Machine learning for prediction of daily sea surface dimethylsulfide concentration and emission flux over the North Atlantic Ocean (1998–2021)

Mansour, Karam; Decesari, Stefano; Čeburnis, Darius; Ovadnevaitė, Jurgita; Rinaldi, Matteo

doi:10.1016/j.scitotenv.2023.162123

Cited by 11 publications

(18 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For this reason, we use the sea-toair DMS flux (FDMS) as a predictor of MSA and SO4 concentrations. Mansour et al (2023a) used an ML predictive algorithm based on Gaussian process regression (GPR) to simulate the distribution of daily seawater DMS concentrations and related FDMS in the NA areas from 35° to 66° N and from 0° to 55° W at 0.25° × 0.25° spatial resolution. We extended the GPR model within the NA to encompass the NAAMES measurements, which are essential because they cover the western most section of the study area.…”

Section: Dimethylsulfide Flux Datamentioning

confidence: 99%

“…S2 displays the main differences between the two domains. Simply, the GPR was trained once more, utilizing the same approach of Mansour et al (2023a), with a higher number of data points and yielded an enhanced R 2 value up to 0.77 on the independent test dataset. The daily sea-to-air FDMS was calculated using the gas transfer velocity (Goddijn-Murphy et al, 2012) and the DMS derived from GPR predictions.…”

Section: Dimethylsulfide Flux Datamentioning

confidence: 99%

See 1 more Smart Citation

IPB-MSA&SO₄: a daily 0.25° resolution dataset of In-situ Produced Biogenic Methanesulfonic Acid and Sulfate over the North Atlantic during 1998–2022 based on machine learning

Mansour,

Decesari,

Ceburnis

et al. 2023

Preprint

View full text Add to dashboard Cite

Abstract. Accurate long-term marine-derived biogenic sulfur aerosol concentrations at high spatial and temporal resolutions are critical for a wide range of studies including climatology, trend analysis, model evaluation, accurate investigation of their contribution to aerosol burden, or to elucidate their radiative impacts and to provide boundary conditions for regional models. By applying machine learning algorithms, we constructed the first, publicly available, daily gridded dataset of in-situ produced biogenic methanesulfonic acid (MSA) and sulfate (SO4) concentrations covering the North Atlantic Ocean. The dataset is of high spatial resolution of 0.25° × 0.25°, spanning 25 years (1998–2022), far exceeding what observations alone could achieve both space- and time-wise. The machine learning models were generated by combining in-situ observations of sulfur aerosol data at Mace Head research station, west coast of Ireland, and from NAAMES cruises in the NW Atlantic, combined with the constructed sea-to-air dimethylsulfide flux (FDMS) and ECMWF-ERA5 reanalysis datasets. To determine the optimal method for regression, we employed four machine learning model types: support vector machines, ensemble, Gaussian process, and artificial neural networks. A comparison of the mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2) revealed that the Gaussian process regression (GPR) was the most effective algorithm, outperforming the other models in simulating the biogenic MSA and SO4 concentrations. For predicting daily MSA (SO4), GPR displayed the highest R2 value of 0.86 (0.72) and the lowest MAE of 0.014 (0.10) µg m–3. The GPR partial dependence analysis suggests that the relationships between predictors and MSA and SO4 concentrations are complex rather than linear. Using the GPR algorithm, we produced a high-resolution daily dataset of In-situ Produced Biogenic MSA and SO4 sea-level concentrations over the North Atlantic, which we named IPB-MSA&SO4. The obtained IPB-MSA&SO4 data allowed us to analyze the spatiotemporal patterns of MSA, SO4, and the ratio between them (MSA:SO4). A comparison with the existing CAMS-EAC4 reanalysis suggests that our high-resolution dataset reproduces with high accuracy the spatial and temporal patterns of the biogenic sulfur aerosol concentration and has high consistency with independent measurements in the Atlantic Ocean. The IPB-MSA&SO4 is publicly available at https://doi.org/10.17632/j8bzd5dvpx.1 (Mansour et al., 2023b).

show abstract

Section: Dimethylsulfide Flux Datamentioning

confidence: 99%

Section: Dimethylsulfide Flux Datamentioning

confidence: 99%

IPB-MSA&SO₄: a daily 0.25° resolution dataset of In-situ Produced Biogenic Methanesulfonic Acid and Sulfate over the North Atlantic during 1998–2022 based on machine learning

Mansour,

Decesari,

Ceburnis

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Efforts are being made to create hybrid ML-NM models that maximize the advantages of both approaches while minimizing their disadvantages. ML has been demonstrated to be effective in the analysis of environmental processes including predicting sea surface DMS concentrations 71,72 , aerosol-cloud interactions 73 , CCN concentrations 70,74 , and global particle number concentrations 75 . In the Arctic, k-means clustering has been utilized at several sites to understand the dynamics of aerosol particle number size distributions [76][77][78][79][80] and their links to environmental conditions (e.g., open water extent, transport patterns) as well as their sources and climate-relevant properties.…”

Section: Introductionmentioning

confidence: 99%

Pan-Arctic Methanesulfonic Acid Aerosol: Source regions, atmospheric drivers, and future projections

