Reconstructing Flow Dynamics of Temporary Streams Using Citizen Science Data and a Hierarchical Network Structure

Scheller, Mirjam; Durighetto, Nicola; Fischer, Anja; Furrer, Reinhard; Vis, Marc; Seibert, Jan; Botter, Gianluca; van Meerveld, Ilja

doi:10.5194/egusphere-2026-4479

Preprints

https://doi.org/10.5194/egusphere-2026-4479

Preprints

24 Aug 2026

| 24 Aug 2026

Status: this preprint is open for discussion and under review for Hydrology and Earth System Sciences (HESS).

Reconstructing Flow Dynamics of Temporary Streams Using Citizen Science Data and a Hierarchical Network Structure

Mirjam Scheller, Nicola Durighetto, Anja Fischer, Reinhard Furrer, Marc Vis, Jan Seibert, Gianluca Botter, and Ilja van Meerveld

Abstract. Temporary (i.e., non-perennial) streams form a large fraction of the global stream network and are important habitats. However, they are generally excluded from official monitoring networks. Visual observations by citizen scientists provide a valuable opportunity to collect temporary stream data, but these observations are irregular in space and time. We used observations by citizen scientists to derive the hierarchical structure (i.e., fixed order) of stream wetting and drying for a forested hill in Zurich, Switzerland. The dataset consists of 2 to 235 (median: 64) observations for 55 sites, and at least one observation for 402 days over the three-year period. In addition, an expert determined the stream state for 59 sites on 24 days during a six-month period. The hierarchical structure was more uncertain for the citizen science dataset than for the expert data, likely due to the irregularity of the observations. However, the hierarchical structures were comparable when considering only the 26 overlapping sites for the overlapping observation period. The hierarchical structure could be used to predict the stream state based on observations for other sites, allowing us to fill in data gaps. The correlation between the percentage of sites with flowing water and cumulative net precipitation, furthermore, allowed us to extend the prediction of stream states to days for which there were no observations. This study highlights the usefulness of citizen science-based observations of intermittent streams for understanding stream network dynamics. Information on the accuracy of the hierarchy can be used to determine which sites are particularly useful to observe for gap filling and thus to give feedback to citizen scientists.

Received: 24 Jul 2026 – Discussion started: 24 Aug 2026

Competing interests: At least one of the (co-)authors is a member of the editorial board of Hydrology and Earth System Sciences.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2709 KB)

Supplement (2529 KB)

Download & links

Mirjam Scheller, Nicola Durighetto, Anja Fischer, Reinhard Furrer, Marc Vis, Jan Seibert, Gianluca Botter, and Ilja van Meerveld

Status: open (until 30 Oct 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-4479', Anonymous Referee #1, 10 Sep 2026 reply
This is a review of the manuscript titled “Reconstructing Flow Dynamics of Temporary Streams Using Citizen Science Data and a Hierarchical Network Structure” submitted to Hydrology and Earth System Science. Overall, the manuscript was interesting, suitable for the journal, and easy to follow what was done and why. The introduction sets up the problem appropriately and the questions and goals align well with the manuscript. I only have a few overall comments for the authors to consider to help strengthen the manuscript and improve clarity in parts. Please note that I am not a hydrology but a broadly trained ecologist who specialized in citizen science. So please take these comments from this perspective. The comments are in no particular order.

I’m curious why your question II is stated as does expert improve the structure from citizen science data. Could you ask the opposite question: “Does citizen science observations improve the hierarchical structure obtained by an expert?” Would it change anything in your analyses? I think both are perfectly valid questions, but I found the framing presented as somewhat interesting and wondering if the motivation for this framing could be better presented/backed up in the introduction.

I think the methods present a fair bit of ‘results’. In terms of the citizen science participation etc. but I think a better structure would be to present this in one overall ‘citizen science participation’ (or something) section in the results. Since, to me, those are part of the results.

Also, the finding that most of the observations come from one participant is interesting and aligns with many other citizen science platforms.

For a 4km2 area, 57km of roads and paths seems like A LOT. I’m curious what this looks like? Could this be described better.

A lot of the results and downstream analyses rely on how a ‘site’ is defined. It would be nice to see some better clarification of this and how a site was defined. And accordingly, if the results are robust to this. Especially in a relatively small area. Just eyeballing the Figure 1a, some of the ‘sites’ look very similar to another. I.e., the blue and orange points look almost overlapping, but presumably someone could be standing and looking at a slightly different angle. So all in all, more treatment of the spatial resolution is warranted and how it does (or does not) impact the results.

