Decompose
Separate each predictor into time-localised frequency components with a discrete wavelet transform.
WASP Overview · WAvelet System Prediction
Find the signal across scales.
WASP refines a predictor's spectral representation before modelling. It isolates informative frequency bands, adjusts their influence, and reconstructs predictors that better express hydroclimatic predictability.
Methodology
Decompose → Identify → Modulate → Reconstruct.
Identify
Estimate the predictive contribution of components at different scales using the target series.
Modulate
Amplify informative components and attenuate components that contribute little predictive signal.
Reconstruct
Combine the modulated components into spectrally refined predictors for downstream modelling.
Online workflow
Prepare a transparent experiment.
CSV input
Upload a UTF-8 comma-separated file with 30–5000 complete observations.
- Select one predictand (Y) and 1–50 predictors (X) after loading
- Only selected Y/X columns must be numeric, finite, and non-constant
- Unused date, identifier, or text columns are ignored
- Resource limit: 51 total columns, including unused columns
- Maximum CSV file: 10 MB (the proxy allows 11 MB including multipart overhead)
Run parameters
wavelet: db1 (Haar), db2, db4, db8, or db16level: decomposition depth; 0 selects automaticallytest_size: 20%, 25%, 33%, or 50% chronological holdoutmodel: Linear Regression, K-Nearest Neighbors, or XGBoost
Open-source software
Choose your research environment.
WASP
The reference research implementation and established workflow.
Open R repository ↗WASP Python
A Python implementation for scientific and machine-learning pipelines.
Open Python repository ↗WASP MATLAB
A MATLAB implementation for signal-processing research workflows.
Open MATLAB repository ↗Foundational publications
Method, software and applications.
Water Resources Research · 2020
Refining predictor spectral representation using wavelet theory for improved natural system modelling
The methodological foundation for using wavelet-scale information to refine predictors.
Read the WRR method paper ↗Environmental Modelling & Software · 2021
WASP: a wavelet-based tool to modulate variance in predictors for improved prediction
The software paper describes the open implementation and practical modelling workflow.
Read the software paper ↗Journal of Hydrology · 2021
Variable transformations in the spectral domain – Implications for hydrologic forecasting
An application and extension of spectral variance transformation for long-lead hydrologic forecasting.
Read the hydrologic forecasting paper ↗Run WASP in your browser.
Load the demonstration data or upload a compatible CSV, select the transform settings, and compare WASP predictions with a raw-predictor baseline.