El Niño Project (Part 3)
climate-networksel-nino-predictiontime-seriescross-correlationnetwork-theory
Abstraction: Climate network cross-correlation method for early El Niño prediction
Key points:
- Ludescher et al. (PNAS Feb 2014) claimed 75% probability of El Niño by end of 2014, predicted ~6+ months early using climate networks
- Network uses 14 grid points inside the El Niño basin and 193 outside; link strength = (max - mean) / std of time-delayed cross-correlation for lags up to ±200 days
- Average link strength threshold of 2.82 triggers prediction; threshold chosen in a learning phase (1950-1980s) and tested on later data
- Niño 3.4 index: area-averaged sea surface temperature anomaly in 5°S-5°N, 170°-120°W; El Niño declared when > 0.5°C for 5+ months
- Mathematical nuance: Ludescher et al.'s normalization formula contains nested running averages inconsistent with their data start date (graph shows 1950, but formula requires 930 days of data from 1948 start)
- Data source: NCEP Reanalysis Daily Averages (144×73 grid, 2.5° resolution, 1948 to present) in NetCDF format
Connections: Azimuth Project · Noaa · Climate Networks · Time Series Analysis · Cross Correlation
Source: http://johncarlosbaez.wordpress.com/2014/07/01/el-nino-project-part-3/