of Spatio-temporal variations in lateral and atmospheric carbon ﬂuxes from the Danube Delta

Abstract. River deltas, with their mosaic of ponds, channels and seasonally inundated
areas, act as the last continental hot spots of carbon turnover along the
land–ocean aquatic continuum. There is increasing evidence for the important role of riparian wetlands in the transformation and emission of terrestrial carbon to the atmosphere. The considerable spatial heterogeneity of river deltas, however, forms a major obstacle for quantifying carbon emissions and their seasonality. The water chemistry in the river reaches is defined by the upstream catchment, whereas delta lakes and channels are dominated by local processes such as aquatic primary production, respiration or lateral exchange with the wetlands. In order to quantify carbon turnover and emissions in the complex mosaic of the Danube Delta, we conducted monthly field campaigns over 2 years at 19 sites spanning river reaches, channels and lakes. Here we report on the greenhouse gas fluxes (CO2 and CH4) from the freshwater systems of the Danube Delta and present the first seasonally resolved estimates of its freshwater carbon emissions to the atmosphere. Furthermore, we quantify the lateral carbon transport of the Danube River to the Black Sea. We estimate the delta's CO2 and CH4 emissions to be 65 GgC yr−1 (30–120 GgC yr−1, a range calculated using 25 to 75 percentiles of observed fluxes), of which about 8 % are released as CH4. The median CO2 fluxes from river branches, channels and lakes are 25, 93 and 5.8 mmol m−2 d−1, respectively. Median total CH4 fluxes amount to 0.42, 2.0 and 1.5 mmol m−2 d−1. While lakes do have the potential to act as CO2 sinks in summer, they are generally the largest emitters of CH4. Small channels showed the largest range in emissions, including a CO2 and CH4 hot spot sustained by adjacent wetlands. Thereby, the channels contribute disproportionately to the delta's emissions,
considering their limited surface area. In terms of lateral export, we
estimate the net total export (the sum of dissolved inorganic carbon, DIC, dissolved organic carbon, DOC, and particulate organic carbon, POC) from the Danube Delta to the Black Sea to be about 160 ± 280 GgC yr−1, which only marginally increases the carbon load from the upstream river catchment (8490 ± 240 GgC yr−1) by about 2 %. While this contribution from the delta seems small, deltaic carbon yield (45.6 gC m−2 yr−1; net export load/surface area) is about 4 times higher than the riverine carbon yield from the catchment (10.6 gC m−2 yr−1).



S1.1 Information on sampling stations
Table S1 Name, waterscape and geographical location of all investigated sampling stations. R indicates river stations, C are channel stations and L represent lake stations. Station 16 was sampled only once due to insufficient accessibility by motor boat and is therefore excluded from the dataset in this paper. a Due to the border with Ukraine, this station had to be shifted to a small side branch of the Chilia branch. Comparative measurements showed no difference from the main branch. b The Sulina station was shifted upstream end of February 2016, since a continuously measuring sensor indicated influence of the Black Sea water: * old location, ** new location S1.2 Evaluation of CH4 concentrations and fluxes S1.2.1 comparability of CH4 concentration measurements In October 2017, we conducted a comparison of CH4 measurement procedures using the GC and the Los Gatos using field samples from the Danube Delta. We calculated average values for the lab-based GC procedure (n=2) and the field-based LG procedure (n=3). Considering the standard deviation of the samples, only 2 samples deviate from the 1:1 line, however they 5 are still within the 10% measurement uncertainty of our GC system. Based on the results of this comparison, we deemed it appropriate to pool data acquired using the two different methods.

S1.2.2 evaluation of CH4 fluxes
The following examples document the evaluation of CH4 fluxes from discrete samples (Fig S2), which were measured using GC-FID (Aglient Technologies, USA). Continuous measurements (Fig. S3) in the headspace of the chamber were performed 10 with an Ultraportable CH4/N2O analyzer (Los Gatos Research, USA).

Figure 1 Average CH4 concentration measured with lab-based GC method (n=2) versus field-based
LG method (n=3). Error bars show the standard deviation, the orange line symbolizes the 1:1 line.

Figure S2
Flux measurements at location 13 highlighting the 3 different cases encountered during data evaluation. Dotted lines link individual measurements (x), solid lines show the calculated regression line. Green: linear fluxes with R 2 > 0.96, which we interpreted as diffusive CH4 fluxes. Yellow and orange: fluxes that had an R 2 < 0.96 and therefore were categorized to show signs of ebullition. Estimated fluxes were therefore reported as total fluxes. The diffusive contribution was estimated from CH4 water concentration and k600, which we derived from CO2 flux measured at the respective time and location. Dark red: non-monotonous flux, either by capturing a bubble or mislabelling a sample. Such timeseries were not included in the further analysis.

S2.2 Seasonality of carbon import and export
Import and Export loads varied seasonally, which was to a large extent driven by variations in discharge (Fig. S5).

