Westerly Wind Events in the Tropical Pacific, 1986–95

D. E. Harrison NOAA/PMEL and JISAO/Hayes Center, University of Washington, Seattle, Washington

Search for other papers by D. E. Harrison in
Current site
Google Scholar
PubMed
Close
and
Gabriel A. Vecchi School of Oceanography, University of Washington, Seattle, Washington

Search for other papers by Gabriel A. Vecchi in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Based on examination of 10 yr of 10-m winds and wind anomalies from European Centre for Medium-Range Weather Forecasts (ECWMF) analysis, definitions for westerly wind events (WWEs) of eight different types are proposed. The authors construct a composite for each type of event, show that a simple propagating Gaussian model satisfactorily describes the evolution of zonal wind anomaly for each type of event, and determine the scales of each composite event by fitting the model to each composite. The authors discuss the WWEs that occurred during the Tropical Oceans Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) intensive observing period (IOP) and show the extent to which these composite events are able to reproduce the major westerly wind features of the IOP. The frequency of occurrence of each type of WWE for each year of this record and by calendar month are described; the authors find several types of events are negatively correlated with the annual mean troup Southern Oscillation index (SOI), and that the stronger WWEs often have a statistically significant seasonality. Several instances of widespread westerly wind anomaly are identified and described, but these “mega”-WWEs have few features in common. Although the authors’ composites underestimate the peak amplitude of many WWEs and cannot always accurately represent the time evolution of each WWE, the authors believe that they offer a useful framework for representing the sort of westerly wind variability that occurs in the western and central tropical Pacific and can provide a basis for further study of the importance of such winds in the climatological and interannual variability of this part of the World Ocean.

Corresponding author address: Dr. D. E. Harrison, NOAA/PMEL/OCRD, 7600 Sand Point Way, NE, Seattle, WA 98115.

Abstract

Based on examination of 10 yr of 10-m winds and wind anomalies from European Centre for Medium-Range Weather Forecasts (ECWMF) analysis, definitions for westerly wind events (WWEs) of eight different types are proposed. The authors construct a composite for each type of event, show that a simple propagating Gaussian model satisfactorily describes the evolution of zonal wind anomaly for each type of event, and determine the scales of each composite event by fitting the model to each composite. The authors discuss the WWEs that occurred during the Tropical Oceans Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) intensive observing period (IOP) and show the extent to which these composite events are able to reproduce the major westerly wind features of the IOP. The frequency of occurrence of each type of WWE for each year of this record and by calendar month are described; the authors find several types of events are negatively correlated with the annual mean troup Southern Oscillation index (SOI), and that the stronger WWEs often have a statistically significant seasonality. Several instances of widespread westerly wind anomaly are identified and described, but these “mega”-WWEs have few features in common. Although the authors’ composites underestimate the peak amplitude of many WWEs and cannot always accurately represent the time evolution of each WWE, the authors believe that they offer a useful framework for representing the sort of westerly wind variability that occurs in the western and central tropical Pacific and can provide a basis for further study of the importance of such winds in the climatological and interannual variability of this part of the World Ocean.

Corresponding author address: Dr. D. E. Harrison, NOAA/PMEL/OCRD, 7600 Sand Point Way, NE, Seattle, WA 98115.

1. Introduction

In this paper we explore the spatial and temporal characteristics of westerly wind events (WWEs) over the western and central tropical Pacific Ocean, from 1986 through 1995. We find that it is possible to classify the WWEs into eight types, and then evaluate the composite of each type of WWE. From the composite wind fields we evaluate the spatial structure of each type of WWE; a simple analytical model further summarizes the basic properties of each type of event. We use the Tropical Oceans Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) intensive observation period (November 1992–February 1993) to test the applicability of the composite model to individual wind events. Finally, we study the temporal distribution on monthly and interannual timescales of both the frequency and intensity of the WWEs.

The existence of unusual multiday periods of surface westerly winds over the western and central tropical Pacific was noted during the planning for the TOGA program. Some of the most striking instances of these events occur when tropical cyclones pair on either side of the equator (Keen 1982), but westerly wind variability is much more common than are paired cyclones (Luther et al. 1983; Harrison and Geise 1991; Hartten 1996). Harrison and Geise (1991) used 30 yr of near–date line island surface wind observations (1950–80) to characterize the meridional and time scales of WWEs seen at these islands. They suggested that near–date line WWEs could be classified into four relatively distinct types according to distance from the equator of the strongest westerly wind anomaly. They also noted a significant tendency for near–date line WWEs to be associated preferentially with warm El Niño–Southern Oscillation (ENSO) periods and with particular seasons. The limited spatial coverage of the island dataset was inadequate to provide a full characterization of the events.

The large-scale circulation structure associated with WWEs is of interest. Harrison (1984) showed that a dramatic southerly flow, which began over the Tasman Sea (40°S) and later crossed the equator, immediately preceded the onset of strong equatorial westerly winds at the beginning of the 1982–83 ENSO warming period. Others have provided additional examples of a subtropical association to particular equatorial westerly winds, as well as a tendency for cross-equatorial flow from the winter hemisphere to precede the formation of some tropical westerly anomalies (Love 1985; Chu 1988; Chu and Frederick 1990). Hartten (1996) describes an effort to update and extend the analysis of Harrison and Geise (1991) to the entire western tropical Pacific, with particular attention to large-scale aspects of the circulation associated with periods of westerly winds. According to her subjective classification scheme for westerly wind bursts (WWBs), 90% of the synoptic westerly wind variability in the western Pacific can be described using one of nine categories.

Another aspect of westerly wind variability in the western and central Pacific involves relationships with the Madden–Julian oscillation (MJO) [intraseasonal oscillation (ISO)] (Madden and Julian 1972, 1994; Rui and Wang 1990). The ISO is a global scale, rapidly eastward propagating phenomenon associated with large outgoing longwave radiation (OLR) and zonal wind anomalies in the tropical atmosphere above the planetary boundary layer. At the ocean surface, zonal wind variability in this 40–60-day band exists but is not a prominent feature of the available wind data (Harrison and Luther 1990). It is has been suggested that the intensity and frequency of surface westerly wind variability is modulated by the ISO, (Lau et al. 1989; Sui and Lau 1992). The exact connection between episodes of surface westerly wind anomaly and the convectively active phase of the ISO remains unclear, with work on going.

