Volume 127, Issue 11 e2021JG006654
Research Article
Open Access

Spatial Microbial Respiration Variations in the Hyporheic Zones Within the Columbia River Basin

Kyongho Son

Corresponding Author

Kyongho Son

Pacific Northwest National Laboratory, Richland, WA, USA

Correspondence to:

K. Son,

[email protected]

Contribution: Conceptualization, Methodology, Software, Validation, Formal analysis, ​Investigation, Resources, Data curation, Writing - original draft, Writing - review & editing, Visualization

Search for more papers by this author
Yilin Fang

Yilin Fang

Pacific Northwest National Laboratory, Richland, WA, USA

Contribution: Conceptualization, Methodology, Software, Formal analysis, ​Investigation, Data curation, Writing - review & editing

Search for more papers by this author
Jesus D. Gomez-Velez

Jesus D. Gomez-Velez

Vanderbilt University Nashville, Nashville, TN, USA

Environmental Sciences Division, Climate Change Science Institute, Oak Ridge National Laboratory, Oak Ridge, TN, USA

Contribution: Software, Resources, Data curation, Writing - review & editing

Search for more papers by this author
Xingyuan Chen

Xingyuan Chen

Pacific Northwest National Laboratory, Richland, WA, USA

Contribution: Conceptualization, Writing - review & editing, Supervision, Project administration, Funding acquisition

Search for more papers by this author
First published: 08 November 2022
Citations: 1

Abstract

While the hyporheic zone (HZ) accounts for a significant portion of whole stream CO2 concentrations, HZ respiration modeling studies are lacking in quantifying their contributions to the total CO2 at large watershed/basin scales. Quantifying the contribution of anaerobic respiration is also underappreciated. This study used a carbon-nitrogen-coupled river corridor model to quantify HZ aerobic and anaerobic respiration and determined the key factors controlling their spatial variability within the Columbia River Basin (CRB). The modeled respiration patterns showed high spatial variability. Among the nine sub-basins composing the CRB, the Lower Columbia and the Willamette, which receive higher precipitation, had higher respiration. Medium-sized rivers (fourth to sixth orders) produced the highest aerobic and anaerobic respiration among reaches of different sizes. At the basin scale, aerobic respiration is dominant, representing approximately 98.7% of the total respiration across the CRB. While most of the reaches were dominant with aerobic respiration, reaches in agricultural land showed a relatively higher anaerobic respiration (18%) ratio. A variable importance analysis showed that hyporheic exchange flux controlled most of the spatial variability of HZ respiration, dominating over other physical variables such as residence time, stream dissolved organic carbon (DOC), nitrate, and dissolved oxygen (DO). The influence of substrate concentration (DOC and DO) is larger in modeling anaerobic respiration than aerobic respiration. Future efforts will focus on improving the estimation of the HZ exchange flux and the implementation of spatially explicit parameterizations for the reactions of interest to reduce model uncertainty.

Plain Language Summary

Riverbeds generate high amounts of CO2, but their contribution to the total CO2 budgets in rivers is not well quantified via numerical simulation models for large regions. This study used a numerical simulation model to estimate the CO2 emissions from riverbeds into water columns in the presence and absence of oxygen and identified important variables explaining the spatial variation of riverbed CO2 emissions within the Columbia River Basin (CRB). Our modeling study found that CO2 emissions from riverbeds showed high spatial variability. Within the CRB, wetter sub-basins showed higher CO2 emissions than drier sub-basins. Medium-sized rivers generated the highest CO2 emissions. Most channel CO2 emissions occurred in the presence of oxygen. However, reaches in agricultural lands generated relatively high CO2 emissions in the absence of oxygen. The water exchange rate between channel and riverbed can explain the spatial variation of CO2 emissions over other physical variables.

Key Points

  • Hyporheic exchange flux controls the spatial variation of respiration across reaches with different sizes and land uses

  • At the basin scale, hyporheic zone aerobic respiration accounts for about 98.7% of total simulated respiration

  • 18% of total respiration is from anaerobic respiration in agricultural streams

1 Introduction

Carbon dioxide emissions from lotic and lentic inland waters account for a significant portion of the global carbon cycles (Butman et al., 2016; Raymond et al., 2013); however, many carbon accounting studies have misrepresented CO2 emissions from inland waters (Battin et al., 2009). Recent studies using the CO2 gradient between the atmosphere and inland waters through gas exchange velocity estimated CO2 emissions from inland waters at regional and global scales (Butman & Raymond, 2011; Butman et al., 2016; Raymond et al., 2013). Due to the uncertainty in estimating the exchange velocity and CO2 concentration in inland waters, the estimates of the inland CO2 emissions still showed a high uncertainty. For example, the estimates of aquatic carbon fluxes via outgassing at global scale have been steadily revised upwards over the past decade, with the most recent estimate being 3.88 (PgC/yr) (Drake et al., 2018). The uncertainty of CO2 inland water emissions results in incomplete closure of the carbon balance at regional and global scales. Therefore, accurate quantification of CO2 evasion from inland waters is required. In particular, the hyporheic zone (HZ), the interface between groundwater and surface water, generates a significant CO2 into the atmosphere (Battin et al., 2003; Burrows et al., 2017; Fellows et al., 2001). HZ is a source of dissolved CO2 to the stream/river, and that CO2(along with CO2 from other sources—notably water column respiration and delivery from the terrestrial watershed) is then available for emission to the atmosphere. However, due to the difficulty of measuring the CO2 production in streambeds and a high spatial and temporal heterogeneity due to varying hydrologic conditions and streambed hydraulic/substrate conditions, the relative contribution of the HZ to the whole-stream dissolved CO2 concentrations is underexplored, except for a few studies (Battin et al., 2003; Fellows et al., 2001).