I don’t like being the reviewer who says ‘do more analysis’. So I apologize up front. BUT I think there is some interesting low-hanging fruit here. The paragraph that starts on page 23 around line 520 correctly highlights an interesting question about how smaller citizen science datasets may perform similarly or not. I would suggest the authors consider decomposing (randomly subsampling) the dataset they have and rerunning the hierarchical structure analysis to see the ‘correlation’ between the citizen science and expert as a function of ‘sample size’ (i.e., size of the citizen science dataset). This should be relatively straightforward and would make for a nice addition to the manuscript.

Overall, I hope the authors find these thoughts and comments useful. I wish them the best of luck in this and future work!

Reply
Citation: https://doi.org/10.5194/egusphere-2026-4479-RC1
- AC1: 'Reply on RC1', Mirjam Scheller, 06 Oct 2026 reply
  
  We thank the reviewer for the careful and constructive comments and for the overall positive assessment of our manuscript. We appreciate the reviewer’s perspective as an ecologist with expertise in citizen science. The comments will help us improve the manuscript and clarify several aspects of our study. Below, we respond to each comment and describe the changes we are planning to make to the manuscript.
  Comment 1: “I’m curious why your question II is stated as does expert improve the structure from citizen science data. Could you ask the opposite question: “Does citizen science observations improve the hierarchical structure obtained by an expert?” Would it change anything in your analyses? I think both are perfectly valid questions, but I found the framing presented as somewhat interesting and wondering if the motivation for this framing could be better presented/backed up in the introduction.”
  Response: We thank the reviewer for raising this point. We agree that both directions of the question are relevant. We chose for the current framing of question II because our study design is based on an ongoing citizen science project, with citizen science observations forming the main dataset. The general idea is that citizen scientists can provide observations over a large spatial and temporal extent but that this data also has irregularities in spatial and temporal coverage. We were mainly curious if a few measurements by an expert would increase the value of the data for the modeling approach, as this would mean that project leaders need to be able to make additional observations (which limits the application of the approach for areas far away from the project leaders). We will clarify that our question is whether a short period of additional observations by an expert can improve the hierarchical structure derived from the citizen science observations.
  At the same time, we agree with the reviewer that the opposite question is also relevant. In situations where expert observations form the main dataset, citizen science observations could potentially complement these data by providing observations at additional locations or different times, thereby increasing the spatial or temporal extent. We will mention this perspective in the introduction and discussion to make the broader relevance of the two possible questions clearer.
  
  Comment 2: “I think the methods present a fair bit of ‘results’. In terms of the citizen science participation etc. but I think a better structure would be to present this in one overall ‘citizen science participation’ (or something) section in the results. Since, to me, those are part of the results.”
  Response: We agree with the reviewer that the information describing citizen science participation could also be part of the results. We used this information as background information on the data collection as this information is not related to the research questions but will move the relevant information from the methods section into a dedicated subsection in the Results, called “citizen science participation”.
  We will also review the other parts of the methods section to ensure that results are clearly separated from the methods. However, information that is necessary to understand how the analyses were conducted will remain in the Methods.
  
  Comment 3: “Also, the finding that most of the observations come from one participant is interesting and aligns with many other citizen science platforms.”
  Response: Indeed, it is common for citizen science projects that a handful of participants collect the majority of the data. We have previously called these participants super-users (Etter et al., 2023; https://doi.org/10.3389/fenvs.2023.1105682). We will point out this interesting aspect of the dataset in the revised manuscript.
  
  Comment 4: “For a 4km2 area, 57km of roads and paths seems like A LOT. I’m curious what this looks like? Could this be described better.”
  Response: We agree that the density of roads and paths in the study area is rather high. The study area is located on the Zürichberg, which is a frequently used recreational area in Zürich and consequently has a relatively high density of forest roads and trails. We will clarify this in the description of the study area and explain that the 57 km of roads and paths reflect the recreational use and accessibility of the area.
  In addition, we will provide a better visual representation of the roads and trails in the study area, by adding a map of the study area with the roads and paths as a supplementary figure in Figure S3 (see Figure 1 in the attached PDF file).
  
