General comments

The authors highlight the need for a unified dataset that links chlorophyll-a (chla) and nutrient measurements in freshwater as the main motivation behind OLIGOTREND. The resulting dataset focuses on high-income countries, meaning gaps outside North America and Western Europe remain. Still, OLIGOTREND successfully improves the current spatial and temporal coverage of the aforementioned water quality (WQ) parameters, making it applicable for large-scale oligotrophication studies. I found the manuscript structure logical and the writing concise and up to ESSD standards. I particularly appreciate the clarity of the data processing workflow and the inclusion of complementary GIS layers.

Below is a list of my remarks and suggestions, mainly minor technical corrections related to wording. Finally, I recommend a minor revision before the final acceptance of the manuscript in ESSD.

Specific comments

You have listed the summary statistics of the dataset in the abstract (L32-33). I suggest adding the overall temporal coverage of the dataset here, too. Although the time series length depends on the parameter, you can use the earliest and latest years in OLIGOTREND to give a general estimate for users.

I appreciate the well-defined data processing workflow shown in Fig. 1 (section 2). I particularly like that you used “levels” to label the different modification stages of the dataset, e.g. “L0a” for raw data. These make it easier to follow the workflow as it moves through the processing pipeline.

I commend the authors for providing catchment boundaries and attributes along with the WQ measurements (section 2.3), which enhances the applicability of the data for modelling purposes. However, due to coarse resolution, the boundaries of HydroATLAS layers are often quite inaccurate compared to national-level datasets. Did you consider any alternative sources by any chance, e.g. the CAMELS catchments?

You mention providing the GIS data emerging from the data extraction step (L423-425). I understand the GIS data is currently available in the GitLab repository rather than the EDI data portal. I cloned the GitLab repository and accessed the GIS layers, so this was not an issue for me. However, I suggest providing at least some of the GIS data (e.g., catchment boundaries) together with the WQ data in the EDI portal. This would make GIS data access more convenient for the non-technical user.

You plan to continuously update OLIGOTREND in the future (L431-433) in the conclusions. Considering that the dataset mostly consists of measurements from high-income countries, resulting in gaps globally, would these updates also include improving the spatial coverage? For example, would leveraging remote sensing be an option to fill some of these gaps?

Technical corrections

Throughout the manuscript: “timeseries” -> “time series”

L53: “understanding of oligotrophication is still not fully understood” -> Perhaps “our understanding is incomplete” to avoid repetitiveness.

L66: “across-ecosystem” -> “cross-ecosystem”

L80: “geo-spatial” -> “geospatial”

L109: “Kjeldhal” -> “Kjeldahl”

L128: “was” -> “were”

L132-133: “To offer the possibility to OLIGOTREND users to design their own quality check procedure, we did not remove any data in response to data curation (QA/QC).” -> “We did not remove any data in response to data curation (QA/QC) to allow users to design their own quality check procedure.”

L152: Please add the QGIS version.

L195: “Table 2. Overview of La data” -> Should it be “L1 data”?

L353: “nutrients” -> “nutrient”

L396: “inputs” -> “input”

Please unify the mixed use of British and American spelling in verbs like “analyse” (L397) and “characterize” (L398) throughout the manuscript.

This is an important and very well-written paper, presenting an important dataset with broad applicability across environmental and ecological fields. Figures are clear and very informative. Despite its spatial bias, OLIGOTREND offers a solid foundation for understanding patterns of oligotrophication at large scales for different water body types. I also commend the authors for the inclusion of a flexible and reproducible processing workflow, which will facilitate the continued expansion of the database in the future. OLIGOTREND will also help to support the development of analyses that can inform water quality management and monitoring efforts across diverse regions.

I think the authors should consider three main points that require further clarification: 

1.    It was unclear to me what was the temporal coverage of the datasets for different water quality parameters. Can you add a panel in figure 2 showing a violin plot or something similar that depicts the distribution of datasets across years per water quality parameter and classified by water body type?

2.    I think there is a high risk of matching the monitoring stations with some attributes from HydroATLAS, as there may be substantial mismatches between the temporal coverage and the spatial scale of each water quality dataset and layers such as land cover data. Wouldn’t it be better to link characteristics from national datasets when possible, rather than relying solely on HydroATLAS? Or at least you should add a flag indicating if the trend data aligns with the temporal scale of the land cover characteristics and anthropogenic data. This may not be an issue in areas where land cover has remained stable for a long time, but it could affect applications in areas where human pressure has changed over the analysed time frame.  

3.    I also have concerns about the process used to link monitoring stations with HydroRivers. Even if only one river falls within 200 meters of a station, the resolution of the dataset does not guarantee that it is the correct river. You should consider adding flags to indicate the level of uncertainty in the river-station match. Ideally, this matching should be manually verified. However, as manual verification may not be feasible, comparing discharge values reported in HydroRivers with those reported by the station (when available) could help assess the uncertainty associated with each match.

 

Specific comments:

L98-100 Please specify what criteria was used (search terms) to find these datasets. How many papers/reports/books did you include?

L100 Can you give a bit more context of why the database architecture allows “researchers to easily complement it with additional timeseries in the future”. You can tell this by checking the shared repository, but a short text here would be good to understand a bit better strengths and the scope of the database.

L138 What did you do when you found the “obvious mistakes”? The explained flags in L134 don’t fit this category. Did you exclude these mistakes?

L151: Although you acknowledge the uncertainty associated with linking rivers and monitoring stations, I think you should flag this more explicitly. See my main comment above. Also, what did you do with stations that matched more than one river? Did you assign them manually?. 

L231- 234 HydroATLAS land cover data is based on GLC2000, so it is unclear how you can determine associations for stations that do not match the timeframe of that land cover dataset.