Oct 20, 2016 - 1Department of Physical Geography, Stockholm University, ..... tored during the ice-free season. .... and second-order streams (in this study we used Strahler stream ..... Model performance was quan tifi ed with the. Nash-Su tcliffe effi cie ... TOC and CO2 were significant in six of the study streams (C1, C3,.
PUBLICATIONS Journal of Geophysical Research: Biogeosciences RESEARCH ARTICLE 10.1002/2016JG003420
Decoupling of carbon dioxide and dissolved organic carbon in boreal headwater streams
Key Points: • We observed large losses of CO2 in a network of small boreal streams • DOC mineralization was a small source of stream CO2 • The main source of CO2 was supersaturated groundwater
Mattias Winterdahl1,2, Marcus B. Wallin3, Reinert Huseby Karlsen3, Hjalmar Laudon4, Mats Öquist4, and Steve W. Lyon1,2
Supporting Information: • Supporting Information S1 • Table S1
Abstract Streams and rivers emit large quantities of carbon dioxide (CO2) to the atmosphere. The sources of this CO2 are in-stream mineralization of organic carbon (OC) and CO2 input via groundwater inﬂow, but their relative importance is largely unknown. In this study, we quantiﬁed the role of in-stream OC mineralization as a source of CO2 in a number of nested boreal headwater streams. The results showed that mineralization of stream OC contributed 3% of CO2 supersaturation at time scales comparable to the estimated water travel times in the streams ( estimated concentrations) or losses (observed concentrations < estimated concentrations). To be able to assess whether relationships between observed and estimated concentrations in C7, C12, and C16 were signiﬁcantly different from a 1:1 line, we compared slopes of regressions to a 1:1 line by using the Student’s t test (Text S1). If differences between slopes were insigniﬁcant, we also compared intercepts of regressions to 0 (the intercept of the 1:1 line) with the Student’s t test. We also used the one-sample Student’s t test to test if differences between estimated and observed concentrations as well as groundwater inﬂows were signiﬁcantly different from zero. The signiﬁcance level for all tests was 0.05.
3.1. Stream Carbon Concentrations
Stream TOC concentrations were consistently higher than CO2 concentrations, although concentrations varied considerably among streams (Table 2 and Text S4). Streams were supersaturated with
Journal of Geophysical Research: Biogeosciences
CO2 for more than 99.5% of all observations, and median stream partial CO2 pressure was 2300 μatm. CO2 as percent of DIC varied between 99% and 48% due to differences in pH. Both CO2 and TOC showed consistent downstream decreases in concentrations, whereas pH generally increased concurrently (Figure S2 in the supporting information). A MLR model with mire coverage, catchment area, and tree volume explained most of the spatial variability in TOC concentrations (arcsine transformed mire coverage and log-transformed catchment area and tree volume; R2 = 0.87; n = 17). The spatial variability in median CO2 concentrations was not as well predicted by landscape variables, but it was correlated to mire coverage (R2 = 0.31; n = 17). However, the best performing MLR model included catchment area and till coverage and explained 64% of the variability in CO2 concentrations (R2 = 0.64; n = 17). Spatial variability in mean pH was most strongly correlated with catchment coverage of sorted sediments (arcsine-transformed percentage of sorted sediments; R2 = 0.54; n = 17). Combining catchment coverage of sorted sediments, catchment area, and mire coverage in a MLR model explained most of the spatial variability in mean pH (R2 = 0.84; n = 17). 3.2. Stream TOC-CO2 Correlations Log-transformed TOC (logTOC) and CO2 (logCO2) concentrations were signiﬁcantly (p < 0.05) correlated in 10 of the 17 studied streams. The correlations were, however, generally weak as median R2 for all signiﬁcant correlations was 0.04 (R2 range: 0.03–0.31). In the 10 streams where correlations were signiﬁcant, there was a positive correlation between TOC and CO2 in 4 of the streams, while the correlation was negative in the rest. The logarithm of discharge (logQ) was signiﬁcantly correlated to logTOC in 14 of the 17 streams. The only sites where correlations were nonsigniﬁcant were C5, C7, and C22. Median R2 was 0.21 where correlations were signiﬁcant (R2 range: 0.06–0.56). There were positive correlations between logQ and logTOC in all streams except C3 and C4, where catchments are dominated by mires. In nine streams logQ was signiﬁcantly correlated to logCO2. The median R2 was 0.48 with a range of 0.04–0.71. Where correlations were signiﬁcant, logQ was negatively correlated to logCO2. Correlations between discharge-corrected TOC and CO2 were signiﬁcant in six of the study streams (C1, C3, C4, C5, C14, and C20). The correlations were positive in all of these streams except in C1. In streams with signiﬁcant correlations, the median R2 was 0.13 (R2 range: 0.05–0.20). 3.3. Amount of TOC Mineralization Required to Sustain CO2 Supersaturation The median CO2 excess was 6 ± 8% (median ± IQR) of observed TOC concentrations. Median CO2 excess as percent of TOC concentration varied between 2 and 15% among streams. However, the temporal variability was substantially larger (Table 3). The percentage of TOC that would be needed to be mineralized to sustain CO2 excess in C20, for example, varied between 4 and 146%. Two of the streams had periods when CO2 excess exceeded 100% of TOC concentrations. These observations occurred at low-ﬂow conditions when TOC concentrations were low and CO2 concentrations were high. CO2 excess was generally related to both discharge and TOC concentration, although the strength and direction varied among streams. Based on MLRs between CO2 excess, logTOC, and logQ, discharge was a signiﬁcant predictor of CO2 excess in all streams except C1, C5, C20, C21, and C22. In streams where discharge was a signiﬁcant predictor, it was negatively correlated to CO2 excess in all streams except C7. TOC was a signiﬁcant positive predictor in eight streams: C3, C4, C5, C6, C10, C14, C15, and C20. Water travel times in the streams were generally short with a median travel time of 8 ± 11 h (Table 2). The maximum median travel time was found for C13 and was 19 h. However, the distribution of travel times was strongly skewed resulting in the 90th percentile of travel times being 18 ± 25 h. The 90th percentile of travel times was between 17 and 83 h in the larger streams (third and fourth order). By using previously published degradation rates and estimated water travel times in the streams, we found that on average 3 ± 5% of the CO2 excess could potentially originate from TOC mineralization. Although the estimated TOC mineralization varied among streams, median mineralization was below 10% of the CO2 excess in all streams. Furthermore, the mineralization was less than 18% in 97.5% of all observations. There was a downstream relationship of increasing estimated TOC mineralization as median TOC mineralization was positively correlated to both log-transformed stream length (R2 = 0.70; n = 17) and catchment area (R2 = 0.23; n = 17).
