, 8.4, 6.1 and 5.4 °C. Given climate change projections (Chambers and others 2008; Mote and others 2014), temperature regimes could quite conceivably transition between these sites over the next 50 years, for example, the future PFS average temperature could be close to that of present day LoS. Under historical climate conditions, winter precipitation in western mountains was retained as snow, which melted as a fairly discrete pulse in spring. Climate warming will continue to reduce or eliminate water storage in the snowpack (Nayak and others 2010) as shown by the simulations plotted in Figure 8, and thereby shift water availability earlier in the spring (or late winter) as the snowpack melts earlier in the season. An increase in rain at the expense of snow may result in an earlier initiation of the growing season; results from this study suggest that this shift in timing of water availability will play out differently at different elevations depending on whether plant productivity at the site is predominantly water limited.

Lower elevations, where productivity is water supply limited, will likely experience increased plant water stress in the dry summer season with attendant mortality and fire hazard under a warmer climate. Indeed, a +2 °C temperature increase had a negligible impact on cumulative ET and GEP, but did shift ET and GEP earlier in the season (Figure 7). Additionally, the date during the active growing season at which simulated daily evaporative fraction (EF) dropped below 0.5 (that is, less than half of the turbulent fluxes going to evapotranspiration) averaged one to two weeks earlier under the warmer climate at the lower elevation sites. In contrast, there was no change in the average date at the MBS site. Thus, warmer spring temperatures under a climate change scenario may result in a shift to earlier ET, plant production and plant water use, resulting in an earlier onset and longer periods of water stress. Low elevation sagebrush sites are already susceptible to encroachment by invasive annual weeds such as cheatgrass that can take advantage of early season soil water availability (Harris and Wilson 1970; Call and Roundy 1991; Young and Longland 1996; Chambers and others 2007, 2014); water stress occurring earlier in the growing season may exacerbate this encroachment (Polley and others 2013), resulting in an associated increase in wildfire frequency (Young and Longland 1996). Although annual invasive weed encroachment has not historically been prevalent at higher elevations (Chamber and others 2007, 2014), these areas may be put at risk with the shift from snow to rain (Concilio and others 2013) if temperature is a current limitation to encroachment (Bradley and others 2009).

Higher elevations receiving more than 450 mm of precipitation will likely increase productivity under a warmer climate. Simulations presented herein indicating earlier, increased GEP and limited water stress at the highest elevation site under a warmer climate support this conclusion. Findings from this study also suggest that there is limited opportunity for increased GEP under a warmer climate at mid-elevation sites such as the Upper Sheep PFS site without an increase in precipitation. Given that less than 20% of watershed and 34% of the Great Basin lies above the elevation of Upper Sheep, it is likely that only a modest portion of RCEW and Western rangelands as a whole would experience increased productivity.

Most of the streamflow from the snow-dominated mountainous watersheds is generated from the higher elevations, where much of the precipitation input comes as a pulse of snowmelt water in early spring, fills the root-zone soil water storage capacity, and then is lost to streamflow. The earlier growing season and increased ET will result in reduced recharge and streamflow; thus, the increase in productivity under a warmer climate at higher elevations will likely come at the expense of downstream ecosystem services in the form of reduced water supply. The 40 mm (or 10%) increase in simulated ET at MBS is important in Reynolds Creek, and presumably many parts of the northern Great Basin. Streamflow near the MBS site, for example, ranges from 87 to 1064 mm, indicating the ET increase may account for 4–46% of streamflow in the area.

Results presented from the RC CZO Climate Gradient Core Network are from only the first four years of observation having precipitation ranging from −9% to + 33% of the measured 30-year mean. Long-term monitoring across this gradient capturing a range of weather patterns, along with ancillary studies at the sites, will provide the opportunity to assess how these ecosystems respond to different stress conditions and the resultant response of GEP and net carbon flux to the ecosystems. Data from three of the sites (WBS, LoS and MBS) are currently available for download from the AmeriFlux network (https://ameriflux.lbl.gov/). Information gained from these sites will provide the opportunity to advance our understanding and prediction of ecosystem processes, energy and carbon fluxes, and carbon cycling within these sagebrush ecosystems.