A Machine Learning Approach to Estimate Surface Ocean PCO2 from Satellite Measurements

Shuangling Chen; Chuanmin Hu; Brian B. Barnes; Rik Wanninkhof; Wei-Jun Cai; Leticia Barbero; Denis Pierrot

doi:10.1016/j.rse.2019.04.019

A Machine Learning Approach to Estimate Surface Ocean PCO2 from Satellite Measurements

Shuangling Chen, Chuanmin Hu, Brian B. Barnes, Rik Wanninkhof, Wei-Jun Cai, Leticia Barbero, Denis Pierrot

College of Marine Science

Research output: Contribution to journal › Article › peer-review

Abstract

Surface seawater partial pressure of CO 2 ( p CO 2 ) is a critical parameter in the quantification of air-sea CO 2 flux, which further plays an important role in quantifying the global <a title="Learn more about carbon budget from ScienceDirect's AI-generated Topic Pages"> carbon budget </a> and understanding <a title="Learn more about ocean acidification from ScienceDirect's AI-generated Topic Pages"> ocean acidification </a> . Yet, the remote estimation of p CO 2 in coastal waters (under influences of multiple processes) has been difficult due to complex relationships between environmental variables and surface p CO 2 . To date there is no unified model to remotely estimate surface p CO 2 in oceanic regions that are dominated by different oceanic processes. In our study area, the <a title="Learn more about Gulf of Mexico from ScienceDirect's AI-generated Topic Pages"> Gulf of Mexico </a> (GOM), this challenge is addressed through the evaluation of different approaches, including multi-linear regression (MLR), multi-nonlinear regression (MNR), principle component regression (PCR), decision tree, supporting vector machines (SVMs), multilayer <a title="Learn more about perceptron from ScienceDirect's AI-generated Topic Pages"> perceptron </a> neural network (MPNN), and random forest based regression ensemble (RFRE). After modeling, validation, and extensive tests using independent cruise datasets, the RFRE model proved to be the best approach. The RFRE model was trained using data comprised of extensive p CO 2 datasets (collected over 16 years by many groups) and MODIS (Moderate Resolution Imaging Spectroradiometer) estimated <a title="Learn more about sea surface temperature from ScienceDirect's AI-generated Topic Pages"> sea surface temperature </a> (SST), <a title="Learn more about sea surface salinity from ScienceDirect's AI-generated Topic Pages"> sea surface salinity </a> (SSS), surface chlorophyll concentration (Chl), and diffuse attenuation of <a title="Learn more about downwelling from ScienceDirect's AI-generated Topic Pages"> downwelling </a> irradiance (Kd). This RFRE-based p CO 2 model allows for the estimation of surface p CO 2 from satellites with a spatial resolution of ~1 km. It showed an overall performance of a root mean square difference (RMSD) of 9.1 μatm, with a coefficient of determination (R 2 ) of 0.95, a mean bias (MB) of −0.03 μatm, a mean ratio (MR) of 1.00, an unbiased percentage difference (UPD) of 0.07%, and a mean ratio difference (MRD) of 0.12% for p CO 2 ranging between 145 and 550 μatm. The model, with its original parameterization, has been tested with independent datasets collected over the entire GOM, with satisfactory performance in each case (RMSD of ≤~10 μatm for open GOM waters and RMSD of ≤~25 μatm for coastal and river-dominated waters). The sensitivity of the RFRE-based p CO 2 model to uncertainties of each input environmental variable was also thoroughly examined. The results showed that all induced uncertainties were close to, or within, the uncertainty of the model itself with higher sensitivity to uncertainties in SST and SSS than to uncertainties in Chl and Kd. The extensive validation, evaluation, and sensitivity analysis indicate the robustness of the RFRE model in estimating surface p CO 2 for the range of 145–550 μatm in most GOM waters. The RFRE model approach was applied to the Gulf of Maine (a contrasting oceanic region to GOM), with local model training. The results showed significant improvement over other models suggesting that the RFRE may serve as a robust approach for other regions once sufficient field-measured p CO 2 data are available for model training.

Original language	American English
Journal	Remote Sensing of Environment
Volume	228
DOIs	https://doi.org/10.1016/j.rse.2019.04.019
State	Published - Jan 1 2019

