A Nonlinear Local Approximation Approach for Catchment Classification.

Area of Science:

• Hydrology and Environmental Science
• Nonlinear Dynamics and Chaos Theory

Background:

• Catchment classification is crucial for water resource management and environmental applications.
• Existing methods often rely on dimensionality measures from nonlinear dynamics and chaos theory.
• There is a need for alternative approaches to catchment classification.

Purpose of the Study:

• To explore prediction accuracy as a novel measure for catchment classification.
• To apply a nonlinear local approximation prediction method for this purpose.
• To assess the effectiveness of phase-space reconstruction using streamflow data.

Main Methods:

• Utilized a nonlinear local approximation prediction method based on phase-space reconstruction.
• Employed daily streamflow data from 218 Australian catchments.
• Analyzed prediction accuracy across various embedding dimensions and numbers of neighbors.

Main Results:

• Phase-space reconstruction using streamflow data alone yielded good prediction accuracy.
• Optimal predictions were achieved with lower embedding dimensions and fewer neighbors.
• This suggests potentially low dimensionality in streamflow dynamics.

Conclusions:

• Prediction accuracy derived from streamflow data is a viable method for catchment classification.
• The approach successfully identifies catchments with higher predictability.
• Findings have significant implications for streamflow data interpolation and extrapolation.