The WWEs are of interest because they are a unique aspect, so far as we can determine at present, of the synoptic timescale variability of this region, and because they can force substantial oceanic response. While there are periods of strong easterly anomalies in the Pacific, the easterly wind variability is of a different nature than the westerly wind variability: the space and time scales are usually much larger and the periods of sharp variability analogous to WWEs are not prevalent in the record. No such tropical variability has been described in the Atlantic or Indian Oceans. According to high-resolution primitive equation ocean models, the remote response can involve SST changes of more than 0.5°C and zonal current changes of more than 50 cm s−1 (Harrison and Geise 1988; Geise and Harrison 1990, 1991; Kindle and Phoebus 1995). Thus the role of WWEs in the evolution of ENSO events bears careful examination.

Observations of the ocean response to WWEs are now accumulating. The TOGA COARE field experiment in the western Pacific had, as one of its primary goals, collecting the observations needed to better understand the roles of WWEs in the local and remote thermal budgets of the near-equatorial Pacific. The COARE field program took place in 1992–93 and the synthesis of its observations is well under way as this is written. Thanks largely to the TOGA observing system, and the Tropical Atmosphere Ocean (TAO) mooring array in particular, repeated looks at the near-equatorial ocean during and after WWEs are becoming available. Reports of observations following particular events show a wide range of behavior (McPhaden et al. 1992; Delcroix et al. 1993).

It is our goal to extend our knowledge about WWEs and their space and time scales. We think this is now possible because much work has been devoted to improving the operational surface meteorological analyses in the Tropics. Since 1986 the U.S. Navy, National Centers for Environmental Prediction (NCEP), National Oceanic and Atmospheric Administration (NOAA), and the European Centre for Medium-Range Weather Forecasts (ECMWF) have each developed improved tropical surface wind analyses. Further, the TOGA–TAO array has been providing, particularly since about 1992, substantially larger amounts of quality surface wind data as input to the analyses than have ever before been available. The TAO array was fully deployed for the first time in late 1994.

After considerable evaluation, we consider the ECMWF surface wind analysis to be sufficiently plausible to provide a basis for this study. We have used it to prepare a description of the xyt characteristics of WWEs during the period 1986–95. Section 2 describes the datasets used; section 3 describes our screening process and the compositing technique we have used in this study. We present our composite results and describe a simple mathematical model structure for these composite WWEs in section 4. The westerly wind variability during the TOGA COARE intense observation period (IOP) and the ability of our composite events to characterize the IOP variability is presented in section 5. Section 6 discusses the seasonal and interannual distribution of WWEs and their correlation with the Southern Oscillation index, and describes the extent to which WWEs occur in particular sequences. Section 7 offers some summary and discussion.

2. Data

We used as our wind dataset the ECMWF 10-m operational 12-h wind analysis, on their 2.5° × 2.5° global grid (ECMWF 1989). Our attention was on the region from 100°E to 100°W, by 30°S to 30°N and over the years 1986–95. The ECMWF analysis was significantly improved in the middle 1980s, hence our choice of 1986 for the beginning of this analysis effort. Other changes to their operational analysis were implemented after 1986, but we found no dramatic changes in the fields sufficient to deter us from this project. We constructed a monthly climatology from the 12-h ECMWF surface wind analysis of the entire 10 yr of record (1986–95), using a time axis centered on the midday of each month. We defined our anomalies as the difference between the instantaneous wind and the climatology, linearly interpolated in time.

In order to assess the utility of the ECMWF 10-m wind field, we carried out two comparisons. First we looked at the large space and time scale aspects of the circulation and compared our ECMWF monthly climatology to the Comprehensive Ocean–Atmosphere Data Set (COADS) (Woodruff et al. 1987) monthly climatology from 1946 through 1993. Then we examined the short timescale variability by comparing ECMWF time series with data from the TOGA–TAO buoys (McPhaden 1993).

We compare our ECMWF climatology with the COADS climatology for 2 months, March and September (Figs. 1–3). Contour plots are shown for the ECMWF climatology from 1986 through 1995, for the COADS climatology from 1946 through 1993 and for the difference ECMWF–COADS, for zonal wind (Fig. 1), meridional wind (Fig. 2), and wind divergence (Fig. 3). For all these plots, the COADS climatology is smoothed with a five-point triangle filter (half-power point at 18°) in the zonal direction and a three-point triangle smoother (half-power point at 11°) in the meridional direction.

The ECMWF climatology reproduces the main aspects of the large-scale circulation appropriately (Figs. 1 and 2). The zonal wind cores of the SE and NE trade winds are clear in Fig. 1, and their magnitudes are comparable to the ship-based winds of COADS. The small area of western Pacific equatorial westerlies in March, and the comparable band of westerlies from SE Asia to 160°W along 10°N in September are present in the ECMWF fields. The difference fields of zonal wind (Figs. 1e,f) show considerable spatial structure, but the large-scale difference is seldom more than 1.5 m s−1 and is typically closer to 1 m s−1. Near the equator, ECMWF winds tend to be weaker (less easterly) than the COADS winds west of 120°W and stronger (more easterly) east of 120°W. The SE trades tend to be weaker (less easterly) in ECMWF, but the NE trades are in some places weaker and in others stronger in ECMWF. The meridional wind comparisons of Fig. 2 also show considerable large-scale similarity. The strong seasonal variations between 10°S and 15°N east of the date line are well reproduced, as is the broad area of southerly wind east of Australia. However, there is much small space scale structure in ECMWF in the western tropical Pacific that has no counterpart in the COADS climatology. This is particularly evident in the meridional wind difference results (Figs. 2e,f). The large-scale meridional wind differences are typically less than 1 m s−1, but NE of Australia they can exceed 1.5 m s−1. The ECMWF climatology in general overestimates the meridional wind speed north of the intertropical convergence zone (ITCZ) and underestimates it to the south.

It is possible that the generally weaker near-equatorial easterlies in ECMWF result from their being a recent average, over a period in which ENSO conditions have been prominent. We have not pursued this possibility, because we judge the differences found here to be small enough to be of little concern for our interests. The WWEs typically have peak zonal wind anomalies in excess of 10 m s−1, so a 1 m s−1 difference is not troublesome.

