Introduction

The Baltic Sea is suffering from severe anthropogenic nutrient input causing anoxic conditions (Kotta et al. 2020), with permanent anoxia in deep bottom waters and periodic hypoxia in shallower coastal regions (Conley et al. 2011). The focus of research over the past couple of decades has been to identify and to remediate the anoxic conditions caused by excess nitrogen and phosphorus run off (Kotta et al. 2020). However, hypoxia in the Baltic Sea persists, even if management efforts have successfully reduced the nutrient loading (Gustafsson et al. 2012; HELCOM 2021). This has led current research into investigating the effects of increased transport of terrestrial organic matter on coastal marine environments (Soares et al. 2019). The utilization of bioavailable carbon by bacteria, a process which requires oxygen, may contribute to a lower oxygen concentration in coastal areas (Dagg et al. 2008). Rivers flowing into the Baltic Sea supply organic matter and as a consequence 86% of the DOM in the Baltic Sea is thought to be terrestrially derived (Deutsch et al. 2012). The transported organic matter is made up of 45% organic carbon, with around 90% of the carbon existing as DOC and a small fraction as particulate organic matter (POC) (Maciejewska and Pempkowiak 2014).

The molecular composition of DOC depends on the source of carbon (Autio et al. 2016), which affects the reactivity and utilization by microbes (Reyes and Crisosto 2016). Autochthonous carbon which is produced by autotrophs within the system (Kritzberg et al. 2004) has a low molecular weight and aromaticity and is bioavailable to bacteria (Autio et al. 2016), but has previously been believed to be completely taken up in the riverine system before reaching coastal waters (Catalán et al. 2016). Allochthonous carbon which is carbon transported from the catchment area to the system (Kritzberg et al. 2004), was earlier believed to be refractory to microbial degradation, but several recent studies have shown microbial utilization of terrestrial derived organic carbon (Figueroa et al. 2016; Kritzberg et al. 2004; Kuliński et al. 2016). The quantity and composition of allochthonous carbon in rivers is affected by the characteristics and land-use of the catchment area (Agren et al. 2007; Reader et al. 2014), but also on precipitation, anthropogenic alterations such as clear cutting, and soil type (Sepp et al. 2019). In addition, composition is related to water residence time (WRT) with longer WRT allowing more time for photodegradation and microbial utilization, resulting in less bioavailable carbon (Autio et al. 2016). The source of allochthonous carbon determine the bioreactivity, for example peat land and boreal forest have been found to export carbon with high molecular weight and low bioavailability (Autio et al. 2016). Whereas agricultural land has shown to export carbon with a low molecular weight and high bioavailability (Autio et al. 2016).

Previous studies that focus on the transport of bioavailable carbon into the Baltic Sea have either investigated only a few sampling sites (Reader et al. 2014) or have sampled only during seasons with highest productivity (spring and autumn) (Asmala et al. 2013; Soares et al. 2019) and therefore lack an understanding of seasonal changes and how that affects the reactivity and loading (Autio et al. 2016). In this research we investigated how biological oxygen demand (BOD), which was converted to carbon utilization rate (R = ΔC/Δt (mg C L−1 d−1), and the bioavailable fraction (BFc) of riverine carbon (% of DOC utilized by bacteria within 24 h) is influenced by catchment composition and how it varies during different seasons. We hypothesize that carbon utilization rate and the BFc will be related to carbon composition and concentrations of nitrogen and phosphorus. Further we hypothesized that carbon utilization rate and the BFc peak in spring/early summer during highest terrestrial productivity. Lastly, we believe that rivers with agricultural dominated catchments contain a higher bioavailable fraction compared to forest dominated catchments. The results from this study provide insight into the seasonality of allochthonous carbon inputs and BOD relative to hypoxia in the Baltic Sea.