The hyporheic exchange rate, residence time of the riverbed hydrologic exchange flux within the HZ, nutrients such as oxygen, organic matter, and nitrogen, and temperature and biogeochemical reaction rates can all control the relative contribution of the HZ to the whole-stream CO2 (Battin et al., 2003). Battin et al. (2003) showed that the hyporheic respiration accounts for 41% of whole-stream respiration, with the averaged HZ respiration of 0.38 (gC/m2/d), and its rate increased from 46% to 59% with increased baseflow. Also, Fellows et al. (2001) showed that the HZ's contribution to whole-stream respiration ranged from 49% to 93%. The higher transient storage in streams with higher surface-groundwater exchange was attributed to higher stream respiration, hyporheic respiration, and the percentage of contribution to hyporheic respiration. Hotchkiss et al. (2015) showed the relative contribution of the terrestrially derived CO2 and internal productions to CO2 evasion from running waters by compiling the previously published measurements of net ecosystem production from 187 streams and rivers across the Contiguous United States (CONUS) and using estimated CO2 emission from flowing waters. The study showed that about 72% of CO2 evasion (3.09 gC/m2/d) is from the terrestrially derived CO2. While the significant contribution of HZ respiration is proven, many HZ respiration studies focused on aerobic respiration processes by using oxygen data (Demars et al., 2015); anaerobic respiration is understudied (Stanley et al., 2016). Although anaerobic respiration can be dominant under conditions of low oxygen, lower rate of hyporheic exchange, and longer residence time, it is uncertain how much anaerobic respiration contributes to stream respiration, and how its contribution becomes larger or smaller. Therefore, it is critical to quantify the relative contribution of HZ aerobic and anaerobic respiration along streams with varying hydrologic and biogeochemical conditions.

The HZ is recognized as a biogeochemical hot spot and has high potentials for nutrient removal and attenuation of pollutants, so it can regulate water quality in the river and carbon cycle from the reach scale to regional and global scales (Findlay, 1995). Thus, lack of a solid understanding of the underlying mechanisms, and our inability to predict where high-process rates occur in a large river basin, impedes effective river/watershed management (Dwivedi et al., 2018; McClain et al., 2003). Most HZ respiration modeling tools have been limited to reach scales (Zarnetske et al., 2011), or numerical experiments (Kaufman et al., 2017; Newcomer et al., 2018) have been conducted, except for recent watershed/basin-scale modeling applications (Fang et al., 2020; Jan et al., 2021; Painter, 2021). Jan et al. (2021) developed and tested an integrated surface and subsurface model (Amanzi-ATS), coupling with the ADEKS (advection-dispersion equation with Lagrangian subgrids) that allows computing aerobic respiration and denitrification in the HZ. However, this study is still limited to demonstrating the capability of the watershed model to simulate the HZ processes and their impacts on streamwater quality in an agriculture-dominant watershed. Fang et al. (2020) developed a SWAT-MRMT-R model to simulate the impact of the HZ on stream nitrate concentration for the upper Columbia–Priest Rapids watershed in the Columbia River Basin (CRB). The model coupled a carbon-nitrogen microbially driven reaction model with SWAT (a watershed water quality model). However, the study focused on the nitrogen cycle rather than CO2 production from the HZ.

Many experimental studies (Battin et al., 2003; Burrows et al., 2017; Fellows et al., 2001) and detailed small-scale numerical experiments (Dwivedi et al., 2018; Kaufman et al., 2017; Newcomer et al., 2018) have improved our understanding of mechanisms of the HZ. However, it is still challenging to obtain a generality across systems and provide an organizing framework to simplify intersystem comparisons (Findlay, 1995). Our study developed a coupled carbon-nitrogen river corridor model (RCM) that adopted the reaction network model from the SWAT-MRMT-R (Fang et al., 2020) to study the spatial variation of HZ aerobic and anaerobic respiration at the basin scale. We used the CRB to examine general HZ aerobic/anaerobic respiration patterns across reaches with different sizes and land uses. Specifically, we ask two questions:

  1. How does HZ aerobic and anaerobic respiration vary at the sub-basin and reach scales within the CRB? We used the modeled aerobic and anaerobic respiration estimates to quantify their spatial variation at the reach and sub-basin scales. Also, we calculated the ratio of aerobic respiration to total respiration (sum of aerobic respiration and anaerobic respiration).

  2. What are the key factors controlling the spatial variation of HZ aerobic and anaerobic respiration in streams/rivers of the CRB? We used a variable importance analysis (Grömping, 2006) to identify which factors control the spatial variation of HZ respiration, and tested how the variable importance changed with reaches sizes and land uses and variables of interest (aerobic and anaerobic respiration).

2 Methods

2.1 Study Site: Columbia River Basin

The CRB is a large river basin in the CONUS, and has drainage areas from a portion of seven states (Idaho, Montana, Wyoming, Washington, Oregon, Utah, and Nevada) in the United States, and the British Columbia Province in Canada (Figure 1). The total basin size is approximately 620,000 km2; however, our modeling study focuses on US-CRB, with a size of approximately 570,413 km2. Elevation ranges from 0 to 4,154 m, and the US-CRB is composed of nine sub-river basins: (a) Lower Columbia, (b) Middle Columbia, (c) Upper Columbia, (d) Lower Snake, (e) Middle Snake, (f) Upper Snake, (g) Kootenai-Pend Oreille-Spokane, (g) Willamette, and (i) Yakima River. The climate over the US-CRB varies widely. The mean annual precipitation ranges from 159 to 5,230 mm, and the mean air temperature ranges from −1.7 to 12.4°C. The western sides of Washington and Oregon have a humid continental climate; and the eastern sides of Washington and Oregon and all of Idaho have a semi-arid steppe climate. The Cascade Range in Washington and Oregon, and the Rocky Mountains in Idaho, Montana, and Wyoming have an alpine climate. The seasonal patterns of precipitation are consistent over the basin; most precipitation occurs in the winter, while the precipitation phases (rain vs. snow) vary with elevation (or temperature).

Details are in the caption following the image

Location of the Columbia River Basin (CRB), its sub-basins, and precipitation/temperature and land cover/use and elevation: (a) mean annual precipitation (mm); (b) mean annual air temperature(°C), (c) 2016 NLCD; and (d) elevation map (meter). The nine sub-basins: Lower Columbia (LC), Lower Snake (LS), Middle Columbia (MC), Middle Snake (MS), Kootenai-Pend Oreill-Spokane (KO), Upper Snake (US), Upper Columbia (UC), Willamette (WM), and Yakima (YK).

