, Volume 126, Issue 3, pp 363–378 | Cite as

Observations of net soil exchange of CO2 in a dryland show experimental warming increases carbon losses in biocrust soils

  • Anthony Darrouzet-Nardi
  • Sasha C. Reed
  • Edmund E. Grote
  • Jayne Belnap


Many arid and semiarid ecosystems have soils covered with well-developed biological soil crust communities (biocrusts) made up of mosses, lichens, cyanobacteria, and heterotrophs living at the soil surface. These communities are a fundamental component of dryland ecosystems, and are critical to dryland carbon (C) cycling. To examine the effects of warming temperatures on soil C balance in a dryland ecosystem, we used infrared heaters to warm biocrust-dominated soils to 2 °C above control conditions at a field site on the Colorado Plateau, USA. We monitored net soil exchange (NSE) of CO2 every hour for 21 months using automated flux chambers (5 control and 5 warmed chambers), which included the CO2 fluxes of the biocrusts and the soil beneath them. We observed measurable photosynthesis in biocrust soils on 12 % of measurement days, which correlated well with precipitation events and soil wet-up. These days included several snow events, providing what we believe to be the first evidence of substantial photosynthesis underneath snow by biocrust organisms in drylands. Overall, biocrust soils in both control and warmed plots were net CO2 sources to the atmosphere, with control plots losing 62 ± 8 g C m−2 (mean ± SE) over the first year of measurement and warmed plots losing 74 ± 9 g C m−2. Between control and warmed plots, the difference in soil C loss was uncertain over the course of the entire year due to large and variable rates in spring, but on days during which soils were wet and crusts were actively photosynthesizing, biocrusts that were warmed by 2 °C had a substantially more negative C balance (i.e., biocrust soils took up less C and/or lost more C in warmed plots). Taken together, our data suggest a substantial risk of increased C loss from biocrust soils with higher future temperatures, and highlight a robust capacity to predict CO2 exchange in biocrust soils using easily measured environmental parameters.


Autochamber Biological soil crusts Castle valley Utah Soil respiration Gap filling Global climate change Net soil exchange NSE Net ecosystem exchange NEE 


Dryland responses to global change have the potential to affect carbon (C) cycling and climate at the global scale; for example, recent analyses suggest that the interannual variability in the planet’s terrestrial C sink is driven by dryland ecosystems (Poulter et al. 2014). Drylands are also Earth’s largest biome (Schimel 2010) and, although these arid and semiarid landscapes (with an aridity index of <0.65) typically store less C per unit area than other biomes (Safriel et al. 2005), their large global extent translates into dryland pools and fluxes that are highly relevant to global C cycling (drylands make up 41 % of the terrestrial surface). Understanding the response of dryland soil C fluxes to climate change is also crucial to considerations of soil health in arid and semiarid systems, as organic matter content is a fundamental control on soil fertility, stability, and thus susceptibility to desertification (Lal 2003; Schlesinger et al. 1990).

A key feature of drylands biogeochemistry is the thin soil surface known as the “mantle of fertility”, which is dominated by biological soil crusts (biocrusts; Collins et al. 2008; Garcia-Pichel and Belnap 2003; Pointing and Belnap 2012) and where a great deal of soil C cycling and storage occur. Biocrusts are a complex consortium of organisms that includes multiple species of lichens, cyanobacteria, mosses, and a diverse (mostly uncharacterized) array of bacteria and microfungi (Belnap and Lange 2003). Biocrusts cover most dryland soil surfaces that are not occupied by stems of vascular plants, rocks, or active disturbance (Harper and Marble 1988; West 1990), and can account for 70 % or more of a dryland ecosystem’s living cover (Belnap 1995). Beyond their many functions related to soil stability, fertility, and hydrological cycling (Belnap and Lange 2003), biocrusts are also an important source of C uptake due to their autotrophic constituents (Beymer and Klopatek 1991; Elbert et al. 2012; Lange 2003). In addition to their importance, biocrust organisms may be especially vulnerable to climate change (Ferrenberg et al. 2015; Grote et al. 2010; Maestre et al. 2013; Reed et al. 2012). Previous results from the same site we focus on in this study, a semiarid ecosystem on the Colorado Plateau, have shown that changes in temperature and rainfall regimes can have large effects on the composition of biocrust communities (Coe et al. 2012; Ferrenberg et al. 2015; Reed et al. 2012; Zelikova et al. 2012), and that these effects have feedbacks that significantly modify C and nutrient cycling (Reed et al. 2012). Specifically, frequent small artificial precipitation events were shown to rapidly kill mosses (Reed et al. 2012), while temperature had large effects on both lichen and moss populations but over a much longer timescale (Ferrenberg et al. 2015). Similar reductions in biocrust cover were seen in a climate manipulation experiment in Spain in response to 2–3 °C warming (Maestre et al. 2013). Despite the vulnerability of these key organisms to climate change and their importance in the C cycle, relatively few experiments have addressed how climate change will affect C balance in soils dominated by biocrusts.

In this study, we will refer to dryland soils that support healthy surface communities of biocrusts as “biocrust soils”. These soils include not only the photosynthetic biocrust organisms residing at the soil surface, but also a suite of soil heterotrophs functioning throughout the soil profile, as well as vascular plant roots. Because autotrophic biocrust organisms take up CO2 via photosynthesis and because biocrusts, heterotrophs, and plant roots respire CO2, we refer to our measurements of CO2 fluxes from the soil profile as net soil exchange (NSE) of CO2. NSE is the difference between the gross fluxes of CO2 uptake (via biocrust photosynthesis) and CO2 release (via respiration from multiple soil sources), and is distinguished from net ecosystem exchange (NEE) of CO2 by the fact that it does not include vascular plant photosynthesis.

Temperature and moisture are key controls on NSE in biocrust soils (Fernandez et al. 2006; Grote et al. 2010). In laboratory settings, increased temperatures have been shown to cause higher C losses in biocrust organisms when soil moisture levels are kept constant (Grote et al. 2010; Lange et al. 1997, 1998). Furthermore, data from the climate manipulation in Spain suggested significant increases in C loss from biocrust soils in warmed plots (Maestre et al. 2013). However, as with many assessments of soil respiration or NSE, those data were based on single measurements taken during the daytime every 1–4 months, and thus cannot be used to integrate total fluxes over longer periods, such as annually. Due to substantial daily and seasonal variations in NSE and NEE that have been documented in dryland ecosystems (Bowling et al. 2010, 2011), measurements of these fluxes at high temporal resolution are necessary to accurately describe dryland C cycling. In sum, the data we do have suggest that predicted global warming may have large consequences for biocrust soil NSE; however, we lack a record for a full year’s NSE for any well-developed biocrust soils and no studies have combined temperature manipulations with high temporal resolution measurements of C exchange in biocrust soils.

We installed automated soil CO2 flux chambers and measured NSE at an hourly time scale for 21 months to assess the effect of +2 °C warming on the C balance in biocrust soils. Our research questions were: (1) What are the diurnal and seasonal patterns in the NSE of CO2 for biocrust soils in a semiarid ecosystem? (2) What is the summed annual NSE? (3) Can environmental controls such as temperature, moisture, and light be used to predict NSE? and (4) How does NSE respond to a +2 °C warming treatment? Based on the previous research discussed above, we predicted that our warming treatment would stimulate soil respiration more than biocrust photosynthesis, resulting in increased soil CO2 losses under warming.