Materials and methods

Sampling and set-up

Sampling of nine rivers draining into the Baltic Sea was performed five times during 2020 (January, April, June, August, and October). At each time point 10 L of water were collected close to the river mouths of Emån, Alsterån, Ljungbyån, Lyckebyån, Mörrumsån, Helgeå, Skivarpsån, Kävlingeån, and Råån (Fig. 1). Rivers were selected to encompass the varying landscapes and land-use of catchment areas across Sweden (Table 1). Rivers 1–6 have catchment areas dominated by forest, while catchments of rivers 7–9 are dominated by agriculture land (Table 1; Fig. 1). The water from the rivers were filtered through type A/E glass fiber filters (142 mm, PALL Corporation) to remove particles and collected in clean 10 L bottles.

Fig. 1
figure 1

Map of the sampling locations, 1-Emån, 2-Alsterån, 3-Ljungbyån, 4-Lyckebyån, 5-Mörrumsån, 6-Helgeå, 7-Skivarpsån, 8-Kävlingeån, 9-Råån

Full size image
Table 1 Sampling location and composition of catchment area for each river
Full size table

Nutrients and DOC

Nutrient and DOC samples were subsampled from the 10-liter bottles and collected in acid-washed polyethylene bottles (50 mL and 100 mL respectively) and then frozen. For samples collected in January, nitrogen concentrations were processed at Denmark Technical University using standard protocols (Hansen 1999) on a SmartChem 2000 discrete analyzer (AMS Alliance). Nitrogen concentrations in April, June, August, and October, as well as all DOC samples were analyzed by the department of Biology, Lund University, using a TOC V-CPH/CPN (Shimadzu) with the method Non-Purgeable Organic Carbon (NPOC). Total phosphorus was analyzed manually using MRP-method (Menzel & Corwin 1965).

Carbon utilization rate and bioavailable fraction

Biological oxygen demand (BOD) was measured and converted to carbon utilization rate. Water from each river was used to start triplicate incubations in 2 L glass bottles. The bottles were filled carefully to avoid bubble formation and to remove all the headspace and then sealed with gas tight stoppers. In addition to the nine rivers, three controls with milliQ water were also included, resulting in a set-up with 30 incubation bottles. The incubation bottles were incubated in darkness for 5 days in a temperature controlled room at 20 °C (Wilhelm 2009). Oxygen (mg/L) was measured every 30 min using optical oxygen sensors (optodes) and three 10-channel fiber oxygen transmitters (oxy-10, PreSens®). It took a few hours for the system to reach 20 °C. The first 15 h were therefore removed in the data analysis. However, in January stabilization took longer and the first 24 h were removed. The slope for oxygen consumption was calculated using excel (R = ΔO2/Δt (mg O2 L−1 d−1), Fig. S1). The oxygen consumption was converted to carbon units using a molar respiratory quotient of 1.0 (Soares et al. 2019) which is close to estimates for rivers (Berggren et al. 2012). BFc was calculated by dividing the carbon utilized in 24 h by the DOC concentration.

Absorbance

A subsample of the filtered water from each river was used for absorbance measurements and to calculate absorbance ratios for carbon size, watercolor, and aromaticity as well as specific ultraviolet absorbance (SUVA) (De Haan HaDB 1987; Joo et al. 2008; Li et al. 2016; Reyes and Crisosto 2016; Uyguner and Bekbolet 2005). Absorbance was measured every 1 nm from 200 to 800 nm on a UV-2600, UV–VIS spectrophotometer (Shimadzu) using a 1 cm quartz cuvette and with milli-Q water as a reference and a blank. The ratio a250/a365 provides an index of the size of the carbon molecules, where a low value indicates larger molecules, and a high value indicates smaller molecules (Reyes and Crisosto 2016). Watercolor was calculated based on a420 (Pennanen 1986). Specific UV-absorbance (SUVA (m2 g−1 C) was calculated as the ratio between a254 nm (m−1) and total organic carbon (mg L−1) (Goodman et al. 2011). SUVA is used as an indicator of the aromatic content of DOC and provides information about the source and reactivity of the carbon (Goodman et al. 2011). The high SUVA value is related to high aromatic content (Goodman et al. 2011). Aromaticity of carbon molecules is also described by the ratio a254/a204, with a high value indicating high aromaticity (Reyes and Crisosto 2016). This absorbance ratio provides understanding of the composition of carbon but in contrast to SUVA it does not account for total organic carbon concentration.

