Distinct hydrologic response patterns and trends worldwide revealed by physics-embedded learning

Area of Science:

• Hydrology
• Climate Science
• Environmental Modeling

Background:

• Global Water Models (GWMs) struggle to represent dynamic hydrologic system responses due to limited data assimilation capabilities.
• Accurately tracking rapid changes in water resources requires models that capture characteristic hydrologic response patterns (signatures).

Purpose of the Study:

• To introduce a novel, high-resolution, physics-embedded, big-data-trained model for improved hydrologic signature analysis.
• To assess global shifts in water balance components and their ecological and management implications.

Main Methods:

• Development of a physics-embedded, big-data-trained model integrating high-resolution data.
• Realistic representation of long-term water balance, including green-blue-water partitioning and baseflow ratios.
• Analysis of shifts in hydrologic response patterns over 20 years.

Main Results:

• Widespread shifts exceeding 20% in water partitioning and baseflow ratios observed globally over two decades.
• Identified contributions of these shifts to increased flood risks (northern mid-latitudes), water supply stress (subtropical regions), and declining freshwater inputs (Europe).
• Demonstrated superior accuracy in simulating monthly and daily runoff responses, including elasticity and flashiness, especially in arid/semi-arid regions.

Conclusions:

• The next-generation model provides substantially more accurate simulations than current operational systems.
• Highlighted regions facing management challenges due to water supply variability and climate sensitivity.
• The advanced tool offers reliable forecasting of seasonal water availability for global water management.