Materials and methods

Study site

The site was located in a cool desert ecosystem on the Upper Colorado Plateau (36.675 N, −109.416 W; near Castle Valley, UT) at an elevation of 1310 m above sea level. Mean annual temperature for the surrounding area is 13 °C and mean annual precipitation is 269 mm, of which 65 % comes in the winter and spring, (based on 1981–2010 data; WRCC 2014). Soils at the site are relatively shallow (with exposed bedrock in places) and classified as sandy loam, calcareous, Rizno series Aridisols. The vegetation is dominated by a native C3 perennial grass, Achnatherum hymenoides; a native C4 perennial grass, Pleuraphis jamesii; a native C4 perennial shrub, Atriplex confertifolia; and the exotic invasive C3 grass, Bromus tectorum. Biocrust communities are dominated by the cyanobacterium Microcoleus vaginatus, the moss Syntrichia caninervis, and the cyanolichens Collema tenax and Collema coccophorum.

Warming treatment

In 2005, five control and five warmed (ambient +2 °C) plots were established at this site in a randomized block design. The larger experimental design also includes a set of associated altered monsoonal precipitation plots and plots receiving both altered precipitation and warming (Zelikova et al. 2012), but these treatments are not part of the analysis described here. Also, we note that although temperature treatments eventually affected biocrust community composition (Ferrenberg et al. 2015), this effect occurred many years after the measurement time of this study. The plots used were 2 × 2.5 m and oriented with the long side running east to west. Because the field site is located on a natural drainage gradient of approximately 10 % slope, the experimental plots were arranged in a randomized blocked design, with the five blocks located along the drainage gradient. The plots within the each block were randomly assigned to either control or warmed (ambient +2 °C at 5 cm soil depth). The plots were warmed using one 800 W infrared radiant (IR) heat lamp (Kalglo Model MRM-2408) per plot. Lamps were turned on in October 2005. Each heater was placed 1.3 m above the soil surface and oriented east to west to minimize shading of the plot (Harte et al. 1995; Kimball 2005). The control plots have identical ‘dummy lamps’ placed in the same configuration as the plots that receive warming, but the lamps have no heating elements. Due to high microsite variability in soil surface temperatures, the heating treatment was regulated weekly by altering the voltage supplied to each lamp. The treatment was effective at heating the soils most of the year, although the effect of the lamp was reduced beginning at temperatures higher than 27 °C; this reduction continued to the point where there was little warming when soils were at their hottest temperatures of ~40 °C (Figure S1). There were also a few times during which the temperature effect was inverted (i.e., colder soil in the plots receiving warming) due to the melting of snow and subsequent loss of thermal insulation (Figure S1).

Environmental variables

Soil moisture was monitored in each plot using horizontally oriented water content reflectometer probes (CS616, Campbell Scientific, Inc., Logan, UT) at both 5 cm and 15 cm depth. Due to microsite differences in the presence of these probes, site averages within each treatment were used. These moisture data were also temperature-corrected to account for temperature effects on CS616 readings. Soil temperature was measured in each plot at 5 cm and 15 cm depth using 4-tipped thermopiles made from Type-T thermocouple wire (Omega Engineering Inc.). Air temperature and other basic weather variables were measured with a CR10X datalogger-based weather station on site (Campbell Scientific Inc.). Light availability was measured with photosynthetic photo flux density (PPFD) sensors (Apogee Instruments Inc. model # MQSO-S). Two of these sensors were placed 150 cm above the ground and their results were averaged and used as site-wide PPFD values.

Automated chambers for CO2 flux

Ten automated CO2 flux chambers (one per plot) were used to make hourly measurements of NSE in control and warmed plots from January 1, 2006–September 20, 2007. The chambers followed the design of Riggs et al. (2009) and Bowling et al. (2011). In the Bowling et al. (2011) paper, multiple methods were used to measure CO2 and showed consistent results with the automated chambers. During chamber measurements, transparent flux chamber lids were pneumatically lowered onto PVC soil collars once per hour. The collars were 30 cm tall × 38 cm inner diameter and contained a soil surface area of 0.11 m2. Collars were installed to a depth of 27 cm in the soil, leaving ~3 cm of the chamber protruding above the soil surface. The collars were placed in plot locations containing biocrusts but no vascular plants. Chambers were controlled by two centralized systems each equipped with a Campbell CR10X data logger and a PP Systems WMA-4 infrared gas analyzer. For each hourly measurement, the transparent chamber lid was closed for 3 min. The chamber caused a slight reduction in photosynthetic photon flux density (10–20 %) and a slight increase in temperature (~1 °C) during the 3 min closed-chamber interval (Bowling et al. 2011). The flow rate through the system is ~4L/min with ~0.5 L/min going through the infrared gas analyzer (IRGA). The air that is split off to run through the IRGA is then returned to the main path to go back out to the chamber, creating a closed-path system in which CO2 builds up over the course of the measurement. CO2 concentration, temperature, pressure and water vapor of the gas stream were measured at 2 s intervals with 10 s averages recorded by the data loggers. CO2 fluxes including corrections for dilution and band broadening due to water vapor were calculated in post processing using Matlab (Mathworks, Inc.). Negative values indicate net photosynthesis (C gain to the ecosystem from the atmosphere; i.e., when biocrust soils were a sink for C) and positive values indicate net respiration (C loss from the ecosystem to the atmosphere; i.e., when biocrust soils were a source of C).

Statistical analysis

Our NSE data set consisted of data from five control chambers and five warmed chambers collecting hourly measurements for a total of 15,048 h (January 1, 2006–September 20, 2007). Analyses were done using three different versions of these data, with the calculations for each version described in Appendix 1 in supplementary material. These data sets were: (1) fully imputed data sets (control and warmed) used to estimate annual and seasonal NSE values that included values for all 15,048 h in the sampling period; (2) a partially imputed data set to examine correlations in which time periods when less than two of the five treatment chambers were working were excluded (87 % of hourly values remained); and (3) a daily sums data set in which we summed values from the fully imputed data set, but then eliminated days during which more than 4 h of data were missing (82 % of daily values remained). See Appendix 1 in supplementary material for all details related to these data sets. For seasonal analyses, we divided our measurement period of January 1, 2006 through September 20, 2007 into seven quarterly intervals with cutoff dates of April 1, July 1, and October 1. The final interval, summer 2007, was ten days short of being a full quarter, while the others were each about 91 days long. For annual analyses, we focused on 2006, as it was our first complete year of data. We also distinguished time periods in which we observed active photosynthesis. We defined these periods as days during which the minimum NSE value was < −0.2 µmol CO2 m−2 s−1.

To assess the predictability of NSE based on easily measureable environmental variables, we fit a random forest model of mean NSE from the five chambers using air temperature, soil moisture at 5 cm, photosynthetically active radiation (PAR), and time of day as predictors and examined the correlation coefficient, denoted r2, and calculated as 1—m.s.e./var(y), where m.s.e. is the mean squared error of the model-predicted values and var(y) is the variance of the response variable (Breiman 2001). Treatment differences were assessed on cumulative values from a given time interval (i.e., annual, seasonal, or active photosynthesis days) using unpaired t tests with no assumption of equal variance (Welch t test; Ruxton 2006). Results of t tests are reported as mean differences with the associated confidence intervals. Percent differences between control and treatment plots were calculated as [(treatment − control)/control] with 95 % confidence intervals propagated from treatment and control means using root sum-of-squares. To examine relationships between NSE and environmental variables, we fit local regression curves (loess, with span = 0.5; Cleveland 1993) to the relationships between environmental variables and subsets of hourly and daily NSE measurements (e.g., NSE vs. air temperature in wet daytime conditions). The local regressions were well-suited to illuminate the nonlinear relationships among these variables without imposing a model structure on the data. To allow for reproducibility and transparency of these analyses, the raw data and analysis code (all analyses were done in R 3.0.3; R Core Team 2014) for all figures and calculations are included as supplementary files.



In 2006, the site received slightly above-average (109 %) precipitation of 294 mm. The first nine months in 2007 had below-average precipitation of 146 mm, which is 81 % of the long-term average for those months. Daily high temperatures in 2006 and 2007 were slightly hotter than the long term mean (20.3 °C and 21.6 °C vs. long term mean of 19.8 °C), with 2007 hotter than 2006.

Temporal trends in NSE in control plots

Hourly measurements of soil NSE in the five control chambers ranged from −0.94 to 1.97 µmol CO2 m−2 s−1, with a median value of 0.12 µmol CO2 m−2 s−1. After integrating these hourly spot measurements to represent one hour of each day and imputing missing values, our data showed that over the first complete year (2006), the flux of C from soils to the atmosphere was 62 ± 8 g C m−2 (mean ± SE; n = 5). The size of this flux varied by season (Fig. 1; Table 1), with the highest fluxes observed in spring 2006 (30 ± 4 g C m−2) and spring 2007 (28 ± 4 g C m−2). The second highest fluxes were seen in summer 2006 (17 ± 2 g C m−2). Substantially lower fluxes were seen in fall 2006 (7 ± 1 g C m−2), winter 2006 (7 ± 1 g C m−2) and winter 2007 (8 ± 1 g C m−2).
Fig. 1

Daily average net soil exchange (NSE) of CO2-C in control plots from the daily sums data set over the 21-month measurement period of the study. Positive values indicate net loss of C from soils. Red curve is a local regression (loess) with span = 0.2. (Color figure online)

Table 1

Treatment averages for control and warmed plots and treatment differences between control and warmed plots expressed as absolute magnitude and as percentage difference

Measurement period

Control (mean ± SE) (g C m−2)

Warmed (mean ± SE) (g C m−2)

Difference from control (mean ± 95 %CI) (g C m−2)

Difference from control (mean ± 95 %CI) (%)

Winter 2006

7.5 ± 1.3

9.7 ± 1.7

2.2 ± 5.1

29 ± 89

Spring 2006

30 ± 4

32.6 ± 3.4

2.6 ± 12

9 ± 50

Summer 2006

17.4 ± 1.9

19.8 ± 2.5

2.4 ± 7.2

14 ± 52

Fall 2006

6.7 ± 1.1

11.4 ± 2.1

4.7 ± 5.7

69 ± 115

2006 total

61.6 ± 8

73.5 ± 8.5

11.9 ± 26.9

19 ± 57

Winter 2007

7.7 ± 1.4

11.9 ± 2.1

4.2 ± 6

55 ± 110

Spring 2007

27.8 ± 3.8

30.3 ± 3.8

2.5 ± 12.4

9 ± 56

Summer 2007

12.2 ± 0.9

14 ± 1.3

1.8 ± 3.7

15 ± 38

Winter 2006 (day)

5.4 ± 0.8

7.2 ± 0.8

1.8 ± 2.6

34 ± 67

Spring 2006 (day)

21.7 ± 2.5

23.7 ± 1.7

2 ± 7.1

9 ± 41

Summer 2006 (day)

14.4 ± 1.3

15.5 ± 1

1.1 ± 3.8

7 ± 33

Fall 2006 (day)

1.9 ± 0.7

4.8 ± 0.6

2.9 ± 2.1

157 ± 270

2006 (day) total

43.4 ± 4.8

51.2 ± 3.5

7.8 ± 13.9

18 ± 42

Winter 2007 (day)

3.8 ± 0.8

6.9 ± 0.9

3.1 ± 2.8

82 ± 124

Spring 2007 (day)

18.7 ± 2.3

20.1 ± 2.1

1.5 ± 7.2

8 ± 48

Summer 2007 (day)

8.7 ± 0.5

10 ± 0.6

1.3 ± 1.7

14 ± 25

Winter 2006 (night)

2.1 ± 0.6

2.5 ± 1

0.4 ± 2.7

18 ± 158

Spring 2006 (night)

8.3 ± 1.5

8.9 ± 1.7

0.6 ± 5.3

8 ± 80

Summer 2006 (night)

3 ± 0.8

4.4 ± 1.6

1.4 ± 4.3

45 ± 178

Fall 2006 (night)

4.9 ± 0.9

6.6 ± 1.5

1.7 ± 4.2

36 ± 108

2006 (night) total

18.3 ± 3.5

22.4 ± 5.2

4.1 ± 14.8

23 ± 102

Winter 2007 (night)

3.9 ± 0.7

5 ± 1.2

1.1 ± 3.3

29 ± 107

Spring 2007 (night)

9.2 ± 1.6

10.1 ± 1.7

1 ± 5.4

11 ± 74

Summer 2007 (night)

3.5 ± 0.5

4 ± 0.8

0.5 ± 2.2

15 ± 76

Net photosynthesis

3 ± 0.9

7.3 ± 1

4.3 ± 3.1

142 ± 218

21-month total

109.3 ± 13.7

129.7 ± 15.1

20.4 ± 47

19 ± 56

Biocrust soils in the chambers were almost always net C sources to the atmosphere, but when soils were wet, we observed crust autotrophs generating enough photosynthesis to make the biocrust soils a net C sink (Fig. 2; blue bars indicate rainfall events). Of the daily summed NSE values, 43 ± 4 days (mean ± SE; n = 5) out of the 627-day measurement period (7 ± 1 %) were less than zero, indicating net photosynthesis on those days. On an additional 33 ± 7 days (5 ± 1 %), there was substantial net photosynthesis during daylight hours; however, it was counterbalanced by higher respiration at night, resulting in net C loss for the day (Fig. 2). Times during which net photosynthesis was observed were always associated with precipitation events, including measurable photosynthesis occurring under melting snow in winter at freezing or near-freezing air temperatures. For example, substantial subnivean photosynthesis was observed during the snow event on March 8–12, 2006, with NSE rates as low as −0.6 µmol C m−2 (Figs. 1, 3a; and see Figure S2 for an image of the snow cover at the site during that event). Not all precipitation events were associated with C uptake by the biocrust soils though: when air temperatures were high during summer months, precipitation events such as the one on June 6, 2006 resulted in some of the highest C losses, with few to no observations of net C gain.
Fig. 2

Hourly measurements of NSE of CO2 in biocrust soils throughout the 21-month measurement period, divided into quarter years and labeled by season. Values shown are the mean of the five chambers in control plots. Positive values indicate net loss of C from soils. Points showing individual hourly values are colored to indicate whether or not the means included imputed values. Black points are calculated from no imputed values, purple points are calculated using one to three imputed values of the five. These values are well constrained due to strong correlations among the five chambers (Appendix 1 in supplementary material). The red points are calculated using four to five imputed values and are the least certain. Most of red points are from times when power was lost at the site and all chambers were inoperative. Days on which substantial net photosynthesis (defined as minimum NSE < −0.2 µmol m−2 s−1) was measured are shaded in green: dark green indicates days on which the daily total NSE was below 0 while light green indicates net photosynthesis values were observed, but there was still net C loss for the day. Hourly air temperatures (some of which are imputed, typically for the same days on which NSE was imputed) are colored by temperature for easier visual comparison of different temperature ranges among panels. Sums of daily precipitation quantity are shown as blue bars. A snow event on March 8–12, 2006 during which substantial subnivean photosynthesis was observed is marked with a blue asterisk. See Figure S2 for an image of the snow cover during this event. Cumulative NSE (mean ± SE) for each season is shown in the upper right of each panel. (Color figure online)