Pernov,

Harris,

Volpi

et al. 2024

Preprint

View full text Add to dashboard Cite

Natural aerosols are an important, yet understudied, part of the Arctic climate system. Natural marine biogenic aerosol components (e.g., methanesulfonic acid, MSA) are becoming increasingly important due to changing environmental conditions. In this study, we combine in situ aerosol observations with atmospheric transport modeling and meteorological reanalysis data in a data-driven framework with the aim to (1) identify the seasonal cycles and source regions of MSA, (2) elucidate the relationships between MSA and atmospheric variables, and (3) project the response of MSA based on trends extrapolated from reanalysis variables and determine which variables are contributing to these projected changes. We have identified the main source areas of MSA to be the Atlantic and Pacific sectors of the Arctic. Using gradient-boosted trees, we were able to explain 84 % of the variance and find that the most important variables for MSA are indirectly related to either the gas- or aqueous-phase oxidation of dimethyl sulfide (DMS): shortwave and longwave downwelling radiation, temperature, and low cloud cover. We project MSA to undergo a seasonal shift, with non-monotonic decreases in April/May and increases in June-September, over the next 50 years. Different variables in different months are driving these changes, highlighting the complexity of influences on this natural aerosol component. Although the response of MSA due to changing oceanic variables (sea surface temperature, DMS emissions, and sea ice) and precipitation remains to be seen, here we are able to show that MSA will likely undergo a seasonal shift solely due to changes in atmospheric variables.

show abstract

“…Later, DMS values were estimated across the oceanic biomes as a function of estimated DMSP and the satellite-based data of photosynthetically available radiation (PAR) using a similar regression analysis (Galí et al, 2018). An upgrade to this method is using machine learning, such as an artificial neural network (ANN) (Wang et al, 2020b) or Gaussian process regression (GPR) (Mansour et al, 2023) to create the parameterization. The climatology in these cases is created by training the machine learning algorithms in data-rich regions.…”

Section: Introductionmentioning

confidence: 99%

“…However, these estimations need a large dataset across different biogeochemical provinces to train the models. Another machine learning known as Gaussian Process Regression (GPR) was recently applied byMansour et al(2023) which was able to address ~71 % DMS variability at high temporal and spatial scale in North Atlantic Ocean as compared to observations. With fewer DMS points (~ 2236) the model results show that this can be an efficient tool for obtaining seawater DMS concentration and it may be successful in other oceanic regions or entire global ocean as well.…”

mentioning

confidence: 99%

Dimethyl sulfide (DMS) climatologies, fluxes and trends – Part A: Differences between seawater DMS estimations

Joge,

Mahajan,

Hulswar

et al. 2024

Preprint

View full text Add to dashboard Cite

Abstract. Dimethyl sulfide (DMS) is a naturally emitted trace gas that can affect the Earth's radiative budget by changing cloud albedo. Most models depend on regional or global distributions of seawater DMS concentrations and sea-air flux parameterizations to estimate its emissions. In this study, we analyze the differences between three estimations of seawater DMS, one of which is an observation-based interpolation method (Hulswar et al., 2022 (hereafter referred to as H22)) and two are proxy-based parameterization methods (Galí et al., 2018a (G18); Wang et al., 2020 (W20)). The interpolation-based method depends on the distribution of observations and the methods used to fill data between observations, while the parameterization-based methods rely on establishing a relationship between DMS and environmental parameters such as chlorophyll a, mixed layer depth, nutrients, sea surface temperature, etc., which can then be used to predict DMS concentrations. On average, the interpolation-based methods show higher DMS values compared to the parameterization-based methods. Even though the interpolation method shows higher values than the parameterization-based methods, it fails to capture mesoscale variability. The regression-based parameterization method (G18) shows the lowest values compared to other estimations, especially in the Southern Ocean, which is the high DMS region in Austral summer. The parameterization-based methods suggest significant positive long-term trends in seawater DMS (6.94 ±1.44 % decade-1 for G18 and 3.53 ±0.53 % decade-1 for W20). Since large differences, often more than 100 %, are observed between the different estimations of seawater DMS, the derived sea-air fluxes and hence the impact of DMS on the radiative budget are very sensitive to the estimate used.

show abstract

Machine learning for prediction of daily sea surface dimethylsulfide concentration and emission flux over the North Atlantic Ocean (1998–2021)

Cited by 11 publications

References 33 publications

IPB-MSA&SO₄: a daily 0.25° resolution dataset of In-situ Produced Biogenic Methanesulfonic Acid and Sulfate over the North Atlantic during 1998–2022 based on machine learning

IPB-MSA&SO₄: a daily 0.25° resolution dataset of In-situ Produced Biogenic Methanesulfonic Acid and Sulfate over the North Atlantic during 1998–2022 based on machine learning

Pan-Arctic Methanesulfonic Acid Aerosol: Source regions, atmospheric drivers, and future projections

Dimethyl sulfide (DMS) climatologies, fluxes and trends – Part A: Differences between seawater DMS estimations

Contact Info

Product

Resources

About

Machine learning for prediction of daily sea surface dimethylsulfide concentration and emission flux over the North Atlantic Ocean (1998–2021)

Cited by 11 publications

References 33 publications

IPB-MSA&SO4: a daily 0.25° resolution dataset of In-situ Produced Biogenic Methanesulfonic Acid and Sulfate over the North Atlantic during 1998–2022 based on machine learning

IPB-MSA&SO4: a daily 0.25° resolution dataset of In-situ Produced Biogenic Methanesulfonic Acid and Sulfate over the North Atlantic during 1998–2022 based on machine learning

Pan-Arctic Methanesulfonic Acid Aerosol: Source regions, atmospheric drivers, and future projections

Dimethyl sulfide (DMS) climatologies, fluxes and trends – Part A: Differences between seawater DMS estimations

Contact Info

Product

Resources

About

IPB-MSA&SO₄: a daily 0.25° resolution dataset of In-situ Produced Biogenic Methanesulfonic Acid and Sulfate over the North Atlantic during 1998–2022 based on machine learning

IPB-MSA&SO₄: a daily 0.25° resolution dataset of In-situ Produced Biogenic Methanesulfonic Acid and Sulfate over the North Atlantic during 1998–2022 based on machine learning