Predicting lake dissolved organic carbon at a global scale

Kaire Toming,Tiit Kutser,Evelyn Uuemaa,Jonne Kotta,Lars J Tranvik,Sebastian Sobek

doi:10.1038/s41598-020-65010-3

Abstract

The pool of dissolved organic carbon (DOC), is one of the main regulators of the ecology and biogeochemistry of inland water ecosystems, and an important loss term in the carbon budgets of land ecosystems. We used a novel machine learning technique and global databases to test if and how different environmental factors contribute to the variability of in situ DOC concentrations in lakes. In order to estimate DOC in lakes globally we predicted DOC in each lake with a surface area larger than 0.1 km2. Catchment properties and meteorological and hydrological features explained most of the variability of the lake DOC concentration, whereas lake morphometry played only a marginal role. The predicted average of the global DOC concentration in lake water was 3.88 mg L−1. The global predicted pool of DOC in lake water was 729 Tg from which 421 Tg was the share of the Caspian Sea. The results provide global-scale evidence for ecological, climate and carbon cycle models of lake ecosystems and related future prognoses.

Highlights

The pool of dissolved organic carbon (DOC), is one of the main regulators of the ecology and biogeochemistry of inland water ecosystems, and an important loss term in the carbon budgets of land ecosystems
We used the novel machine learning technique Boosted Regression Trees (BRT)[20], HydroLAKES v. 1.021,22 and WorldClim v. 2.023 databases, and a compilation of globally distributed lake DOC concentration data[9] (1) to test if and how different factors contribute to the variability of in situ DOC concentrations in lakes and (2) based on the established relationships to predict DOC in lakes globally
Temperature is positively related to DOC concentration up to ~7 °C, but did not have any effect >7 °C is similar to what Laudon et al.[24] observed for stream DOC across regions

Summary

Introduction

The pool of dissolved organic carbon (DOC), is one of the main regulators of the ecology and biogeochemistry of inland water ecosystems, and an important loss term in the carbon budgets of land ecosystems. Regional DOC models seem to have a limited potential in predicting DOC in other geographical areas of the world but the proportion of wetlands in the watershed and lake elevation are good predictors of lake DOC concentration across regions[16] This is due to a hierarchical regulation of DOC in lakes, where climatic and topographic characteristics set the regional range of DOC concentrations, and catchment and lake properties define the DOC concentration in each individual lake[9,17]. In another study the lake total organic carbon (TOC) was only weakly related to morphological characteristics whereas climatic controls described nearly half of variability in TOC18 Based on these relationships the estimated global mean concentrations and storage of TOC in lake water are 5.58 mg L−1 and 984 Tg18. We expect that the BRT modelling captures the complex patterns of DOC in lakes and improves our understanding of the causes of that variation

Methods

Results

Conclusion