  Comment 5: “A lot of the results and downstream analyses rely on how a ‘site’ is defined. It would be nice to see some better clarification of this and how a site was defined. And accordingly, if the results are robust to this. Especially in a relatively small area. Just eyeballing the Figure 1a, some of the ‘sites’ look very similar to another. I.e., the blue and orange points look almost overlapping, but presumably someone could be standing and looking at a slightly different angle. So all in all, more treatment of the spatial resolution is warranted and how it does (or does not) impact the results.”
  Response: We thank the reviewer for highlighting the importance of the spatial coverage of the sites. We agree that this is an important aspect because the location of the sites is relevant for derivation of the hierarchical structure. However, the citizen scientists selected the location of the stream sites themselves. Therefore, we have limited knowledge about their selection processes. Nonetheless, it is obvious that the citizens selected sites near road crossings where the streams were easy to observe. We have provided information about the sites in the Supplementary Material to illustrate the spatial distribution and characteristics of the sites (including Figures S1-S3, Table S1 and Figure S5). The table provides the estimated catchment areas, topographic wetness indices, and locations.
  We will add extra information about why sites that appear close to each other in Figure 1 can nevertheless represent different stream sections and flow dynamics. As we repeatedly visited the study area during the study period, we observed ourselves that nearby stream sections can exhibit substantially different flow dynamics. These sites with different flow dynamics were purposefully included in the selection of sites for the Expert dataset.
  As the photo of the site is included in the app and most sites are at road crossings, we don’t think that citizens accidentally uploaded data for the wrong site in the app, even if these sites were only a few meters apart. Furthermore, there is usually a circle included in the first photo to indicate the exact location for which the flow state needs to be determined (see Figure 2 in the attached PDF file). Although mistakes cannot be ruled out, the good correspondence between the citizen science data and the expert data suggests that there are few mistakes in the data as well.
  As the number of sites is large and the density of sites is fairly high, we don’t think that the exact selection of sites affects any of the answers to the research questions and objectives, nor the conclusions that citizen science data can be used to derive the hierarchical structure, and that the hierarchical model can be used to fill in the gaps in the citizen science data.
  
  Comment 6: “I don’t like being the reviewer who says ‘do more analysis’. So I apologize up front. BUT I think there is some interesting low-hanging fruit here. The paragraph that starts on page 23 around line 520 correctly highlights an interesting question about how smaller citizen science datasets may perform similarly or not. I would suggest the authors consider decomposing (randomly subsampling) the dataset they have and rerunning the hierarchical structure analysis to see the ‘correlation’ between the citizen science and expert as a function of ‘sample size’ (i.e., size of the citizen science dataset). This should be relatively straightforward and would make for a nice addition to the manuscript.”
  Response: We thank the reviewer for this interesting suggestion. We agree that investigating how the performance of the hierarchical structure changes by decreasing the citizen science sample size could be interesting and would provide further insight into the number of data(sites) required for this type of analysis. However, this would increase the length of the manuscript substantially and depend on which sites are included in the network.
  Most likely, inclusion of different sites only will lead to an improvement in the accuracy as there will be less nodes at the same level in the network (next to each other in the hierarchy shown in Figure 2 a), whereas having only sites that are responding very similar would decrease the accuracy as there are more sites with the same position in the hierarchy. In part, we have done this by comparing the hierarchies for the citizen science data (55 sites) and those from the Expert and citizen scientists combined (88 sites). These results suggest that the number and location of sites have an influence on the accuracy of the hierarchy but that this effect is not very large.
  Even though the number of sites does not directly affect the accuracy of hierarchy (but the variation in sites does), the number of sites included in the dataset will determine how well the variability in intermittent stream dynamics for the study area is captured by the data. An extreme case with only two sites: one perennial and one nearly ephemeral would clearly not show all the variation in drying and wetting across the study area.
  In reality, the number of sites in a study area will depend on how well accessible the area is and what part of the study area is crossed frequently by citizens, as well as how involved the citizens are in submitting their observations.
  As the analysis suggested by the reviewer would substantially extend the length of the manuscript and the scope of the study beyond its main objectives, we prefer to not include an additional subsampling analysis in the revised manuscript. Still, we will more explicitly highlight the aspect of the relationship between dataset size and the resulting hierarchical structure in the text and mention the value of further analyses as an interesting topic for future research. We will also clarify that the present dataset provides an opportunity for such an analysis, but that systematically evaluating different sample sizes would constitute a separate research question.
  
  We thank the reviewer again for the constructive and helpful comments. We will use these suggestions to improve the clarity of the manuscript, particularly regarding the framing of the research questions, the description of citizen science participation, and the interpretation of the study area and dataset.
  
  Reply
  
  Citation: https://doi.org/10.5194/egusphere-2026-4479-AC1
RC2:
'Comment on egusphere-2026-4479', Anonymous Referee #2, 07 Oct 2026 reply
The study applies the graph theory to derive the wetting and drying order of several stream points over a small area in Switzerland using citizen science observations of stream state characterized by non perennial flows. The stream states are divided in 2 categories: dry/wet. The hiearchical structure obtained with the citizen science datasets is compared with the structure obtained with a data set composed of expert observations. The expert observations are not use to test the effect of the quality of observations on the hiearchical structure as expert and citizen sciences are very consistent whenever/wherever they are comparable. The expert observations are used to test the effect of irregularity in observation points in the hierachical structure, and in the associated uncertainty and accuracy. The results show that despite irregularity and possible biaises in the citizen science observations, they are valuable to identify a hierarchical structure of stream network drying/rewetting that can be used in combination with meteorological data (here effective precipitation) to fill the gap in the citizen science observations, leading to daily stream states for the 55 sites over the 4 km2 study area and over a three year period.
The results of the study are highly significant because it opens very nice perspectives to investigate and understand the mechanisms underlying hydrographic network expansion and contraction. It provides also new insights on the use of citizen science data sets. I think it is fully targeting the journal scope. I found the manuscrit well written, nicely illustrated.
I have noticed a few typos, please find below my technical comments and questions, which I hope will be helpful to polish the manuscript and/or for further researches
Questions I and II could be better distinguish: I is a comparison between the structures obtained from citizen science observations vs from expert observations while II is about the effect of the combination of both on the performance?

