Reviewer 1

We would like to thank Reviewer 1 for the comprehensive analysis of our work, pointing out the quality of the manuscript while identifying important ways to improve its form and content. We are glad to address all of Reviewer1’s concerns.

We agree with this statement. We will better introduce and integrate the bedrock potential origin of POM, and its influence on the refractory terrestrial source, given isotopic and elemental signatures. Effectively, some soil and rock-derived POM may be distinguished by their δ¹³C or C/N values.

However, studies focused on petrogenic POM that measured directly this source showed a great variability of C/N ratio depending on the geology (at Taiwan island scale, C/N ratio ranges from 3 for sandstone to 17 for turbidites, and from 3.8 to 9.7 for rock-derived river sediment POM; Liu et al. 2017; Hilton et al. 2010).

In addition, the soil POM C/N ratio is correlated to the mineral-associated POM C/N ratio, and can also span a great range of values (Europe scale, 2 to 100 vs. 2 to 30, respectively; Cotrufo et al. 2019).

Finally, as rock-derived OM dominates the riverbed sediment OM in major rivers like the Amazon River (Sun et al. 2017) or the Rhône River (Harmelin-Vivien et al. 2010), and as riverbed sediment is subject to resuspension, we are keen to better address these geological pathways resulting in complex refractory terrestrial suspended POM sources.

Cotrufo, M.F., Ranalli, M.G., Haddix, M.L. et al. Soil carbon storage informed by particulate and mineral-associated organic matter. Nat. Geosci. 12, 989–994 (2019). https://doi.org/10.1038/s41561-019-0484-6

Harmelin-Vivien, M., Dierking, J., Bănaru, D. et al. Seasonal variation in stable C and N isotope ratios of the Rhone River inputs to the Mediterranean Sea (2004–2005). Biogeochemistry 100, 139–150 (2010). https://doi.org/10.1007/s10533-010-9411-z

Robert G. Hilton, Albert Galy, Niels Hovius, Ming-Jame Horng, Hongey Chen, The isotopic composition of particulate organic carbon in mountain rivers of Taiwan, Geochimica et Cosmochimica Acta, Volume 74, Issue 11, 2010, Pages 3164-3181, ISSN 0016-7037, https://doi.org/10.1016/j.gca.2010.03.004.

Liu, K. K., and S. J. Kao, 2007: A three end-member mixing model based on isotopic composition and elemental ratio. Terr. Atmos. Ocean. Sci., 18, 1067-1075, doi: 10.3319/TAO.2007.18.5.1067(Oc).

Sun, S., Schefuß, E., Mulitza, S., Chiessi, C. M., Sawakuchi, A. O., Zabel, M., Baker, P. A., Hefter, J., and Mollenhauer, G.: Origin and processing of terrestrial organic carbon in the Amazon system: lignin phenols in river, shelf, and fan sediments, Biogeosciences, 14, 2495–2512, https://doi.org/10.5194/bg-14-2495-2017, 2017.

 Indeed, our data acquisition strategy of 1-year river samplings does not allow for capturing the inter-annual hydrological dynamics of the rivers and is not representative of their overall hydrological dynamics. This was in fact not our aim. Our aim, however, was to describe 23 sets of POM sources dynamics and compare them with 23 sets of environmental forcings to establish a typology of POM dynamics. For this reason, we do not pretend to bind a typology to these particular 23 rivers, but rather bind POM sources dynamics with typical settings of temperate environmental forcings. This would imply that a river may belong to a type in a given year but to another type in another year.

For performing the fuzzy clustering in the manuscript (Fig. 5), we used, when a river was sampled during two or more years, only one of the sampled years. The idea was not to over-represent these rivers compared to one-year sampled rivers in the analysis. To check 1) if river type depends on sampling year and 2) for a possible bias between 1-year sampled rivers and multi-year sampled rivers due to over-representation of the latter, we performed another fuzzy clustering (same caption as Fig. 5), using all the sampling years. It results that the river typology is very similar to that of Fig. 5 and that each of the rivers that were sampled for more than a year belongs to the same type, whatever the sampling year, except the Leyre River. The Leyre River switched from Type III in 2008 to Type I in 2009. Note that all the southwestern rivers belong to types I and III. The Leyre River is a southwestern river.

(See the figure in Supplement)

This will be better explained in the manuscript, and thanks to this comment, we will to better state the scope of our work.

We understand Reviewer 1’s concern about the risk of representing the eight spatially closed little rivers. To test this potential bias, we performed again the fuzzy clustering, removing seven of these eight rivers (keeping only the Milieu River, representative of type I). It results that the typology is very similar to that of Fig. 5 and that Type I still stands with the Seudre and Milieu Rivers that still belong to this type. This indicates that this clustering is a robust analysis for performing such a typology because it is poorly dependent on sampling year and rivers, at least at the time and space scale of the study.  This will be explained in section 2.6 of the revised manuscript.

