Estimation of pCO₂ Using Random Forest in the Seto Inland Sea, Japan

In this study, we developed a model to estimate <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{pCO}_{2}$</tex> from commonly measured parameters via the random forest model. The estimation accuracy was verified by comparing the model-estimated <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{pCO}_{2}$</tex> values with the measured values. The Root mean square error (RMSE) of the random forest method was (28.7 µatm ) approximately half that of the multiple linear regression method (52.7 µatm), indicating that the random forest method was more accurate and applicable. By building separate models for distinct areas, the accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{pCO}_{2}$</tex> estimation was improved. On the basis of the Seto Inland Sea Comprehensive Water Quality Survey data, we could estimate <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{pCO}_{2}$</tex> with at most a 10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> relative error, suggesting that estimation was possible with small errors. Four commonly measured parameters could be used to estimate <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{pCO}_{2}$</tex>.