Responses of trace gases to hydrologic pulses in desert floodplains



[1] Pulsed hydrologic inputs interact with antecedent moisture conditions to shape biogeochemical dynamics in many ecosystems, but the outcomes of these interactions remain difficult to predict. Hydrologic pulses may influence biogeochemical activity through several mechanisms: by providing water as a resource, providing limiting nutrients or substrates that fuel particular biogeochemical pathways, or determining redox conditions. Antecedent moisture conditions may modify the relative importance of each of these potential mechanisms, by influencing accumulation of labile carbon and nutrients, the severity of water limitation to biological processes, and longer-term effects on abiotic conditions, including redox. We experimentally applied hydrologic pulses of different sizes (1-cm and 20-cm events) to soils of desert floodplains and assessed responses of trace gases (CO2, CH4, NO, and N2O) in dry and monsoon seasons to test these mechanisms. Size of the hydrologic pulse strongly interacted with antecedent soil-moisture conditions to determine emissions of some trace gases. Following dry antecedent conditions, water addition stimulated emissions of CO2, CH4, and NO, but not N2O, and larger experimental pulses resulted in larger fluxes. In the monsoon season, responses to water addition were muted and size of the hydrologic pulse had no effect, except for CH4emission, which increased in response to the 20-cm event. Seasonal contrasts indicated that antecedent moisture conditions constrain the effects of hydrologic pulses on biogeochemical processes, whereas contrasts among responses of different trace gases demonstrated that mechanisms controlling emissions of particular gases are water limitation (CO2), in situ production of nitrogen substrates (NO), or redox conditions (CH4). Strong and predictable interactive effects of water inputs and antecedent conditions indicate that extended droughts may cause elevated emissions of gaseous C and NO following the return of precipitation, whereas larger floods or longer wet seasons are expected to dampen gaseous fluxes, which may contribute to conserving soil C and nutrients within floodplains.

1. Introduction

[2] Hydrologic pulses, defined as temporary increases in water inputs, organize biogeochemical processes in many ecosystems by establishing physical conditions, influencing resource availability to biota, and transporting materials across landscapes. For example, precipitation triggers bouts of productivity in deserts [Noy-Meir, 1973], floods rearrange physical and biotic templates in floodplains [Junk et al., 1989], and snowmelt mobilizes solute transfer from soils to streams [Hornberger et al., 1994]. However, ecosystem responses to water inputs will vary, since both characteristics of the pulse and antecedent conditions may shape biogeochemical outcomes. In uplands, for example, where precipitation is the primary hydrologic vector, wet-dry cycles can cause strong and short-lived biogeochemical responses, particularly for xeric ecosystems [Austin et al., 2004]. In contrast, wetland soils are inundated for long periods of time, often by fluvial or groundwater inputs, which may result in anoxia and diminished rates of nutrient mineralization [Bridgham et al., 1998].

[3] Conceptualizing ecosystems along a continuum of hydrologic regimes, with aquatic and terrestrial states defined by availability and residence time of water [Grimm et al., 2003], may improve predictions of biogeochemical responses to hydrologic inputs (Figure 1). Aquatic and terrestrial ecosystem states contrast in antecedent moisture conditions that shape the resources available to biota, and may determine responses to limiting resources delivered by hydrologic pulses. Accumulation of labile C and nutrients during dry periods provides a well studied example of this effect [Borken and Matzner, 2009; Davidson et al., 1993; Valett et al., 2005]. Characteristics of the hydrologic pulse, such as magnitude and intensity, in turn influence the biogeochemical pathways activated and duration of responses [Belnap et al., 2005; Sponseller, 2007; Sánchez-Andrés et al., 2010]. Here we apply this conceptual model to floodplain ecosystems, which occupy transition zones along the aquatic-terrestrial continuum and may shift along this axis in time. Although floodplain soils may be dry for much of the year, periodic inundation by precipitation, overland flow, tributary inputs, or overbank floods alter water availability, transport materials, and influence redox conditions [Meixner et al., 2007; Ocampo et al., 2006]. Thus, myriad potential hydrologic inputs combined with strong temporal variability in antecedent moisture conditions make biogeochemistry of floodplains difficult to predict.

Figure 1.

Conceptual model describing responses of hypothetical wetland and upland ecosystems to pulsed hydrologic inputs. Width of triangles represents the relative size of available resource pools within the ecosystem. Vertical arrows depict hydrologic pulses and are labeled with resources introduced by the pulses that are hypothesized to have the greatest influence on biogeochemical responses within each ecosystem. Floodplain responses to hydrologic inputs may shift along the wetland-upland gradient due to antecedent moisture conditions.

[4] Water plays multiple roles in ecosystems and differentiating these roles may improve mechanistic understanding of biogeochemical responses to hydrologic pulses. Plants and soil microorganisms depend directly upon water for diffusion of solutes across cell membranes and sufficient hydration [Huxman et al., 2004; Schimel et al., 2007]. This direct effect of water as a resource should produce positive relationships between the size of hydrologic pulses and biogeochemical responses. Alternatively, energy- or nutrient-limited processes may respond to missing reactants delivered by hydrologic pulses [Belnap et al., 2005; McClain et al., 2003], release of protected substrates by re-wetting [Borken and Matzner, 2009], or stimulation of other biogeochemical pathways that generate available substrates [Burgin et al., 2011]. These indirect effects of water may also underlie a positive relationship between rates of biogeochemical processes and pulse size, but the relationship may be confounded by antecedent conditions. Finally, water indirectly affects C and N transformations via a negative correlation of soil water and redox potential [Renault and Sierra, 1994]; thus, larger hydrologic pulses may suppress aerobic processes and promote anaerobic processes [Yahdjian and Sala, 2010].

[5] We measured emissions of trace gases (CO2, CH4, NO, and N2O) to evaluate how interactions between hydrologic pulses and antecedent moisture conditions shape biogeochemistry of desert floodplains. As end products or intermediates of biogeochemical transformations, gas emissions integrate biological activity within soils. Emission of carbon dioxide (CO2) results from respiration by roots and heterotrophic soil microbes [Raich and Schlesinger, 1992]. Nitrous oxide (N2O) is produced during the aerobic processes of nitrification and nitrifier denitrification, as well as heterotrophic denitrification carried out under anoxic conditions [Conrad, 1996; McLain and Martens, 2006b; Weier et al., 1993]. Methane (CH4) emission results at low redox potential, when production of CH4 by methanogenesis outweighs consumption by methanotrophy [Le Mer and Roger, 2001]. Finally, biogenic production of nitric oxide (NO) is attributed to nitrification, an aerobic process that can proceed even when soil moisture is relatively low [Stark et al., 2002].