(see the figure in Supplement)

We thank Reviewer 1 for the wise tips on new parameters to take into account. We will acquire the necessary data and add these parameters to perform RDAs (slope and river connectivity).

Indeed, all metrics used to perform multivariate analyses need to be justified, here as translators of potential environmental forcings to the POM dynamics.

In the case of river-specific RDAs, environmental forcing dynamics may explain local seasonal dynamics, as seasonal meteorological or hydrological dynamics (e.g., high wind energy may be linked to resuspension processes). They also served to assess the robustness of the mixing model by linking known parameters to be influenced by the source (e.g., chlorophyll a is a direct proxy of the living phytoplankton biomass, or pH may be indirectly lowered by living phytoplankton biomass).

In the case of the multi-rivers RDA, environmental forcings dynamics only explain the spatial dynamics between rivers, as the data input is only the annual mean. It permits to consider stable watershed forcings (on an annual scale), as soil properties and uses, or geographical coordinates, that may translate a latitudinal temperate climatic gradient (from degraded oceanic to Mediterranean climates).

We will consider this comment and improve the understandability of the different environmental drivers to POM dynamics.

Will be considered.

Will be considered.

Will be considered.

Will be considered.

Indeed, our work is restricted to Western Europe. It will be considered.

See the 4-metrics issue answer.

Technically, datasets for soil organic carbon data were retrieved as a raster, Corine Land Cover as a polygon map, and erosional soil data as a punctual map. Each dataset was cut according to each watershed, then averaged individually. This pattern ensures that each value is a real average of each watershed and that there is harmony between the resulting parameters.

We will add this information to the manuscript.

Will be considered.

PN/SPM ratio was used to distinguish sources for rivers where consensual discriminators were not available (chlorophyll a and phaeopigments). Although it has not been documented as a phytoplankton descriptor, the PN/SPM ratio seemed appropriate for this source in our study.

The pertinence of parameters used to qualify the isotopic variability of certain sources is discussed, as for the nitrates (that Reviewer 1 noted l.288f) at l.189-191 of the manuscript.

Will be considered.

We understand the cumbersome nature of this notation. However, we would like to avoid any confusion about the mol/mol nature of the C/N ratio, while it cohabits with C/N ratios expressed in g/g in the literature (sometimes implicitly).

Chlorophyll a, phaeopigments, or the sum of both. Will be considered.

See the answer of the 1-Litospheric origin. We do not consider the refractory terrestrial POM originating from fresh terrestrial plants (or fresh litter; C/N ratio always > 12 mol/mol), but from eroded mineral-dominated soils or resuspended riverbed sediment, possibly associated with bacteria (same Fig. in Lamb et al. 2006). In that case, the low C/N ratio signatures of refractory terrestrial POM are consistent with those of these sources (l. 501 to 508 of the manuscript).

We agree with Reviewer 1 that high SPM is usually driven by flood, so high river flows. The slight decoupling shown in the redundancy analyses has multiple causes.

In the Milieu River, since there is a dominance of terrestrial matter (so a weak source dynamics), SPM is strongly correlated with the POC dynamics, so both POC sources, while river flow induced more POC of terrestrial origin, where the minimum of phytoplankton proportion was found.

In the Charente and Rance Rivers, if river flow is correlated with more labile terrestrial POM than SPM, the latter is also linked with phytoplankton bloom, responsible for bulk POC and SPM peaks. Moreover, high phytoplankton proportions may also be linked with low river flows.

In the Hérault River, even if they are closer to one terrestrial source, both river flows and SPM are correlated with both terrestrial sources. This slight decoupling may come from processes of resuspension that can be triggered by other forcings than high river flows (as wind, poorly shown here).

We will consider this commentary to improve the RDAs discussion.

The multi-site RDA is set as an average year, so without any temporal information (see 4-metrics answer). Here, the rivers with the highest average temperature are the Mediterranean ones, logically, the only ones where we considered refractory terrestrial POM. Since this bias is understood, no real environmental information can be extracted from this correlation.

We agree. A discussion will be added about the environmental drivers to the refractory terrestrial dynamics across rivers, given the new RDA performed with more pertinent parameters (slope, connectivity).

In addition to the precedent answer, we assure that we will revise our hypothesis depending on the new results, better integrating watershed processes.

Indeed, correlations between the environmental parameters are not treated here. We will consider adding statistical univariate correlations when necessary.

After checking, SPM was removed from the multi-site RDA for its weakness in the variance representation (short arrow). It seems that rivers with an average phytoplankton dominance induced as high SPM concentrations as rivers with an average terrestrial dominance.

You are right. Since the Loire River sources proportions and RDA had already been published (Ferchiche et al. 2024), we decided not to duplicate these specific results in this manuscript.