Another measure of the large-scale aspects of the ECMWF analysis is the monthly mean divergence field. The March and September divergence patterns for the ECMWF and COADS climatologies are compared in Fig. 3. The two climatologies have similar divergence patterns, with maximum convergence in the ITCZ and the South Pacific convergence zone (SPCZ). Generally the location and meridional scale of each convergence zone are similar, but the ECMWF convergence is substantially weaker than COADS. The divergence patterns along the equator are rather different in March, with ECMWF indicating a clear band of divergence from the date line to about 110°W, but with COADS indicating a broad area of weak equatorial convergence east of 140°W. Because this is a region of limited data in COADS, we suggest more observations should be made as part of the Pan American Climate Studies (PACS) program. We see no large-scale differences in the central and western tropical Pacific sufficient to merit concern.

We now describe our comparisons with the TAO buoys. We computed the rms difference between the daily averaged ECMWF wind analysis, interpolated to each buoy location (when necessary), and the daily averaged buoy winds. No trends were removed. Note that the period over which it was possible to compute the rms difference varies from buoy to buoy, so there can be significant change from one longitude to another. The buoys along 110°W, 140°W, and 165°E had the longest time series; other buoys had data for as little as a year, because they were recently deployed. The results for a large sampling of the buoys are presented in Fig. 4. Overall, the zonal and meridional wind results were comparable. Maximum rms differences were about 2.5 and minimum differences about 1 m s−1. The largest values tended to be in the ITCZ and SPCZ regions, although the differences along 165°E were 2 m s−1 or more, from 5°N to 9°S. The smallest values tended to be in the eastern SE trade winds.

The implication of these values comes following comparison with the magnitude of the WWE signal (which we later shall show is 8 to 20 m s−1) and with the accuracy of the TAO measurement. The wind sensors themselves are claimed to be accurate to about 0.2 m s−1 predeployment and are estimated to drift by up to 0.5 m s−1 during 1 yr of activity (Mangum et al. 1994). It should also be noted that the wind measurements on the buoys are done at 4 m from the surface, while the ECMWF winds are at 10 m. According to a stability correction algorithm developed by W. G. Large (Large and Pond 1981) to adjust wind speed from a height z to a height of 10 m, we found that the wind speed correction is typically 10% or less. Thus without height adjustment we may expect errors of about 0.5 m s−1 on a wind of 5 m s−1, or 1 m s−1 for 10 m s−1 wind. Since we cannot height-adjust the TAO winds because the information required to make the stability correction is typically not available, we suggest a plausible expected mean error on the TAO data, to be 0.7 to 1.5 m s−1 under typical moderate WWE conditions. Thus the disagreement between the TAO data and the ECMWF data is similar to the uncertainty in the TAO data, and, since WWEs typically have maximum zonal winds in excess of 10 m s−1, use of the ECMWF analysis gives us an acceptable signal to noise ratio.

Ideally, the TAO array would be dense enough to permit characterization of the structure of WWEs itself. However, the area coverage and the spatial separation of buoys significantly limit its ability to provide the needed spatial resolution. In its present configuration the TAO array is useful in that it provides data to the wind analyses (and thus should improve their realism) and it provides the information needed to define the likely errors in the analysis. But the ECMWF wind analysis itself is required to carry out the work we describe below.

3. Methods

In this section, we describe how we classify the WWEs in our dataset and how we compute the composite WWEs. We first describe how we came to define our different types of WWEs and then introduce the quantitative classification criteria used for each. We then present the number of events of different types that follow from applying these classification criteria to the 1986–95 ECMWF wind anomaly fields. Finally, we describe the details of how we carried out the WWE compositing. Description of our analysis to determine the statistical significance of the features of the composites is included in the appendix.

a. Classification

We began by looking at vector plots of the wind and wind anomalies over the tropical Pacific, for every 12-h analysis period in our dataset. We then highlighted the vectors with westerly anomaly and looked at all the anomaly plots again. We found that westerly wind anomalies of substantial scale tended to occur in particular regions and not to show major translation during their lifetimes. This suggested that a good classification scheme could be based upon the location of maximum westerly wind anomaly.

Based on the wind anomaly patterns we saw, we defined eight regions to serve as the framework for our classification scheme. These regions, which cover most of the ocean from 120°E to 150°W and 15°S to 15°N, were named according to their location relative to each other. The zonal dimension of these regions is about 30° longitude, and the meridional dimension is 10° latitude. Each contains 5 ECMWF grid points in the meridional direction, and 12 to 13 points in the zonal direction. To help the viewer place them in the context of the large- scale environment of the region, we show the regions superimposed upon the ECMWF climatological (1986–95) 10-m wind vectors, the COADS climatological (1946–95), SST and sea level pressure (SLP) for March (Fig. 5) and for September (Fig. 6). Note that there are substantial changes in the environment in each region between March and September.

In order to systematically identify and analyze WWEs, we introduced a quantitative definition for their existence, within the context of our classification scheme. We defined a WWE of type X as any period of 3 or more days for which the 10-m zonal wind anomaly, averaged over region X and smoothed by a three-point triangle filter in time (half-power point at 2.75 days), exceeded 2 m s−1. To label and organize the events, an event’s center day was defined to be the day for which the zonal wind anomaly, averaged over the region, was greatest. With these classification criteria, we identified all the westerly wind events, by region (type) and center date.

We find it useful to describe the intensity of the WWEs according to their “duration,” “maximum averaged anomaly,” “maximum point anomaly,” and“wind measure.” We defined the duration of each WWE to be the time span between when the identification criteria are first met and when the identification criteria are no longer satisfied. We defined the maximum averaged anomaly to be the zonal wind anomaly averaged over the classifying region on the center day of the event. The maximum point anomaly is the maximum zonal wind anomaly within the classifying region, on the center date. We defined the wind measure as the time integral of the zonal wind anomaly, averaged within the classifying region and smoothed by a three-point triangle filter in time (half-power point at 2.75 days), over the event duration.

Not every period of westerly wind anomalies fit perfectly into our classification scheme. In particular, sometimes substantial westerly wind anomalies occurred in adjacent regions at the same time. This led us to make an additional subclassification of WWEs into overlapping and nonoverlapping events. We defined overlapping events to be events that were identified in two regions that shared more than half an edge and whose center dates were within 3 days of each other. For each pair of overlapping events, we compared the maximum averaged anomaly for each event and classified the event as an event of the type for which the maximum averaged anomaly was greatest. We then had two lists of dates, one with all the events, which we shall refer to as the“complete event list,” and another with the all the nonoverlapping events plus the overlapping events classified according to our secondary classification scheme, which we will refer to as the “nonoverlapping event list.” Table 1 contains the number of events identified for each type of event in each list.

b. Compositing

The basis of our compositing is the identification of the center day [day(0)] for each event of a type X. We compute the day(0) field for our composite type X event by averaging together the wind or wind anomaly field for all the type X event days(0). Composite day(±n) is computed similarly from each of the day(±n) for type X events. We evaluate the composite for day(−10) through day(10) for each of the eight types of WWEs, resulting in a 21-day composite event. The composites are evaluated using all the events from each type of WWE.