Keywords

Surface pCO2
SST
SSS
Chlorophyll
Kd
Satellite remote sensing
Gulf of Mexico

Disciplines

Life Sciences

Access to Document

10.1016/j.rse.2019.04.019License: CC BY

https://doi.org/10.1016/j.rse.2019.04.019

Cite this

@article{a9adac1fc0b24c3c945afc99604c6ab7,

title = "A Machine Learning Approach to Estimate Surface Ocean PCO2 from Satellite Measurements",

abstract = " Surface seawater partial pressure of CO 2 ( p CO 2 ) is a critical parameter in the quantification of air-sea CO 2 flux, which further plays an important role in quantifying the global carbon budget and understanding ocean acidification . Yet, the remote estimation of p CO 2 in coastal waters (under influences of multiple processes) has been difficult due to complex relationships between environmental variables and surface p CO 2 . To date there is no unified model to remotely estimate surface p CO 2 in oceanic regions that are dominated by different oceanic processes. In our study area, the Gulf of Mexico (GOM), this challenge is addressed through the evaluation of different approaches, including multi-linear regression (MLR), multi-nonlinear regression (MNR), principle component regression (PCR), decision tree, supporting vector machines (SVMs), multilayer perceptron neural network (MPNN), and random forest based regression ensemble (RFRE). After modeling, validation, and extensive tests using independent cruise datasets, the RFRE model proved to be the best approach. The RFRE model was trained using data comprised of extensive p CO 2 datasets (collected over 16 years by many groups) and MODIS (Moderate Resolution Imaging Spectroradiometer) estimated sea surface temperature (SST), sea surface salinity (SSS), surface chlorophyll concentration (Chl), and diffuse attenuation of downwelling irradiance (Kd). This RFRE-based p CO 2 model allows for the estimation of surface p CO 2 from satellites with a spatial resolution of ~1 km. It showed an overall performance of a root mean square difference (RMSD) of 9.1 μatm, with a coefficient of determination (R 2 ) of 0.95, a mean bias (MB) of −0.03 μatm, a mean ratio (MR) of 1.00, an unbiased percentage difference (UPD) of 0.07%, and a mean ratio difference (MRD) of 0.12% for p CO 2 ranging between 145 and 550 μatm. The model, with its original parameterization, has been tested with independent datasets collected over the entire GOM, with satisfactory performance in each case (RMSD of ≤~10 μatm for open GOM waters and RMSD of ≤~25 μatm for coastal and river-dominated waters). The sensitivity of the RFRE-based p CO 2 model to uncertainties of each input environmental variable was also thoroughly examined. The results showed that all induced uncertainties were close to, or within, the uncertainty of the model itself with higher sensitivity to uncertainties in SST and SSS than to uncertainties in Chl and Kd. The extensive validation, evaluation, and sensitivity analysis indicate the robustness of the RFRE model in estimating surface p CO 2 for the range of 145–550 μatm in most GOM waters. The RFRE model approach was applied to the Gulf of Maine (a contrasting oceanic region to GOM), with local model training. The results showed significant improvement over other models suggesting that the RFRE may serve as a robust approach for other regions once sufficient field-measured p CO 2 data are available for model training. ",

keywords = "Surface pCO2, SST, SSS, Chlorophyll, Kd, Satellite remote sensing, Gulf of Mexico",

author = "Shuangling Chen and Chuanmin Hu and Barnes, {Brian B.} and Rik Wanninkhof and Wei-Jun Cai and Leticia Barbero and Denis Pierrot",

year = "2019",

month = jan,

day = "1",

doi = "10.1016/j.rse.2019.04.019",

language = "American English",

volume = "228",

journal = "Remote Sensing of Environment",

}

TY - JOUR

T1 - A Machine Learning Approach to Estimate Surface Ocean PCO2 from Satellite Measurements

AU - Chen, Shuangling

AU - Hu, Chuanmin

AU - Barnes, Brian B.