Major land use/cover includes 33.7% forest land (33% evergreen forest, and about 0.3% and 0.4% deciduous forest and mixed forest), 33% shrub land, 12% agriculture land (10% cropland and 2% hay and pasture), and 2.3% urban land. Among the nine sub-basins, the Lower Columbia and Willamette have higher annual precipitation (2,116 and 1,668 mm) than other sub-basins (from 455 to 833 mm). The two sub-basins have the highest proportion of forest land use, and the second dominant land use/cover is shrub (10.5%) and agriculture (21.5%), respectively. Among the nine sub-basins, Willamette has the highest urban area (8%). The drier sub-basins tend to have a higher proportion of shrub area; for example, shrub lands in Middle Columbia, Middle Snake, and Upper Snake occupy more than 30% of these areas. Detailed information about the mean annual precipitation and dominant land use/cover for each sub-basin is summarized in Table S5 in Supporting Information S1. The total number of stream reach is 189,656, defined in National Hydrography Dataset PLUS (NHDPLUS v2) (https://nhdplus.com/NHDPlus/). The largest river of the CRB is a ninth order stream.

2.2 River Corridor Model

The RCM used in this study combines empirical substrate models derived from observations and three microbially driven reactions to compute respiration of the HZ for each NHD reach within the CRB (Figure 2). The reactions in HZs of each NHD reach include anaerobic respiration and two-step anaerobic respiration via denitrification (Table 1). Our HZ respiration estimates are limited to the lotic (or flowing) stream/river systems, and do not account for the respiration process in water column. Note that the RCM only simulates the HZ's contribution to the dissolved CO2 concentrations in the streams, and the CO2 emissions to the atmosphere are not modeled.

Details are in the caption following the image

Conceptual diagram for building the river corridor model (RCM) and its key processes. The RCM requires key input data including long-term annual mean of NEXSS-based hyporheic exchange flux (ql and qv) and residence time (τl and τv), and three substrate concentrations (dissolved organic carbon [DOC], dissolved oxygen [DO], and urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0001) derived from the SPARROW model outputs and empirical regression models. The model computed the HZ respiration components (aerobic and anaerobic respiration) via lateral and vertical exchanges between stream and HZ at reach scale.

Table 1. Aerobic Respiration and Two-Steps of Anaerobic Respiration Reactions
Reaction process Reaction equations
Aerobic respiration R1 urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0002
Anaerobic respiration R2 urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0003
Anaerobic respiration R3 urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0004
The model computes at hourly timesteps because of the fast reaction rates (Tables 1 and 2). The key input data of the model, that is, exchange flux, residence time, and stream solute (dissolved organic carbon [DOC], dissolved oxygen [DO], and urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0005) concentrations, are constant over time and represent long-term averaged values (Figure 3). The results reported in this study represent long-term mean annual estimates per NHD reached when the reactions reach dynamic equilibrium. This model computes the solute exchange between stream and HZ (expressed in Equations 1 and 2). There is no interaction between upstream and downstream reaches; that is, the vertical and lateral exchange zones in each reach are treated as independent batch reactors. Even though the model is able to accommodate multiple HZs per stream, our modeling study only considered one HZ for vertical and lateral exchange per stream reach for simplicity. The following equations are used to calculate the concentration change in the HZ through the mass exchange between the stream and HZ (Equation 1) as well as microbial reactions in the HZ (Equation 2).
urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0006(1)
where τ is the residence time of hydrologic exchange flux, Cs,i is the stream “i” solute concentration (DOC, urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0007, and DO), and Ci,t is the hyporheic “i” solute concentration at the t timestep. μi,j is the stoichiometric coefficient of solute i in reaction j. Rj is the reaction rate the jth reaction.
urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0008(2)
where mi,t is “i” solute cumulative production/consumption amount (urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0009 and urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0010) at “t” timestep, Vt is hyporheic exchange volume (q(t) × w × l × τ) and q(t) is hyporheic exchange flux (m/s), w is stream width (m), l is stream length (m), and τ is residence time (s).
Table 2. Reaction Parameter Values and Initial Substrate Concentrations in the HZ
Parameter R1 R2 R3
Reaction rates fi urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0011 0.65 0.99
ki (mole/L/hr) 3 × 1.17 1.17 0.97
Kd,i (mmole/L) 0.25 0.25 0.25
Ka,i (mmole/L) 0.001 0.001 0.004
Initial concentrations (mole/L) in hyporheic zone DOC urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0012 DO
6.37e−5 7.92e−5 2.87e−4
  • Note. R1 is aerobic respiration (O2CO2), R2 urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0013 and R3 urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0014 are a two-step anaerobic respiration.
Details are in the caption following the image

Key model inputs: (a) annual mean stream dissolved organic carbon (DOC) (mg/L); (b) annual mean stream urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0015 (mg/L); (c) annual mean stream dissolved oxygen (DO) (mg/L); (d) annual mean residence time (log10(s)); and (e) annual mean hyporheic exchange flux (log10(m/s). To construct the annual mean stream DOC/DO concentration, we used samples collected in the periods of January 1, 1980 to December 31, 2021, and January 1, 2007 to December 31, 2021, respectively. Stream urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0016 is based on the 2012 SPARROW model outputs (Wise et al., 2019). Hyporheic exchange flux and residence time are based on the NEXSS model estimates (Gomez-Velez et al., 2015).

The reaction rate law (Equation 3) of R1, R2, and R3 and the associated parameters are obtained from Table 2 in Song et al. (2018). Tables 1 and 2 include the three reactions and their associated model parameter values.
urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0017(3)
urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0018(4)
urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0019(5)
where ki (mol/L/hr), Ka,i (mol/L), and Kd (mol/L) denote the maximum specific uptake rate of organic carbon, half-saturation constants of the electron acceptors, and half-saturation constants for the electron donors. ai is the concentration of electron acceptor (mol/L), di is the concentration of electron donor (mol/L), and BM is the concentration of biomass (mol/L). Reaction rate Ri (mol/L3/s) is computed using unregulated effect (a Monod-type kinetics coefficient urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0020 in Equation 4, and regulated effects (ei) in Equation 5.