Statistics

All statistical analysis were performed in Prism graphpad 9.2.0. One-way anova was used to compare carbon utilization rate and bioavailable fraction between sampling months for each river. Rivers were grouped into forest dominated and agricultural dominated catchment areas based on land use data from SMHI (Table 1). Unpaired t tests were used to compare carbon utilization rate and bioavailable fraction between rivers with catchment areas dominated by agriculture or forest. Multiple linear regressions were used to investigate which variables (DOC, carbon composition, concentration of total nitrogen (TN) and concentration of total phosphorus (TP)) explain the variation in carbon utilization rate and the BFc. However, the results in the manuscript are based on single regressions because the multiple linear regressions did not yield any significant results.

Results

Seasonal and spatial variations in nutrient concentrations and carbon composition

Concentration of DOC, total N and total P

Concentrations of DOC, TN and TP varied both seasonally and between rivers with catchment areas dominated by agriculture and forest, respectively. The highest DOC concentrations were found in April, June, or October in rivers with catchment areas dominated by forest, whereas rivers with catchment area dominated by agricultural land had highest concentrations of DOC in August or October. Rivers with forested catchments generally had lowest DOC concentrations in August. The only exception was Alsterån which had the lowest concentration in January (Table 2). In January and April DOC was generally higher in rivers with forest dominated catchments (14.9–27.7 mg/L) than in rivers with agricultural dominated catchment areas (5.5–9.5 mg/L) (Table 2, t test, p = 0.0013 (January), p = 0.0018 (April)). The TN concentration was significantly higher in April compared to other months in all rivers (ANOVA, p < 0.0001). On the other hand, TP was highest in June or August (Table 2). In all months except January and April, TP concentrations were higher in rivers with agricultural dominated catchments than in rivers with forest dominated catchments (ANOVA, p < 0.001, Table 2).

Table 2 Carbon composition and nutrient concentrations at each sampling location over the year
Full size table

Carbon composition

Carbon composition also varied both seasonally and between rivers with catchment areas dominated by agriculture or forest. In rivers with catchment areas dominated by forest, the values of the ratio a250/a365 (molecular size) was generally lowest in January and highest in August (Table 2). However, for rivers with catchment area dominated by agriculture no clear pattern was found (Table 2). Values of the ratio a254/a204 (aromatic content) was generally lower in rivers with agriculture dominated catchment areas than in rivers with forest dominated catchment areas (Table 2). Watercolor (a420) was significantly lower in rivers with agriculturally dominated catchment areas over the whole year compared to rivers with forest dominated catchment areas (Table 2, Agriculture, 20–92.5 mg Pt/L, Forested, 30–285 mg Pt/L, t test, p < 0.001). In addition, the seasonal variations in watercolor were lower in rivers with agricultural dominated catchment areas and the lowest values are detected in April. In contrast, watercolor in forest dominated rivers was lowest in August or October Highest values were detected in January or April for all rivers with forest dominated catchment areas except Alsterån where watercolor was highest in June (Table 2). SUVA fluctuated over the year with highest and lowest values found in different months for the different rivers (Table 2).

Comparison between rivers with forest and agricultural dominated catchments

To investigate the effect of catchment area on carbon utilization rate and BFc, the rivers were divided into two groups based on dominating land use activity in the catchment area (Table 1). There was a significant difference in carbon utilization rate between rivers with forest dominated catchment areas and agricultural dominated catchment areas in August (t test, p = 0.04, Fig. 2), and at this timepoint the carbon utilization rate was highest in rivers with catchment areas dominated by agricultural land. However, the BFc was significantly higher in agricultural dominated catchments than forested dominated catchments in January (t test, p = 0.016), April (t test, p = 0.001), June (t test, p = 0.007) and August (t test, p = 0.009, Fig. 2).