For each type of event, then, our composite is evaluated according to
i1520-0442-10-12-3131-eq1
where U is the composite vector wind anomaly, u is the instantaneous vector wind anomaly, x is the zonal location, y is the meridional location, {τi} are the center days for the individual events, tn ∈ [−10, 10] is the event day, and N is the number of events to be composited. To study the effect of overlap in the averaging process, we generated two sets of composite westerly wind events: one using the dates in the complete event list, the other using the dates in the nonoverlapping event list. The two composites were not different in any significant qualitative aspect. Thus we present only the composites generated from the complete event list. Also, we similarly computed the composite wind fields for each type of event.

4. Composite results

In this section we present the results of our composite analysis. We show vector wind anomaly maps for selected days of each type of event. We also present vector maps of the wind field on the center day [day(0)] for each type of event. We shall show that a simple model structure provides a convenient way to define the zonal, meridional, and temporal scales of the zonal wind anomalies for each of our composites, and we present the scales of each event from this perspective.

a. Vector maps

We computed daily wind anomaly vector maps for days (−10) through (+10) for each of the eight types of WWEs and show the full set of maps in Vecchi and Harrison (1997). Here we present only selected maps to illustrate the patterns when the wind anomalies are substantial. For each type of event, we show the wind anomaly on day(−2), day(0), and day(+2) (Figs. 7–14). Note in these figures that the region used to define the featured type of event is outlined. The anomalies highlighted as bold vectors have zonal component significant at the 99% level, and those highlighted by a shaded background have a meridional wind component comparably significant. Statistical significance was determined by a Student’s t-test (appendix).

We describe qualitatively the features of the patterns that seem to us to merit note and follow the atmospheric convention that an easterly wind is a wind from the east, etc. It is helpful to introduce composite wind magnitude labels as follows: weak (u ≤ 2m s−1), moderate (2 m s−1 < u ≤ 4 m s−1), and strong (u > 4 m s−1). We find it is simplest to identify common elements in the composite events by sorting them into the northern regions (NW, N, and NE), the equatorial regions (W, C, and E), and the southern regions (S and SE).

1) Northern regions: NW, N, and NE (Figs. 7–9)

Each of these events has strong meridional and zonal wind anomalies. Although the areas of maximum significance are principally occupied by moderate to strong westerly and southwesterly anomalies, there are clear suggestions that these events are associated with anomalously cyclonic circulations, with moderate to strong easterlies and northerlies to the north and west of the regions. All of these events also show a modest translation of the area of maximum anomalies during the event. The type N and NW events show northwestward displacement throughout the event, while the type NE event shows an eastward displacement throughout the event.

The type NW and N events share additional common features. Both types of events have periods of distinct weak to moderate cross-equatorial inflow from day(−3) until day(2) for the type NW event, and from day(−3) until day(0) for the type N event. There is also well- defined inflow from west of the cyclonic feature from day(−2) to day(3) for the type NW event, and from day(−2) to day(2) for the type N event. The circulation is more complicated than a simple cyclonic flow.

The type NW and NE events share periods of weak to moderate equatorial westerly anomalies. For the type NW events, there is a period of equatorial westerlies to the south of the classifying region prior to the center day, lasting 8 days, centered around day(−5), when the maximum equatorial westerlies occur; there are also weak equatorial westerlies from day(2) through day(5), to the southeast of the region. For the type NE events, equatorial westerlies to the south of the classifying region occur during the period of maximum anomaly in the region, last for 3 days, and have their peak amplitude on the center day of the event.

2) Equatorial regions: W, C, and E (Figs. 10–12)

For each of the equatorial WWEs, the maximum anomalies are primarily zonal, with periods of moderate meridional inflow to the western part of the regions. The equatorial westerlies in all three events are meridionally contained within the defining region, but their zonal extent can be substantially larger than the defining region. The meridional inflow typically can be found at least 20 degrees away from the equator but can extend as far as 30 degrees [the southerly inflow on day(0) of the type C event]. For the type C composite event, the zonal wind anomalies are close to symmetric with respect to the equator, while for the types W and E event there is a pronounced asymmetry. The type W composite event has symmetric anomalies in the early part of its evolution, while as day(0) is approached the anomalies become more pronounced to the north of the equator. The type E composite event has its largest anomalies to the south of the equator.

For the type W event, there is moderate southerly inflow during event days(−5) and (−4) (Vecchi and Harrison 1997), followed by a moderate northerly inflow that begins on the center day, and an anomalous cyclonic circulation subsequently develops. For the type C event, there is moderate northerly and southerly inflow beginning on day(−1) and ending after day(1). For the type E event, there is moderate southerly inflow beginning on day(−2) that continues until the center day. Both the type W and the type E events exhibit a translation of the are of maximum anomaly. The type W event has a translation toward the northwest after the center day. The type E event has a westward translation after the center day.

3) Southern regions: S and SE (Figs. 13 and 14)

Both of these composite events have a moderate to strong meridional component as well as strong westerly anomalies near day(0). Although most of the significant anomalies are westerly and northwesterly, there are also moderate to strong easterly and southerly anomalies suggesting a pattern of anomalously cyclonic flow. The events also exhibit eastward translation of the area of maximum anomaly throughout the event.

Both of the composite events exhibit moderate cross- equatorial inflow on and immediately preceding the center day. The cross-equatorial inflow for the type S event occurs through the northwestern corner of the region as well as the eastern half of the northern edge of the region, from day(−2) through day(0). For the type SE event, the cross-equatorial flow is most persistent north of the eastern corner of the region, from day(−2) through day(0). Both events exhibit strong inflow from the west on the days before and during the period of maximum anomaly. The westerly inflow for the type S event begins on day(−9) and extends as far west as the Indian Ocean. So the structure of the circulation of these events, like that of the type N and NW events, is not just a simple cyclonic one.

