Precipitation Forecasts for the Tropical Pacific Islands Using Canonical Correlation Analysis (CCA)

contributed by Yuxiang He and Tony Barnston

Climate Prediction Center, NOAA, Camp Springs, Maryland

Canonical correlation analysis (CCA) identifies linear relationships between multicomponent predictors and multicomponent predictands, e.g. pattern-to-pattern relationships in space and/or time. Like simpler forms of linear regression, CCA minimizes squared errors in hindcasting the predictands from the predictors. During the last decade, CCA has started being used increasingly in the atmospheric sciences (e.g. Barnett and Preisendorfer 1987; Graham et al. 1987a, 1987b; Barnston and Ropelewski 1992; Barnston 1994, Barnston and He 1996). Here, CCA is used to predict 3-month precipitation anomalies in the Pacific Islands out to a year in advance, as described in He and Barnston (1996). Because rainfall in the tropical and subtropical Pacific is strongly related to ENSO (Ropelewski and Halpert 1987, 1996), it is reasonable to expect usable skill in seasonal Pacific rainfall forecasts, and thus worthwhile to establish a real-time prediction system for the benefit of commercial interests in the Pacific Islands.

The predictor fields used for the forecasts include quasi-global sea surface temperature (SST), Northern Hemisphere 700 mb geopotential height, and the predictand precipitation itself (at 33 island stations) at an earlier time. CCA sensitivity experiments indicate that the SST field is the most valuable predictor field, with 700 mb heights and prior precipitation somewhat helpful. Further details about the skill, the underlying relationships, and the predictors are provided in He and Barnston (1996). The set of predictors is configured as four consecutive 3-month periods prior to the time of the forecast, followed by a variable lead time, and then a single 3-month predictand period. The predictand includes 3-month total rainfall at 33 Pacific Island stations within 25^oN-30^oS, including 4 Hawaiian stations (see Fig. 1 of any of the 1997 issues of this Bulletin). The lead time is defined as the time between the end of the final (fourth) predictor period (i.e., the time of the forecast) and the beginning of the 3-month predictand period. The set of stations predicted is planned to be increased in the near future. The rainfall data and climatology for the larger set of stations has been described in a recently published atlas (He et al. 1998).

The expected skill of the forecasts was estimated using a 1-year-out cross-validation methodology (see He and Barnston 1996). These skill estimates indicate that at 1 month lead time the highest correlation skill across the Pacific Islands occurs in Jan-Feb-Mar at 0.44 (0.29) averaged over all stations north (south) of the equator, and the lowest occurs from September through December at about 0.15 (0.30) for stations north (south) of the equator. At 4 months lead, the skill is only slightly lower except for the Jan-Feb-Mar average skill north of the equator which drops to 0.26. Skill for only warm or cold ENSO years is higher.

Figure 1 shows forecasts of the standardized precipitation anomaly (X100) for 33 Pacific Island stations using data through August 1998. The top panel shows the forecast for Oct-Nov-Dec 1998 (1 month lead), and the middle and bottom panels for Jan-Feb-Mar and Apr-May-Jun 1999 (4 and 7 months lead), respectively. The expected skill for these forecasts, based on cross-validation, is shown by the size of the numerals (as opposed to their value, which is the forecast itself): Small numerals indicate low skill (correlation below 0.3), medium sized numerals indicate usable but modest skill (correlation between 0.3 and 0.45), and large numerals indicate moderate or better skill (0.45 and higher). A tendency toward enhanced rainfall at many of the off-equator locations is forecast for Oct-Nov-Dec and Jan-Feb-Mar. While not shown due to the lack of stations along the immediate equator near and east of the date line, (i.e. Kiribati) suppressed rainfall would be forecast. This pattern is associated with the La Niña condition that developed in May and June this year. At Hawaii, however, weak residual effects of the previous El Niño are forecast to remain for a while longer. Confidence in the above normal rainfall at many of the off-equator stations is proportional to our confidence that the cold episode will persist through the remainder of 1998. By late boreal spring 1999, climate anomalies associated with the La Niña condition are expected to dissipate. While the SOI and the trade wind strength have not been very high in recent months (they have been only average or slightly higher), a tendency toward continuation of the cold conditions exists in view of the cold subsurface waters. The more the atmosphere "cooperates", the stronger the negative SST anomaly will be along the equatorial Pacific. Most statistical and dynamical models continue to indicate a moderate La Niña through the end of 1998.