Fig. 2
figure 2

Carbon utilization rate (A) and the bioavailable fraction (B) of rivers with forest dominated catchments and agricultural dominated catchments during different sampling months. *Indicate significant difference between forest and agriculture dominated catchment areas

Full size image

Seasonal variation in carbon utilization rate and bioavailable fraction

Carbon utilization rate and the BFc varied seasonally. In rivers with forest dominated catchment areas carbon utilization rate and the BFc was highest during spring or early summer (April and June) as well as in October (Fig. 3). Ljungbyån showed the largest seasonal variations for carbon utilization rate and the BFc, with values comparable to the other rivers for January - August (2.23–6.85 µmol C L−1 Day −1, 0.13–0.45% BFc) but significantly higher in October (31.97 µmol C L−1 Day −1, 2.45% BFc)(October carbon utilization rate vs. January, April, June, and August carbon utilization rate, ANOVA, p < 0.01) (October BFc vs. January, April, June, and August BFc, ANOVA, p < 0.01, Fig. 3). All forest dominated rivers had lowest values in January and August, except Mörrumsån where the BFc was lowest in January and June (Fig. 3).

Fig. 3
figure 3

Carbon utilization rate (blue) and bioavailable fraction (red) of the 9 rivers over the year (**Axis is doubled compared to other graphs)

Full size image

Skivarpsån showed the same pattern as the rivers with forest dominated catchment area, where carbon utilization rate and the BFc increased from January to June, decreased in August, then increased again in October (Fig. 3). No significant differences were found for the carbon utilization rate, but the BFc yielded significant differences between January and June respectively June and August (ANOVA, p = 0.044 and 0.022, Fig. 3). Carbon utilization rate and BFc was stable throughout the year in Kävlingeån whereas in Råån carbon utilization rate and the BFc increased from January through August and then declined in October (Fig. 3). In Råån, carbon utilization rate was significantly higher in August compared to January, April, June, and October (ANOVA, p < 0.05, Fig. 3).

Seasonal variations in factors explaining total utilization and bioavailable fraction

Linear regressions were used to investigate which factors explain the variation in carbon utilization rate and the BFc. Based on data from all rivers and all months, there were no significant factors explaining the carbon utilization rate or BFc (Table S1).

Examining each month separately, different factors significantly explained differences in carbon utilization rate and the BFc during different months. Watercolor and SUVA were positively related to the carbon utilization rate in April whereas in August phosphorus concentration was positively related to carbon utilization rate (Table 3). Focusing on the BFc, there were positive relations to nitrogen concentration and a250/a365, and negative relations to DOC concentration, watercolor, a254/a204, and SUVA in April (Table 3). In June phosphorus was positively related, while DOC and a254/a204 were negatively related (Table 3). In August nitrogen concentration is positively related to BFc (Table 3).

Table 3 Results from linear regressions (Regression coefficient, p-value, and R2-value) investigating which factors explain the variation in carbon utilization rate (µmol C L−1 Day −1) and the bioavailable fraction (%) at each separate month. p-values for significant relations are shown in bold and indicated by *
Full size table

Discussion

Carbon transport from terrestrial environments to aquatic systems has been a major player in coastal hypoxia over the years (Conley et al. 2011; Reader et al. 2014). This topic has been examined in many studies, but these primarily focused on bacterial utilization rate during spring and autumn (Ågren et al. 2008; Autio et al. 2016). Few studies have explored the complex nature of carbon utilization and the BFc over the year, which would further expand our understanding of how seasonality effects transport. In addition, there is a gap in understanding how catchment composition influences carbon utilization rate and the BFc.

