Using Random Forests to Compare the Sensitivity of Observed Particulate Inorganic and Particulate Organic Carbon to Environmental Conditions

The balance between particulate inorganic carbon (PIC) and particulate organic carbon (POC) holds significant importance in carbon storage within the ocean. A recent investigation delved into the spatial distribution of phytoplankton and the physiological mechanisms governing their growth. Employing random forests, a machine learning technique, this study unveiled apparent relationships between POC and 10 environmental fields. In this work, we extend the use of random forests to compare how observed PIC and POC respond to environmental conditions. PIC and POC exhibit similar responses to certain environmental drivers, suggesting that these do not explain differences in their distribution. However, PIC is less sensitive to iron and more sensitive to light and mixed layer depth. Intriguingly, both PIC and POC display weak sensitivity to CO2, contrary to previous studies, possibly due to the elevated pCO2 in our data set. This research sheds light on the underlying processes influencing carbon sequestration and ocean productivity. This study looks at how different types of carbon, specifically tiny particles of chalk (particulate inorganic carbon, PIC) and organic carbon from microscopic marine plants (particulate organic carbon, POC), are distributed in the ocean and how they respond to environmental conditions. The ratio between PIC and POC has a big impact on how carbon is stored in the ocean. We used a machine learning technique to analyze how patterns in these fields estimated from satellite were related to drivers such as light and nutrients. We found that PIC and POC react similarly to some environmental factors (such as ammonium) but differently to others (such as iron and light). Surprisingly, both types of carbon showed less sensitivity to CO2 than expected from previous work, possibly because of high CO2 levels in the data set. Particulate inorganic carbon (PIC) and particulate organic carbon (POC) estimated from satellites can be robustly related to environmental conditions Random forests produce similar nonlinear relationships between some environmental factors (i.e., ammonium) and PIC and POC PIC is less sensitive to iron and more sensitive to light and mixed layer depth than POC