WINTERDAHL ET AL.
DECOUPLING OF CO2 AND DOC IN STREAMS
Journal of Geophysical Research: Biogeosciences
Table 3. Estimated Median CO2 Excess and TOC Mineralization in the Studied Streams Catchment C1 C2 C3 C4 C5 C6 C7 C9 C10 C12 C13 C14 C15 C16 C20 C21 C22
CO2 Excess (mg C/L)
CO2 Excess (% of TOC)
TOC Mineralization (mg C/L)
TOC Mineralization (% of CO2 Excess)
0.4 (0.0–4.4) 1.7 (0.4–13.2) 4.3 (1.2–12.2) 3.9 (0.9–11.7) 2.0 (0.3–13.4) 0.8 (0.3–2.7) 0.8 (0.2–5.4) 0.8 (0.4–3.2) 0.9 (0.3–3.7) 0.5 (0.0–3.0) 1.5 (0.4–7.9) 0.9 (0.2–4.1) 0.4 (0.3–1.5) 0.6 (0.1–2.5) 1.3 (0.3–4.9) 0.9 (0.3–3.8) 1.8 (0.5–9.6)
2.3 (0.1–36.0) 9.2 (1.2–90.0) 11.4 (2.8–72.6) 12.1 (5.1–64.5) 8.3 (1.2–115.1) 4.3 (1.4–33.7) 3.5 (1.0–24.3) 4.7 (2.0–28.8) 4.9 (2.2–38.4) 2.7 (0.2–28.6) 8.4 (2.0–80.0) 7.5 (1.3–34.4) 3.1 (1.9–16.5) 5.4 (1.4–45.3) 14.7 (3.8–145.5) 6.1 (2.1–26.8) 10.1 (3.1–51.8)
0.1 (0.0–0.3) 0.1 (0.0–0.4) 0.3 (0.0–0.6) 0.2 (0.1–0.4) 0.1 (0.1–0.2) 0.1 (0.0–0.2) 0.1 (0.0–0.3) 0.1 (0.1–0.2) 0.1 (0.0–0.3) 0.1 (0.0–0.5) 0.1 (0.0–0.3) 0.1 (0.0–0.2) 0.1 (0.0–0.2) 0.1 (0.0–0.2) 0.1 (0.0–0.3) 0.1 (0.0–0.3) 0.1 (0.1–0.2)
9.1 (0.6–36.5) 3.0 (0.5–18.1) 0.5 (0.1–1.5) 0.6 (0.1–3.2) 0.1 (0.0–0.8) 3.0 (0.8–9.6) 4.5 (23.3–30.0) 4.1 (6.5–48.4) 3.7 (1.0–24.9) 6.5 (1.0–279.1) 4.6 (0.8–20.7) 5.5 (1.4–73.7) 7.5 (5.6–75.3) 3.2 (1.2–24.1) 0.6 (3.8–7.0) 0.8 (0.1–3.7) 2.5 (0.5–21.5)
Values are the medians with minimum and maximum values in parentheses. Negative CO2 excess indicates undersaturation, i.e., stream concentrations lower than what would be expected if the water was in equilibrium with the atmosphere.
3.4. Mass Balance Calculations 3.4.1. Site C7 The estimated TOC concentrations from the mass balance in the second-order stream C7 were slightly higher than observed concentrations (Table 4). Further, estimated TOC concentrations were strongly correlated to observed concentrations, and there was no signiﬁcant difference between the regression slope of this relationship and the slope of the 1:1 line (Figure 2 and Table 4; for a residual analysis of the regression, see Text S2). There was, however, a signiﬁcant difference in intercept (p = 0.02; t = 2.33; d.f. = 126). The estimated median TOC loss along the streams was 0.8%, which was signiﬁcantly different from 0 (p < 0.0001; t = 4.30; d. f. = 127). The calculations also indicated that a signiﬁcant portion (p < 0.0001; t = 15.06; d.f. = 122) of the CO2 was lost downstream (Figure 3). On average 60% of the CO2 from C2 and C4 were lost before reaching C7 with a maximum loss of 96%. Consequently, estimated concentrations were higher than observed concentrations (Table 4). In addition, the correlation between observed and estimated CO2 concentrations was weak; a
Table 4. Results of Mass Balances of TOC, CO2, and Si
TOC Site C7 C12 C16
Observed TOC (mg/L) 21.6 ± 8.5 16.7 ± 7.6 10.5 ± 6.0
Estimated TOC (mg/L) 23.1 ± 10.2 20.7 ± 8.2 13.4 ± 4.9
n 128 94 113