Comparison between rivers with forest and agricultural dominated catchment areas

Catchment composition plays an important role when it comes to the quality of carbon that is leached into riverine systems (Sepp et al. 2019). Boreal forest, when compared to agricultural dominated catchments, leads to loadings of DOM that are higher in humic content and are a major driver for watercolor (Reader et al. 2014; Sepp et al. 2019). In our study, the 3 rivers with agricultural dominated catchments had significantly lower watercolor throughout the year compared to rivers with forest dominated catchments (t test, p < 0.0001, Table 2). Similarly, in these rivers a254/a204 values were consistently lower (Table 2). Data on watercolor and a254/a204 therefore suggest that carbon exported from agricultural catchments is more bioavailable (Reader et al. 2014). This agrees with a previous study showing that agricultural catchments in Europe export greater amounts of reactive carbon compared to boreal forested catchments (Autio et al. 2016). In addition, studies have shown that high TP and TN concentrations close to river mouths lead to a higher bioreactivity due to primary production being mostly P limited (Soares et al. 2019). This phosphorus allows for increased primary production which leads to a release of labile carbon (Soares et al. 2019). Our study found elevated nutrient concentrations for the three agricultural dominated rivers compared to the rivers with forested catchments, most likely caused by lowland agriculture and urban environments leaching nutrients (Soares et al. 2019). Based on carbon composition and nutrient concentrations we therefore expected higher carbon utilization rates in rivers with catchment area dominated by agricultural activity compared to forest.

However, we only found a significant difference in carbon utilization rate in August, and at this time point carbon utilization rate was highest in rivers with agricultural catchments (Fig. 2). In contrast, the BFc was significantly higher in rivers with catchment areas dominated by agriculture at all timepoints except October. One reason for the lack of difference in carbon utilization rate at most time points is probably related to rivers with forested catchments having higher DOC concentrations but a lower BFc compared to rivers with an agricultural dominated catchment. This is in agreement with a previous study showing that bioavailability and carbon utilization rate depend on both the quality and quantity of carbon (Berggren et al. 2018).

Seasonal variation in utilization rate and bioavailable fraction

This study sheds light on the seasonal variation in carbon transport and shows that carbon utilization rate and the BFc increased from winter to spring followed by a decrease in August (Fig. 1). This study is in general in agreement with a previous study which found that carbon utilization rate and the BFc are lowest during the summer and are higher during spring and autumn (Reader et al. 2014). The low carbon utilization rate and BFc during winter is most likely due to low water temperatures, hindering bacterial processes and primary production (Scavia and Laird 1987; Wetzel 2001). In agreement it has previously been described that low winter temperatures preserve organic matter in the catchment due to low metabolic rates and a low flow of water through soils (Reader et al. 2014). During spring when the soils thaw, carbon utilization rate and the BFc increases which is related to increased carbon mobility as well as bacterial respiration.

Forested catchments showed a significant decrease in carbon utilization rate and BFc between June and August, while agricultural catchments had a more stable carbon utilization rate and BFc (Fig. 2). The decrease in carbon utilization has previously been explained by that many of these rivers experience a relatively dry summer which decreases loading to the aquatic systems (Reader et al. 2014). However, in our study this is probably related to low flow rates during the summer months (Table S2). Agricultural catchments are not as effected by less loading possibly due to a majority of the carbon being produced within the system by primary producers (Soares et al. 2019). In conclusion we found that forested catchments transport more DOC with a lower BFc, while agriculture catchments transport less DOC with a higher BFc.

Rivers with forested catchments showed large variability for both carbon utilization rate and BFc in October. Ljungbyån, Lyckebyån, and Mörrumsån had their highest carbon utilization rate and BFc during October, and for Lyckebyån and Mörrumsån the carbon utilization rate was twice as high compared to other months (Fig. 2). Possible explanations could be the increase in DOC concentration and a decrease in SUVA and/or a254/a204 between August and October (Table 2), suggesting loading of more reactive carbon. Another possible explanation could be increased loading from precipitation. In Ljungbyån there was 62.4 mm of rainfall in October (Table S3) which is 26 mm higher than the historical average (SMHI 2021). Mörrumsån experienced 87.3 mm (Table S3) of rainfall which is 14 mm higher than the historical averages (SMHI 2021). Similarly, it has previously been shown that forested catchments export bacterial productive carbon during flooding (Ågren et al. 2008) resulting in a delivery of reactive material increasing the BOD (Reader et al. 2014). Our results show different seasonal loading patterns for rivers with forested catchments and agricultural catchments, understanding these seasonal fluctuations as well as the driving factors are important for future management efforts.

