the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 01 Sep 2025
A centennial-scale data of suspended particulate matter over the global ocean
Abstract. Suspended Particulate Matter (SPM) is a crucial indicator of marine nutrient transport, the carbon cycle, and land-sourced pollution. However, SPM research has been constrained by limited data. While marine remote sensing has helped fill this gap since late 20th century, pre-2000 data remains scarce, particularly for global datasets with high spatiotemporal resolution. To address the shortage of in-situ SPM observations, we developed two empirical models for estimating SPM concentrations: one based on remote sensing reflectance (Rrs) and the other on Secchi depth (Zsd). Model evaluation shows that retrieval uncertainties do not exceed 35 %. To mitigate errors in highly turbid waters (>100 mg l-1), we applied data quality constraints by excluding samples above this threshold. The resulting dataset merges in-situ SPM observations, providing global coverage of SPM concentrations from 1890 to 2020. Temporal and spatial analysis reveals an overall increase in global SPM concentrations, with a sharp rise between 1920 and 1942, followed by a rapid return to pre-1920 levels. The underlying causes of this fluctuation warrant further investigation. This unique dataset holds considerable potential for marine resource management, climate modeling, human activity assessment, and environmental protection.
- Preprint
(2371 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
- RC1: 'Comment on essd-2025-484', Anonymous Referee #1, 16 Nov 2025
-
RC2: 'Comment on essd-2025-484', Anonymous Referee #2, 24 Dec 2025
The paper by Chen and coauthors presents a dataset for ocean particulate matter.
From the beginning (title and abstract), I was expecting a global reconstruction ofÂ
SPM, almost complete as satellite Rrs is mentioned. However, the files provided (what isÂ
the difference between .zip and .rar files?) provide pixel-wise values (from ships only?),Â
and it is spatially incomplete. Interestingly, there are some values outside the oceans (maybe lakes?).
If annual statistics are provided, then instantaneous Rrs/Zsd/SPM data which would have value
are not available. The origin of data needs to be provided as a flag in the files.
In the current state, I cannot appreciate the value of the authors' efforts.Â
General commentsFigure 1 and 2 are quite confusing. Where SPMmea goes in the workflow is ambiguous to me,
and anyway the workflow should be described more clearly.
At the moment I understand that Zsd and Rrs go into an empirical model giving SPM for theÂ
Secchi disk and satellite radiances. Then the Secchi disk data creates a centennial dataset.
I must have misunderstood something.I have concerns on the simulated dataset. SPM values seem to be mostly below 100 mg/l (Fig. 6),
which I understand means moderately turbid water. However, if larger values are meaningful,
the distribution of numerical simulated data should have been different. Presumably you could
consider using the observed distribution rather than an uniform one.
It is also unclear to me which forward models were used to generate synthetic satellite Rrs.
As far as I know, the retrieved radiance depends on a multitude of factors including atmospheric
conditions, which are not even mentioned. How the 12000 data points are used is unclear to me.The results of Fig. 7d are suspect. This kind of variability might mostly reflect observational
changes rather than real changes in the SPM (here called "TSM"... why?).
Is there any previous work supporting such large interannual changes?
Having opened the dataset, I understand this is a collection of existing datasets SPM datasets,
but no information on the source is given there.
Validation with peer-datasets (e.g. https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2021JC017303
or https://data.marine.copernicus.eu/product/OCEANCOLOUR_GLO_BGC_L3_NRT_009_101/description)
is required.In general the figures are confusing, with captions not very clear, odd labeling choices
(such as order a/b/d/c) and poor visual choices. This is not helpful to the reader.
typo 'Longtitude' in table 2 and elsewhere
Specific comments17 what is ">100 mg/l"?
24 first and last keywords are way too generic
67 this seems to suggest that Rrs and Zsd are equivalent. I don't think that's the case, limitations in Rrs retrieval needs to be stressed
112 this reads a bit random, I would more generally speak about trends and human forcings, in case
130 there are presumably wavelengths (not monochromatic, presumably!), but this must be explained
138 I guess no measurement can ever be?
154 How do you reach this figure?
Fig. 2c I don't get it, are you showing Rrs measurements before the satellite era? Please revise the caption and add titles throughout.