The model requires three stream substrate concentrations (DOC, DO, and urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0021) as inputs. Annual mean stream DOC and DO concentrations were estimated by developing multilinear regression models with the measured stream DOC/DO concentration at the U.S. Geological Survey gage stations, and the watershed/stream variables (Schwarz et al., 2018). To construct the annual mean stream DOC concentration, we used sampled data from 1/1/1980 to 12/31/2021. The developed annual mean stream DOC concentration model is a function of percentage of basin/catchment shrub area (tshrub and logshrub) and basin agriculture area (logtargc) (stream DOC = −0.03(tshurb) +0.45(logtargc) −0.12(logshurb) +3.15). In addition, to construct the annual mean DO concentration data, we used the samples collected in the period from January 1, 2007 to December 31, 2021 since the DO sensor had some accuracy issues before the 2007 year. The stream DO concentration model is a function of the soil bulk density (TOT_BDAVE), topographic wetness index (TOT_TWI), basin drainage area (TOT_BASIN_AREA), and catchment dam storage (logCAT_NID) (stream DO = −2.85(TOT_BDAVE) −0.49(TOT_TWI) +0.31(logTOT_BASIN_AREA) +0.12(logCAT_NID)). The annual mean stream urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0022 concentrations at the reach scale were estimated via the output of the SPAtially Referenced Regressions on Watershed attributes (SPARROW) model (Wise et al., 2019). The SPARROW model output represents the 10 years averaged flow and nutrient flux estimates (suspended solid sediment, total nitrogen, and total phosphorous), based on a base year (2012). The annual mean stream urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0023 concentration was computed by dividing the total nitrogen (TN) annual mean loading with the annual mean streamflow estimate from the output of the SPARROW model, and multiplying the ratio of urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0024 to TN. The ratio was computed based on the measured urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0025 and TN concentrations at the U.S. Geological Survey gauge stations in the CRB. Detailed descriptions of the developed substrate model are included in the Supporting Information S1. Exchange flux and residence time are results from the Networks with Exchange and Subsurface Storage (NEXSS) model (Figure 3). The NEXSS model couples NHDPLUS-based geomorphologic model with physics-based surrogate models for hyporheic exchange to compute the exchange flux, residence time distribution, and median residence time via a bedform-driven (or vertical) and sinuosity-driven (or lateral) exchange (Gomez-Velez et al., 2015; Gomez-Velez & Harvey, 2014). Details of the NEXSS model can be found in the papers of (Fang et al., 2020; Gomez-Velez & Harvey, 2014).

The model is simulated at the hourly timestep and computes the mean annual CO2 production (gC/day) via aerobic and anaerobic reactions for each reach of the NHDPLUS stream, and scales by stream surface area (m2). The stream surface areas were calculated using stream lengths from NHDPLUS (Schwarz et al., 2018) and stream width derived by the power relationship between measurement of instantaneous flow and bankfull width and NHD cumulative drainage area (Gomez-Velez et al., 2015). The model separately calculates the CO2 production via vertical and lateral hyporheic exchange. To test the variation of annual CO2 production between years, we ran the model over 10 years, and found that after the second year simulation, the annual production amounts reached a dynamic steady state (Figure S10 in Supporting Information S1). For our modeling analysis, the second-year simulation results were used. The detailed modeling results can be found in the Supporting Information.

2.3 Spatial Analysis of Modeled HZ Respiration

2.3.1 Reach, Sub-Basin, and Basin-Scale Modeled HZ Aerobic/Anaerobic Respiration Variation Within the CRB

We quantified the spatial variation of the HZ mean annual aerobic/anaerobic respiration at the reach, sub-basin, and basin scales within the CRB. Also, we quantified the difference in the HZ aerobic/anaerobic respiration (gC/d/m2) in reaches with different orders (sizes) and different land uses (forest, shrub, urban, and agriculture). This study classified the channel sizes in three groups based on Strahler's ordering system: (1a) small-sized streams (1–3rd); (b) medium-sized rivers (4–6th); and (c) large-sized rivers (7–12th). While the largest order in the CRB is the ninth river, the large-sized rivers include seventh to ninth order streams in our analysis. To determine the dominant land use for each stream reach, we calculated the percentage of each land use within the total upstream routed accumulated area. If the percentage of the drainage area for each land use type is larger than 80%, we assign that as the dominant type. The National Land Cover Database 2001 was used to calculate the percentage of each land cover. To simplify the classification, forest land use includes mixed, deciduous, and evergreen forest types; and urban land use includes developed open spaces and developed low-density, medium-density, and high-density areas. Agriculture land use includes pasture, hay, and cultivated crop areas. We observed that the key inputs for the model vary with the stream orders and dominant land uses (Figures S11 and S12 in Supporting Information S1). For example, hyporheic exchange flux has a unimodal relationship with channel orders; and medium-sized rivers show the highest flux. While the stream urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0026 concentrations vary with the channel orders (or sizes) in the CRB, stream DOC and DO increase with the channel orders. The reaches with different land types also have varying hydrologic and substrate conditions. The reaches in forest land show the highest exchange flux and DO concentrations, and lowest urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0027. The reaches in agricultural lands show the lowest exchange flux, and the highest substrate DOC/urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0028. The detailed results can be found in the Supporting Information.

2.3.2 Variable Importance Analysis of Key Input Variables

To disentangle the relative importance of the hydrologic and substrate variables in explaining the spatial variation of modeled HZ aerobic and anaerobic respiration in the reaches across different sizes and land uses, respectively, we used the Linderman-Merenda-Gold (LMG) (Grömping, 2006) algorithm. The LMG is implemented in the relaimpo R package. The algorithm quantifies the contribution of different correlated variables (hydrologic variables vs. substrate variables) in multilinear regression models. The LMG is based on the sequential R2s but accounts for the effects of ordering additional variables in the multilinear regression model by averaging over the ordering. We used all five variables (exchange flux, residence time, stream DOC, urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0029, and DO concentrations) to develop multilinear regression models. All variables were normalized to minimize effects of the magnitude of each variable.