Fig. 3

Relationships between hourly NSE values (means of 5 control plots) and environmental conditions. Positive values indicate net loss of CO2 from soils. Red lines are local regressions (loess, span = 0.5). Several points that were known to be peaks in subnivean photosynthesis are circled in green in a. An exponential curve is fitted in blue on b for the temperature range of 5–25 °C. The average times of peaks in net photosynthesis and net respiration are shown in h. The cutoff between wet soil and dry soil is 5 % VWC and the cutoff between daylight and dark is PAR = 72 µmol m−2. (Color figure online)

The random forest model of mean NSE from the five chambers using air temperature, soil moisture at 5 cm, PAR, and time of day as predictors explained 79 % of the variation. This strong predictability was based on clear patterns in the relationships between NSE and environmental conditions. In high soil moisture conditions, we observed positive relationships between air temperature and NSE during both daylight (PAR > 72 µmol m−2) and dark conditions (Fig. 3a, b). This same relationship was much less prevalent when soils were dry (Fig. 3c, d). The Q10 for nighttime NSE values in wet soils between 5 °C and 25 °C in dark conditions was 2.4. During daylight hours, a negative relationship was observed between NSE and moisture, driven by photosynthesis when soil moisture was sufficient, a relationship that was absent at night (Fig. 3e, f). Across all hourly measurements, PAR was positively correlated with NSE, with few net photosynthetic values under very high light conditions (Fig. 3g). The correlations between NSE and environmental variables were manifest as daily cycles with peak activity midday (Fig. 3h). On days in which net photosynthesis was observed, the peak negative flux occurred at 10:59 MST. On days in which net respiration dominated, the average peak positive flux was at 13:32 MST (Fig. 3h).

In the spring, there was a hump-shaped relationship between daily NSE values and air temperature, with maximal C loss when average daily highs were around 30 °C (Fig. 4). In the summer, we did not observe a similar relationship, despite similarly dry conditions during both measurement seasons. Moisture events in the summer were associated with “puffs” of CO2 loss, likely due to the duration of soil wetting being sufficient for abiotic loss and respiration of C, but not for the biocrust organisms to reach their net compensation point for C uptake. This contrasted with more prolonged wet periods in the spring, which resulted in a reduction in NSE due to substantial photosynthetic activity. In winter and fall, NSE and air temperature were positively correlated, though in winter, sub-freezing temperatures were associated with slightly higher NSE values than those just above freezing, whereas negative values at some times of day and from photosynthesis in some conditions pushed NSE values downwards.
Fig. 4

Correlations between summed daily NSE values (means of 5 control plots) and maximum daily air temperatures. Positive values indicate net loss of C from soils. Colors indicate moisture level (VWC volumetric water content). Red lines are local regressions (loess). There are fewer points in the fall quadrant because we only had data from one fall period, whereas two years of data were available for all of the other seasons. (Color figure online)

Effect of warming on net soil exchange of CO2

When values from the control and warmed chambers were compared, variation was high and there was substantial overlap between the treatments (Fig. 5). The annual summed NSE during calendar year 2006 was 61.6 ± 8.0 (mean ± SE) mg C m−2 in the control plots and 73.5 ± 8.5 mg C m−2 in the warmed plots, with the warmed plots never showing lower NSE relative to controls (Table 1). The difference between the treatments was 11.9 ± 26.9 mg C m−2. Whereas more replicates would be needed to constrain the size of the treatment effect to a tighter interval, these data show that warming is more likely than not to cause increased C loss under the relatively small degree of warming applied here.
Fig. 5

Cumulative NSE for 5 control chambers (black line) and 5 warmed chambers (dotted red line). Mean differences between treatments from Table 1 are shown for the first full measurement year (2006) and the entire measurement period. (Color figure online)

Analyses of the effects of the warming treatment on soil NSE during particular seasons and subsets of conditions showed statistically significant increases in NSE by warming treatments in several cases. During daylight hours in fall 2006, the +2 °C warming treatment increased NSE from 1.9 ± 0.7 g C m−2 (mean ± SE) in control plots to 4.8 ± 0.6 g m−2 in warmed plots, a difference of 2.9 ± 2.1 g m−2 (mean ± 95 %CI). Similarly in winter 2007, NSE increased from 3.8 ± 0.8 g m−2 to 6.9 ± 0.9 g m−2, a difference of 3.1 ± 2.8 g m−2 (Fig. 6; Table 1). We saw a hint of a similar effect during the first winter of the experiment, 2006, but it was not as pronounced as it was during the second winter (NSE in warmed chambers was 1.8 ± 2.6 higher than the control value of 5.4 ± 0.8 g C m−2).
Fig. 6

Treatment differences in cumulative NSE between control and warmed plots. Confidence intervals from Welch t tests are shown for various data subsets defined by year, season, and day/night. Four of these subsets are shown in detail, with points indicating individual chambers, colored by treatment, and mean ± SE shown as black lines and gray bars. Subsets were chosen to show statistically significant differences and, in one case (summer 2006; top), a relatively well constrained lack of difference between treatments. (Color figure online)

Over the course of the experiment, we also observed an increase in NSE on the 12 ± 1 % of measurement days on which substantial net photosynthesis values (NSE < −0.2 µmol m−2 s−1) were observed in the control plots. Warmed chambers showed an increase C loss of 4.3 ± 3.1 g C m−2 over the control chamber values of 3.0 ± 0.9 g C m−2 (Fig. 6). There were also 25 ± 22 % fewer days on which substantial net photosynthesis occurred in the warmed plots compared to the controls. Most of the rest of the seasonal and environmental condition subsets we examined were poorly constrained due to high variation among replicate chambers in the most active spring and summer months. However, we did see a relatively well-constrained lack of difference during summer 2006 days coinciding with a time during which our heating treatment was less effective (Figure S2). During that time, the mean difference between treatments was 1.1 ± 3.8 g C m−2, with control values being 14.4 ± 1.3.


Net CO2 loss from soils including biocrusts

Biocrust soils in our control plots were net CO2 sources to the atmosphere over the course of the study period, losing 62 ± 8 g C m−2 during the first measurement year (2006; equivalent to 0.17 ± 0.02 g C m−2 day−1). Although other dryland CO2 studies have suggested overall C losses for dryland soils, with a similar range of hourly flux magnitudes (Bowling et al. 2011; Maestre et al. 2013), this estimate of annual NSE is the first of its kind for biocrust soils. The hourly flux magnitudes are comparable to magnitudes recorded using a similar autochamber system from two nearby sites (Corral Pocket and Squaw Flat; Bowling et al. 2011). At those sites, the highest NEE value was 2 µmol m−2 s−1, the same as at our site. However, the lowest values at the nearby sites were −0.5 µmol m−2 s−1 indicating the crusts at our site had higher photosynthetic capacity. This finding is expected as the crusts at the Bowling et al. (2011) site were highly degraded due to heavy grazing (Bowling et al. 2011; Zaady et al. 2000). Studies of respiration or NSE in other semiarid ecosystems show hourly flux values in a similar range to our net respiration values. For example, a study in a similar ecosystem near ours showed maximal respiration fluxes in the month of March that were ~2 µmol m−2 s−1 and measurements of NEE in crust-dominated soil in Spain showed fluxes ranging from −0.5 to 1.2 µmol m−2 s−1 (Fernandez et al. 2006; Maestre et al. 2013).