I also can't understand which is column 3 you are referring to
169 is there also a private SPM?
178 nonsensical citation
189-90 probability distributions of these parameters should be shown
200 discussion of Fig. 3b needs enhancement. Why there are basically just two groups, i.e. SPM 0.5 and SPM 10?
You referred to 6000 points before, how many are shown here?
207 is this the simulated or an observational dataset?
218 unnecessary sentence, please provide something supported and specific for this problem
221 this was already mentioned, but how accurate is this synchronization? Within a 5x5 box and one day? Or what?
235 if we are treating the Zsd errors, why speaking about satellite bands?
Fig. 5b showing histograms as shaded overlapping areas makes them difficult to see, revise
240 Well, if there is no NIR band, this should just be mentioned in conclusions to avoid distraction
Fig. 6d this does not seem like an histogram, and what is the definition of X?
259 not knowing how frequent this value is, it is difficult to make sense of this choice
Fig. 7 I do not understand what is plotted here, not least because title and labels are generic
269 is this a global average? What is the purple line and what does the shading represent?
280 please provide concrete examples or remove speculationsCitation: https://doi.org/10.5194/essd-2025-484-RC2
Status: closed
-
RC1: 'Comment on essd-2025-484', Anonymous Referee #1, 16 Nov 2025
The work of A centennial-scale data of suspended particulate matter over the global ocean merges multi-souce dataset into synthesis to reveal the temporal vairation in last one centrury. Abundant datasets, including remote sensing reflectance and other on Secchi depth and field bottled-sampled observation, are used to derive the dataset. This work has significant novelty in the era of big data in ocean studies. However, it still requires more details and analysis regarding the data quality and interpretation. I recommend this manuscript can be accepted, pending on major revision.
Â
The detailed comments are listed below:
- Ln145: How to make the numerical simulation? More details are required, such as how to determine SPM from transparency and Rrs.
- The caption of Figure 2 is not correct.
- Ln230: This paragraph can be moved to discussion.
- Ln276: The reason of thes increase and decline of SPM in cenntial scale should be carefully and thoghouhly analyzed. Author should cross-check these results with previous studies, such as Milliman and Farnsworth’s River Discharge to the Coastal OceanA Global Synthesis. The decline trend looks like starting from 1990s, not 2010. Does this agree with Milliman and Farnsworth’s anaylsis? Additionally, I suggest authors to conduct some regional calculation with abundant river discharge and strong variations. Does this dataset reveal some temporal patterns like the riverine SPM discharge into oceans?
- The spikes in 1920s should be analyzed, since the concentration has largest magnitude in this study. I suggest the authors compare their results against Milliman and Farnsworth’s work.
- Ln275: I am afraid I cannot agree with authors about the SPM increase from 1940-2010 were caused by infrastructure projects. There were strong damming activities in this period, which should result in the decrease of sediment discharge into oceans. Why is there SPM increase?
- Can author make some spatial plot of SPM in the typical times, such as 1920s, 1990s.
- The SPM dynamics are more signifcant in mega estuires and deltas.
- The limitation of this work should be fully discussed. One of these limitation is the cut-off threshold of SPM 100mg/L, which can not fully represent the global riverine sediment discharge into oceans. This magnitude of 100 mg/L is easily interefered by local resuspension or other physical dynamics such as wind-induced waves. Does the temporal variation have the relationship with atmospheric osscillations?
Citation: https://doi.org/10.5194/essd-2025-484-RC1 -
RC2: 'Comment on essd-2025-484', Anonymous Referee #2, 24 Dec 2025
The paper by Chen and coauthors presents a dataset for ocean particulate matter.
From the beginning (title and abstract), I was expecting a global reconstruction ofÂ
SPM, almost complete as satellite Rrs is mentioned. However, the files provided (what isÂ
the difference between .zip and .rar files?) provide pixel-wise values (from ships only?),Â
and it is spatially incomplete. Interestingly, there are some values outside the oceans (maybe lakes?).
If annual statistics are provided, then instantaneous Rrs/Zsd/SPM data which would have value
are not available. The origin of data needs to be provided as a flag in the files.