The type SE event has persistent weak to moderate equatorial westerly anomalies on the days surrounding the center day. The equatorial westerlies are stronger to the south of the equator than to the north of it. The type SE event also has moderate southerly inflow from higher latitudes on the days preceding and on the center day.

b. Center day wind maps

Figures 15–17 show vector plots of the center day of the total wind composite, for each type of WWE. As before, it is helpful to discuss the composites by grouping the events by northern, equatorial, and southern regions. We shall refer to many of the features described in the previous section; see Figs. 7–14 as needed.

1) Northern regions: NW, N, and NE (Fig. 15)

The anomalously cyclonic circulations in the type NW and N anomaly composites manifest themselves as cyclonic wind patterns in our wind composites. The type N wind pattern indicates a reversal of the northeasterly trades within the classifying region. The situation differs for the type NE event, for which the wind composite indicates a trade wind break, but not a cyclonic wind pattern, because of the strength of the northeasterly trades in that region. The translation of the area of maximum anomaly, noted in the anomaly plots, is also apparent in the full sequence of wind composite figures (Vecchi and Harrison 1997).

The cross-equatorial inflow noted in the type NW and N event anomalies is also apparent in the wind composites, but the anomaly inflow from the west is not clear in the winds themselves. The equatorial westerly anomalies in the type NW event appear as actual equatorial westerly winds to the south of the region. The equatorial westerly anomalies noted in the type NE events do not appear as equatorial westerly winds here.

2) Equatorial regions: W, C, and E (Fig. 16)

For the type W and C events, the primarily zonal westerly anomalies in the region appear as primarily zonal westerly winds. However, for the type E event, the wind field has a noticeable meridional component. Also, while the anomalies for the type C event were centered about the equator, the winds are asymmetrical with the greatest westerly amplitudes south of the equator [this is due to the seasonality of the strong type C events (section 6a), which tend to occur in boreal winter when the climatological winds south of the equator are westerly]. The equatorial asymmetry of the type E zonal wind anomalies is also noticeable in the wind composite.

The southerly inflow anomaly in the type W and C events appears as a strengthening of the already prominent southeasterly trade winds to the east of Australia. For the type E event, southerly inflow in apparent and is linked to a cyclonic circulation pattern to the southwest of the region, which translates westward. For the type W event, the northerly inflow anomaly is discernible in the winds, so is the cyclonic anomaly circulation, which tends to move northwestward. For the type C event, the northerly inflow anomaly is overwhelmed by the strong trade winds, but a cyclonic wind circulation forms, centered to the south of the region, which does not propagate from day(−4) when it first appears to day(4) when it dissipates.

3) Southern regions: SE and S (Fig. 17)

Both of these composite events display strong meridional winds, as well as zonal winds in the classifying regions. The type SE composite winds shows a reversal of the southeasterly trades. The winds also have a cyclonic pattern to them, which translates to the east as it decays for the type S events, and decays quickly for the type SE event. The cross-equatorial inflow anomaly for the type S event produces cross-equatorial flow in the winds. For the type SE event, there is no cross-equatorial flow in the area of cross-equatorial anomalies. For the type SE event, the anomalously southerly inflow enhances the southeasterly trade winds to the west of the trade wind reversal.

c. Scales

Now we turn to the problem of defining the scales of our composite WWEs. Table 2 shows the duration, maximum averaged anomaly, wind measure, and maximum point anomaly for each of the composite WWEs (section 3a). The events have a duration in the range of 4.5–5.5 days, except the type NE event (4 days). The maximum averaged anomaly is in the range of 3.9 to 4.7 m s−1, the strongest events are type SE (4.7 m s−1) and type N (4.6 m s−1), and the weakest is type W (3.9 m s−1). The wind measure for all the events is between 1.1 and 1.6 × 106 m; the strongest event is type N (1.6 × 106 m). The events have maximum point anomaly between 6.3–7.2 m s−1, except the type E event (5.5 m s−1).

Histograms for duration, maximum point anomaly, maximum averaged anomaly, and wind measure for all the events over the entire record are shown in Fig. 18. The frequency distributions for the individual events are similar to those for the entire record, so we do not present distributions for each type of event; see Vecchi and Harrison (1997) for breakdown by event type. From Fig. 18, we see that approximately 50% of the WWEs in the record have a duration greater than 3 and less than 6 days (recall the composite WWE durations vary between 4 and 5.5 days). For maximum averaged anomaly, approximately 50% of the events have values greater than 3.5 m s−1 and less than 5 m s−1, and we find the composite WWE values to be between 3.9 and 4.7 m s−1. For wind measure, approximately 40% of the events have wind measure between 1 × 106 m and 2 × 106 m (composite WWE wind measures vary between 1.1 and 1.6 × 106 m). Since the previous three quantities are large space scale linearly averaged quantities, they are not affected a great deal by the smoothing effects of compositing. For the maximum point anomaly, the smoothing effects of the averaging process are evident. The maximum point anomaly of the composite events varies between 5.5 and 7.2 m s−1, but fewer than 20% of the WWEs had maximum point anomaly values that are that low; this is because much of the small-scale structure of the individual events has been smoothed out by the averaging.

To characterize the behavior of the zonal wind anomaly field as simply as possible, we formulated a simple analytical model for it. We constructed xt and yt contour plots of the zonal wind anomaly for each type of event [see Vecchi and Harrison (1997) for the full set of these plots] and found the zonal wind anomaly to be sharply bounded in space and time. Examples of type C and SE composite suffice to illustrate the typical situation (Fig. 19).

We use a Gaussian in (x, y, t) with translating center to define the scales of our events:
i1520-0442-10-12-3131-eq2
where U is the model zonal wind anomaly field, Uo is the maximum point anomaly, (Xo, Yo) is the geographic center, (cx, cy) is the translational velocity, (Lx, Ly) are the spatial e-folding scales, T is the temporal e-folding scale, and x, y, and tn are as in section 3b. We computed the scales for our Gaussian model as follows. We defined the geographic center of the event as the location of the maximum zonal wind anomaly. The spatial e-folding scales were calculated as half the distance between the two closest e-folding points on either side of the center, along the axis in the direction of interest. Since the center of the event generally moves in time through the region, we defined the instantaneous center of the event as the point with the largest zonal wind anomaly within 50° in the zonal and 30° in the meridional direction from the geographic center. We calculated the temporal e- folding scale as half the time between the two closest e-folding points on either side of the center day. Using the instantaneous centers for our “path” data, the translational velocity is the mean velocity with which the instantaneous center moved, during the time when the zonal wind anomalies at the instantaneous centers were above the e-folding level.