The one eddy flux tower study of NEE (i.e., including vascular plants) done in a similar ecosystem on the Colorado Plateau did not report an annually summed value (Bowling et al. 2010), but we can infer that annual NEE in that study was closer to zero than our NSE value of 62 ± 8 g C m−2. A magnitude closer to zero in that study is likely because they observed periods of substantial net vascular plant CO2 uptake during the spring, which offset respiratory losses during other parts of the year (Figure S3). Reported NEE values in other drylands vary substantially, from overall atmospheric C sources on the order of 150 g C m−2 year−1 (Emmerich 2003; Mielnick et al. 2005) to overall C sinks on the order of 100 g C m−2 year−1 (Serrano-Ortiz et al. 2012; Wohlfahrt et al. 2008). Due to these varying results, it is difficult to place our NSE rates into the context of NEE, but our data do indicate that biocrust soils (which include soil crusts, heterotrophs, and plant roots) are generally a source of CO2 to the atmosphere, similar to what would be observed with a dark-chamber respiration system (e.g., Billings et al. 2004). This does not mean the crust organisms themselves are operating at a net C loss, as we cannot separate crust fluxes from the soil column on which they live. It does however suggest that, for the years we assessed, biocrust photosynthesis does not outpace the multiple sources of respiration (i.e., biocrusts, below-crust heterotrophs, and plant roots).

There are several possible sources for the overall net loss of C that we observed: (1) biocrust organisms may have contributed to the losses, indicating a net decline in biocrust C storage during the measurement period; (2) soil heterotrophs may have contributed to the losses, indicating a net decline in soil organic C (SOC) in these soils; (3) CO2 may have been abiotically lost from pedogenic carbonates (Emmerich 2003); and (4) CO2 fixed by adjacent vascular plants may have been lost beneath our soil chambers via (a) root respiration (Wang and Guo 2006) or (b) microbial respiration of C compounds from rhizodeposition (Pendall et al. 2004). Due to the magnitude of the net C losses, while biocrust organisms may have contributed, they are not likely to have been solely responsible for the losses. The partitioning of the remaining soil CO2 efflux is uncertain, with at least some plant and microbial contributions likely. Though the chambers were not placed underneath plant canopies, there is substantial plant cover near the chambers and shrub roots may well be deeper than the chambers (27 cm), making plant contributions plausible. In contrast, we consider pedogenic carbonates to be an unlikely source of CO2 because there is little evidence that they efflux CO2 even when exposed to the atmosphere (Serna-Pérez et al. 2006), and the patterns of CO2 losses from our soils are not consistent with dissolution of CaCO3 during wetting events. Our results do show that partitioning the sources of these fluxes is an important future direction in investigations of dryland C cycling. To aid in considering these future efforts and to further clarify the potential sources of CO2, we have included a more detailed discussion of the role of each of these processes in the supplementary materials (Appendix 2 in supplementary material).

Associations between NSE of biocrust soils and environmental conditions

The overall correlative structure of our data suggests that simple models can be effective for predicting NSE in biocrust soils. Our random forest model showed that 79 % of the variation in the mean hourly flux from our five control plot autochambers could be predicted with the easily measureable inputs of PAR, soil moisture, soil temperature, and time of day. While extensive modeling of the data to find parsimonious non-ensemble-tree relationships among these variables is beyond the scope of this study, the strong predictive capability of the random forest model used here has important implications (Sandri and Zuccolotto 2006). For example, the data suggest that these few abiotic variables are the dominant controls over CO2 fluxes from these biocrust soils, even over relatively fine timescales (i.e., hourly). Such strong relationships may provide a way forward in including these fluxes in ecosystem- and global-scale modeling work. Although biocrust soils account for a large fraction of dryland CO2 flux due to their abundance, they are currently not explicitly represented in any Earth System Model. Relationships such as those found here illustrate that parameters for biocrust soils may be relatively straightforward to include in future modeling efforts. That said, additional work is needed to understand the substantial variability among the chambers, as well as to determine how similar or different such relationships would be at different sites. Interestingly, visible crust cover (i.e., proportional cover of lichen, moss, and cyanobacteria) was not a strong covariate with net CO2 flux (data not shown). We suspect that belowground features of the soil column—such as depth to bedrock or caliche—that influences soil water, nutrient, and root distribution (Fernandez et al. 2006) and deeper root biomass (Cable et al. 2011), could help to explain the spatial heterogeneity, as could more detailed information about the at- and below-ground composition of soil organisms in each individual chamber (Steven et al. 2015). Below we describe some of the key relationships observed between environmental conditions and net fluxes of CO2.

When air temperatures were below freezing, NSE values were typically close to zero, likely due to the cold conditions slowing all biological activity and reducing availability of liquid water. However, during several snow events, we observed substantial net photosynthesis rates under snow, a phenomenon that has been documented for lichens in Antarctica and at European alpine sites (Kappen 1993) and arctic plants in Alaska (Starr and Oberbauer 2003), though to our knowledge has not been shown for any biocrust in a dryland ecosystem. Snow in cool and cold desert settings is a common occurrence and the ability of biocrusts to photosynthesize under such conditions suggests a high degree of adaptation to the environments of these globally extensive regions. As climates warm and winter precipitation transitions from snow to rain (Garfin et al. 2014), this tight coupling between biocrusts and their environment could prove less beneficial to those biocrust organisms that are adapted to both cold and hot conditions.

At volumetric soil water contents of >10 %, nearly all measured values during daylight showed net photosynthetic fluxes for the soil columns, suggesting that very wet (for drylands) soils are optimal for biocrust C gain. Increasing soil water content generally correlated with NSE becoming increasingly negative, likely due to photosynthesis becoming more active with wetter conditions. We observed peaks of net photosynthesis occurring at ~10:30 MST on the 12 % of days during which substantial photosynthetic rates were apparent. Laboratory measurements of C exchange with isolated lichens, including C. tenax, the dominant lichen in our plots, similarly show optimal photosynthesis and C gain at high water contents of ~80 % of their dry mass (Lange et al. 1998; Lange et al. 1997). Spot measurements of daytime soil respiration from a nearby site showed that respiration rates (i.e., gross CO2 loss) were highest when soils had >10 % moisture and were between 10 and 16 °C (Fernandez et al. 2006). These data are consistent with our data in suggesting that wet biocrust soils with >10 % moisture have both high photosynthesis and high respiration, with a net gain of C overall.

Nighttime CO2 efflux values showed three notable patterns. First, when soil volumetric water contents were >5 %, we saw an exponential relationship between efflux and air temperature. The positive relationship in nighttime values was describable with an exponential equation having a Q10 of 2.4 for temperatures between 5° and 25 °C, a parameter estimate well within the expected range for soil respiration (Fierer et al. 2006; Xu and Qi 2001). Second, we observed reduced rates of CO2 loss when soils were hot (>25 °C) and dry (<5 % VWC) (see Fig. 3d). The lower respiration rates at higher temperatures may have been due to either substrate limitation at high temperatures (Davidson et al. 2006) or suppression of microbial activity (possibly including microbial death) in hot and dry conditions (Cable et al. 2011; Parker et al. 1983). Finally, there was no clear relationship between nighttime NSE and soil moisture, and even very wet (>10 % VWC) soils respired at low levels (~0.2 µmol m−2 s−1) during the night (see Fig. 3f). This suggests that even when there is substantial C gain during the day, there is not a larger proportion of C lost at night. This pattern was unexpected, as moisture, in addition to temperature, should be a strong control on respiration rates (Fernandez et al. 2006). However, the strong temperature dependence of C flux we found may provide a clue as to why wet soils are not respiring more at night. Colorado Plateau ecosystems have large diurnal temperature fluctuations, and organisms adapted to function at high temperatures during the day may greatly down-regulate respiratory activity when temperatures drop at night, regardless of moisture conditions. Diurnal measurements of respiration in a study of Chihuahuan desert soils similarly showed greatly reduced respiration and lack of moisture sensitivity in nighttime measurements, providing some support for this hypothesis (Parker et al. 1983).