More detailed forecasts for 9 U.S. affiliated and 18 non-U.S. affiliated Pacific Island stations are shown in Fig. 2, in the form of long-lead rainfall forecasts from 1 to 13 seasons lead (solid bars) along with their expected skill (lines). The horizontal axis reflects the lead time, whose corresponding actual target period for this forecast is indicated in the legend along the top of the figure (e.g. 1=Oct-Nov-Dec 1998). The same ordinate scale is used for both the forecasts and the skill (standardized anomaly and temporal correlation coefficient, respectively). Sometimes skill may increase as the lead is increased because a more forecastable target season has been reached. The forecasts and their skill differ not only due to their differing ENSO-responsiveness caused by general location differences in the Pacific basin, but also due to differences in orientation with respect to the local orography (if any).

Except for Hawaii, the U.S. affiliated stations are generally forecast to have above normal rainfall this boreal fall and into winter. South of the equator at the non-U.S. affiliated islands, the same forecast generally applies. While not shown here, locations close to the equator from 170E and eastward (e.g. Kiribati) are forecast to be dry, and should be making mitigation efforts.

The CCA modes (not shown; He and Barnston 1996) clearly show ENSO as the dominant (nearly exclusive) influence on tropical Pacific climate, especially during Nov-Dec-Jan-Feb-Mar-Apr-May (and even earlier than Nov along the immediate equator near and somewhat east of the dateline). The current forecasts show a moderate (not severe) influence from the La Niña conditions.

References:

Barnett, T.P. and R. Preisendorfer, 1987: Origins and levels of monthly and seasonal forecast skill for United States surface air temperatures determined by canonical correlation analysis. Mon. Wea. Rev., 115, 1825-1850.

Barnston, A.G., 1994: Linear statistical short-term climate predictive skill in the Northern Hemisphere. J. Climate, 7, 1513-1564.

Barnston, A.G. and C.F. Ropelewski, 1992: Prediction of ENSO episodes using canonical correlation analysis. J. Climate, 5, 1316-1345.

Barnston, A.G. and Y. He, 1996: Skill of CCA forecasts of 3-month mean surface climate in Hawaii and Alaska. J. Climate, 9, 2579-2605.

Graham, N.E., J. Michaelsen and T. Barnett, 1987a: An investigation of the El Niqo-Southern Oscillation cycle with statistical models. 1. Predictor field characteristics. J. Geophys. Res., 92, 14251- 14270.

Graham, N.E., J. Machaelsen and T. Barnett, 1987b: An investigation of the El Niqo-Southern Oscillation cycle with statistical models. 2. Model results. J. Geophys. Res., 92, 14271-14289.

He, Y. and A.G. Barnston, 1996: Long-lead forecasts of seasonal precipitation in the tropical Pacific islands Using CCA. J. Climate, 9, 2020-2035.

He, Y. A.G. Barnston and A.C. Hilton, 1998: NCEP/Climate Prediction Center Atlas No. 5: A precipitation climatology for stations in the tropical Pacific basin; effects of ENSO. U.S. Dept. of Commerce, NOAA, 280pp.

Ropelewski, C.F. and M.S. Halpert, 1987: Global and regional scale precipitation patterns associated with the El Niqo/Southern Oscillation. Mon. Wea. Rev., 115, 1606-1626.

Ropelewski, C.F., and M.S. Halpert, 1996: Quantifying Southern Oscillation-precipitation relationships. J. Climate, 9, 1043-1059.

Fig. 1. CCA-derived precipitation standardized anomaly forecast (X100) for 33 Pacific Islands stations for Oct-Nov-Dec 1998, and Jan-Feb-Mar and Apr-May-Jun 1999. Latest data for these forecasts is August 1998. The cross-validated skill expected for the forecasts is indicated by the size of the numerals (not their value, which shows the forecast itself). Small numeral size indicates correlation skill of less than 0.30, considered unusable; medium size is used for 0.30<skill<0.45 which is modest but usable; large size indicates skill>0.45, considered a relatively more reliable forecast.

Fig. 2. Time series of CCA-based long-lead precipitation anomaly forecasts, and their expected skills, out to one year into the future for 9 U.S.-affiliated Pacific Island stations (first page) and 18 non-U.S.-affiliated stations (page two, page three). The bars indicate the forecast values (as standardized anomalies) and the lines indicate the associated skills (as temporal correlation coefficients). Both forecasts and skills use the same ordinate scale. The target season is indicated on the abscissa, ranging from 1 (Oct-Nov-Dec 1998) through 13 (Oct-Nov-Dec 1999); see the legend at top.