Seasonal variations in explanatory factors for carbon utilization rate and the bioavailable fraction

Carbon utilization rate has previously been shown to be correlated with both quantity and quality of carbon (Kritzberg et al. 2004) as well as nutrient concentrations (Soares et al. 2019). In our study none of the included parameters (DOC concentration, carbon composition or nutrient concentration) could explain the carbon utilization rate or the BFc in all rivers over the whole year (Table S1). We therefore analyzed the data for carbon utilization rate and BFc for each sampling time separately and found that different factors were important for carbon utilization rate and BFc during different seasons (Table 3). In April carbon utilization rate was positively correlated with both watercolor and SUVA. Generally, more aromatic carbon does not increase bacterial utilization since more complex carbon has been shown to be difficult for bacteria to degrade (Kritzberg et al. 2004). However, bacteria can produce ectoenzymes to break down recalcitrant DOC into sources of energy and nutrients that are more easily utilized (Chróst 1990; Kritzberg et al. 2004). If only a small fraction of the carbon is labile this might be an important strategy to obtain nutrients and carbon. Although DOC concentration did not significantly correlate with carbon utilization rate (Table 3), we found that DOC is positively correlated with watercolor and SUVA (p < 0.0001, p 0.0491). This indicates that the DOC concentration and not the composition is important for the carbon utilization rate. In August, nitrogen concentrations have a significant positive effect on carbon utilization rate (Table 3). A potential explanation for the correlation with nitrogen could be the rivers being nitrogen limited, although other studies have shown that increased phosphorus concentrations resulted in higher rates of microbial growth (Soares et al. 2019). However, the fraction of nitrogen and phosphorus that is bioavailable to bacterioplankton varies over space and between seasons (Muscarella et al. 2014; Soares et al. 2017; Stepanauskas et al. 1999). It has been shown that for dissolved organic nitrogen (DON) anywhere from 2 to 75% (Pellerin et al. 2006) of the nitrogen pool was biologically available whereas for phosphorus the estimates vary from less than 5 to 50% of the TP pool.

Most of the rivers contained their lowest BFc in January whereas the highest BFc was found in June, August, or October and the variation was significantly explained by several factors. Carbon composition has previously been shown to affect carbon utilization rate and the BFc (Berggren et al. 2018). In April, the BFc was negatively correlated to DOC concentration, watercolor (a420), a254/a204 (aromatic content), and SUVA and positively related to a250/a365 (molecular size), whereas in June BFc was negatively correlated to a254/a204 (aromatic content). In general, these correlations show that BFc is lower when the carbon has a higher molecular complexity (negative correlations to watercolor, aromatic content, and SUVA) which is in agreement with that carbon becomes more bioavailable as colored DOM (CDOM) is broken down by bacteria and photodegradation (Goodman et al. 2011).

The BFc was significantly related to nutrient concentration. In April and August there was a positive correlation to TN and in June to TP. During April and June this seems to be related to the loading of nutrients; nitrogen concentrations increased between January and April followed by an increase in phosphorus concentration between April and June (Table 2). As mentioned above the variation in which nutrient is correlated to the BFc might be related to seasonal variations in the bioavailability of the nitrogen and phosphorus pools (Pellerin et al. 2006).

To summarize, the results from the linear regressions show that carbon quantity and nutrient concentrations seem most important for carbon utilization rate whereas carbon quality and nutrient concentrations are more important for BFc.

Conclusion and implications

Our data provides insight to the complex nature of carbon transport and how carbon composition and nutrient concentrations are driving carbon utilization rate and BFc. We have found that while rivers with agricultural catchments have a lower DOC concentration, they have a higher BFc. Contrary, rivers with forested catchments have a higher DOC concentration containing a lower BFc. However, the discharge of bioavailable carbon will be higher from rivers with forested catchments since the flow rate is generally higher than for rivers with agricultural dominated catchments. Knowing which form is more detrimental when it reaches coastal zones is difficult and more studies are needed to investigate the fate of carbon. Climate change, with heavier and more frequent precipitation events, will undoubtedly increase carbon transport and thereby further lower the status of coastal water. A recommendation for management would be to focus efforts during spring and autumn when carbon utilization rate and BFc is highest. A potential way of doing this could be by increasing water retention time with the implementation of ponds and wetlands along the river continuum.