2.3.3 Sensitivity of Aerobic/Anaerobic HZ Respiration to Available Substrate Concentration

This study used the mean annual substrate concentration as one of the key inputs for simulating HZ respiration. Since accurately predicting representative mean substrate concentrations of streams/rivers are limited by the available measured concentration data in the CRB, we acknowledge that there may be a high uncertainty in estimating the substrate concentrations and this uncertainty may affect our modeling results. Thus, we tested the impact of the substrate concentrations on the modeled HZ aerobic and anaerobic respiration by (a) comparing individual model estimates using spring and summer average substrate concentrations(Figure S3, S4, S7, and S8 in Supporting Information S1) and annual average substrate concentrations (Figure 3), and (b) evaluating the sensitivity of modeled estimates to increasing/decreasing substrate concentrations across streams/rivers. For stream DO, we applied decreasing stream DO concentrations (25% and 50%) scenarios because many reaches are already close to the saturation limit, and a 100% decreasing DO concentration scenario creates an unrealistic DO concentration (<0 mg/L) in reaches. On the other hand, because stream DOC and urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0030 are low enough to limit respiration processes, we applied increasing stream DOC and urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0031 concentrations (25%, 50%, and 100%) scenarios. The detailed descriptions of constructing the spring and summer DOC and DO concentrations are included in the Supporting Information S1. This analysis can evaluate how the uncertainty of estimated substrate concentration impacts the modeled HZ respiration at the reach and sub-basin scales.

3 Results

3.1 Spatial Variation of Modeled HZ Respiration at Reach, Sub-Basin, and Basin Scales

We computed mean annual HZ aerobic and anaerobic respiration (gC/m2/day) for each reach of the NHDPLUS with the CRB (Figure 4). The modeled HZ aerobic respiration had similar spatial patterns with the modeled HZ anaerobic respiration. For the CRB, HZ aerobic respiration ranged from 0 to 1,060.3 (gC/m2/day) with a median value of 2.86, and HZ anaerobic respiration ranged from 0 to 265.8 (gC/m2/day) with a median value of 0.015. Aerobic respiration was dominant at all scales. At the reach scale, ratio of aerobic respiration to total respiration ranged from 66.0% to 99.9% with a median value of 99.5%. At the basin scale, the ratio of aerobic respiration is also about 98.7%. Across the sub-basins, the ratio of aerobic respiration ranged from 98.0% to 99.3%(Figure 5). The Lower Columbia had the highest total respiration (sum of aerobic and anaerobic respiration). Generally, sub-basins with higher precipitation tended to have higher respiration than sub-basins with lower precipitation. The sub-basin average aerobic respiration has higher spatial correlations (R2 = 0.98) with mean annual precipitation, compared to the sub-basin averaged anaerobic respiration (R2 = 0.68). The lower correlation in the relationship between anaerobic respiration and mean annual precipitation is due to lower anaerobic respiration in the Lower Columbia with the largest mean annual precipitation (2,120 mm). The correlation coefficient increases to 0.96 without the Lower Columbia. The Lower Columbia sub-basin has the largest aerobic respiration, but the anaerobic respiration is the second largest. The Willamette, with the second largest precipitation (1,660 mm), has the largest anaerobic respiration.

Details are in the caption following the image

Modeled mean annual hyporheic zone respiration within the Columbia River Basin: (a) aerobic respiration (log10(gC/m2/day)); (b) anaerobic respiration (log10(gC/m2/day); (c) total (aerobic + anaerobic) respiration (log10(gC/m2/day)); and (d) ratio of anaerobic respiration to total respiration.

Details are in the caption following the image

Sub-basin scale modeled hyporheic zone (HZ) respiration within the Columbia River Basin (CRB): (a) sub-basin scale respiration (gC/m2/day); and (b) relationship between mean annual precipitation (mm) and sub-basin averaged aerobic/anaerobic respiration (gC/m2/day).

At the reach scale, both the modeled aerobic respiration and anaerobic respiration had the highest values in the medium-sized rivers. The ratio of anaerobic respiration to total respiration did not change with the channel orders; the aerobic respiration was always dominant along different orders. Among the reaches with different land uses, forest lands showed the highest aerobic and anaerobic respiration. Reaches in urban lands showed the second-highest aerobic and anaerobic respiration. While aerobic respiration was still dominant between reaches with different land uses, the ratio of anaerobic respiration to total respiration largely varied with the dominant land uses. For example, reaches in forest lands had the largest anaerobic respiration, but the ratio of anaerobic respiration was lower (less than 0.6%) than those in the agriculture and urban lands. While reaches in urban lands had the second largest anaerobic respiration, the anaerobic respiration ratio was about 8% of total respiration. Reaches in agricultural lands had the highest anaerobic respiration ratio (18%). Reaches in shrub lands had the lowest aerobic and anaerobic respiration and the lowest anaerobic respiration ratio (less than 0.5%).

3.2 Key Factors Controlling the Spatial Variation of HZ Respiration

We observed high spatial variation of aerobic and anaerobic HZ respiration along reaches with different orders and dominant land uses (Figure 6). To determine which factors control the spatial variation of modeled HZ respiration at reach scale, we evaluated the relative importance of the key model input variables in explaining the spatial variation of aerobic and anaerobic respiration across reaches with different sizes and land uses using the LMG.

Details are in the caption following the image

Variation of modeled hyporheic zone respiration (gC/m2/day) in the streams with different stream orders and dominant land use: (a) aerobic respiration (log10 (gC/m2/day)) with different stream orders; (b) anaerobic respiration (log10 (gC/m2/day)) with different stream orders; (c) ratio of anaerobic respiration to total respiration with different stream orders; (d) variation of aerobic respiration in streams with different land uses (log10 (gC/m2/day)); (e) variation of aerobic respiration in streams with different land uses (log10 (gC/m2/day)); and (f) ratio of anaerobic respiration to total respiration with different land uses.

