Forecasting El Niño and La Niña events using decision tree classifier

• Published in: Theoretical and applied climatology 2022-05, Vol.148 (3-4), p.1279-1288
• Main Authors: Silva, Karita Almeida, de Souza Rolim, Glauco, de Oliveira Aparecido, Lucas Eduardo
• Format: Article
• Language: English
• Subjects: Accuracy; Aquatic Pollution; Atmospheric models; Atmospheric Protection/Air Quality Control/Air Pollution; Atmospheric Sciences; Classifiers; Climate change; Climate models; Climate science; Climatology; Computer applications; Computing time; Crop yield; Decision trees; Earth and Environmental Science; Earth Sciences; El Nino; El Nino forecasting; El Nino phenomena; El Nino-Southern Oscillation event; Forecasting; Global climate; Global weather; La Nina; La Nina events; Meteorological satellites; Modelling; Ocean models; Ocean temperature; Ocean temperature measurements; Oceans; Original Paper; Precipitation (Meteorology); Precipitation patterns; Southern Oscillation; Temperature measurement; Tropical climate; Waste Water Technology; Water Management; Water Pollution Control; Weather
• ISSN: 0177-798X; 1434-4483
• DOI: 10.1007/s00704-022-03999-5

The El Niño-Southern Oscillation (ENSO) phenomenon affects the global climate by changing temperature and precipitation patterns mainly in tropical climatic regions and median latitudes. Such event strongly influences agricultural activities and crop yields. The Niño Oceanic Index (ONI) of the US National Oceanic and Atmospheric Administration (NOAA) describes and monitors ENSO intensity from ocean temperature measurements. When ONI in the Niño 3.4 region was + 0.5 °C above normal or − 0.5 °C below normal for 5 consecutive 3-month running averages, El Niño (EN) or La Niña (LN) events, respectively, were established. The prediction of ENSO events is made by modeling at major global weather centers by atmosphere–ocean coupling models; however, no articles were found using decision tree classifier (DTC) for ENSO forecasting purposes. This modeling approach requires much less computational time and capacity. Furthermore, DTC can be sufficiently accurate for agricultural purposes. Thus, the objective of this research was to forecast as early as possible the El Niño and La Niña yearly events using a DTC technique from ONI data from 1950 to 2020. We used as input variables for DTC quarterly ONI values from 15 quarters prior the data of forecasting. The DTC showed an accuracy of 89%, 84%, and 78% to predict La Niña, El Niño, and neutral years, respectively, without training period. For validation, the accuracy was 100%, 79%, and 79% for La Niña, El Niño, and neutral years, respectively. The selected ONI quarters were July–August-September, January–February-March, and February–March-April of the previous year and January–February-March of the current year, allowing an 8-month advance forecast with an average accuracy of 78% (validation).