S2.3 Visual results of Kruskal-Wallis Test
We visualized the results of the multiple comparison test after Dunn-Sidak based on statistical data obtained from the Kruskal-Wallis test using the multcompare function in Matlab. Most parameters are significantly different in the channels and lakes within the Danube delta compared to the river reaches outside. The time series of POC seems to be similar in lakes and the river, however closer inspection indicates that this coincidence may be caused by different POC sources, specifically the spring 5 peak during high sediment transport in the river and the summer peak during algal bloom in the delta lakes.
10 Figure S6 Visualization of results from Kruskal-Wallis non-parametric test for equal medians followed by Dunn-Sidak multiple comparison testing. Circles indicate the group mean, non-overlapping horizontal bars show significant difference: The channels and lakes in red are significantly different from the blue river group. Grey indicates no significant difference. The x-axis shows the average group ranks from the Kruskal-Wallis test.

S3.1 CO2 and CH4 fluxes upscaled for three different waterscapes
In the main text, Figure 5 illustrates the CO2 and CH4 fluxes from the different waterscapes to the atmosphere estimated from our monthly measurements. Below, Table S2 displays the numbers that are underlying the bar graphs in Figure 5, where the height of the bar indicates results based the median fluxes, while the minimum and maximum extension of the black lines 5 show the range resulting from calculations with 25 and 75 percentiles, respectively. S3.2 Respiratory CO2 production rate vs CO2 flux CO2 production rates were estimated from incubations for O2 community respiration and compared to measured CO2 fluxes for river (Fig. S7), channel (Fig. S8) and lake stations (Fig. S9). CO2 fluxes (FCO2) marked with a triangle were calculated using the median k600 from the measurements in the respective waterscape. Dark purple bars represent measured respiration rates (R), light purple bars indicate the effect of a correction with a factor of 2.7 for measurement limitations using BOD bottles 5 (RcorBOD) according to Ward et al. (2018). Station numbers refer to the sampling locations displayed in Figure 1 in the main text and documented further in Table S1.
While respiration rate exceeds CO2 flux in most cases in the River stations, the opposite is the case at many occasions in the channels, especially in the first half of the year until July. Respiration rates lower than CO2 flux indicate an additional CO2 source CO2 source. The channels are in close contact with adjacent reed beds, and lateral inflow from these wetlands could 10 explain the CO2 excess.

S3.3.1 Estimating the effect of potential hot spot channels
One of the channel stations (location 10, Fig. 1 main paper) showed very high CO2 and elevated CH4 concentrations during our observations. This location is draining a region with reed stands that are classified as strictly protected area by the Danube Delta Biosphere Reserve Authority (Fig. S10). The observation station itself is located in the buffer area surrounding the 5 strictly protected area. Considering these two characteristics, i) more or less west-east drainage and ii) connection to either strictly protected area or buffer area, we estimate the fraction of potential hotspot channels to up to 2 % (Fig. S10). This approach assumes the hotspot conditions prevail along the whole channel and not just in the sampled cross-section. Figure S10 Channel system of the Danube Delta. Potential hotspot channels are indicated in purple. 1) The channel at location 10, where we observed high CO2 concentrations. 2) EXO2 sensor data (not shown in this paper) indicated that this channel also has very 10 high CO2 concentrations. 3) and 4) channels within strictly protected area draining more or less west-east direction. Red and green shaded areas indicate core protection zones (red) and buffer areas (green) of the Danube Delta Biosphere Reserve between the three main branches of the Danube River. Protection zones redrawn from a map of the Danube Delta Biosphere Reserve Authority (2018), shape files for map creation in QGIS adapted from mapcruzin.com (Contains information from www.openstreetmap.org, which is made available here under the Open Database License (ODbL), https://opendatacommons.org/licenses/odbl/1.0/).

15
Assuming that 2 % of the channels' surface area are hotspots, the upscaling exercise suggests that the hotspots could contribute up to ~20 % to the CO2 and total CH4 fluxes from the channels (Fig. S11 and Table S3). However, the median flux for the channels decreased slightly in this example (cf. Table 2 Table   5 S3).

S3.3.2 Estimating the effect of potentially isolated lakes 10
There are several reports in the literature that refer to some of the delta lakes as "isolated", receiving water from the reed bed and having a long residence time compared to the highly connected lakes. We used the references in Tudorancea and Tudorancea (2006), Covaliov and Coops (2003), the assessment of lake regimes in Oosterberg et al. (2000) and a land use map indicating isolated lakes presented in a report by Doroftei et al. (2013) for our estimation of the surface area of these isolated lakes.

5
The total surface area of the lakes presented as isolated in the above-mentioned studies (see Fig. , & ) amounts to 99 km 2 .
Studies on very small lakes with surface areas < 0.2 km 2 were not present in the literature. While small lakes represent the dominant type of lakes in the delta (75% of the lakes), their surface area amounts to only 7% (17.8 km 2 ). These lakes are likely also isolated, yet we do not consider them in the following scenario analysis. Attributing hotspot like flux characteristics to the isolated lakes (A = 99 km 2 ), indicates that they could strongly increase both CO2 and CH4 fluxes from the lakes and also 10 reverse the potential sink capacity of the lakes in 2017 (Fig. 5 main text and Fig. ). It is however unlikely that all isolated lakes have fluxes as high as the hotspot, so the evaluation of this scenario represents an upper boundary to the potential fluxes from these poorly characterized, isolated systems.