Figure 7 shows the relative importance of each variable with modeled aerobic and anaerobic HZ respiration for different stream/river types (sizes and land uses). Overall, two hydrologic variables, exchange flux and residence time, had higher relative importance values than other substrate variables. Between the two hydrologic variables, exchange flux had a higher relative importance value than residence time. The difference of the relative importance between exchange flux and residence time varied with the stream/river types and variable of interest. While the relative importance of substrate concentration was low, stream DO and DOC showed higher relative importance in aerobic respiration and anaerobic respiration than stream urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0032, respectively. Stream DO showed slightly higher importance for aerobic respiration in the small-sized streams, medium-sized rivers, and forest/shrub streams than other substrate concentrations. Stream DOC showed slightly higher relative importance for anaerobic respiration in the small-sized/large-sized streams and forest/urban streams. Larger importance of the DOC concentration in the small-sized/large-sized streams and rivers may be due to relatively lower exchange change since the modeled exchange flux shows the highest rate in the medium-sized rivers. The importance of DOC in the urban streams was larger than the importance of residence time. In summary, the relative importance measures showed that hyporheic exchange had the highest importance values across all stream/river types for both aerobic and anaerobic respiration. While the importance of the substrate concentration was limited, it is larger for modeled anaerobic respiration than for the modeled aerobic respiration, especially for the stream DOC.

Details are in the caption following the image

Relative importance of modeled aerobic/anaerobic respiration in streams with different sizes and land uses: (a) aerobic respiration; (b) anaerobic respiration; (c) aerobic respiration; and (d) anaerobic respiration. Dissolved oxygen (DO): stream dissolved oxygen concentration, dissolved organic carbon (DOC; mg/L): stream DOC concentration (mg/L), HZ_Flux: hyporheic zone (HZ) exchange flux (m/s), HZ_RT: HZ residence time (s), and NO3: stream urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0033 concentration (mg/L).

3.3 Sensitivity of Modeled HZ Respiration to Substrate Availability

We tested the impact of seasonal average DOC/DO on the modeled HZ respiration at reach, sub-basin, and basin scales (Table 3). Comparing annual average substrate with spring average substrate, aerobic respiration was increased from 26.57 to 30.77 (gC/m2/day) at the basin scale, and with summer average substrate, anaerobic respiration was increased from 0.33 to 0.51 (gC/m2/day). The effects of seasonal average substrate concentration on aerobic and anaerobic respiration vary with the sub-basins (Figure 8). Among the nine sub-basins, Willamette has the largest variation of anaerobic respiration with using different substrate concentrations. For example, comparing annual mean substrate concentration with summer average substrate, anaerobic respiration was increased from 1.86 to 3.02 (gC/m2/day) due to lower stream DO concentration. However, the effect of seasonal average substrate concentration on reach-scale aerobic and anaerobic respiration is minor (Figure 9). Aerobic respiration with spring and summer average substrate concentrations have very high correlation value (≥0.99) with those of annual average substrate concentration, and anaerobic respiration with spring average substrate concentrations has slightly lower correlation values (≥0.96) with those of annual average substrate concentration.

Table 3. Effect of Seasonal Substrate Concentration on Modeled HZ Aerobic and Anaerobic Respiration at Sub-Basin and Basin Scales
Aerobic respiration (gC/m2/day) Anaerobic respiration (gC/m2/day)
Sub-basin/Basin Annual Spring Summer Annual Spring Summer
Lower Columbia 103.48 96.96 104.97 0.94 0.66 1.20
Lower Snake 31.31 41.19 31.76 0.40 0.53 0.59
Middle Columbia 13.42 16.65 14.06 0.14 0.25 0.19
Middle Snake 7.33 10.38 7.43 0.06 0.08 0.11
Kootenai-Pend Oreille-Spokane 34.76 40.77 34.83 0.40 0.47 0.60
Upper Snake 12.97 16.31 13.60 0.14 0.17 0.24
Upper Columbia 19.06 24.59 19.13 0.13 0.24 0.18
Willamette 90.99 79.17 91.23 1.86 0.84 3.02
Yakima 27.92 34.25 28.20 0.18 0.28 0.24
CRB 26.57 30.77 26.92 0.33 0.34 0.51
Details are in the caption following the image

Spatial maps of the modeled mean annual hyporheic zone aerobic/anaerobic respiration with seasonal/annual average substrate concentrations: (a), (b), and (c) modeled aerobic respiration with annual average substrate concentrations, spring average substrate concentrations, and summer average substrate concentrations, respectively; and (d), (e), and (f) modeled anaerobic respiration with annual average substrate concentrations, spring average substrate concentrations, and summer average substrate concentrations, respectively.

Details are in the caption following the image

Variation of the modeled mean annual hyporheic zone respiration: (a) variation of the modeled aerobic respiration with average annual, spring, and summer substrate concentrations; (b) relationship between modeled aerobic respiration with annual average substrate concentrations and ones with spring and summer average substrate concentrations; (c) variation of modeled anaerobic respiration with annual, spring, and summer average substrate concentrations; and (d) the relationship between modeled anaerobic respiration with annual average substrate and ones with spring and summer average substrate concentrations.

Also, the sensitivity of the modeled respiration to increasing/decreasing substrate concentration varies with substrate and variable of interest (Figure 10). Among the substrates, increasing DOC and decreasing DO showed the highest increase of the modeled anaerobic respiration, respectively. However, increasing urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0034 had very minor change in modeled aerobic and anaerobic respiration. Also, decreasing DO has very minor change in modeled aerobic respiration. Increasing DOC resulted in higher modeled aerobic respiration, but the change is smaller than modeled anaerobic respiration with increasing DOC.

Details are in the caption following the image

Sensitivity of the modeled aerobic/anaerobic respiration with increasing substrate concentrations: (a), (b), and (c) modeled aerobic respiration with increasing dissolved organic carbon (DOC), urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0035, and decreasing dissolved oxygen (DO) concentration, respectively, and (d), (e), and (f) modeled anaerobic respiration with increasing DOC, urn:x-wiley:21698953:media:jgrg22354:jgrg22354-math-0036, and decreasing DO concentration, respectively.