[6] Desert floodplains receive hydrologic pulses that vary in size and frequency [Harms and Grimm, 2010; Jacobs et al., 2007]. Precipitation inputs to floodplains provide small pulses of variable frequency, whereas floods are large inputs of water that may also vary in frequency, but are generally constrained to time periods of highest precipitation. We experimentally manipulated hydrologic pulses to test the alternative hypotheses that hydrologic pulses influence biogeochemical fluxes by: 1) providing water as a resource, 2) providing limiting nutrients or substrates that fuel particular biogeochemical pathways, or 3) determining redox conditions (Table 1). Wetting experiments were conducted during the dry (low-frequency hydrologic pulses) and monsoon (large, high-frequency hydrologic pulses) periods of summer, which provide strong contrasts in hydrologic conditions while maintaining similarity in temperature and productivity. These temporal contrasts were used to test the effects of antecedent conditions on responses to hydrologic pulses, allowing us to determine how biogeochemical responses of desert floodplains compare to a conceptualized wetland-upland gradient (Figure 1).

Table 1. Hypothesized Mechanisms Describing Trace-Gas Responses to Antecedent Conditions and Hydrologic Pulses in Desert Floodplains, and Predicted Patterns Resulting From Experimental Water Addition
H1: Hydrologic pulses stimulate gas emissions by alleviating water limitation to soil microorganismsP1a: Experimental hydrologic pulses will result in rapid increases in gas emission rate (i.e, ‘responses’)
 P1b: Experimental pulses will result in significantly greater responses (post-pulse versus pre-pulse) under dry antecedent conditions than under wet antecedent conditions
 P1c: Post-pulse patterns in gas emission will be temporally coherent with those of soil moisture
 P1d: Gas emissions will be correlated with pulse size
H2: Pulses stimulate gas emissions because they increase availability of some other resource (e.g., organic C, NO3, NH4+)P2a: Post-pulse temporal patterns in gas emission will not be temporally coherent with those of soil moisture
 P2b: Responses of different gases will not be temporally coherent
 P2c: Labile C and N will be scarce under wet compared to dry antecedent conditions, which will correspond to smaller responses to pulses
H3: Pulses stimulate gas emissions because they lower redox potential, creating conditions favoring anaerobic processesP3a: Experimental hydrologic pulses will result in rapid increases in gas emission rate only for end products or intermediates of anaerobic processes (CH4+, N2O [in part])
 P3b: Temporal patterns of responses for some gases will be correlated with those of soil redox potential.
 P3c: Pulses will stimulate greater production of CH4 and N2O following wet compared to dry antecedent conditions

2. Materials and Methods

2.1. Study Site

[7] The study was conducted in floodplains of the San Pedro River during the dry and monsoon seasons in summers of 2005 and 2007. The upper basin of the San Pedro (∼7,600 km2) is drained by an unregulated, perennially flowing river near the transition of the Sonoran and Chihuahuan deserts (Figure 2). Field sampling was conducted in replicate experimental plots within two floodplains (>250 m wide) along hydrologically gaining stream reaches [Baillie et al., 2007; Pool and Coes, 1999]. Results did not differ significantly between the two floodplains and no further distinctions are made between them. The floodplains lie at approximately 1300 m elevation and receive 300 to 750 mm of precipitation annually [Arizona Department of Water Resources, 2009], largely confined to two distinct seasons: winter rains (December–March) and summer monsoon storms (July–September). A hot, dry period (May–July) precedes the monsoon season. Median of daily stream discharge values during the dry season of both years was 0.05 m3 s−1 (Figure 3; USGS gauge #09471000 located ∼3 and 10.5 km downstream from the study floodplains). Median discharge for the monsoon season was 1.42 m3 s−1 in 2005 and 0.58 m3 s−1 in 2007. Flash floods of instantaneous discharge >30 m3 s−1 inundated the floodplains during both study years. The floodplains feature shallow groundwater, abundant hydric plant species, and productive vegetation. Such sites occupy approximately half of the riparian area along the length of the Upper San Pedro River [Brooks and Lemon, 2007; Stromberg et al., 2007]. Abundant plant species include cottonwood trees (Populus fremontii), seepwillow shrubs (Baccharis salicifolia and B. emoryi), native bunchgrass (Sporobolus wrightii) and two invasive grasses: Johnson grass (Sorghum halepense) and Bermuda grass (Cynodon dactylon).

Figure 2.

Location of study sites in the Upper San Pedro basin. Data from the stream and rain gauges are shown in Figure 3.

Figure 3.

Mean daily stream discharge and total daily precipitation during the study. Vertical gray bars indicate sampling periods. Discharge data are from USGS gauge #09471000; precipitation data are from the Walnut Gulch Experimental Watershed [Goodrich et al., 2008].

2.2. Experimental Method

[8] We used experimental addition of water to evaluate the effects of hydrologic pulses, and addressed antecedent moisture conditions by conducting experiments in the dry and monsoon seasons. The summer dry season precedes the monsoon season and encompasses the lowest frequency of precipitation annually, whereas the monsoon season includes sequential convective storms and floods. Experimental wetting treatments were applied to replicate plots located ∼10 m to 200 m from the stream; plots were selected to capture observed spatial heterogeneity in soil texture and chemistry [Harms and Grimm, 2008]. Although the experimental approach provided a small number of hydrologic pulses, such events are also infrequent under ambient conditions, and thus the temporal scope of manipulations is consistent with that of the hydrologic regime.

[9] Experimental and analytical approaches differed somewhat between the two years. In 2005, emission was measured in response to precipitation treatments from 10 plots for NO, and 16 plots for CO2. Water was applied to 1-m2plots using a watering can to simulate a 1-cm event, an amount typical of precipitation received during convective monsoon-season storms [Goodrich et al., 2008]. This event water was synthesized by adding salts to distilled water to match regional precipitation chemistry (National Atmospheric Deposition Program,; Central Arizona Long-Term Ecological Research Program,, and contained 1.5 mg/L N and C in the form of NO3, NH4+, and DOC (as dextrose). The experiment was repeated in the monsoon season. In 2007, emissions of CO2, N2O, and CH4were monitored following application of 2 treatments that varied pulse size. The 16 experimental plots were divided into two sub-plots; one 0.5 m2sub-plot received 1 cm of water, and the other received 20 cm, creating conditions similar to those occurring during overbank floods. The 20-cm treatments were applied within the base of chambers used for gas sampling (0.14 m2; described below). To facilitate application of the larger event size, water applied to all plots in 2007 was collected from the stream (0.08 mg/L DIN, 1–2 mg/L DOC). C and N applied in experimental event water in both 2005 (15 mg C/m2, 30 mg N m−2) and 2007 (20–400 mg C/m2, 1.6–32 mg N m−2) constituted a small fraction of the organic matter contained in soils (0.5–4 kg C m−2), but at higher concentrations could have enhanced N availability, at least in shallow soils (soil inorganic N: 0.5–1224 mg N m−2). Estimates of gas flux were made for each plot prior to application of water and at 0.5 and 24 h following treatment. Some experiments were additionally monitored at 2, 4, and/or 48 h. Air and soil temperatures were measured during flux measurements, and soils were collected immediately following each flux measurement for analysis of soil moisture content. Estimates of soil redox potential (see Soil analyses) were also made following each flux measurement in 2007.