Carbon losses in dry soils peaked several hours later than photosynthesis (~13:30 MST), which is consistent with a higher temperature optimum for respiration (Grote et al. 2010). However, the timing of CO2 production and CO2 efflux from the soil may not necessarily coincide. Other dryland studies have shown that daily cycles in CO2 transport fluxes are driven by physical phenomena such as thermal expansion of soils and soil air (Hamerlynck et al. 2013; Parsons et al. 2004). The daily trends we observed, particularly the increasing CO2 loss with increasing temperatures, suggest similar mechanisms could be influencing the hourly rates at our site. While this physical effect on CO2 could not explain the net C losses observed (i.e., the flux must have a source), physical phenomena could play a role in the observed relationships between temperature and the rate of CO2 efflux.

During the day when soils were drier (<10 % VWC) and air temperatures were above freezing, a positive relationship was observed between NSE and temperature (Fig. 3a). Interestingly, no net photosynthesis was observed at temperatures >25 °C. Laboratory studies using well-developed biocrusts from a nearby site showed that higher temperatures increased both gross photosynthesis and dark respiration of biocrust organisms up to 25 °C for intermediately wet soils, at which point photosynthetic rates began to decline while respiration continued to climb (Grote et al. 2010). These relationships may help to explain why small summer precipitation events appeared to cause “puffs” of CO2 and net C loss by biocrust soils. These natural events that we observed in the control plots are consistent with the mechanism of moss death described in Reed et al. (2012) in which mosses repeatedly experienced net C loss when subjected to small artificial precipitation events. This relationship suggests that, in the absence of acclimation, a warmer future climate could restrict the number of days in which biocrusts can gain C, and that any increase in small precipitation events during hot conditions would be detrimental to biocrust organisms.

Effect of 2 °C warming on NSE

Overall, our data suggest that 2 °C warming at our study site drove NSE values upwards, thus causing more overall C loss from biocrust soils, especially during daylight hours in winter and fall seasons. However, the high variation among individual chambers observed during spring, the season with the most active C fluxes, precludes conclusive evidence of an annual effect. Summer showed the least warming effect, which is consistent with the reduced effectiveness of the IR lamps when air temperatures were >30 °C (Figure S1). While higher replication would be necessary to better evaluate the year-long difference in NSE between treatments, the 19 ± 57 % difference in means and the clear differences during winter and fall, driven by differences on photosynthetically active days, suggest that 2 °C warming does lead to higher C losses from these soils, at least at certain times of year. This type of effect may help to explain why higher aridity is associated with reduced soil C storage in drylands on a global scale (Delgado-Baquerizo et al. 2013). It is also consistent with higher net C losses observed during point measurements for a warming experiment in another biocrust-dominated arid ecosystem (Maestre et al. 2013) and with higher soil respiration rates found with warming in many other non-arid ecosystems (Rustad et al. 2001).

We observed the most obvious effects of warming during daylight hours in fall 2006 and winter 2007. During those times, NSE was substantially higher (higher CO2 loss from soils) in warmed plots than in control plots (Fig. 6). These seasonal effects were associated with a large number of days during which soils were wet and biocrusts were photosynthesizing, suggesting that increased CO2 loss in warmed plots in NSE was driven by increased respiration that was not fully compensated by increasing crust photosynthesis. Increased respiration at this time was also likely driven by the biocrust organisms, because during these periods biocrusts are active and plants are less active or dormant. Below a temperature threshold, biocrust organisms generally increase both their photosynthetic and respiration rates as temperature increases. A biocrust examined near the study site showed a peak in net C uptake around 15–20 °C (Grote et al. 2010). However, we observed increased CO2 loss in the warming treatment compared to the control during fall and winter when temperatures were on the cooler side of the optimal uptake curve, suggesting that warming did not increase photosynthesis more than respiration as the laboratory patterns would suggest. Indeed, if this were the case, we would see a trend opposite to our data: that is, we would have seen greater C uptake in warmed plots and not the larger CO2 losses we observed. One possible influence is snow, which melted earlier in the warmed plots. This may have led to increased respiration relative to photosynthetic activity in the biocrusts. More detailed observations of the timing of snowfall and melt would be necessary to provide further evidence for this possibility. In addition to crust respiration, heterotrophic respiration is also temperature-sensitive when soils are wet; thus, their respiration likely contributed to the difference between control and warmed plots. CO2 losses from carbonates might also be expected to be highest during these wet periods, and temperature sensitivities of this process could also contribute to increased C loss (Mielnick et al. 2005). Overall we cannot totally partition the CO2 sources of this trend, and more information on the effect of warming on each gross flux would better identify the processes driving increased CO2 loss during periods when crusts are most active. Regardless of mechanism, increased CO2 losses with warming indicate a vulnerability of these biocrust organisms to future warming.

Our results are interesting in the context of another global change experiment in a semiarid ecosystem. At a Wyoming rangeland site with greater rainfall (384 mm annual precipitation) and much greater plant biomass, monthly NEE chamber measurements (which included vascular plants and soils) in a warming × elevated CO2 experiment suggested the opposite effect of warming on NEE than we observed in the NSE of biocrust soils during photosynthetically active periods. Namely, warming reduced ecosystem C losses (Pendall et al. 2013). This difference may have been driven by differences in moisture responses between the two experiments, with more substantial warming-induced reductions in soil moisture at the Wyoming site. It could also be related to the Wyoming measurements including vascular plants and thus incorporating the effects on the main producers in the system (i.e., vascular plants). Nevertheless, the differences between this study and ours suggest that climate change drivers and their interactions may affect specific dryland biota and ecosystems idiosyncratically. Therefore, future CO2 and warming increases may have variable effects on C flux in drylands, with the possibility that C losses in some regions are offset by gains in others.


We have provided the most comprehensive look at CO2 exchange between biocrust soils and the atmosphere to date. Our results showed that biocrust soils were a net source of CO2 to the atmosphere >90 % of the time, but multi-day moisture events allowed for substantial net photosynthesis, even beneath snow cover in the winter. While we saw substantial spatial variation in NSE among chambers, the daily and seasonal patterns we observed were well explained by simple environmental variables, indicating that incorporation of biocrust soils into terrestrial ecosystem models is tractable. Finally, we saw clear evidence of increased C loss in the 2 °C warming treatment during times when biocrust soils were wet and actively photosynthesizing. This led to greater C losses during fall 2006 and winter 2007. However, the annual response of NSE to our warming treatment remains uncertain due to high chamber to chamber variation in NSE during the spring, which may in part be controlled by other CO2-producing components of the soil column such as plant respiration. As such, better partitioning of C losses among biocrust organisms, sub-crust microbes, plant roots, and abiotic sources is a clear research priority for understanding the dryland C cycle. Taken together, our data suggest that future warming may negatively affect biocrust organisms and thus impact the ecosystem services they provide and the C balance of dryland ecosystems.



This material is based upon work supported by the U.S. Department of Energy Office of Science, Office of Biological and Environmental Research Terrestrial Ecosystem Sciences Program, under Award Number DE-SC-0008168, as well as US Geological Survey’s Climate and Land Use and Ecosystems Mission Areas. Thanks are given to the many technicians who have and are currently working on this project, Sue Phillips and Dave Housman who directed the early setup of the site, and several anonymous reviewers whose contributions have improved the manuscript. Any trade, product, or firm name is used for descriptive purposes only and does not imply endorsement by the U.S. Government.

Supplementary material

10533_2015_163_MOESM1_ESM.docx (1.5 mb)
Supplementary material 1 (DOCX 1564 kb) (24 mb)
Supplementary material 2 (ZIP 24,618 kb)


  1. Belnap J (1995) Surface disturbances: their role in accelerating desertification. Environ Monit Assess 37(1–3):39–57CrossRefGoogle Scholar
  2. Belnap J, Lange OL (eds) (2003) Biological soil crusts: structure, function, and management. Springer-Verlag, BerlinGoogle Scholar
  3. Beymer RJ, Klopatek JM (1991) Potential contribution of carbon by microphytic crusts in pinyon-juniper woodlands. Arid Soil Res Rehabil 5:187–198Google Scholar
  4. Billings SA, Schaeffer SM, Evans RD (2004) Soil microbial activity and N availability with elevated CO2 in Mojave Desert soils. Global Biogeochem Cy 18(1):GB1011CrossRefGoogle Scholar
  5. Bowling DR, Bethers-Marchetti S, Lunch CK, Grote EE, Belnap J (2010) Carbon, water, and energy fluxes in a semiarid cold desert grassland during and following multiyear drought. J Geophys Res 115(Go 4026):16Google Scholar
  6. Bowling DR, Grote EE, Belnap J (2011) Rain pulse response of soil CO2 exchange by biological soil crusts and grasslands of the semiarid Colorado Plateau, United States. J Geophys Res 116(G3):G03028Google Scholar
  7. Breiman L (2001) Random forests. Mach Learn 45(1):5–32CrossRefGoogle Scholar
  8. Cable J, Ogle K, Lucas R, Huxman T, Loik M, Smith S, Tissue D, Ewers B, Pendall E, Welker J, Charlet T, Cleary M, Griffith A, Nowak R, Rogers M, Steltzer H, Sullivan P, Gestel N (2011) The temperature responses of soil respiration in deserts: a seven desert synthesis. Biogeochemistry 103(1–3):71–90CrossRefGoogle Scholar
  9. Cleveland WS (1993) Visualizing data. AT and T Bell Laboratories; Published by Hobart Press, Murray HillGoogle Scholar
  10. Coe KK, Belnap J, Sparks JP (2012) Precipitation-driven carbon balance controls survivorship of desert biocrust mosses. Ecology 93(7):1626–1636CrossRefGoogle Scholar
  11. Collins SL, Sinsabaugh RL, Crenshaw C, Green L, Porras-Alfaro A, Stursova M, Zeglin LH (2008) Pulse dynamics and microbial processes in aridland ecosystems. J Ecol 96(3):413–420CrossRefGoogle Scholar
  12. Davidson EA, Janssens IA, Luo Y (2006) On the variability of respiration in terrestrial ecosystems: moving beyond Q10. Global Change Biol 12(2):154–164CrossRefGoogle Scholar
  13. Delgado-Baquerizo M, Maestre FT, Gallardo A, Bowker MA, Wallenstein MD, Quero JL, Ochoa V, Gozalo B, Garcia-Gomez M, Soliveres S, Garcia-Palacios P, Berdugo M, Valencia E, Escolar C, Arredondo T, Barraza-Zepeda C, Bran D, Carreira JA, Chaieb M, Conceicao AA, Derak M, Eldridge DJ, Escudero A, Espinosa CI, Gaitan J, Gatica MG, Gomez-Gonzalez S, Guzman E, Gutierrez JR, Florentino A, Hepper E, Hernandez RM, Huber-Sannwald E, Jankju M, Liu J, Mau RL, Miriti M, Monerris J, Naseri K, Noumi Z, Polo V, Prina A, Pucheta E, Ramirez E, Ramirez-Collantes DA, Romao R, Tighe M, Torres D, Torres-Diaz C, Ungar ED, Val J, Wamiti W, Wang D, Zaady E (2013) Decoupling of soil nutrient cycles as a function of aridity in global drylands. Nature 502(7473):672–676CrossRefGoogle Scholar
  14. Elbert W, Weber B, Burrows S, Steinkamp J, Budel B, Andreae MO, Poschl U (2012) Contribution of cryptogamic covers to the global cycles of carbon and nitrogen. Nat Geosci 5(7):459–462CrossRefGoogle Scholar
  15. Emmerich WE (2003) Carbon dioxide fluxes in a semiarid environment with high carbonate soils. Agric Forest Meteorol 116(1–2):91–102CrossRefGoogle Scholar
  16. Fernandez D, Neff J, Belnap J, Reynolds R (2006) Soil Respiration in the Cold Desert Environment of the Colorado Plateau (USA): abiotic regulators and thresholds. Biogeochemistry 78(3):247–265CrossRefGoogle Scholar
  17. Ferrenberg S, Reed SC, Belnap J (2015) Climate change and physical disturbance cause similar community shifts in biological soil crusts. Proc Natl Acad SciGoogle Scholar
  18. Fierer N, Colman BP, Schimel JP, Jackson RB (2006) Predicting the temperature dependence of microbial respiration in soil: A continental-scale analysis. Global Biogeochem Cy 20(3):GB3026CrossRefGoogle Scholar
  19. Garcia-Pichel F, Belnap J (2003) Small-scale environments and distribution of biological soil crusts. In: Belnap J, Lange O (eds) Biological soil crusts: structure, function, and management. Ecological Studies Series 150, Springer-Verlag, Berlin, pp 193–201Google Scholar
  20. Garfin G, G, Franco HB, Comrie A, Gonzalez P, Piechota T, Smyth R, Waskom R (2014) Ch. 20: Southwest. In: Melillo JM, Richmond TC and Yohe GW (eds) Climate change impacts in the United States: The Third National Climate Assessment. U.S. Global Change Research Program, pp 462–486Google Scholar
  21. Grote EE, Belnap J, Housman DC, Sparks JP (2010) Carbon exchange in biological soil crust communities under differential temperatures and soil water contents: implications for global change. Global Change Biol 16(10):2763–2774CrossRefGoogle Scholar
  22. Hamerlynck EP, Scott RL, Sánchez-Cañete EP, Barron-Gafford GA (2013) Nocturnal soil CO2 uptake and its relationship to subsurface soil and ecosystem carbon fluxes in a Chihuahuan Desert shrubland. J Geophys Res 118(4):2013JG002495CrossRefGoogle Scholar
  23. Harper KT, Marble JR (1988) A role for nonvascular plants in management of arid and semiarid rangelands. In: Tueller PT (ed) Vegetational science applications for rangeland analysis and management. Kluwer Academic Press, Dordrecht, pp 135–169CrossRefGoogle Scholar
  24. Harte J, Torn MS, Chang F-R, Feifarek B, Kinzig AP, Shaw R, Shen K (1995) Global warming and soil microclimate: results from a meadow-warming experiment. Ecol Appl 5:132–150CrossRefGoogle Scholar
  25. Kappen L (1993) Plant activity under snow and ice, with particular reference to lichens. Arctic 46:297–302CrossRefGoogle Scholar
  26. Kimball BA (2005) Theory and performance of an infrared heater for ecosystem warming. Global Change Biol 11(11):2041–2056Google Scholar
  27. Lal R (2003) Soil erosion and the global carbon budget. Environ Int 29(4):437–450CrossRefGoogle Scholar
  28. Lange OL (2003) Photosynthesis of soil-crust biota as dependent on environmental factors. In: Belnap J, Lange OL (eds) Biological soil crusts: structure, function, and management. Ecological studies series. Springer-Verlag, Berlin, pp 217–240Google Scholar
  29. Lange OL, Belnap J, Reichenberger H, Meyer A (1997) Photosynthesis of green algal soil crust lichens from arid lands in southern Utah, USA: role of water content on light and temperature responses of CO2 exchange. Flora 192:1–15Google Scholar
  30. Lange OL, Belnap J, Reichenberger H (1998) Photosynthesis of the cyanobacterial soil-crust lichen Collema tenax from arid lands in southern Utah, USA: role of water content on light and temperature responses of CO2 exchange. Funct Ecol 12(2):195–202CrossRefGoogle Scholar
  31. Maestre FT, Escolar C, De Guevara ML, Quero JL, Lazaro R, Delgado-Baquerizo M, Ochoa V, Berdugo M, Gozalo B, Gallardo A (2013) Changes in biocrust cover drive carbon cycle responses to climate change in drylands. Global Change Biol 19(12):3835–3847CrossRefGoogle Scholar
  32. Mielnick P, Dugas WA, Mitchell K, Havstad K (2005) Long-term measurements of CO2 flux and evapotranspiration in a Chihuahuan desert grassland. J Arid Environ 60(3):423–436CrossRefGoogle Scholar
  33. Parker LW, Miller J, Steinberger Y, Whitford WG (1983) Soil respiration in a chihuahuan desert rangeland. Soil Biol Biochem 15(3):303–309CrossRefGoogle Scholar
  34. Parsons AN, Barrett JE, Wall DH, Virginia RA (2004) Soil carbon dioxide flux in Antarctic dry valley ecosystems. Ecosystems 7(3):286–295CrossRefGoogle Scholar
  35. Pendall E, Mosier AR, Morgan JA (2004) Rhizodeposition stimulated by elevated CO2 in a semiarid grassland. New Phytol 162(2):447–458CrossRefGoogle Scholar
  36. Pendall E, Heisler-White JL, Williams DG, Dijkstra FA, Carrillo Y, Morgan JA, LeCain DR (2013) Warming reduces carbon losses from grassland exposed to elevated atmospheric carbon dioxide. PLoS ONE 8(8):e71921CrossRefGoogle Scholar
  37. Pointing SB, Belnap J (2012) Microbial colonization and controls in dryland systems. Nat Rev Microbiol 10(8):551–562CrossRefGoogle Scholar
  38. Poulter B, Frank D, Ciais P, Myneni RB, Andela N, Bi J, Broquet G, Canadell JG, Chevallier F, Liu YY, Running SW, Sitch S, van der Werf GR (2014) Contribution of semi-arid ecosystems to interannual variability of the global carbon cycle. Nature 509(7502):600–603CrossRefGoogle Scholar
  39. R Core Team (2014) R: A language for Statistical Computing. In. R Foundation for Statistical Computing, Vienna, AustriaGoogle Scholar
  40. Reed SC, Coe KK, Sparks JP, Housman DC, Zelikova TJ, Belnap J (2012) Changes to dryland rainfall result in rapid moss mortality and altered soil fertility. Nat Clim Change 2:752–755CrossRefGoogle Scholar
  41. Riggs AC, Stannard DI, Maestas FB, Karlinger MR, Striegl RG (2009) Soil CO2 flux in the Amargosa Desert, Nevada, during the El Niño and La Niña years 1998–1999. U.S. Geol Surv Sci Investig Rep 2009–5061:25Google Scholar
  42. Rustad L, Campbell J, Marion G, Norby R, Mitchell M, Hartley A, Cornelissen J, Gurevitch J, Gcte N (2001) A meta-analysis of the response of soil respiration, net nitrogen mineralization, and aboveground plant growth to experimental ecosystem warming. Oecologia 126(4):543–562CrossRefGoogle Scholar
  43. Ruxton GD (2006) The unequal variance t test is an underused alternative to Student’s t test and the Mann–Whitney U test. Behav Ecol 17(4):688–690CrossRefGoogle Scholar
  44. Safriel U, Adeel Z, Niemeijer D, Puigdefabregas J, White R, Lal R, Winslow M, Ziedler J, Prince S, Archer E, King C, Shapiro B, Wessels K, Nielsen T, Portnov B, Reshef I, Thonell J, Lachman E, McNab D (2005) Dryland systems. In: Mohammed El-Kassas EE (ed) Millenium Ecosystem Assessment. World Resources Institute, Washington, pp 623–662Google Scholar
  45. Sandri M, Zuccolotto P (2006) Variable selection using random forests. In: Zani S, Cerioli A, Riani M, Vichi M (eds) Data analysis, classification and the forward search. Studies in classification, data analysis, and knowledge organization. Springer, Heidelberg, pp 263–270CrossRefGoogle Scholar
  46. Schimel DS (2010) Drylands in the earth system. Science 327(5964):418–419CrossRefGoogle Scholar
  47. Schlesinger WH, Reynolds JF, Cunningham GL, Huenneke LF, Jarrell WM, Virginia RA, Whitford WG (1990) Biol Feedbacks Global Desertif. Science 247(4946):1043–1048CrossRefGoogle Scholar
  48. Serna-Pérez A, Monger HC, Herrick JE, Murray L (2006) Carbon dioxide emissions from exhumed petrocalcic horizons. Soil Sci Soc Am J 70(3):795–805CrossRefGoogle Scholar
  49. Serrano-Ortiz P, Sánchez-Cañete EP, Oyonarte C (2012) The carbon cycle in drylands. In: Recarbonization of the biosphere. Springer, Dordrecht, p 347–368Google Scholar
  50. Starr G, Oberbauer SF (2003) Photosynthesis of Arctic evergreens under snow: implications for tundra ecosystem carbon balance. Ecology 84(6):1415–1420CrossRefGoogle Scholar
  51. Steven B, Kuske CR, Gallegos-Graves LV, Reed SC, Belnap J (2015) Climate change and physical disturbance manipulations result in distinct biological soil crust communities. Appl Environ Microbiol 81:7448–7459CrossRefGoogle Scholar
  52. Wang W, Guo J (2006) The contribution of root respiration to soil CO2 efflux in Puccinellia tenuiflora dominated community in a semi-arid meadow steppe. Chin Sci Bull 51(6):697–703CrossRefGoogle Scholar
  53. West NE (1990) Structure and function of microphytic soil crusts in wildland ecosystems of arid to semi-arid regions. Adv Ecol Res 20:179–223CrossRefGoogle Scholar
  54. Western Regional Climate Center (2014) Castle Valley Utah Inst.—Climate Summary: Accessed 22 May 2014
  55. Wohlfahrt G, Fenstermaker LF, Arnone JA (2008) Large annual net ecosystem CO2 uptake of a Mojave Desert ecosystem. Global Change Biol 14(7):1475–1487CrossRefGoogle Scholar
  56. Xu M, Qi Y (2001) Spatial and seasonal variations of Q10 determined by soil respiration measurements at a Sierra Nevadan Forest. Global Biogeochem Cy 15(3):687–696CrossRefGoogle Scholar
  57. Zaady E, Kuhn U, Wilske B, Sandoval-Soto L, Kesselmeier J (2000) Patterns of CO2 exchange in biological soil crusts of successional age. Soil Biol Chem 32(7):959–966CrossRefGoogle Scholar
  58. Zelikova TJ, Housman DC, Grote EE, Neher DA, Belnap J (2012) Warming and increased precipitation frequency on the Colorado Plateau: implications for biological soil crusts and soil processes. Plant Soil 355(1–2):265–282CrossRefGoogle Scholar

Copyright information

© US Government 2015

Authors and Affiliations

  • Anthony Darrouzet-Nardi
    • 1
    • 2
  • Sasha C. Reed
    • 1
  • Edmund E. Grote
    • 1
  • Jayne Belnap
    • 1
  1. 1.U.S. Geological SurveySouthwest Biological Science CenterMoabUSA
  2. 2.University of Texas at El PasoEl PasoUSA

Personalised recommendations