AU - Wanninkhof, Rik

AU - Cai, Wei-Jun

AU - Barbero, Leticia

AU - Pierrot, Denis

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Surface seawater partial pressure of CO 2 ( p CO 2 ) is a critical parameter in the quantification of air-sea CO 2 flux, which further plays an important role in quantifying the global carbon budget and understanding ocean acidification . Yet, the remote estimation of p CO 2 in coastal waters (under influences of multiple processes) has been difficult due to complex relationships between environmental variables and surface p CO 2 . To date there is no unified model to remotely estimate surface p CO 2 in oceanic regions that are dominated by different oceanic processes. In our study area, the Gulf of Mexico (GOM), this challenge is addressed through the evaluation of different approaches, including multi-linear regression (MLR), multi-nonlinear regression (MNR), principle component regression (PCR), decision tree, supporting vector machines (SVMs), multilayer perceptron neural network (MPNN), and random forest based regression ensemble (RFRE). After modeling, validation, and extensive tests using independent cruise datasets, the RFRE model proved to be the best approach. The RFRE model was trained using data comprised of extensive p CO 2 datasets (collected over 16 years by many groups) and MODIS (Moderate Resolution Imaging Spectroradiometer) estimated sea surface temperature (SST), sea surface salinity (SSS), surface chlorophyll concentration (Chl), and diffuse attenuation of downwelling irradiance (Kd). This RFRE-based p CO 2 model allows for the estimation of surface p CO 2 from satellites with a spatial resolution of ~1 km. It showed an overall performance of a root mean square difference (RMSD) of 9.1 μatm, with a coefficient of determination (R 2 ) of 0.95, a mean bias (MB) of −0.03 μatm, a mean ratio (MR) of 1.00, an unbiased percentage difference (UPD) of 0.07%, and a mean ratio difference (MRD) of 0.12% for p CO 2 ranging between 145 and 550 μatm. The model, with its original parameterization, has been tested with independent datasets collected over the entire GOM, with satisfactory performance in each case (RMSD of ≤~10 μatm for open GOM waters and RMSD of ≤~25 μatm for coastal and river-dominated waters). The sensitivity of the RFRE-based p CO 2 model to uncertainties of each input environmental variable was also thoroughly examined. The results showed that all induced uncertainties were close to, or within, the uncertainty of the model itself with higher sensitivity to uncertainties in SST and SSS than to uncertainties in Chl and Kd. The extensive validation, evaluation, and sensitivity analysis indicate the robustness of the RFRE model in estimating surface p CO 2 for the range of 145–550 μatm in most GOM waters. The RFRE model approach was applied to the Gulf of Maine (a contrasting oceanic region to GOM), with local model training. The results showed significant improvement over other models suggesting that the RFRE may serve as a robust approach for other regions once sufficient field-measured p CO 2 data are available for model training.

AB - Surface seawater partial pressure of CO 2 ( p CO 2 ) is a critical parameter in the quantification of air-sea CO 2 flux, which further plays an important role in quantifying the global carbon budget and understanding ocean acidification . Yet, the remote estimation of p CO 2 in coastal waters (under influences of multiple processes) has been difficult due to complex relationships between environmental variables and surface p CO 2 . To date there is no unified model to remotely estimate surface p CO 2 in oceanic regions that are dominated by different oceanic processes. In our study area, the Gulf of Mexico (GOM), this challenge is addressed through the evaluation of different approaches, including multi-linear regression (MLR), multi-nonlinear regression (MNR), principle component regression (PCR), decision tree, supporting vector machines (SVMs), multilayer perceptron neural network (MPNN), and random forest based regression ensemble (RFRE). After modeling, validation, and extensive tests using independent cruise datasets, the RFRE model proved to be the best approach. The RFRE model was trained using data comprised of extensive p CO 2 datasets (collected over 16 years by many groups) and MODIS (Moderate Resolution Imaging Spectroradiometer) estimated sea surface temperature (SST), sea surface salinity (SSS), surface chlorophyll concentration (Chl), and diffuse attenuation of downwelling irradiance (Kd). This RFRE-based p CO 2 model allows for the estimation of surface p CO 2 from satellites with a spatial resolution of ~1 km. It showed an overall performance of a root mean square difference (RMSD) of 9.1 μatm, with a coefficient of determination (R 2 ) of 0.95, a mean bias (MB) of −0.03 μatm, a mean ratio (MR) of 1.00, an unbiased percentage difference (UPD) of 0.07%, and a mean ratio difference (MRD) of 0.12% for p CO 2 ranging between 145 and 550 μatm. The model, with its original parameterization, has been tested with independent datasets collected over the entire GOM, with satisfactory performance in each case (RMSD of ≤~10 μatm for open GOM waters and RMSD of ≤~25 μatm for coastal and river-dominated waters). The sensitivity of the RFRE-based p CO 2 model to uncertainties of each input environmental variable was also thoroughly examined. The results showed that all induced uncertainties were close to, or within, the uncertainty of the model itself with higher sensitivity to uncertainties in SST and SSS than to uncertainties in Chl and Kd. The extensive validation, evaluation, and sensitivity analysis indicate the robustness of the RFRE model in estimating surface p CO 2 for the range of 145–550 μatm in most GOM waters. The RFRE model approach was applied to the Gulf of Maine (a contrasting oceanic region to GOM), with local model training. The results showed significant improvement over other models suggesting that the RFRE may serve as a robust approach for other regions once sufficient field-measured p CO 2 data are available for model training.

KW - Surface pCO2

KW - SST

KW - SSS

KW - Chlorophyll

KW - Kd

KW - Satellite remote sensing

KW - Gulf of Mexico

UR - https://digitalcommons.usf.edu/msc_facpub/2034

UR - https://doi.org/10.1016/j.rse.2019.04.019

U2 - 10.1016/j.rse.2019.04.019

DO - 10.1016/j.rse.2019.04.019

M3 - Article

VL - 228

JO - Remote Sensing of Environment

JF - Remote Sensing of Environment

ER -

A Machine Learning Approach to Estimate Surface Ocean PCO2 from Satellite Measurements

Abstract

Keywords

Disciplines

Access to Document

Other files and links

Cite this