2.3. Gas Sampling and Analysis

[10] NO flux was measured from 10 replicate plots in 2005 only, using a portable chemiluminescence analyzer (LMA-3D, Unisearch Associates, Canada), following the instrument configuration and protocol described byHall et al. [2008]. Briefly, sample gas is passed through a nafion dryer system and through an ozone filter, and the concentration of NO2 in the sample stream is determined via chemiluminescence following reaction with Luminol II solution. The sample may be passed through an additional CrO3 filter, which converts all NO in the sample to NO2 for analysis of NO+NO2. Since measured NO2 concentration was never greater than 2 ppb at the study sites, concentrations measured as NO+NO2 were interpreted as NO. Chambers for NO sampling were constructed of white molded PVC (surface area 508.6 cm2) and placed over PVC bases, with a total chamber volume of approximately 10 L. Exact volumes for each measurement were calculated using height of the base plus chamber from the ground surface. Chambers were connected to the analyzer via Teflon tubing and air from the chamber headspace was pumped into the analyzer at a rate of 625 to 1025 mL/min. After establishment of a baseline, concentrations of NO were recorded every 15 s for 5 min. A final baseline value in air was collected, and shift in baseline conditions was removed from the flux time series by subtracting the slope of the pre- and post-sampling baseline. The instrument was calibrated against a certified NO standard (0.1 ppm, Scott-Marrin Inc., Riverside, California) prior to conducting flux measurements each day and after changes in environmental conditions. Standard concentrations were corrected for changes in O2mixing ratios caused by mixing of standard gas and the scrubbed-air carrier-gas stream. Fluxes over the linear range during chamber incubations were calculated by:

equation image

as presented by Venterea et al. [2003], where FNO is the flux of NO (μg N m−2 h−1), dC/dt is the rate of change in NO concentration (μg N m−3) in the chamber over time (h) determined by linear regression, V is the total chamber volume (m3), A is the soil surface area (m2), CA is the average NO concentration during the full measurement period, and Q is the airflow rate through the chamber (m3 h−1). Estimates of NO flux were obtained from 8 to 10 plots at each time point following treatment during the dry season, and a maximum of 6 plots in the monsoon season due to limited site accessibility following floods.

[11] CO2flux was measured from 16 plots in 2005, including the 10 plots monitored for NO, using an infrared gas analyzer (EGM-2, PP Systems). The instrument chamber was coupled to pre-installed PVC collars via a rubber gasket to prevent gas leaks between the chamber and soil surface. Three replicate collars were installed within each experimental plot, and the mean of these analytical replicates was used in subsequent analyses. Instrument-calculated fluxes were based on measurements of CO2concentration in the chamber recorded at 4-s intervals over 60–120 s. The IRGA was calibrated in the field using a certified gas standard (380 ppm, Matheson) prior to sampling, and following changes in ambient temperature.

[12] To facilitate multiple, simultaneous measurements on 16 experimental replicates, emissions of CO2, CH4, and N2O were measured in 2007 using static chambers, with analysis performed by gas chromatography. Insulated stainless steel chambers were attached to stainless steel bases, with vacuum grease applied to a vinyl seal at the chamber-base junction. Total chamber volume was approximately 20 L, with exact volume calculated using height of each installed chamber and base. Chambers were incubated for 45 min, with gas samples withdrawn every 15 min. A 20-mL sample was collected into a syringe via a septum in the chamber lid and air returned passively via a vent to the chamber headspace. Headspace gas samples were immediately transferred to evacuated 10-mL vials that had been flushed with N2. Overpressurizing vials prevented atmospheric contamination and facilitated retrieval of sample for analysis. Certified standards were similarly collected and transported to the field site to assess sample preservation, with no significant changes observed over the <2-week storage interval. Trace-gas concentrations were measured on a Varian CP-3800 gas chromatograph equipped with thermal conductivity, electron capture, and flame-ionization detectors for analysis of CO2, N2O, and CH4, respectively. Repeated injection of certified gas standards indicated variation of 2% or less. Gas flux was determined as the linear change in headspace concentration over time.

[13] A relatively long incubation time was necessary for accurate determination of N2O and CH4 as these gases are present at low concentration. Underestimation of CO2 flux may occur during static chamber incubations due to a decreasing concentration gradient between the soil air and chamber headspace, resulting in reduced flux of CO2 across the soil–air interface [Norman et al., 1997; Pumpanen et al., 2004]. Although all incubations of static chambers produced linear changes in CO2concentration within the chamber headspace, caution should be applied toward direct comparison of values from the infrared gas analyzer (2005) and static chambers (2007), as side-by-side comparison of the two methods showed a fourfold underestimation of fluxes using static chambers compared to the IRGA (T. K. Harms, personal observation, 2008). Absolute values from the 2 methods are not directly compared across years, but qualitative comparisons are used to assess similarities in responses (i.e., differences among treatments and across study years). Quantitative comparison of CO2 flux with other data sets should be made with dynamic measurements from 2005.

2.4. Soil Analyses

[14] Prior to wetting experiments, one soil sample was collected from each plot for determination of nutrient and organic matter content, texture, and initial water content by standard methods [Robertson et al., 1999]. Soils were collected manually using a copper ring (5 cm diameter) inserted to a depth of 8 cm. Soils for nutrient determination were kept in a cooler until extraction, which was performed within 24 h of sample collection. Available nutrient content was determined by extraction of soil in 2 M KCl by vigorous shaking for 1 h. Resulting extracts were filtered through pre-leached, ashless Whatman 42 filter papers and frozen until analysis for NH4+-N (phenol-hypochlorite method) and NO3-N (cadmium reduction method) on a Lachat flow-injection autoanalyzer. Soil moisture was determined gravimetrically by drying subsamples at 105°C for 48 h. Organic matter content was determined as ash-free dry mass by combustion of subsamples at 550°C for 4 h. The density hydrometer method was used to determine soil texture following dispersion of clays in sodium hexametaphosphate. Sand content was determined gravimetrically by rinsing subsamples through a 53-μm sieve. All analyses of soils were conducted on the <2-mm fraction.