The characteristics of the events are summarized in Table 3. We present the maximum point anomaly, geographic center on day(0), spatial and temporal e-folding scales, and translational velocities, for each composite WWE. Using the parameters in the table, we see that the maximum point anomaly (Uo) is between 6 and 7 m s−1 and the timescale (T) is close to 3 days for each type of event. All but two types of events have meridional length scale (LY) of about 700 km (NE has LY ∼ 1100 km and W has LY ∼ 400 km) and zonal length scale (LX) between 1400 and 1900 km (NW has LX ∼ 2500 km and S has LX ∼2200 km). Note that the atmospheric baroclinic Rossby radius of deformation varies from (assuming a wave speed of 20–80 m s−1) 650 to 1300 km (Gill 1982), which is in general agreement with the meridional length scales summarized in the table. All but one type of event (C) has zonal translation of its center (values ranging from −5.1 for N and 3.8 for SE); only type NW, N, and W events have meridional translation (northward in each case). For the type of events that appear clearly associated with Northern Hemisphere cyclonic circulation patterns (NW, N, and W), the translational direction is consistent with the direction expected of a Northern Hemisphere tropical cyclone (Lau and Lau 1992; Joint Typhoon Warning Center 1994a,b; Tsutsui and Kasahara 1996).

5. WWEs and the TOGA COARE intensive observations period

So far we have concentrated on describing our composite WWEs (section 4a and 4b) and have proposed a simple kinematical model for them (section 4c). A natural subsequent question is the extent to which our composite WWEs usefully describe the wind field when particular WWEs are taking place. The TOGA COARE IOP (Webster and Lukas 1992; Lukas et al. 1995) offers an example period (November 1992–February 1993) when WWEs were the focus of a major oceanographic and meteorological field program. The IOP occurred during the warm phase of ENSO, ISOs passed over the western pacific during the program (Chen et al. 1996; Lin and Johnson 1996), and there were some tropical cyclones as well (McBride et al. 1995). So conditions were very favorable for the appearance of WWEs.

Table 4 lists the WWEs that occurred during the IOP according to our classification scheme. There were three type NE and C events; two type NW, N, E, and SE events; and one type W and S event during the IOP. Maximum point anomalies are ≥ 15 m s−1 in six events, between 10 and 15 m s−1 in nine events, and less that 10 m s−1 in one event. The maximum averaged anomaly offers another indicator of the overall intensity of the event. By this standard the 1 January 1993 type S (10.4 m s−1) and the 2 January 1993 type C (6.4 m s−1) events are the most intense of the IOP. The next most intense are the 19 November 1992 type NE and the 6 February 1993 type SE events. The remaining events have maximum averaged anomalies of 5 m s−1 or less.

The final column of Table 4 lists the wind measure of each event, as defined in section 3. Three “primary” events have wind measure values near 4 × 106 m, four“secondary” events have values between 2.5 × 106 m and 3 × 106 m, and the rest have values less than or equal to 1.3 × 106 m. The three primary events are the type S and C events centered on 1 and 2 January 1993, and the type SE event centered on 6 February 1993. The secondary events are the type NW event centered on 20 November 1992, the type SE event centered on 3 January 1993, the type NE event centered on 5 January 1993, and the type C event centered on 31 January 1993. Late December 1992 and early January 1993 was the period of primary WWE activity according to the wind measure. The next most active period was late January 1993 through early February 1993.

These WWE statistics are consistent with westerly wind periods that have been discussed elsewhere. Eldin et al. (1994) reported strong westerly winds and found large eastward surface currents from their ship track along 156°E, during early January 1993 and early February 1993. The large westerly wind activity during early January 1993 and during early February 1993 coincided both times with the passage of the convectively active phase of an ISO (Lin and Johnson 1996; Chen et al. 1996). Four WWEs seem to be associated with named tropical cyclones: the 31 October 1992 type N event (Cyclones Dan and Carrie), the 20 November 1992 type NW event (Cyclone Hunt), the 1 January 1993 type S event (Cyclones Nina and Kina), and the 6 February 1993 the SE event (Cyclone Mick) (McBride et al. 1995).

How well do our composite WWEs describe the periods of WWEs during the IOP? We have compared the IOP anomaly fields with the wind anomaly fields obtained by simply superimposing the composite anomaly WWEs for the events listed in Table 4. Vecchi and Harrison (1997) show the results of this comparison for the full IOP period. We find a number of instances in which the simple composite anomaly field is a good first approximation to the real anomaly field, and several instances where the simple field is not so satisfactory. In the interest of brevity, we present here only three examples of each. These six event center days are shown for both the modeled anomaly wind field and the daily wind anomaly field, with bold vectors indicating zonal component greater than 3 m s−1 (Figs. 20 and 21).

First, we consider the three center days when the composite description is a reasonable approximation to the actual anomaly field (9 December 1992, N; 5 January 1993, NE and SE; 29 January 1993, W and C). Figure 20 shows that generally the composite description is more diffuse than the daily wind anomaly fields;amplitudes are somewhat reduced and length scales are somewhat larger, even in the defining regions. Outside the defining regions there can be substantial differences, but in these areas the wind anomalies are substantially smaller than in the defining regions.

Now consider the three examples in which the composite fit is least satisfactory (20 November 1992, NW and NE; 29 November 1992, E; 6 February 1993, SE), shown in Fig. 21. The 29 November 1992 type E event is quite weak and has substantial (≥3 m s−1) westerly anomalies only in the western third of the defining region; our composite type E event has westerlies over a much greater zonal extent. This is the only situation in the IOP when the composite does not reproduce the basic features of the wind within the defining region. This is also the weakest event during the IOP, based on the wind measure, and its maximum averaged anomaly is only barely over the identification threshold of 2 m s−1. In the other two examples, the substantial shortcoming of the composite fit concerns the winds outside the defining regions, particularly near the equator. We find that these IOP off-equatorial events have less equatorial westerly wind than one would expect from the composite WWEs.

