A Data-driven approach with reanalysis and geospatial data for chloride deposition prediction

Accurately estimating atmospheric chloride deposition can offer important insights into metal corrosion. Corrosion of metals causes high associated costs in engineering work. Hence, there are standardized models to estimate the corrosivity of environments. One entry variable of such models is airborne salinity that comes from the sea, collected with the wet candle method and measured in chloride ion dry deposition rate. This method has some limitations because it is complex and demands long periods of exposure to environments. This work proposes a new data-driven approach for the prediction of dry deposition of chloride ions using reanalysis and geospatial data. Hence, we have developed a machine-learning regression model based on Random Forests for the prediction of airborne salinity. This work also analyzes variables and compares the proposed approach with other available models. The Random Forest obtained the best results with an r \(^{2}\) $_{}^2$ of 0.82.