[15] Following application of water to plots, soil samples (2 cm depth) were collected manually from the treated area following gas sampling (i.e., within 30 min after water addition) using a 5-cm diameter core, stored in airtight plastic bags, and analyzed for water content in the laboratory. In 2007, redox potential was additionally measured in soils following gas collection. Probes were constructed of a platinum electrode and copper wire soldered to a brass rod. The Pt-brass junction was insulated with marine-grade epoxy and heat-shrink tubing. Electrodes were checked for accuracy and precision in the laboratory using Light solution and water, respectively [Wafer et al., 2004]. Pt electrodes were inserted 2 cm into the soil at a 45° angle prior to each experiment. Salt bridges constructed of perforated PVC filled with 2M KCl in agar ensured electrical connection with soil. For each measurement, 2M KCl was added to the salt bridge, a Calomel reference electrode was placed in contact with the agar, and the reference and Pt electrodes were connected to a current meter. Field measurements were corrected to the standard H2 electrode by adding 250 mV to each measurement [Urquhart and Gore, 1973].

2.5. Statistical Methods

[16] Ambient emissions of trace gas and soil characteristics prior to wetting experiments were compared between the dry and monsoon seasons using the Wilcoxan sign-rank test (due to non-normality), with the exception of NO flux, which was compared using a two-sample t-test. These tests were performed separately for each year. To evaluate temporal trends in gas flux within each treatment and season, we used repeated-measures ANOVA followed by post hoc comparisons among times that were corrected for multiple tests using the Sidak procedure in Systat v. 12. Repeated-measures ANOVA was again used to compare responses to experimental wetting across treatments (event size) and seasons (dry and monsoon) within each year. Response variables were cumulative fluxes determined by integration of flux versus time following wetting, and fluxes measured 0.5 h following treatment, which was selected to capture the “peak” response of most variables. This mixed-effects model was analyzed using PROC MIXED in SAS v. 9, with plot as a random factor and treatment and season (2 levels each) as fixed factors. Tukey's multiple comparisons post-hoc test was applied to examine differences among combinations of treatment and season.

[17] Factors influencing trace gas fluxes at the plot-level were assessed using regression. Soil characteristics (texture, inorganic N, moisture, temperature, and organic matter) were used in multiple regression models to predict ambient emissions and cumulative flux for the 24-h period following wetting treatments. Additionally, soil moisture and redox potential were used in regressions to predict emissions of trace gases at each time step following experimental wetting. All resulting significant models contained only single predictors. We report results of regressions for NO in 2005, but for clarity, all other results reflect data from 2007. However, models for CO2 emission in 2005 and 2007 were qualitatively similar. Statistical tests were considered significant when P < 0.05.

3. Results

3.1. Seasonal Patterns in Ambient Conditions

[18] Soil organic matter, inorganic N, and water content differed significantly between dry and monsoon seasons (Table 2). Across both study years, soil moisture content increased at least fivefold from the dry to monsoon seasons. Although sampling in the dry season of 2007 followed a rain event, this did not alter soil moisture conditions at the time of sampling, and dry season soil moisture content was equivalent across both study years (Table 2). Whereas soil organic matter content increased significantly from the dry to monsoon season in 2007, it declined in 2005. This decline was accompanied by a significant decline in silt and clay content of surface soils in the monsoon season of 2005. Soil inorganic N varied greater than 10-fold across the plots, but despite this high spatial variability, the NO3-N pool decreased following the monsoon season in most plots, resulting in a significant seasonal contrast (Table 2). A decrease in soil NH4+ also occurred following the monsoon season, but this change was not significant. Declines in soil N and organic matter following the monsoon season are consistent with Prediction (P) 2c, with the exception of soil organic matter in 2007.

Table 2. Mean, Standard Deviation, Minimum, and Maximum Values of Soil Characteristics and Ambient Trace Gas Emissions for the Experimental Plots in Dry and Monsoon Seasonsa
  • a

    N = 16, except N = 15 for CH4 in monsoon season (as described in text), NO dry season N = 8, NO monsoon season N = 6. Significant differences between seasons: one asterisk indicates P < 0.01, and two asterisks indicate P < 0.001. Note that CO2 flux was estimated by a dynamic method in 2005 and using static chambers in 2007, producing absolute values that are not directly comparable between years.

Air temperature (C)35.21.833.337.826.8**0.825.927.5
Soil temperature (C)34.54.328.943.323.6**1.421.326.9
Soil moisture (%)**7.3915.9134.67
Organic matter (%)**1.900.616.58
silt+clay (%)**10.26.337.3
CO2 (mg C m−2 h−1)292.49118.63178.18457.270.18**0.410.000.91
NO (μg N m−2 h−1)1.150.94−0.142.531.671.790.645.26
Air temperature (C)32.53.527.139.829.2**2.226.333.2
Soil temperature (C)26.14.621.437.021.3**1.019.722.9
Eh (mV)5002844655548614310625
Soil moisture (%)5.013.320.3413.1830.23**15.771.2063.66
Organic matter (%)3.681.471.196.144.56*1.631.056.99
NO3-N (mg/kg dry soil)21.3852.020.25192.21.24*
NH4+-N (mg/kg dry soil)1.561.560.145.780.880.9403.63
Silt+clay (%)61.529.87.098.468.731.713.998.7
N2O (μg N m−2 h−1)0.713.38−5.2910.502.257.82−3.6529.59
CO2 (mg C m−2 h−1)66.5935.0421.01143.6076.7831.587.78137.63
CH4 (μg C m−2 h−1)−2.458.22−24.659.7417.3044.37−20.05137.38

[19] Despite seasonal contrasts in soil chemistry and physical attributes, ambient rates of CH4, N2O, and NO flux (i.e., before experimental hydrologic pulses) were statistically indistinguishable across seasons (Table 2). Ambient rate of CO2 emission, however, was much greater in the dry than the monsoon season of 2005. In 2005, sampling immediately followed a large, overbank flood whereas the 2007 sampling followed a much smaller flood (Figure 3), consistent with P1d, even after accounting for methodological differences between years. CH4 and N2O emission tended to increase from the dry to monsoon seasons, but these trends were not statistically significant. During the monsoon season, ambient flux of CH4 from one plot was 1652.6 μg CH4-C m−2 h−1, >5 SD above the mean, and it was removed from all analyses, although responses to experimental wetting were qualitatively similar to the other plots. This point corresponded with the lowest measured ambient Eh value.

[20] Ambient fluxes of N2O, CH4, and NO varied by more than a factor of 10 across experimental plots during both seasons, with greater spatial variation during the monsoon season (Table 2). However, multiple regression models indicated that little of this variation could be explained by measured soil attributes. In the dry season, a significant, negative relationship occurred between CO2 flux and soil temperature, and soil moisture explained a significant portion of variation in ambient NO flux (Table 3).