4 Discussion

4.1 Reach-Basin and Sub-Basin-Scale Spatial Patterns of HZ Respiration in the CRB

This study used a coupled carbon-nitrogen RCM to quantify the spatial variation of aerobic and anaerobic respiration in the HZ across streams/rivers of the CRB. Modeled HZ respiration exhibited high spatial variations at the stream reach and sub-basin scales. Aerobic respiration accounts for approximately 97% of the total respiration at both scales. Its pattern remains between the reaches with different orders; however, the reaches with different land uses showed varying patterns. For example, reaches in forest lands showed only 0.5% of anaerobic respiration in total respiration, but urban and agricultural reaches showed the relatively higher contribution of anaerobic respiration (about 8% and 18%, respectively) of the total respiration. This result suggests that accurately quantifying anaerobic respiration in the HZ is important for watershed-scale and basin-scale carbon estimates, especially for reaches with agricultural components. Also, future expansion of the agricultural land use regions (Spangler et al., 2020) will increase the importance of considering the anaerobic respiration in the stream and HZ.

In addition, our modeling study demonstrated the difference between aerobic and anaerobic respiration across different stream and river types and nine sub-basins in the CRB, and revealed the dominant controlling mechanism. The variable importance analysis using LMG (Grömping, 2006) suggests that the HZ exchange is the most important variable in explaining the spatial patterns of HZ aerobic and anaerobic respiration, and the effect of substrate concentration is larger in the anaerobic respiration. Thus, the HZ exchange flux can explain most spatial patterns of the HZ respiration at sub-basin and reach scale. At the sub-basin scale, HZ aerobic and anaerobic respiration has a positive, high-linear relationship with the mean annual precipitation, which is highly correlated with HZ exchange. Also, lower correlation value (R2 = 0.68) in the anaerobic respiration is due to relative low substrate in the Lower Columbia. The Lower Columbia, with the largest precipitation, has lower anaerobic respiration than the Willamette, with the second largest precipitation due to lower DOC concentrations. This result is similar to the modeling result of Butman and Raymond (2011). They showed that CO2 evasion rates from streams and rivers are positively correlated with annual precipitation across CONUS streams/rivers (R2 = 0.86).

At the reach scale, medium-sized rivers and reaches in forest lands produced the highest aerobic and anaerobic respiration amounts due to the highest exchange flux. The highest respiration in medium-sized rivers is mainly due to the highest exchange flux that is associated with sediment grain size. While the substrate concentrations (DOC and DO) vary with the stream orders, its impact on the spatial distribution of the respiration, including aerobic and anaerobic, is minor. Battin et al. (2003) and Fellows et al. (2001) support our modeling results; higher exchange flux or higher baseflow contributes to increasing the relative contribution of HZ to the whole-stream metabolism. On the other hand, our modeling results suggest that land use also affects the spatial patterns of the respiration. The reaches in urban lands showed the second-highest aerobic and anaerobic respiration. The high anaerobic respiration of urban reaches is due to the combined effects of exchange flux and substrate conditions. Urban reaches have the second largest HZ exchange, the lowest DO, and the second largest DOC values.

Our model results are based on the long-term averaged substrate concentrations and exchange flux/residence time. The substrate concentrations were empirically driven, and the accuracy of the products is highly dependent on the quality of the observed concentration data. The sensitivity of the modeled results to the substrate concentrations showed varying effects depending on the substrate type and variable of interest. However, most of the spatial patterns of aerobic and anaerobic respiration at sub-basin and reach scales remain with varying substrate concentrations. For example, a positive linearity of the HZ respiration and precipitation is observed using seasonal substrate concentrations. At the reach scale, the medium-sized river still has the highest HZ aerobic and anaerobic respiration. The reach in forest lands has the highest aerobic respiration. Therefore, we believe the major findings of our modeling results are still valid, even though the effects of uncertainty of the substrate concentration exist.

In summary, many previous measurements (Fellows et al., 2001) and modeling studies (Dwivedi et al., 2018; Kaufman et al., 2017; Newcomer et al., 2018; Song et al., 2018) of hyporheic respiration were limited to the stream reach scale, where local conditions may control respiration. It is challenging to synthesize the varying effects of hydrologic and substrate conditions on the HZ respiration in a large-river system (Findlay, 1995). Our modeling study is the first effort to quantify basin-scale HZ respiration, and its variation among sub-basins and stream/river types, including sizes and land uses. Previous basin-scale modeling frameworks have attempted to simulate HZ denitrification potentials (Gomez-Velez et al., 2015) or in-stream nutrient uptake (Alexander et al., 20072009). However, basin-scale HZ respiration modeling studies are relatively limited. Also, our modeling demonstrated that the combined effects of hyporheic exchange and substrate concentration determine spatial variation of HZ respiration at the reach and sub-basin scales within the CRB. Therefore, our modeling framework can be a hypothesis testing tool for different river systems (climate, biomes, etc.) and a sampling design tool that proposes potential hot spots of respiration processes in large-river systems. A recent study (https://www.pnnl.gov/projects/river-corridor) used our RCM outputs and machine-learning approaches (cluster analysis and random forest model) to guide streambed respiration measurements in the Yakima River Basin. The collected data will test a model-based hypothesis that controls the spatial variation of streambed respiration in the basin. We can improve the current model structure and its associated parameters by comparing the spatial patterns of modeled respiration and field measurements.

