A Verified Estimation of the El Nino Index Nino-3.4 since 1877

Decadal and longer time-scale variabilities of the best known El Nino-Southern Oscillation (ENSO) indexes are poorly correlated before 1950, and so knowledge of interdecadal variability and trend in ENSO indexes is dubious, especially before 1950. To address this problem, the authors constructed and compared physically related monthly ENSO indexes. The base index was El Nino index Nino-3.4, the sea surface temperature (SST) anomaly averaged over the equatorial box bounded by 5 degrees N, 5 degrees S, 170 degrees W, and 120 degrees W; the authors also constructed indexes based on the nighttime marine air temperature over the Nino-3.4 region (NMAT3.4) and an equatorial Southern Oscillation index (ESOI). The Nino-3.4 index used the "uninterpolated'' sea surface temperature data from the Second Hadley Centre Sea Surface Temperature dataset (HadSST2), a dataset with smaller uncertainty and better geographical coverage than others. In constructing the index, data at each point for a given month were weighted to take into account the typical considerable spatial variation of the SST anomaly over the Nino-3.4 box as well as the number of observations at that point for that month. Missing monthly data were interpolated and "noise'' was reduced by using the result that Nino-3.4 has essentially the same calendar month amplitude structure every year. This 12-point calendar month structure from April to March was obtained by an EOF analysis over the last 58 yr and then was fitted to the entire monthly time series using a least squares approach. Equivalent procedures were followed for NMAT3.4 and ESOI. The new ESOI uses Darwin atmospheric pressure in the west and is based on theory that allows for variations of the atmospheric boundary layer depth across the Pacific. The new Nino-3.4 index was compared with NMAT3.4, the new ESOI, and with a record of delta O-18 from a coral at Palmyra, an atoll inside the region Nino-3.4 (Cobb et al.). Correlation coefficients between Nino-3.4 and the three monthly indexes mentioned above before 1950 are 0.84, 0.87, 0.73 and 0.93, 0.86, 0.73 for decadal time scales. These relatively high correlation coefficients between physically related but independent monthly time series suggest that this study has improved knowledge of low-frequency variability. All four indexes are consistent with a rise in Nino-3.4 SST and the weakening of the equatorial Pacific winds since about 1970.