Table 3. Results of Significant Linear Regressions of Trace-Gas Emissions With Soil and Environmental Variables as Predictorsa
  • a

    Ambient refers to pre-manipulation rates and time since water addition is noted in parentheses for experiments.

  • b

    ln transformed.

  • c

    Results are presented for 2007. Relationships in 2005 were qualitatively similar.

  • d

    Two values near the lower limit of instrument detection were not included in the analysis.

  • e

    One negative outlier removed, noted in text.

Dry Season
   CO2b,cair temperature16−0.0320.330.019
   NOdsoil moisture60.2830.890.004
1-cm event     
   CO2 (0.5 h)soil moisture16−6.3340.300.028
   NO (0.5 h)soil moisture9−2.8990.480.039
Monsoon Season
1-cm event     
   CO2 (48 h)soil moisture161.9810.540.001
20-cm event     
   CH4 (0.5 h)b,esoil moisture150.0330.67<0.001
   CH4 (24 h)bEh16−0.0050.68<0.001

3.2. Responses to Experimental Wetting

[21] Soil moisture increased in response to both experimental event sizes during the dry season and returned to ambient values within 24–48 h (Figures 4 and 5). During the monsoon season, only the 20-cm addition resulted in significant increases in soil moisture (Figure 5). Thus, a significant season*treatment effect (F1,28 = 9.40, P = 0.005) was observed when comparing treatment effects on soil moisture at 0.5 h. Ehdeclined significantly following the 1-cm, but not the 20-cm experimental event in the dry season, and returned to ambient by 4 h following water addition (Figure 5). Water addition had no effect on Eh during the monsoon season, but Eh increased significantly 24–48 h after experiments. No significant differences in Eh were found among treatments and seasons measured at 0.5 h following experimental water addition.

Figure 4.

Soil moisture (N = 7–10) and emission of CO2(N = 12–16) and NO (N = 8–10) following an experimental 1-cm hydrologic pulse in the dry and monsoon seasons (N = 2–6, all variables for monsoon season) of 2005. Measurements at t = 0 reflect fluxes prior to addition of water. Points are means (±1 SE). Letters represent significant differences (P < 0.05) among measured time points using the Sidak procedure to correct for multiple comparisons following repeated measures ANOVA. Time points lacking letters were not included in statistical tests due to reduced sample size.

Figure 5.

Redox potential, soil moisture, and trace gas emission following experimental hydrologic pulses in the dry and monsoon seasons of 2007. Measurements at t = 0 reflect fluxes prior to addition of water. Points are means (±1 SE), N = 14–16. Lower case letters indicate significant differences (P < 0.05) among measured time points following a 1-cm event, whereas numbers indicate differences following the 20-cm event. Significance was determined using the Sidak procedure to correct for multiple comparisons following repeated measures ANOVA. Observations collected 48 h following treatment in the monsoon season were not included in analysis due to reduced sample sizes. Capital letters designate significant differences across treatments (1-cm and 20-cm events) and seasons for fluxes measured 0.5 h following wetting.

[22] CO2flux increased significantly in response to experimental water addition during the dry season of both study years, and in 2007 fluxes were greatest following the 20-cm event (Figures 4 and 5), consistent with P1d. The pulsed pattern of CO2 flux following experimental wetting appeared to approximately match the timing of changes in soil moisture during the dry season, supporting P1c (Figures 4 and 5). Water addition resulted in no change in CO2 flux in the monsoon season (Figures 4 and 5), consistent with P1b. Accordingly, comparison of peak-response measurements at 0.5 h following water addition yielded a significant season*treatment interaction (F1,28 = 6.01, P = 0.021), supporting P1b for CO2. Cumulative flux of CO2 following wetting differed only between seasons (2007 data; F1,28 = 36.8, P < 0.001). In the dry season, variation in fluxes of CO2across experimental plots was negatively correlated with soil moisture during the peak (0.5 h) response time following the 1-cm event. There were no significant relationships among soil characteristics and emission of trace gases following the 20-cm event during the dry season. During the monsoon season, flux of CO2at 48 h following the 1-cm event was positively related to soil moisture (Table 3). This occurred when CO2emission had returned to the pre-manipulation rate.

[23] Peak emission (measurements at 0.5 h following treatment) of CH4was greater in response to the 20-cm than the 1-cm treatment (F1,28 = 14.05, P < 0.001), but this effect was driven by responses during the dry season (Figure 5; P1b and d). Cumulative fluxes of CH4 following experimental wetting varied across treatments and seasons (season*treatment F1,26 = 5.32, P = 0.029). A pulsed response of CH4flux was observed in response to the 20-cm, but not the 1-cm event in both seasons, partially supporting P1d. During the dry season the pulse in CH4 emission declined more rapidly than soil moisture following treatment, whereas CH4 emission remained elevated for a longer period than soil moisture in the monsoon season, contrasting with P1c. Highest emissions of CH4 occurred at the same time as lowest Ehduring the monsoon-season experiments, consistent with P3b. In the monsoon season, regression models indicated that spatial variation in emission of CH4across plots was positively correlated with soil moisture at 0.5 h following addition of the 20-cm event. CH4 emission was negatively correlated with Eh24 h following the 20-cm event (Table 3; P3b).

[24] NO flux increased significantly following addition of experimental water in the dry season. Fluxes remained elevated 24 h following wetting, a period longer than observed for responses of both CO2 flux and soil moisture (Figure 4; supports P2a). Spatial variation in flux of NO among plots was explained by a negative correlation with soil moisture 0.5 h following experimental wetting during the dry season (Table 3). NO flux was unchanged in response to experimental wetting during the monsoon season, partially supporting P1b and P2c.

[25] Responses of N2O to water addition were muted compared to other gases (Figure 5). Neither peak nor cumulative responses of N2O to experimental treatments were significant and wide variation in responses among plots contributed to this lack of significance. However, this variation among plots could not be explained by measured soil attributes.

4. Discussion

[26] We experimentally manipulated hydrologic inputs to desert floodplains to evaluate influences of hydrologic pulses on trace-gas production in the context of antecedent moisture conditions. We predicted that pulse responses would be large under dry antecedent conditions and weak, if present, during the monsoon season owing to alleviation of either water (Hypothesis [H]1) or substrate (H2) limitation (Table 1). Strong, pulsed responses of CO2, NO, and CH4 to experimental hydrologic inputs during the dry season demonstrated a primary role of water in initiating biogeochemical processes. In contrast, muted biogeochemical responses to water input following wet antecedent conditions indicated that water no longer functioned as a limiting resource, and that temporary increases in its availability caused changes in abiotic conditions that were relevant only to production of CH4. Seasonal contrasts in responses of biogeochemical processes to hydrologic pulses are consistent with a shift along a conceptual aquatic-terrestrial continuum (Figure 1), wherein dry antecedent conditions result in pulsed responses that mirror uplands and wet antecedent conditions constrain effluxes of trace gas products of C and N cycling, similar to patterns observed in wetlands.

