Predicting lake dissolved organic carbon at a global scale.

Area of Science:

• Environmental Science
• Biogeochemistry
• Ecology
• Machine Learning

Background:

• Dissolved organic carbon (DOC) is a key regulator of inland water ecology and biogeochemistry.
• DOC influences carbon budgets in terrestrial ecosystems.
• Understanding global DOC variability in lakes is crucial for climate and carbon cycle modeling.

Purpose of the Study:

• To investigate environmental factors driving variability in in situ DOC concentrations in lakes globally.
• To predict DOC concentrations and pools in lakes worldwide using machine learning.

Main Methods:

• Utilized a novel machine learning technique.
• Employed global databases for environmental and lake data.
• Predicted DOC for lakes larger than 0.1 km².

Main Results:

• Catchment properties, meteorological, and hydrological factors were primary drivers of lake DOC variability.
• Lake morphometry played a minor role in DOC concentration.
• The predicted global average DOC concentration was 3.88 mg L⁻¹.
• The global predicted DOC pool in lakes was 729 Tg, with the Caspian Sea contributing 421 Tg.

Conclusions:

• Environmental factors significantly explain global lake DOC variability.
• Machine learning provides a robust method for estimating global lake DOC.
• Findings offer critical data for ecological, climate, and carbon cycle models.