The global freshwater system: Patterns and predictability of green-blue water flux partitioning

TL;DR

This study maps global patterns of freshwater flux partitioning, revealing dominant green water fluxes, and develops a machine learning model to predict future water availability under climate and land-use changes.

Contribution

It provides the first comprehensive global analysis of water flux partitioning and introduces a machine learning model for predicting future green and blue water resources.

Findings

Green water fluxes are higher than blue in most regions.

Land-use changes will increase green water flux, reducing blue water availability.

The ML model can predict future water flux partitioning under climate scenarios.

Abstract

The partitioning of precipitation (P) water input on land between green (evapotranspiration, ET) and blue (runoff, R) water fluxes distributes the annually renewable freshwater resource among sectors and ecosystems. We decipher the worldwide pattern and key determinants of this water flux partitioning (WFP) and investigate its predictability based on a machine learning (ML) model trained and tested on data for 3,614 hydrological catchments around the world. The results show considerably higher WFP to the green (ET/P) than the blue (R/P) flux in most of the world. Land-use changes toward expanded agriculture and forestry will increase this WFP asymmetry, jeopardizing blue-water availability and making it more vulnerable to future P changes for other sectors and ecosystems. The predictive ML-model of WFP developed in this study can be used with climate model projections of P to assess future blue and green water security for various regions, sectors, and ecosystems around the world.