More than half the IOP WWEs had a duration shorter than our composite events, so the composite fit produces near equatorial westerly wind anomalies when they were not present in the wind fields. During periods when the WWEs have a duration more like that of the composite events, the fit is much better. This is because while the WWEs identified and averaged to generate the composite WWEs have similar general characteristics, the differences in the details result in a smudged composite WWE.

6. Temporal distribution

In this section we describe various temporal relationships for WWEs. We present the climatological distribution of WWEs by month. We also show the year-to- year variation of WWE occurrence and intensity. Finally, we offer some statistics on the extent to which WWEs occur in particular relationships to each other.

a. Single event distributions

Consider first the climatological distribution by month. Figure 22 shows, for each region, the 10-yr (1986–95) total number of WWEs that occurred during each calendar month. No striking variation is seen in the month-by-month comparison, but there is a clear seasonal preference for types NW, N, and S, which is statistically significant to the 99% level using the test described in the appendix. A seasonal description for the distribution of the type NW WWEs is that of an“on” season from July through October, two transition seasons (May–June and November–December), and an“off” season from January through April. For the type N and S WWEs, the simplest seasonal description is that of a 6 month “on” season and a 6 month “off” season. The “on” seasons are July through December for the type N event, and December through May for the type S event. We cannot define an “on” or “off” season for type NE, W, C, E, or SE events that is statistically significant at the 90% level.

The distribution of moderate or stronger WWEs is somewhat more striking; defining a moderate or stronger event as an event whose wind measure (see section 3a) exceeds 2 × 106 m leads to Fig. 23. Boreal summer and fall are the primary seasons for the type NW (July–December) and N (July–October) events, November–January is the primary season for type C events, and boreal winter and spring are the primary seasons for type S (December–April) and SE (December–March) events. While the type C events are evenly distributed across the months, most of the events that occur between November and February are of moderate or greater strength. All these seasonal distributions are significant to the 97% level. No seasonal distribution, significant even at 70%, exists for the strong type NE, W, and E events.

Now consider the year-to-year distribution of WWEs. Figure 24 shows the number of each type event, year by year from 1986 through 1995. The mean number of events is indicated by the thick horizontal line. It is also instructive to compare these distributions with the troup Southern Oscillation index (SOI). The 12-month running mean of (−1) × SOI is plotted along with the mean (−1) × SOI for the 10-yr period 1986–95, as the lower left-hand panel of Fig. 24. The SOI values have been negative much of this period, indicating warm (ENSO) conditions for the tropical Pacific; the 10-yr average is −4.5.

We see that the fewest number of events occurred in 1988 for every type of event, and in most regions 1989 was also a year of few events. Here, 1988–89 was the only period of persistently positive SOI in this record. There is no comparably strong connection between years of negative SOI and the more than normal total number of WWEs. However, lagged correlations between the SOI and the number of events in each region revealed that some statistically significant relationships exist. The strongest correlation (−0.80) exists for type C events at zero lag and is 99% significant. Zero lag correlations significant at the 90% level exist for type NE(−0.60), and SE(−0.68) events. Type NW and W regions have 90% significance level correlation (−0.61 and −0.63, respectively), but they lead the SOI by 1 yr. Overall, then, we find that WWEs in our easternmost off-equatorial regions and our C region are negatively correlated with zero lag to SOI. The WWEs in our westernmost regions are negatively correlated at a 1-yr lead with SOI. Only type N, E, and S events do not show a significant correlation with SOI.

There is another WWE–SOI relationship of interest, which is evident in Fig. 25. In this figure we have plotted a scatterplot of the wind measure of all the events in our 10-yr record. It is clear that strong events (wind measure >3 × 106 m) associate well with periods of most negative SOI and that no strong events occur when SOI is positive.

b. Multievent distributions

We have emphasized the independence of WWEs throughout this study. Typically WWEs form, develop, and decay without any strong relationship with other events. We base this on having studied the sequencing of events with center days within 3 days of each other, looking to determine if there were any statistically significant patterns. In a few instances of substantial statistical significance (99%) different types of WWEs seem to be related. We find 18% of the time a type N event will evolve into a type NW event (5 times), 17% of the time a type W event will evolve into a type NW event (6 times), and 18% of the time a type S event will evolve into a type SE event (7 times). These three relationships are consistent with the translation speed of the original event. It must be noted that it is not typical for the first event to form out of the second. Two other statistically significant relationships exist in our record, for which we cannot propose mechanisms. In our record 18% of the time a type N event will precede a type SE event (5 times) and 14% of the time a type E event precedes a type NE event (6 times). These last relationships might be coincidental, or there might be some mechanism that accounts for their occurrence. Longer records, containing additional WWEs are needed to determine the robustness of these relationships; only a very few instances exist in our 10-yr record.

Another aspect of interest in the frequency distribution of westerly winds was identified by computing the average zonal wind anomaly over the area spanned by all eight of our WWE regions. Recall that we define a WWE to exist in any one of our WWE regions when the area average zonal wind anomaly exceeds 2 m s−1 for 3 days. If we define a “mega”-WWE (MWWE) to exist when the average over all eight regions meets the WWE criterion, we find there are 11 MWWEs in our analysis period. Figure 26 provides a cartoon of the spatial structure and wind measure of each MWWE.

The number of MWWEs by year is 2—86, 1—90, 1—91, 3—92, 2—93, and 2—94; MWWEs only occur during years when the annual average SOI was negative. Of the 11 MWWEs, one involves three regions, two involve four regions, six involve five regions, and two involve six regions. One MWWE does not involve any of the three equatorial regions, four involve only one, five involve two, and one involves all three equatorial regions. In the MWWEs, we find that the component WWE of largest wind measure occurs five times in a northern region, four times in an equatorial region, and two times in a southern region. In none of the MWWEs does the W or NW region have the event of largest wind measure. Apart from having events of strongest measure in the central or eastern regions, no clear geographical patterns emerge for MWWEs.

We find a suggestion of seasonality in the meridional frequency of the component WWEs in each MWWE. The five MWWEs for which the dominant component WWEs were Northern Hemisphere WWEs occurred between the middle of August and the beginning of November; the three MWWEs that had Southern Hemisphere WWEs as their primary component WWEs all occurred between January and May; the three MWWEs that had equatorial WWEs as their main component WWEs occurred between late November and late April. However, there are too few MWWEs for us to do meaningful statistics.

