wass2s — West Africa Seasonal Forecasting System

wass2s is a Python library that streamlines the full pipeline of seasonal climate forecasting over West Africa and the Sahel. Starting from raw satellite, reanalysis, and GCM data, it lets you build, validate, and operationalise statistical and machine-learning-based forecasts in a reproducible way.

The library implements the new-generation seasonal forecasting framework promoted by the World Meteorological Organization (WMO), covering everything from predictor preparation to probabilistic tercile-probability maps (Below-Normal / Near-Normal / Above-Normal).

Note

A companion set of Jupyter notebooks is available for hands-on walkthroughs of each major workflow.

Key capabilities

• Automated data acquisition — ERA5, CHIRPS, TAMSAT, NMME, C3S GCMs, and AgERA5 agro-indicators in a single unified interface.

• Agroclimatic predictand computation — onset, cessation, dry/wet spells, ETCCDI extreme indices, and heat-wave metrics from daily gridded or station data.

• Bias correction and data merging — quantile mapping (QUANT, RQUANT, SSPLIN, PTF, DIST) and station–gridded merging (Kriging, Regression Kriging, Neural-Network Kriging, Multiplicative Bias).

• Statistical and ML models — OLS, Ridge, Lasso, ElasticNet, MARS, SVR, MLP, stacking ensembles, EOF/PCR, CCA, and analog methods, all sharing a common compute_model / compute_prob / forecast interface.

• Multi-model ensemble post-processing — weighted averaging, BMA, NGR, Extended Logistic Regression, Random Forest / XGBoost / ELM / MLP super-ensembles.

• Leakage-free cross-validation — a custom leave-one-out splitter with a symmetric exclusion window, wired to every model family automatically.

• Forecast verification — deterministic (KGE, Pearson, NSE, RMSE, MAE) and probabilistic (GROC, RPSS, Brier, Ignorance, ROC, Reliability) metrics.

Getting started

• Installation
• Quick Start

User guide

• Data Download
  • 1. Seasonal GCM Forecasts
  • 2. Reanalysis Data
  • 3. Observational Data
• Preprocessing
  • Input data format conventions
  • Computing Agroclimatic Predictands
  • Merging Gridded Data with Station Observations
  • Bias Correction
  • Data Transformation and Distribution Fitting
• Models and Cross-Validation
  • Linear Regression Models
  • Advanced Machine-Learning Models
  • EOF Analysis and Principal Component Regression
  • Canonical Correlation Analysis
  • Analog Forecasting
  • Cross-Validation
• Verification
  • Deterministic Metrics
  • Probabilistic Metrics
  • Ensemble Metrics
  • Year-by-Year Validation
  • Tercile Classification Utilities
• Multi-Model Ensemble Methods
  • 1. Data Preparation
  • 2. Weighted Ensemble Averaging
  • 3. Min et al. (2009) Probabilistic MME
  • 4. Machine-Learning Ensemble Methods
  • 5. Calibration and Post-Processing
• Implementing WAS-NextGen Approaches
  • Overview of the WAS-NextGen pipeline
  • Recommended model hierarchy
  • Running multiple models and selecting the best
  • Multi-model ensemble of the best performers
  • Operational forecast dissemination
  • Agroclimatic forecast products

API reference

• API Reference

Indices and tables

• Index

• Module Index

• Search Page