Direct and indirect simulating and projecting hydrological drought using a supervised machine learning method

Area of Science:

• Hydrology
• Artificial Intelligence
• Climate Change Research

Background:

• Artificial Intelligence (AI) is increasingly used in hydrological simulations for tasks like river discharge and drought prediction.
• Assessing different AI application concepts is crucial for improving hydrological drought simulations and projections.

Purpose of the Study:

• To evaluate two distinct AI-based approaches for simulating and projecting hydrological drought using the Standardized Runoff Index (SRI).
• To compare a direct SRI simulation method against an indirect method involving river discharge simulation.

Main Methods:

• Standardized Runoff Index (SRI) was simulated and projected using Artificial Neural Networks (ANNs).
• Predictors included temperature, precipitation, and the Standardized Precipitation Index (SPI).
• Two approaches were assessed: direct SRI simulation and indirect simulation via river discharge.

Main Results:

• The indirect approach demonstrated superior performance in SRI simulations across four discharge stations in the Odra River Basin (2000-2019).
• Significant discrepancies were observed between the two approaches in projecting hydrological drought under the RCP8.5 scenario for near and far future horizons.
• Both methods indicated similar drought conditions for future projections based on run theory.

Conclusions:

• The indirect AI approach for hydrological drought simulation, incorporating river discharge, is more effective than direct SRI simulation.
• Future hydrological drought projections show notable differences between direct and indirect AI methods, though run theory suggests similar outcomes.
• This research provides insights into optimizing AI applications for accurate hydrological drought assessment and forecasting.