4.1. Dry Season Responses to Hydrologic Inputs

[27] Pulsed emission of CO2, NO, and CH4 following experimental wetting during the dry season indicated a stimulatory effect of water, with emissions responding positively to event size. A stimulatory effect of water may be due directly to alleviation of water limitation, or delivery of substrates in the experimental pulse. Observed pulses in trace gas emissions were brief, and emissions of CO2 were temporally coherent with the increase in soil moisture. Increased soil organic C, inorganic N, and microbial biomass following the first precipitation events of the monsoon season [Harms and Grimm, 2008] suggest that water inputs have a direct effect on the metabolism of microorganisms, of which CO2 is a dominant byproduct, providing support for H1. Several lines of evidence indicate that pulsed CO2 emission was not due to resource input by hydrologic pulses. First, emissions of CO2increased by 1.5 times following a 1-cm event in 2005 and by a factor of 1.8 in 2007 despite greater inputs of labile organic C and inorganic N in experimental water applied in 2005. Further, using estimates of mass balance, bulk soil organic C (0.5–4 kg C m−2) exceeded the cumulative respiratory consumption of C following experimental events by several times, even when considering a wide range of microbial efficiencies. In contrast, the CO2flux exceeded dissolved organic C introduced in event water (1-cm event: 10–15 mg DOC m−2, 20-cm: 200 mg DOC m−2), suggesting that mineralization of soil organic matter in part supported the pulse of respiration. C supporting respiration likely derives from a small, labile pool and although estimates of C lability concurrent with measures of gas fluxes are not available, previous work at this site demonstrated that microbial communities metabolize more complex C compounds during the summer drought and actively use carbohydrates during the wet season [Harms and Grimm, 2008]. This pattern may occur due to release of protected C substrates following water input [Borken and Matzner, 2009]. Additionally, UV-caused breakdown of detritus during dry periods may have resulted in CO2 production [Brandt et al., 2009], but this mechanism is likely constrained in desert floodplains by high quality litter and canopy cover in summer [Uselman et al., 2011]. If the hydrologic pulse increased availability of C substrates in soils, this effect was dependent upon water, given that CO2 flux declined concurrently with soil moisture, and suggesting that any effect of C availability was indirect. Therefore, increased respiratory response to the larger hydrologic pulse was unlikely caused by increased C or nutrient delivery in the larger pulse, and instead was supported by soil C. Additionally, deeper infiltration by a larger volume of water may have activated a larger microbial community [Austin et al., 2009; Schwinning and Sala, 2004], contributing to a larger CO2 flux. Thus, water stimulated soil respiration in the dry season by alleviating water limitation (H1).

[28] Although emissions of CH4 were also stimulated by addition of water, temporal patterns in emissions were not consistent with those of soil moisture, providing evidence that additional factors influenced CH4 flux. Despite a greater response of CH4to the 20-cm compared to 1-cm pulse, redox potential was similar following both sizes of experimental events, suggesting that redox conditions alone did not cause elevated flux of CH4following the 20-cm event. Wetting of deeper soil horizons during larger pulses may have activated sufficient methanogenesis to overcome a CH4 sink caused by methanotrophs in shallower soils [McLain and Martens, 2006a]. Supply of organic C was an unlikely constraint for methanogenesis as organic C in either soil or experimental water was several orders of magnitude greater than that emitted as CH4 following wetting experiments. Finally, although production of CH4 by termites has been indicated in upland, arid soils [Galbally et al., 2010; McLain and Martens, 2006a; Otter and Scholes, 2000], low termite densities in the floodplain study sites (E. Hagen and K. McCluney, personal communication, 2011) provide little support for this mechanism. Thus, observations of CH4 flux do not provide direct support for the hypothesized mechanisms, but indicate that saturation of soils, potentially to depths exceeding surface horizons, supports CH4 flux from desert floodplain soils.

[29] Response of NO also contrasted with patterns expected if water directly controlled flux, with NO flux remaining elevated for several hours longer than soil moisture. Although water inputs were required to stimulate NO production, this extended period of elevated NO flux could have been fueled by newly available substrates following water addition. Nitrification is the likely source of NO emitted from these dryland soils [Stark et al., 2002], and assuming that NO accounts for 1% or less of the total N moving through the nitrification pathway in unfertilized soils, the cumulative flux of NO following the experimental event could not have been sustained by available NH4+ in soil (2–55 mg N m−2), plus NH4+ in water added experimentally (15 mg NH4+-N m−2). Previous work at the study site indicated increased net N mineralization and pools of NH4+ and NO3 following the onset of monsoonal precipitation [Harms and Grimm, 2008]. Thus, tight coupling of nitrification with N mineralization likely provides the required NH4+ [Stark et al., 2002]. Further, NO efflux can continue as soils dry in part because consumption of NO by denitrifiers declines under low soil water potential [Stark et al., 2002]. It is important to note that water addition may promote abiotic generation of NO [Venterea et al., 2005], but this is unlikely a significant contributor to NO flux in N-poor desert soils where soil temperature is <40°C [McCalley and Sparks, 2009]. Thus, patterns of NO flux following water input during the dry season indicate that water initiated production of NO by stimulating N mineralization, enhancing the substrate pool for nitrification (H2).

[30] Mean increases in N2O flux following wetting experiments in the dry season were not significant, implicating additional factors that may influence N2O flux, or suggesting that the amounts of water added were insufficient to stimulate net production of N2O. Emission of N2O results from the net effects of production and consumption by multiple metabolic processes, including autotrophic and heterotrophic nitrification, nitrifier denitrification, and heterotrophic denitrification [Conrad, 1996; McLain and Martens, 2006b; Ullah and Moore, 2011], all of which may respond differentially to water, substrate availability, or redox conditions. Assessment of resources limiting N2O production therefore cannot be conducted here because we do not know which microbial processes contribute to its production in these desert floodplains. However, McLain and Martens [2006b] determined that heterotrophic nitrification produced N2O in sandy upland soils of the San Pedro basin, but respiratory denitrification prevailed once soils were saturated, and this process was limited by organic C.