4.2 Limitations of Current Modeling Study and Future Studies

In this study, our RCM simulated HZ aerobic and anaerobic respiration. Even though the model key inputs—including HZ exchange flux, residence time, and substrate concentrations—were estimated using results from a physical-based model (NEXSS) and measured stream substrate concentrations, there are still high uncertainties in the model inputs. For example, inputs only represented temporally averaged conditions. Among the five key inputs, HZ exchange flux showed the most important variables in explaining the spatial variations of HZ aerobic/anaerobic respiration in all stream/river types (Figure 7). However, the exchange flux used in this study was heavily dependent on the streambed's grain-size estimates of the NEXSS modeling (Gomez-Velez et al., 2015). Therefore, the product quality of the grain size or hydraulic conductivity of streambed sediment may determine the spatial variation of aerobic/anaerobic respiration. A recent study using a machine-learning-based approach attempted to accurately estimate the median grain-sized distribution (D50) across the riverbeds in CONUS (Abeshu et al., 2021). The data will be useful to compare the existing grain-sized products used in NEXSS (Gomez-Velez et al., 2015) and test the impact of using different grain-sized inputs on estimating exchange flux/residence time in NEXSS and respiration in our RCM. In this study, we used empirical-driven stream mean annual concentrations. While these are the best empirical estimates we have, uncertainty at the local level is still high and the concentration products should continue to be refined at finer spatial resolutions for improved local accuracy, as it is demonstrated in our modeling sensitivity analysis that using different seasonal DOC and DO concentrations has large effects on the modeled anaerobic respiration at the reach and sub-basin scales (Figures 10d and 10f; Figure S13b in Supporting Information S1; and Table 3). Also, the substrate concentrations in the stream change dynamically with hydrologic conditions and nutrient availability in the watersheds. Therefore, using dynamic substrate concentration may give more accurate representation of the spatial/temporal respiration amounts in the watershed. Reaction parameters in the RCM (Table 2) were assumed to be spatially constant, and parameter uncertainty associated with the reaction network model was not tested. The default parameter values in the reaction network model were estimated based on batch denitrification data (Li et al., 2017) and literature values (Fang et al., 2020; Song et al., 2018). Therefore, future studies should focus on testing how parameter uncertainty alters the spatial patterns of respiration.

Lastly, our HZ respiration estimates are limited in the lotic (or flowing) stream/river systems, and do not account for the respiration process in water column. Lentic waters, including reservoirs or small ponds and wetlands, impact the downstream water quality and regional nutrient budgets (Harvey & Schmadel, 2021; Schmadel et al., 2019). Schmadel et al. (2019) showed that small ponds influences the retention of nitrogen, phosphorus, and sediment, depending on their spatial location, especially for the headwater catchment. Hotchkiss et al. (2015) suggested that internal productions to CO2 evasion from running waters at CONUS equals about 23%. Therefore, to fully account for the respiration processes in the inland water and its impact on the regional and global cycle, other parts of inland waters including lentic and water column should be incorporated in the future research.

5 Summary and Conclusion

HZ CO2 production accounts for significant portions of the inland waters carbon cycle, but previous stream/river CO2 evasion in regional and global studies did not include the effect of HZ respiration. Thus, this study developed a basin-scale, coupled carbon-nitrogen RCM for reaches of the CRB to quantify spatial variation of HZ aerobic/anaerobic respiration and reveal the dominant process controlling the spatial variation of HZ respiration in reaches with different sizes and land uses. Our modeling showed that aerobic respiration is a dominant component in total respiration at reach and sub-basin scales. At the reach scale, aerobic respiration accounted for about 98.2% of total respiration. At the sub-basin scale, a portion of aerobic respiration in total respiration ranged from 98.0% to 99.3%, and the CRB averaged portion was 98.7%. Among the nine sub-basins composing the CRB, the modeled respiration showed a high spatial variation of HZ respiration amounts at the reach and sub-basin scales. In general, streams/rivers receiving higher precipitation tended to have higher aerobic/anaerobic respiration. Mean annual precipitation can explain the difference of the sub-basin averaged aerobic and anaerobic respiration with correlation coefficients of R2 = 0.98 and 0.63, respectively.

We observed that channel sizes affected the spatial variation of modeled aerobic and anaerobic respiration; among the different-sized channels, medium-sized rivers generated the highest aerobic and anaerobic respiration. Land use also plays a critical role in spatial patterns of respiration. Among reaches with different land uses, reaches with forest lands showed the highest aerobic and anaerobic respiration. Forest reaches showed the lowest level of anaerobic respiration ratios (0.6%), while agricultural reaches showed the highest level (18%). This result suggests that the role of HZ respiration via anaerobic pathway should be incorporated in watershed-scale and regional-scale studies, especially with agricultural components. Hyporheic exchange flux can mostly explain spatial differences of aerobic and anaerobic respiration at the reach scale, while the stream DOC and DO concentrations have a relatively large effect on the anaerobic respiration. Because the modeled hyporheic exchange flux is heavily dependent on the empirical estimates of streambed grain size/hydraulic conductivity in the NEXSS modeling, reducing the uncertainty of estimating streambed hydraulic properties is essential to improve our modeling results. Also, accounting for the seasonal substrate availability across different stream types can improve the model estimates. While our RCM was built based on three microbial reactions, the reaction parameters were assumed to be spatially constant, and their estimates were based on local experimental data (Li et al., 2017) and literature values (Fang et al., 2020; Song et al., 2018). Therefore, the effect of spatial variation of the reaction parameters should be tested in future studies.

In summary, our modeling framework successfully quantified HZ respiration components over multiple scales. It revealed key mechanisms driving the spatial variation of HZ aerobic and anaerobic respiration in reaches with varying hydrologic and substrate conditions. Thus, our modeling study offers a testing hypothesis in different river systems (e.g., climate and biomes) for the HZ respiration processes, and can be used as a sampling design tool for large-scale HZ experimental studies.

Acknowledgments

This research was supported by the Department of Energy (DOE), Office of Science (SC) Biological and Environmental Research (BER) program, as part of BER's Environmental System Science (ESS) program. This contribution originates from the River Corridor Scientific Focus Area (SFA) at Pacific Northwest National Laboratory (PNNL). This research used resources from the National Energy Research Scientific Computing Center, a DOE-SC User Facility. PNNL is operated for DOE by Battelle Memorial Institute under contract DE-AC05-76RL01830. This paper describes objective technical results and analysis. Any subjective views or opinions that might be expressed in the paper do not necessarily represent the views of DOE or the United States Government. The authors thank an associate editor and two anonymous reviewers for providing helpful comments on a previous version of this manuscript.

    Conflict of Interest

    The authors declare no conflicts of interest relevant to this study.

    Data Availability Statement

    The model codes/scripts for this study will be made available on this PNNL Gitlab repository at https://gitlab.pnnl.gov/sbrsfa/hz-respiration, and key model inputs/outputs are freely available at https://doi.org/10.5281/zenodo.6954107.