Ferchiche F, Liénart C, Charlier K, Coynel A, Gorse-Labadie L and Savoye N (2024) Quantifying particulate organic matter: source composition and fluxes at the river-estuary interface. Front. Freshw. Sci. 2:1437431. doi: 10.3389/ffwsc.2024.1437431

Will be considered.

Reviewer 2

We would like to thank Reviewer 2 for providing a thorough analysis of our work, highlighting the manuscript's quality while offering important suggestions to improve its structure and substance. We are glad to address all of Reviewer 2’s concerns.

Indeed, the 23 rivers we chose to gather into this study are clustered in the same temperate climate. We will improve the manuscript to better explain the climatic range of our study in a somewhat restrained geographic area and the broad implications of our approach, at least from a technical point of view.

Reviewer 2 raises an important point: we did not consider any unique petrogenic source in our mixing models. However, we considered that the refractory terrestrial POM sources we distinguished from the bulk POM can be of petrogenic origin.

Indeed, as raised by Reviewer 2, petrogenic OM, according to the geology, can be defined by a δ¹³C between -20 and -26 ‰ (Hilton et al. 2010). Also distinguished by low C/N ratios for a terrestrial source, they can be easily integrated into isotopic mixing models.

For the 23 rivers, rare were the bulk suspended POM values that were higher than -26 ‰ (only a few for the Têt and the Rhône Rivers). Thankfully, our method permitted us to discriminate a refractory terrestrial POM source for these two rivers. Since this, we can consider that for the Têt and Rhône rivers, this source is at least influenced by petrogenic POM.

We will better describe the possible petrogenic origin of this source in the manuscripts with the help of this commentary. Also, see our response to Reviewer 1 (first major issue).

Robert G. Hilton, Albert Galy, Niels Hovius, Ming-Jame Horng, Hongey Chen, The isotopic composition of particulate organic carbon in mountain rivers of Taiwan, Geochimica et Cosmochimica Acta, Volume 74, Issue 11, 2010, Pages 3164-3181, ISSN 0016-7037, https://doi.org/10.1016/j.gca.2010.03.004.

Absolutely, the actual work is focused on temperate Rivers in Western Europe. Nevertheless, the approach that we used in the manuscript can be applicable to other climates and broadened to a greater range of environmental conditions, including different climatic settings (polar to tropical) and geographical locations (catchment properties) (l.655-658 of the manuscript). This will be better detailed in section 5 and in the abstract.

Will be considered.

Reviewer 2 suggests that POM composition mixing models are common in rivers.

Indeed, since the 1990s, a few dozen articles have been published, using mixing models to partially or completely quantify POM sources' proportions in freshwater rivers. However, as for Hilton et al. (2010) previously cited, all these articles (except those of l.102-103 of the manuscript) performed outdated deterministic linear mixing models. These models only measure the distances between end-member signatures and the mixture value to give the proportion of the distances. A part of this literature focuses only on a fraction of the POM, as it only accounts for terrestrial origins, or fails to correctly distinguish phytoplankton source (Hilton et al. 2010, Dalu et al. 2016, Lu et al. 2016).

Bayesian (probabilistic) mixing models published in the 2010s until today calculate proportions using isotopic and elemental signatures and their uncertainties, and introduce priors, allowing to give an interval of confidence for each mixing model, and uncertainties to the source’s proportions. The credibility of the whole model is then measurable, contrary to frequentist models.

For this reason, particular attention should be paid to the methods used to quantify sources of POM. We would like to reassure Reviewer 2 that the use of the Bayesian mixing model is rare (as pointed out by Reviewer 1), and is unique by the integration of our methodology of source discrimination and taking into account isotopic variability by source is a methodological advancement for riverine biogeochemistry.

Robert G. Hilton, Albert Galy, Niels Hovius, Ming-Jame Horng, Hongey Chen, The isotopic composition of particulate organic carbon in mountain rivers of Taiwan, Geochimica et Cosmochimica Acta, Volume 74, Issue 11, 2010, Pages 3164-3181, ISSN 0016-7037, https://doi.org/10.1016/j.gca.2010.03.004.

Dalu, T., Richoux, N.B. & Froneman, P.W. Nature and source of suspended particulate matter and detritus along an austral temperate river–estuary continuum, assessed using stable isotope analysis. Hydrobiologia 767, 95–110 (2016). https://doi.org/10.1007/s10750-015-2480-1

Lu Lu, Hongguang Cheng, Xiao Pu, Jiantong Wang, Qianding Cheng, Xuelian Liu, Identifying organic matter sources using isotopic ratios in a watershed impacted by intensive agricultural activities in Northeast China,Agriculture, Ecosystems & Environment,Volume 222, 2016, Pages 48-59, ISSN 0167-8809, https://doi.org/10.1016/j.agee.2015.12.033.

Reviewer 2 raised a good point. Features common to these rivers will be added.

Will be considered.