Question III "at other sites" confused me : I was wondering if you add external information from neighbouring stations (external from Zuriberg)? If there will be information from a gauged site for instance dowsntream from the studied one ...

Question IV: maybe it is worth to explain directly that you don't have gauging stations over the area or next and that climate data are the only one available ortherwise, one could expect that several sources of information could be tested

lines 125-127: hydrological network has been strongly modified, but your "mapped geomorphic stream" is the observed actual network right? (not a theroetical one from topo map), what is the source of this map?

Fig 1: maybe the name of the 4 sites (A08, H05...) could be indicated in the map next to the white crosses

line 175 : it is interesting to see it is the same number of disagreements within citizen scientists as between them and the expert

line 178-179: then the expert is not the reference here?

lines 194-196: Has the hypothesis of fixed order wetting/drying patterns been verified for winter drying (related to snow) and summer drying (related to low flows) ?

line 231-232: how many nodes are targetted for building the hierarchical structure ?

line 311: a verb missing after "that these nodes"?

line 263-266: but what is used as a reference here to determine if the structure prediction is correct or not?

line 356 : "with a minimum of"? word missing?

line 432-433 : you never discuss alternative use of physical model to provide a temporal prediction of stream extension-contraction, what do you think?

figure 10 : I do not understand anymore the difference between a and b even with the caption

line 491: an "and" missing between "dataset" and "was" ? Higher with Crowdwater than what?

line 493-494: why higher uncertainty would lead to higher accuracy? is it because of lower number of available observations to assess accuracy ?

line 527: then should this lead to recommandations of number of points/dates per km2 or per km of stream length in order to determine a hierarchical structure?

line 534: there are many metrics suggested here to compare the structure performance and consistency !

line 536-541 : maybe you should state more explicitely that expert observations do not increase the quality and do not reduce the biaises of the observations, but they are more systematically repeated in time on a selection of sites. Could the effect of the quality of the observations be tested using the 6 original classes of possible observations instead the 2 wet/dry categories? do we assume the 6 states to occurr successively during drying/rewetting?

line 579-584: analysis of spatial patterns associated to the hierarchical structure would be very interesting to dig the interpretation of underlying hydrological processes! Upstream/ Dowstream areas are consistently least/ most persistant areas but could you relate intermediate nodes temporal dynamics to topographical, morphological characteristics? presence of ditches? presence of wells? type of riparian vegetation?...

line 602: could you use AET estimates from close forests (eg ICOS database?) ?

Best regards!

Reply
Citation: https://doi.org/10.5194/egusphere-2026-4479-RC2

Mirjam Scheller, Nicola Durighetto, Anja Fischer, Reinhard Furrer, Marc Vis, Jan Seibert, Gianluca Botter, and Ilja van Meerveld

Supplement

https://doi.org/10.5194/egusphere-2026-4479-supplement

Mirjam Scheller, Nicola Durighetto, Anja Fischer, Reinhard Furrer, Marc Vis, Jan Seibert, Gianluca Botter, and Ilja van Meerveld

Viewed

Total article views: 348 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
200	89	59	348	83	66	60

HTML: 200
PDF: 89
XML: 59
Total: 348
Supplement: 83
BibTeX: 66
EndNote: 60

Views and downloads (calculated since 24 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	64	12	10	86
Sep 2026	90	47	36	173
Oct 2026	46	30	13	89

Cumulative views and downloads (calculated since 24 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	64	12	10	86
Sep 2026	90	47	36	173
Oct 2026	46	30	13	89

Viewed (geographical distribution)

Total article views: 340 (including HTML, PDF, and XML) Thereof 340 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 26 Oct 2026

Short summary

Many streams do not always contain flowing water. Because these streams are rarely monitored, we tested whether observations by citizen scientists could be used to track whether streams were flowing. Although these observations were irregular in space and time, identifying the order in which sites started or stopped flowing allowed us to fill data gaps and estimate daily flow conditions across the area. This shows that volunteers can effectively monitor streams that sometimes dry up.


Total:	0
HTML:	0
PDF:	0
XML:	0