Pacific ENSO Update

3rd Quarter, 2005 Vol. 11 No. 3

NEW - Experimental Sea level Forecasts
(deviations w. r. t. climatology) for the U.S-affiliated Pacific Islands


In our last issue of this newsletter, we introduced an operational forecasting technique using a Canonical Correlation Analysis (CCA) model. This statistical sea level forecast is based on the teleconnections, or air-sea interactions over great distances, between tropical SST and sea level in the USAPI. Hereafter, we will continue to report results on observed and forecasts values of sea level deviations in the ‘Experimental Sea Level Forecasts’ section. The following sections describe: the observed monthly sea level deviations, and the CCA-based forecasts for sea level deviations for the forthcoming season. This forecast technique does not account for sea level changes created by other atmospheric or geologic conditions such as tropical cyclones, storm surges or tsunamis.

Note that, as per user feedback, the units of the whole sea level values have been converted to inches in this issue.

Observed monthly sea level deviation in JFM, 2005

The monthly time series (April to June) for sea level deviations have been taken from the UH Sea Level Center. The full time series (in mm) is available at: http://ilikai.soest.hawaii.edu/uhslc/woce.html. Deviations are defined here as the difference between the mean sea level for the given month and the 1993 to 2001 mean sea level at that station. Locations of all these stations are shown in Fig. 2.

Table 1 provides the monthly observed sea level deviations (in inches). In the 2nd quarter, Guam continued to show sharp rise in this quarter. The months of April, May, and June displayed 9.4, 10.9, and 12.0 inches of rise from the mean respectively (see Table 1 for monthly values of each of the stations and Fig. 3 for average values of AMJ). Other than Malakal (Palau), the sea level in the northwestern Pacific Island stations (Guam, Yap, and Pohnpei) tended to show positive deviations. Malakal initially recorded lower than mean sea level in April; however, for May and June, no considerable variations were observed. Similarly, the low latitude Pacific Island station (Kapingamarangi) and the central Pacific Island stations (Majuro and Kwajalein) had positive deviations during the same time period. The sea level of the lone south Pacific Island station, Pago Pago (American Samoa) displayed a slight rise during the same time period.

Table 1 : Monthly observed sea level deviations in inches (std deviation in parentheses)

Tide Gauge Station Jan Feb Mar
Marianas, Guam +9.4 (3.7) +10.9 (3.9) +12.0 (3.6)
Saipan, CNMI N/A(3.2) N/A(2.9) N/A(2.8)
Malakal, R. Palau -2.7 (4.6) -.3 (4.6) 0 (4.2)
Yap, FSM -1.7(3.9) +3.9 (3.4) +5.9 (4.1)
Pohnpei, FSM +4.5 (2.0) +4.4 (2.4) +4.4( 2.8)
Kapingamar, FSM +3.2 (3.0) +1.4 (2.9) +.2 (2.6)
Majuro, Marshalls +2.0 (2.0) +3.0 (2.0) +.7 (2.1)
Kwajalein, Marshalls +3.9 (2.0) +5.7 (2.5) +3.8 (2.2)
Pago-Pago, A Samoa -1.6 (3.6) +0.0 (4.2) +2.1 (3.6)
Note: - indicate negative deviations (fall of sea-level from the mean), and + indicate positive deviations (rise of sea-level from the mean), n/a: data not available, Figures in parenthesis are year-to-year SD (standard deviations) for the month. For the forecasted values of sea level deviations, please refer to our previous issue of this newsletter (2 nd Quarter 2005, Vol. 11, No. 2).

 

In the 1st quarter (JFM) of 2005, the average sea level deviations in most of the northwestern Pacific Island stations displayed negative deviations, the 2nd quarter provides a tendency towards enhanced positive deviations (Table 1; Fig. 3). This agrees with the AMJ forecast in the last newsletter (Fig. 3 dotted line). Both the low latitude station (Kapingamarangi) and the central Pacific Island stations ( Kwajalein and Majuro) continued to show rising trends in this quarter. On the contrary, the South Pacific Island station ( Pago Pago ), which displayed negative deviation in the last quarter, tended to show a positive deviation in this quarter. This is somewhat expected pattern for American Samoa . Following a weak El Niño year, American Samoa has been found not to experience any pronounced variation in sea level from July to December (see Pacific ENSO Update, 4th Quarter 2004, Vol. 10, No. 4).

Seasonal Sea Level Forecast for JAS, ASO, and SON, 2005

Forecasts of the sea level anomalies in the USAPI are presented here using CCA statistical model. Based on the independent SST values in JFM 2005, the resulting CCA model was used to forecast the sea level of moving-average season of three consecutive months: Jul-Aug-Sep (JAS), Aug-Sep-Oct (ASO), and Oct-Nov-Dec (OND) (Table 2). 1-season ahead of CCA cross-validation forecast skills (cross-validation is conducted to evaluate the overall forecasting skill of the CCA model) are also presented (Fig. 4). A short summary of qualitative forecast quality is presented (Table 2).

Table 2 : Forecasts of sea level deviation in inches (JAS: Jul-Aug-Sep, ASO: Aug-Sep-Oct, and SON: Sep-Oct-Nov )

Tide Gauge Station JAS ASO SON *Forcst. quality * Forecast quality is a measure of the expected CCA cross-validation skill. In general terms, these kinds of forecasts are thought to be of useful skill (or at least fair skill) if the CCA cross-validation value is greater than 0.3 (Fig.4). Higher skills correspond to greater expected accuracy of the forecasts. Skill levels greater than 0.5 are thought to be good, while skill levels greater than 0.6 are thought to be high.
Guam +4 +3 +2 Good
Saipan +3 +3 * Fair
Malakal +5 +5 +4 Good
Yap +5 +4 +3 Good
Pohnpei +2 +2 +3 High
Kapingamari * * +2 Moderate
Majuro +1 +2 +3 Moderate
Kwajalein +2 +2 +2 Moderate
Pago Pago * +1 +1 Good

Note: For +/-, see notes in Table 1. Any deviations between (0~ ±1) inch is considered as negligible and denoted by *
REMARKS: Deviation of less/above than 2 inches are unlikely to cause any adverse impact. To put in a historical perspective, Yap recorded 6 inches of negative and 7 inches of positive deviations during JAS 1997 and 1998 respectively. Pago-Pago recorded 2.4 inches of positive and 7 inches of negative deviations during the same time period.

From Table 2, it has been found that forecast skill for most of the tide gauge stations varies from moderate-to-good. Pohnpei provided high skill while CNMI at Saipan displayed fair skills (Fig. 3). Other than CNMI, the cross-validation correlation skills for the 3-moving seasons (at 1-season lead time) are reasonably well predicted with a mean skill greater than 0.40 for all the tide gauge stations.

The sea level forecast in JAS, which is based on spring SSTs (AMJ), is slightly less skillful than previous seasons. Climatologists refer to this period as the spring barrier because of the difficulty in predicting SST during the spring. As the year advances, the skill gradually increases until the next spring. For rainfall prediction, this spring barrier has a pronounced effect; however it appears to be a relativily weaker obstacle for SST based sea level predictions. As a result this season has slightly weaker yet still useful skill values for the upcoming seasons (Fig 3).