In the current state, I cannot appreciate the value of the authors' efforts.Â
General commentsFigure 1 and 2 are quite confusing. Where SPMmea goes in the workflow is ambiguous to me,
and anyway the workflow should be described more clearly.
At the moment I understand that Zsd and Rrs go into an empirical model giving SPM for theÂ
Secchi disk and satellite radiances. Then the Secchi disk data creates a centennial dataset.
I must have misunderstood something.I have concerns on the simulated dataset. SPM values seem to be mostly below 100 mg/l (Fig. 6),
which I understand means moderately turbid water. However, if larger values are meaningful,
the distribution of numerical simulated data should have been different. Presumably you could
consider using the observed distribution rather than an uniform one.
It is also unclear to me which forward models were used to generate synthetic satellite Rrs.
As far as I know, the retrieved radiance depends on a multitude of factors including atmospheric
conditions, which are not even mentioned. How the 12000 data points are used is unclear to me.The results of Fig. 7d are suspect. This kind of variability might mostly reflect observational
changes rather than real changes in the SPM (here called "TSM"... why?).
Is there any previous work supporting such large interannual changes?
Having opened the dataset, I understand this is a collection of existing datasets SPM datasets,
but no information on the source is given there.
Validation with peer-datasets (e.g. https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2021JC017303
or https://data.marine.copernicus.eu/product/OCEANCOLOUR_GLO_BGC_L3_NRT_009_101/description)
is required.In general the figures are confusing, with captions not very clear, odd labeling choices
(such as order a/b/d/c) and poor visual choices. This is not helpful to the reader.
typo 'Longtitude' in table 2 and elsewhere
Specific comments17 what is ">100 mg/l"?
24 first and last keywords are way too generic
67 this seems to suggest that Rrs and Zsd are equivalent. I don't think that's the case, limitations in Rrs retrieval needs to be stressed
112 this reads a bit random, I would more generally speak about trends and human forcings, in case
130 there are presumably wavelengths (not monochromatic, presumably!), but this must be explained
138 I guess no measurement can ever be?
154 How do you reach this figure?
Fig. 2c I don't get it, are you showing Rrs measurements before the satellite era? Please revise the caption and add titles throughout.
I also can't understand which is column 3 you are referring to
169 is there also a private SPM?
178 nonsensical citation
189-90 probability distributions of these parameters should be shown
200 discussion of Fig. 3b needs enhancement. Why there are basically just two groups, i.e. SPM 0.5 and SPM 10?
You referred to 6000 points before, how many are shown here?
207 is this the simulated or an observational dataset?
218 unnecessary sentence, please provide something supported and specific for this problem
221 this was already mentioned, but how accurate is this synchronization? Within a 5x5 box and one day? Or what?
235 if we are treating the Zsd errors, why speaking about satellite bands?
Fig. 5b showing histograms as shaded overlapping areas makes them difficult to see, revise
240 Well, if there is no NIR band, this should just be mentioned in conclusions to avoid distraction
Fig. 6d this does not seem like an histogram, and what is the definition of X?
259 not knowing how frequent this value is, it is difficult to make sense of this choice
Fig. 7 I do not understand what is plotted here, not least because title and labels are generic
269 is this a global average? What is the purple line and what does the shading represent?
280 please provide concrete examples or remove speculationsCitation: https://doi.org/10.5194/essd-2025-484-RC2
Data sets
A centennial-scale data of suspended particulate matter over the global ocean Jun Chen https://doi.org/10.5281/zenodo.16991206
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,115 | 112 | 28 | 1,255 | 30 | 40 |
- HTML: 1,115
- PDF: 112
- XML: 28
- Total: 1,255
- BibTeX: 30
- EndNote: 40
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The work of A centennial-scale data of suspended particulate matter over the global ocean merges multi-souce dataset into synthesis to reveal the temporal vairation in last one centrury. Abundant datasets, including remote sensing reflectance and other on Secchi depth and field bottled-sampled observation, are used to derive the dataset. This work has significant novelty in the era of big data in ocean studies. However, it still requires more details and analysis regarding the data quality and interpretation. I recommend this manuscript can be accepted, pending on major revision.
Â
The detailed comments are listed below: