the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Feb 2025
Mapping the global distribution of lead and its isotopes in seawater with explainable machine learning
Abstract. Lead (Pb) and its isotopes are a powerful tool to study the pathways of Pb pollution from land to sea and, simultaneously, investigate biogeochemical processes in the ocean. However, the scarcity and sparsity of in situ measurements of Pb concentrations and isotope compositions do not allow for a comprehensive understanding of Pb pollution pathways and biogeochemical cycling on a global scale. Here, we present three machine learning models developed to map seawater Pb concentrations and isotope compositions leveraging the global GEOTRACES dataset as well as historical data. The models use climatologies of oceanographic and atmospheric variables as features from which to predict Pb concentrations, 206Pb/207Pb, and 208Pb/207Pb. Using Shapley Additive Values (SHAP), we found that seawater temperature, atmospheric dust and black carbon, and salinity are the most important features for predicting Pb concentrations. Dissolved oxygen concentration, salinity, temperature, and atmospheric dust are the most important features for predicting 206Pb/207Pb, while atmospheric black carbon and dust, seawater temperature, and surface chlorophyll-a for 208Pb/207Pb. Our model outputs show that (i) the surface Indian Ocean has the highest levels of pollution, (ii) pollution from previous decades is sinking in the North Atlantic and Pacific Ocean, and (iii) waters characterised by a highly anthropogenic Pb isotope fingerprint are spreading from the Southern Ocean throughout the Southern Hemisphere at intermediate depths. By analysing the uncertainty associated with our maps, we identified the Southern Ocean as the key area to prioritise in future sampling campaigns. Our datasets, models and their outputs, in the form of Pb concentration, 206Pb/207Pb, and 208Pb/207Pb climatologies, are made freely available to the community at Olivelli et al. (2024a, https://doi.org/10.5281/zenodo.14261154) and https://github.com/OlivelliAri/Pb-ML_GEOTRACES.
- Preprint
(7175 KB) - Metadata XML
-
Supplement
(3903 KB) - BibTeX
- EndNote
Status: open (until 15 Apr 2025)
-
RC1: 'Comment on essd-2025-17', Edward Boyle, 18 Mar 2025
reply
Review of “Mapping the global distribution of lead and its isotopes in seawater with explainable machine learning” by Olivelli et al. (essd 2025-17).
Although this reviewer knows something about Pb and Pb isotopes in the ocean, I do not know anything about AI models. So my comments are addressed strictly to what the paper says about Pb and Pb isotopes in the ocean without any critical assessment of the AI model.
My bottom lines on this effort to see if AI is useful in understanding Pb in the ocean is given by my inferred (from this manuscript) answers to two questions:
(1) Is AI a useful way to summarize global 3D patterns of Pb and Pb isotopes in the ocean given very limited data? YES – the maps correspond decently with what is known for sure about Pb in the ocean, both from data and from understood processes. There are some implied features that have no observational basis (e.g. the high surface Pb values in the tropical eastern Pacific), but those can be corrected as future Pb data is published.
(2) Has this AI effort increased our understanding of Pb in the ocean? NO – the conclusions as stated in the abstract and conclusions section were already known from the raw data (and are stated as such in publications), I don’t see any advance in our understanding of Pb in the ocean. However, it does represent an advance to the question “Of what utility can AI have now in studying Pb in the ocean?”
Overall I support publication after relatively minor changes.
Specific comments on sections:
Line 22: “Our model outputs show that…” Baloney, the features that follow were already evident in the raw data. The model is just mimicking the data for the listed features.
Line 38: “did not allow for successful measurement of seawater Pb concentration measurements until 1963 (refs. Tatsumoto and Petterson, 1963)”. I don’t agree that the 1963 data set was correct, and therefore not the first. They are high compared with the coral measurements of Desenfant et al. (2006, Coral Reefs 25:473) and Kelly et al. (2009, EPSL 283:93). The first successful Pb measurements in the Atlantic are the 1979 data of Schaule and Patterson (1983, in Trace Elements in Seawater, eds. C.S. Wong et al.).
Lines 115-116: “Dust was found to be an important source of natural Pb to the surface ocean in several regions, including the … southern Indian Ocean (Lee et al., 2015)”. That is not true. Lee et al. did see high 206Pb/207Pb ratios at 62°S deep waters consistent with crustal Pb, but they did not attribute it to continental dust because there is very little crustal aerosol dust in the Southern Ocean today. It seems more likely that this feature is caused by glacial erosion from the Antarctic continent.
Lines 270- : re the role of oxygen as a fitting device: I don’t think there is any reasonable mechanistic driver for such a correlation, with no obvious indication of “Pb regeneration from sinking organic matter” in the observations. However, for example in the Northeast Atlantic, the water with the lowest O2 has an SF6 age of about 40 years, which was the period of maximum Atlantic region leaded gasoline utilization. So the correlation is simply based on the aging of the water mass (O2 decrease) and the coincidence of that maximum with the period of highest Pb emissions.
Similarly with regard to back carbon as a fitting parameter, yes, Pb and black carbon are both anthropogenic in origin and it isn’t surprising if they often show similar spatial sources (where are people, cars and industrial activities), however, they sometimes should not be correlated (e.g., it isn’t clear the tropical forest burning is much of a Pb source but it is for black carbon).
Lines 422- : “However, historically Pb isotopes have been regarded as tracers of water mass movements and ventilation (refs. Veron et al., 1998, 1999)” – This is only true for the North Atlantic, and only because (in John Edmond’s words) “the Atlantic is a bowling alley” with strong and rapid lateral water mass movements adjacent to recognizable sources. This statement is not true for the Pacific.
Citation: https://doi.org/10.5194/essd-2025-17-RC1 -
RC2: 'Comment on essd-2025-17', Anonymous Referee #2, 22 Mar 2025
reply
General comments
In this manuscript, the authors reconstructed the global distribution of Pb and its isotopes in seawater with machine learning models. I’m not familiar with machine learning at all. So, I make some comments on this manuscript as a sea-going scientist. The models seem unique and provide some interesting points of view for us. For example, black carbon in air is one of the important factors to reconstruct the global Pb distribution in this model. From this result, I realized that the simultaneous determination of black carbon and Pb isotopes in aerosols is a good topic in the field studies. I think this manuscript is challenging but has possibility to give us some insight on the marine biogeochemical cycles of Pb.
Specific comments
- Line 310, “km/m3”: Is this unit correct?
- Line 322 - 325: Is this meaningful description?
- Line 365 – 366, “further studies of the Pb isotope compositions of dust and black carbon sources to the atmosphere”: I could not find so many references for these topics, but at least one reference concerning the topics (e.g. Nizam et al., 2020) should be mentioned.
- Line 370 – 372, “Moreover, black carbon is not only emitted due to fossil fuel consumption, but also from forest fires and coal combustion, which might have substantially different Pb isotope signatures compared to other industrial and urban sources.”: Recently, sources of black carbon in aerosol were discussed from the radiocarbon measurements (Gustafsson et al., 2009; Li et al., 2016). Considering these references, the authors should discuss the sources of the black carbon at this moment.
- Line 385 – 386, “The only exception to this trend is the North Atlantic Ocean, where mapped Pb concentrations at 1000 m and 2500 m are higher than at 10 m.”: From the observational studies, subsurface maxima of Pb were reported in the North Pacific (Wu et al., 2010; Zurbrick et al., 2017; Zheng et al., 2019; Chan et al., 2024; Jiang et al., 2025). Were these features found in this model?
- Line 400 - 403: In the Northwest Pacific, the subduction and ventilation process of the North Pacific mode waters and NPIW were considered to elevate Pb concentrations in the subsurface layers (Jiang et al., 2021). Since numbers of data are relatively small in the Northwest Pacific, this feature might not be captured in this model.
References
Chan, C., L. Zheng, Y. Sohrin (2024). The behavior of aluminium, manganese, iron, cobalt, and lead in the subarctic Pacific Ocean: boundary scavenging and temporal changes. Journal of Oceanography 80, 99 – 115.
Gustafsson, Ö., M. Kruså, Z. Zencak, R. J. Sheesley, L. Granat, E. Engström, P. S. Praveen, P. S. P. Rao, C. Leck, H. Rodhe (2009). Brown Clouds over South Asia: Biomass or Fossil Fuel Combustion? Science 323, 495 – 498.
Jiang, S., J. Zhang, H. Zhou, Y. Xue, W. Zheng (2021). Concentration of dissolved lead in the upper Northwestern Pacific Ocean. Chemical Geology 577, 120275.
Jiang, S., N. Lanning, E. Boyle, J. Fitzsimmons, J. Ramezani, A. G. Wang, J. Zhang (2025). Meridional central Pacific Ocean depth section for Pb and Pb isotopes (GEOTRACES GP15, 152°W, 56°N to 20°S) including shipboard aerosols. Journal of Geophysical Research: Oceans 130, e2024JC021674.
Li, C., C. Bosch, S. Kang, A. Andersson, P. Chen, Q. Zhang, Z. Cong, B. Chen, D. Qin, O. Gustafsson (2016). Sources of black carbon to the Himalayan–Tibetan Plateau glaciers. Nature Communications 7, 12574.
Nizam, S., I. S. Sen, V. Vinoj, V. Galy, D. Selby, M. F. Azam, S. K. Pandey, R. A. Creaser, A. K. Agarwal, A. P. Singh, M. Bizimis (2020). Biomass-Derived Provenance Dominates Glacial Surface Organic Carbon in the Western Himalaya. Environmental Science and Technology 54, 8612 − 8621.
Wu, J., R. Rember, M. Jin, E. A. Boyle, A. R. Flegal (2010). Isotopic evidence for the source of lead in the North Pacific abyssal water. Geochimica et Cosmochimica Acta 75, 460 – 468.
Zheng, L., T. Minami, W. Konagaya, C. Chan, M. Tsujisaka, S. Takano, K. Norisuye, Y. Sohrin (2019). Distinct basin-scale-distributions of aluminum, manganese, cobalt, and lead in the North Pacific Ocean. Geochimica et Cosmochimica Acta 254, 102-121.
Zurbrick, C. M., C. Gallon, A. R. Flegal (2017). Historic and industrial lead within the Northwest Pacific Ocean evidenced by lead isotopes in seawater. Environmental Science and Technology 51, 1203 – 1212.
Citation: https://doi.org/10.5194/essd-2025-17-RC2
Data sets
Data for: Mapping the global distribution of lead and its isotopes in seawater with explainable machine learning Arianna Olivelli, Rossella Arcucci, Mark Rehkämper, and Tina van de Flierdt https://doi.org/10.5281/zenodo.14261154
Model code and software
Pb-ML_GEOTRACES Arianna Olivelli https://github.com/OlivelliAri/Pb-ML_GEOTRACES
Interactive computing environment
Pb-ML_GEOTRACES Arianna Olivelli https://github.com/OlivelliAri/Pb-ML_GEOTRACES
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 167 | 24 | 6 | 197 | 17 | 3 | 4 |
- HTML: 167
- PDF: 24
- XML: 6
- Total: 197
- Supplement: 17
- BibTeX: 3
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1