[31] Mechanisms observable at the whole floodplain extent (km2 extent) explained little of the spatial heterogeneity in fluxes of trace gases across replicate (m2-extent). Negative correlation of soil moisture content with fluxes of CO2 and NO following experimental wetting demonstrates that although water initiated pulses of gas production, higher soil water content limited emissions of these gases. Similar negative relationships of NO with soil moisture have been observed following experimental wetting in other dryland soils and are thought to result from diffusional limitation and subsequent consumption of NO by denitrifiers at higher soil moisture content [Davidson et al., 1993; Stark et al., 2002]. Similarly, high soil moisture may restrict diffusion of CO2 within the soil profile, as observed in poorly drained wetland soils [Ullah and Moore, 2011]. Soil moisture did not explain plot-scale variation in fluxes of N2O and CH4, despite a strong response of CH4to addition of water at the floodplain scale. The magnitude of inter-plot differences in soil moisture therefore appears insufficient to cause local spatial variation in CH4 flux. Finally, soil organic matter and inorganic N pools varied at least fivefold across the study plots (Table 2), yet these characteristics were not correlated with emission of trace gases from floodplain soils before, during, or cumulatively following experimental wetting events, providing no direct evidence for C or N limitation of gas fluxes at the plot scale. In summary, analysis of spatial variation in trace gas emissions suggest that the mechanisms by which water influences fluxes depend on the scale of analysis. Our data demonstrate that water inputs result in predictable changes in gaseous fluxes at the floodplain and event scales, but that determining mechanisms underlying responses at finer temporal and spatial scales will require more focused study that encompasses simultaneous measurement of additional predictor variables. Importantly, the relationships revealed by our data at the floodplain scale are relevant for understanding the contribution of floodplains to C and N export from stream networks and to modeling regional budgets of trace gas emissions.

4.2. Monsoon Season Responses to Hydrologic Inputs

[32] Following several monsoon-season storms that resulted in overbanking floods, experimental hydrologic pulses yielded only small changes in fluxes of trace gases. There were no significant differences in responses to hydrologic pulses of different sizes during the monsoon season, potentially indicating that water no longer functioned as a limiting resource to microorganisms in floodplain soils. For example, ambient soil moisture content in the monsoon season was equal to moisture content during the 20-cm treatment of the dry season (Figure 5), but ambient emissions of trace gases during the monsoon season remained several-fold lower than emissions observed at equivalent soil moisture content during experimental treatments in the dry season. Alternatively, lack of biogeochemical response to hydrologic pulses in the monsoon season may indicate that pulses caused little change in abiotic conditions when soils were already wet.

[33] Substrates to fuel biogeochemical reactions can become depleted in soils during wet seasons (Table 2) [Austin et al., 2004; Davidson et al., 1993; Harms and Grimm, 2008]. Gas emissions then decline following sequential precipitation events [Sponseller, 2007], during wet versus dry seasons [Davidson et al., 1993], or in response to frequent floods [Valett et al., 2005]. In addition to inputs of water, floods may also deposit C- and N-poor sediments on floodplains, effectively decreasing substrate availability to microorganisms in shallow soils [Sponseller and Fisher, 2006]. Significant deposition of coarse sediments was observed in 2005, especially for plots nearest the stream, as indicated by a significant decline in soil organic matter content and fine particles (% silt+clay) in shallow soils. Sediment deposition in 2005 likely suppressed biological activity due to limited colonization by microorganisms of newly deposited sediments, or by causing C limitation of heterotrophic processes. This effect is seen in the very low ambient CO2flux in the monsoon season of 2005, which was measured immediately following a large overbank flood. In contrast, organic matter content of soils increased following monsoon floods in 2007, with no change in soil texture, demonstrating the potential for floods to redistribute organic matter. Thus, inter-annual variation in sediment delivery to floodplains may contribute to longer-term patterns in emission of trace gases from desert floodplains. However, peak annual soil microbial biomass occurs during the monsoon season [Harms and Grimm, 2008], and biomass is positively related to accumulated precipitation [Harms and Grimm, 2010], suggesting that the substrate-limitation mechanism does not explain seasonal-scale patterns.

[34] Hydrologic inputs of high frequency or size, including repeated overbank floods, may result in reducing conditions due to physical displacement of gases in soil pore space or consumption of oxygen by aerobic respiration. Estimates of redox potential did not indicate widespread anoxia in floodplain soils, although sub-oxic conditions (Eh < 200 mV) were detected in several plots following both natural flooding and experimental wetting. Although ambient emission of CH4in the monsoon season was not distinguishable from the dry season, the highest ambient emission of this redox-sensitive gas occurred from a low-lying plot with the lowest measured redox potential (Eh = 10), and CH4 was the only gas to show increased emissions in response to experimental addition of water in the monsoon season. A potential indirect effect of redox is demonstrated by the temporal pattern in NO flux following experimental wetting in the monsoon season. Emission of NO, which results from production by an aerobic process, was observed only 1–2 days following experimental wetting, when soils were drier than at the start of the experiments (Figure 4).

[35] Finally, lower soil and air temperatures in the monsoon compared to the dry season may have influenced the observed seasonal contrasts in ambient fluxes and responses to hydrologic inputs. In upland soils of the San Pedro basin, temperature was a significant predictor of gas fluxes, but only when effects of precipitation or soil moisture were also significant [McLain and Martens, 2006a]. Indeed, the effects of temperature and soil moisture are correlated in many ecosystems and temperature effects are often stronger when soils are wet [Borken and Matzner, 2009; Davidson et al., 1998]. Range in temperature among replicate plots during the dry season of 2007 was 12.7 and 15.6°C for air and soils, respectively, whereas mean differences between dry and monsoon seasons were only 3.3 and 4.8°C, suggesting that our dry and monsoon seasonal comparisons mainly reflect the effects of water availability, not temperature. Despite high variation in temperature across plots, ambient CO2 flux was significantly correlated with temperature only in the dry season, and the effect was negative, as has been observed in other desert soils [Talmon et al., 2011] and poorly drained soils [Ullah and Moore, 2011]. These patterns suggest that temperature is not a primary contributor to the observed seasonal contrasts in emissions of trace gases from floodplain soils. Additionally, it is unlikely that lower average temperatures in the monsoon season would have eliminated the pulse behavior of responses, as temperature was constant over the first several hours of experiments.

[36] In summary, dampened biogeochemical responses to hydrologic pulses in the monsoon season indicate a seasonal shift in the role of water that was induced by antecedent hydrologic conditions. Our data do not directly support the hypothesis that substrate limitation constrained gas fluxes following the monsoon season (H2), but smaller soil C and N pools (relative to the dry season) associated with lack of pulsed responses to water addition provides indirect evidence in support of this mechanism. Additionally, observations of CH4 flux provide some support for the hypothesis that redox conditions are an important influence on biogeochemical cycling following wet antecedent conditions (H3).

4.3. Responses to Hydrologic Pulses Across the Upland-Riparian Interface

[37] Few studies describe ecosystem processes of desert floodplains; therefore, to provide a broader context for experimental data, we compare data from this study with estimates of in situ trace-gas emissions from riparian ecosystems and xeric uplands (seeauxiliary material). All available literature data for trace gas emissions from riparian ecosystems, including those in mesic biomes, are considered here. In general, floodplains of the San Pedro responded to water availability in a manner more consistent with riparian zones than xeric uplands, and rates were generally within the range of those observed in other riparian zones (Figure 6). CO2 emission from riparian soils decreased relative to annual means during wet seasons, whereas xeric upland responses to wet seasons or precipitation were more consistently positive (Figure 6). These contrasts likely arise from differing sizes of hydrologic pulses, with floods or high water tables leading to dampened CO2 fluxes during wet periods in riparian zones [Pacific et al., 2009; Yu et al., 2008].

Figure 6.

Estimates of trace gas fluxes from riparian and xeric upland soils reported in the literature. Riparian ecosystems were defined as fluvially influenced and subject to annual flooding. Xeric uplands included ecosystems receiving <500 mm annual precipitation, as well as ecosystems characterized by seasonal drought. Events include both natural and experimental precipitation or floods. Boxes define inner and outer quartiles, median is represented by centerline, whiskers depict 10th and 90th percentiles, and circles show outliers. Dotted lines represent mean values from this study. For NO, dry season values from floodplains of this study were negative. Note log scale for flux of CO2 and NO. References are listed in the auxiliary material.

[38] Methane flux from the desert floodplains in response to water qualitatively matched other riparian sites, with a strong positive response to hydrologic inputs. Emission of CH4 from the San Pedro and other dryland floodplains [Otter and Scholes, 2000] is unlike that of xeric upland soils (Figure 6). Soils of drylands, including upland terraces at the San Pedro, tend to be sinks for CH4 (Figure 6) [McLain and Martens, 2006a], and the strength of the sink is positively correlated with soil moisture or precipitation [McLain and Martens, 2006a; Otter and Scholes, 2000; Striegl et al., 1992]. Because of this contrast between upland and riparian soils, floodplain extent in xeric basins may be a determinant of regional CH4 emissions.

[39] Similar to experimental results from the present study, N2O emission across riparian zones did not respond consistently to hydrologic pulses or wet seasons (Figure 6). In contrast, xeric upland soils often showed a pulse of N2O production in response to precipitation (Figure 6). Such pulsed N2O emissions reflect microbial use of substrates that have accumulated during dry antecedent periods [Davidson et al., 1993; Scholes et al., 1997]. Although water stimulates biological production of N2O, upland soils typically remain unsaturated, and thus diffusion of N2O from soils to the atmosphere remains high [Chapuis-Lardy et al., 2007; Scholes et al., 1997]. In contrast, temporally variable water and redox regimes characteristic of riparian zones may facilitate multiple N transformation processes [Harms and Grimm, 2010; Pett-Ridge et al., 2006], allowing both production and consumption of N2O [Conrad, 1996; McLain and Martens, 2006b; Ullah and Moore, 2011]. Further, net production of N2O is constrained when N2 is the predominant product of denitrification, including under conditions of abundant C availability, anoxia, and poor diffusion of N2O from saturated soils [Chapuis-Lardy et al., 2007; Davidson et al., 1993; Weier et al., 1993], conditions typical of riparian soils.

[40] Few field-based estimates of NO flux from riparian zones appear in the literature, but NO fluxes from desert floodplains of the present study appeared consistent with those of xeric uplands. A consistent, positive response to hydrologic inputs characterizes NO flux from uplands, but ambient fluxes tended to be greater in upland soils compared to the floodplains of this study (Figure 6). Water appears to limit production of NO in dryland ecosystems [Galbally et al., 2008; Yahdjian and Sala, 2010], whereas N supply has a primary influence in mesic ecosystems [Hall and Matson, 2003]. A laboratory-based comparison of upland and floodplain soils in South Africa indicated greater emissions of NO from floodplain soils, and found that soil moisture, but not soil N content, predicted emissions [Otter et al., 1999].

[41] In summary, data assembled from the literature confirm a stimulatory role of water on emission of CO2and NO following dry antecedent conditions, as demonstrated by data from xeric uplands. Cross-biome patterns in riparian zones suggest that under wet antecedent conditions, desert floodplains conform to patterns typical of mesic riparian zones, but that upland models may be more appropriate during dry conditions—in other words, the desert floodplain shifts seasonally along a wetland–upland continuum (Figure 1). Finally, the literature review revealed a large gap in data sets describing trace-gas emissions from desert riparian ecosystems, particularly for N.

5. Conclusions and Implications

[42] Despite the small spatial extent of floodplains relative to surrounding uplands, their key location within the landscape may result in a disproportionate influence of floodplains on the forms or abundance of C and N exported to downstream ecosystems. In particular, removal of C and N via gaseous pathways limits downstream hydrologic fluxes and influences local budgets of greenhouse gases. Relationships between emissions of CO2, CH4, and NO from floodplains of the San Pedro River and multiple attributes of the hydrologic cycle illustrate strong interactions of the hydrologic cycle and the biogeochemical processes that produce these gases. Because desert floodplains undergo seasonal and inter-annual shifts in biogeochemical dynamics due to dominant climate drivers [Harms and Grimm, 2010], observations from these ecosystems may be used to anticipate general ecosystem responses to altered hydrologic regimes. In particular, droughts are expected to interact with the size of subsequent storms or floods to stimulate microbial processes that turn over soil C and N or generate abiotic conditions conducive to gaseous efflux, resulting in elevated emissions of gaseous C and NO following larger hydrologic pulses. Larger floods or longer wet seasons are expected to dampen gaseous fluxes, which may contribute to conserving soil C and nutrients. Previous ecosystem studies of semi-arid floodplains have concluded that net accumulation of C and nutrients in floodplains will depend upon the interplay of multiple attributes of the hydrologic regime, including length of antecedent dry periods, and size, duration, and intensity of hydrologic inputs [Follstad Shah and Dahm, 2008; Sánchez-Andrés et al., 2010; Valett et al., 2005]. Here we have demonstrated a strong and predictable interaction of the size of hydrologic pulses with antecedent moisture conditions that determines the magnitude and composition of trace gas fluxes, a vector for material loss from ecosystems.


[43] This work was completed while NBG was at the National Science Foundation. Any opinions or findings are those of the authors and do not necessarily reflect the views of the Foundation. We are grateful to members of the Grimm, Hall, and Sabo laboratories at ASU for contributions to data collection. S. Hall, H. Hartnett, and the J. King laboratory provided equipment and advice with field sampling and S. Anderson provided access to a protected field site. S. and S. Martin facilitated completion of this manuscript. Comments from J. Elser, S. Fisher, S. Hall, J. Sabo, and several reviewers and editors improved the manuscript. The Sustainability of Semi-Arid Hydrology and Riparian Areas program, an NSF Science and Technology Center (NSF-OIA-9876800) provided funding.