DroughtED: A dataset and methodology for drought forecasting spanning multiple climate zones

Numerical | Drought | Prevention

DroughtED is a dataset for drought forecasting. It contains 180 daily meteorological observations and geospatial location meta-data for 3,108 U.S. counties and is introduced for multiclass (ordinal) classification.

  • ML task type: Multiclass (ordinal) classification
  • Data Source: Publicly available datasets
  • Size: 180 Daily meteorological observations with geospatial location meta-data for 3,108 counties
  • Timespan: 2000 - 2020
  • Geographical Coverage: Continental US
  • Baseline Information
  • Evaluated on: LSTM, Transformer
  • Metrics used: MAE, macro F1-Scores
  • Results as reported in original paper: MAE: 0.433 (LSTM), F1: 47.5 (LSTM); MAE: 0.435 (Transformer), F1: 46.7 (Transformer)

Christoph D Minixhofer, Mark Swan, Calum McMeekin, and Pavlos Andreadis. DroughtED: A dataset and methodology for drought forecasting spanning multiple climate zones. In ICML 2021 Workshop on Tackling Climate Change with Machine Learning, 2021.