Python framework for short-term ensemble prediction systems.
scores: Metrics for the verification, evaluation and optimisation of forecasts, predictions or models.
Forecast Verification/Validation Tools in Python
Predicting where India's medium-range weather forecasts will bust, district by district, and explaining why. XGBoost + SHAP + FastAPI + React, served for $0. SIH 2026 PS 26079 (Ministry of Earth Sciences).
A verification pipeline for evaluating models and forecasts.
Interactive Streamlit dashboard to verify gridded rainfall/precipitation forecasts (deterministic & ensemble, NetCDF + GRIB2) against observations — POD/FAR/CSI/ETS/HSS, EDI/SEDI/EDS/SEDS, Brier/BSS/CRPS, FSS & CRA, reliability/ROC/rank/PIT/Taylor plots, per-cell score maps, bootstrap CIs, CSV/PNG export. Runs on bundled demo data.
Climate Signals: frozen rare-extreme climate conditions (ENSO, drought, Atlantic and Pacific modes) scored forward against markets, plus a registered paper-portfolio experiment. Live counterpart of Paper 17.
Verification of spatial-structure in ensembles of forecast fields
Evaluation framework for precipitation nowcasting: radar extrapolation, NWP, and blends
Ensemble forecast verification in Python using reliability diagrams for streamflow data.
Radiación solar: IFS y AIFS frente a 34 estaciones SiAR. Resultados, sensibilidad de escala y reproducción sin red.
Daily day-ahead forecasts for European power (DE-LU and Spain: wind, solar, load, price), scored the next day against the TSO's own forecast and persistence. Every file timestamped with OpenTimestamps, nothing edited after the fact.
Temporal adaptation of the gamma index for time series forecast evaluation under joint amplitude and timing tolerances, with applications to renewable energy forecasting.
Public evidence slice of Solstitium's daily climate-forecast record: forecast-outcome verification pairs, model bakeoff, index term sheets, and OpenTimestamps-sealed daily anchors.
Offline space-weather evidence, UTC data quality, leakage-aware forecast verification and accessible analysis by Jason Stys
A stock model that gained 99% in a walk-forward backtest lost 24% in an 11-week frozen live paper test. Code, data and a reproducible statistical post-mortem: block bootstrap, rank IC, calibration, multiple testing, pseudo-replication.
Config-driven forecast verification for NWP and AI weather models
Two years of archived weather forecasts scored against the guess you could make without them.
Do published probabilistic forecasts hold up when you score them? Pre-registered verification of NWS precipitation, AirNow air quality, and CAISO day-ahead power prices — one scoring toolkit, three domains.
Adaptive multi-model weather forecast blending: learns which NWP or AI model to trust by region, lead time, season and weather regime, with uncertainty, IMD-aligned extreme guidance and honest verification. SIH 2026 · SIH26081 · MoES.
To associate your repository with the forecast-verification topic, visit your repo's landing page and select "manage topics."