Driving factors of TOC concentrations in four different types of estuaries (canal, urban, agricultural, and natural estuaries) identified by machine learning technique

Abstract

Mangroves are among the most significant organic carbon sinks on Earth. However, the drivers of mangrove carbon remain poorly understood due to the lack of data on organic carbon across different types of estuaries. In this study, boosted regression tree (BRT) was used to relate Total Organic Carbon (TOC) concentrations to explanatory variables describing water quality, nutrients, and meteorological characteristics of four estuaries in southwest Florida, USA. TOC concentrations were primarily enhanced by low salinity, with elevated NO₂⁻-N concentrations in natural estuary, increased PO₄³⁻-P concentrations in agricultural estuary, decreased PO₄³⁻-P concentrations in canal estuary, and higher temperatures in urban estuary. These findings suggest that terrestrial inputs and photosynthesis play a crucial role in regulating TOC concentrations. Nutrient behavior varied depending on the specific estuarine environment. Based on these relations, a BRT model was developed to predict the annual average TOC fluxes in estuaries from 2002 to 2008. During the wet season, TOC fluxes in the estuaries showed net export, whereas in the dry season, most estuaries exhibited net input, except for the Henderson estuary, which showed net export. Notably, the canal estuary had the highest fluxes of both export (62.81 t/d) and input (−3.27 t/d). These results indicate that estimating TOC fluxes should be based on local environmental factors such as salinity, nutrients (NO₂⁻-N, PO₄³⁻-P), and temperature. Carbon sinks and sources should be more accurately identified by considering the differences in mangrove estuary types and seasonal variability.