The TOGA COARE IOP contained one MWWE, composed of WWEs with center days between 1 January 1993 and 9 January 1993. Only this IOP MWWE and one other had each type of WWE in it with wind measure greater than 2.0 × 106 m. Most MWWEs have several types of WWEs with wind measure less than 2.0 × 106 m. The TOGA COARE IOP thus represented an extreme period of westerly wind variability.

7. Summary and discussion

We have examined ECMWF 10-m surface wind analyses every 12 h between 1986 and 1995 to try to characterize the space and timescales of westerly wind events (WWEs) in the tropical Pacific Ocean. We found that Pacific WWEs can be classified satisfactorily according to the region in which they attain their maximum zonal wind anomalies and that eight adjoining regions are needed to describe the different WWEs that we saw in the wind fields. We name the type of event by the region used to define its existence, where regions are named according to their position relative to each other and to the equator (Figs. 5 and 6): NW, N, and NE are north of the equator; W, C, and E straddle the equator; S and SE are south of the equator. With a quantitative measure to define the existence of a WWE, we find that there are 351 events in this period: 58 NW, 28 N, 36 NE, 35 W, 62 C, 42 E, 39 S, and 51 SE.

We generate a composite event for each type by averaging and find that the zonal wind anomalies associated with each type of event are quite compact in space and time. We modeled the structure of each event as a uniformly translating Gaussian in space and time (Table 3). The typical amplitude is between 6 and 7 m s−1; the typical e-folding timescale is about 3 days (duration is 6 days between the times of e−1 × amplitude). The timescales found by Harrison and Geise (1991) are longer (5–10 day), but the amplitudes obtained in this study compare well to theirs. The meridional e-folding scale varies between event types from 400 to 1100 km, but values are mostly in the 600–700 km range; these meridional e-folding scales are slightly larger to those found by Harrison and Geise (1991). Zonal e-folding scales vary from 1400 to 2500 km, while the equatorial events have e-folding scales of 1700 to 1900 km; Geise and Harrison (1991) estimated a zonal length scale of 20°, that is an e-folding scale of about 1000 km, from the island wind data. Some events translate slowly (largest speed is about 5 m s−1) and others show no significant translational motion. In section 2 we defined four additional quantities to characterize WWEs: duration, maximum point anomaly, maximum averaged anomaly, and wind measure; Table 2 summarizes these characteristics for each type of composite WWE.

Our classification method is similar to an extension of that used by Harrison and Geise (1991), except that we have a two-dimensional surface wind field that covers the entire western and central Pacific, instead of a distribution of islands. We can extend our analysis to the zonal scales and translation characteristics of WWEs throughout the entire tropical Pacific. Hartten (1996) used a subjective classification scheme based on the circulation patterns that are observed in association with westerly wind activity, which is defined in terms of wind and not wind anomaly. Hartten’s analysis, which covers the area west of the date line, finds many of the circulation patterns seen in our composite WWE analysis, such as cross-equatorial flow, cyclonic circulation patterns, and inflow from the west. However, where the trade winds are a strong and persistent feature of the tropical atmospheric circulation, the strong anomalies that are seen in our studies will not usually satisfy her westerly wind burst criterion of 5 m s−1 westerly winds. Also, west of the date line, climatological winds are westerly at 2–3 m s−1 in the equatorial western Pacific from November through February, and in the northwestern tropical Pacific (120°E to 145°E, 5–15°N) from July through September. So Hartten’s definition identifies as westerly wind bursts situations in which the winds are nearly climatological and does not identify all periods where the wind anomalies are strongly westerly.

Occasionally, much of the area spanned by our eight regions experiences westerly wind anomalies of sufficient intensity that the average over the entire area satisfies our criteria for the existence of a WWE. We call such events “mega”-WWEs (MWWEs). Eleven MWWEs occurred in our time period and we describe their features in section 4b. No particular geographical distribution of WWEs was found in these MWWEs. However, a slight indication of seasonality for these events was found, with the MWWEs whose main events were off-equatorial preferentially occurring in the local summer and fall. MWWEs did not occur in the while the troup-SOI (Harrison and Larkin 1996) was positive.

Composites like these are most useful when they reasonably well represent the characteristics of typical events in their region. The TOGA COARE IOP provided a 4-month period (November 1992–March 1993), which contained at least one event of each of our types. We described the WWEs of the IOP period in Table 4. There were 16 WWEs during the IOP; the events with the largest wind measure values occurred late December 1992, early January 1993, and toward the end of January 1993. Another period that was strongly forced was the third week of November 1992. Nine IOP events had wind measure less than 1.3 × 106 m, and seven of the events had wind measure greater than 2.5 × 106 m. Most events had maximum point anomaly values between 11 and 16 m s−1; one event had maximum point anomaly greater than 19 m s−1 and one less than 10 m s−1. There was an MWWE in early January 1993, involving four different types of WWEs (S, C, SE, and NE) in rapid sequence (center days between 1 January 1993 and 5 January 1993); this was one of the most intense periods of westerly wind variability in our 10-yr record. The strong WWEs occurring at the beginning of January 1993 and the beginning of February 1993 were associated with the convectively active phase of an ISO (Lin and Johnson 1996; Chen et al. 1996). Four WWEs were associated with named tropical cyclones, two in the Northern Hemisphere and two in the Southern Hemisphere

We compared the wind anomaly fields produced simply by superimposing our composite anomalies in place of the particular WWEs of the IOP. In many cases the composite representation is reasonable. The most common shortcoming is that the weaker WWEs during the IOP typically were of shorter duration than our composites, particularly near the equator. This means that the composite representation tended to have more near- equatorial westerlies than the IOP wind anomalies indicate. It is not simple to characterize the events that are not well modeled by our composites; some had a small wind measure and others a large one, and there were no event types that were dramatically better (or worse) represented by their composite.

Overall, our composite WWEs offer a useful first characterization of the structure of substantial westerly wind anomalies in this region. The composite events are representative of many WWEs, according to our intensity criteria (Fig. 18). However, there are events more extreme than our composite events. Because aspects of the oceanic response to WWEs depend on the wind stress or some higher power of the wind stress magnitude, it may turn out that these extreme events must be examined separately in order to understand the full range of ocean response to WWEs. The idealized WWEs used in the response studies of Harrison and Geise (1988) and Geise and Harrison (1991) in their WWE experiments were conservative, indicating that WWEs might force the ocean more strongly than suggested by their idealized experiments.

We also examined the distribution of WWEs with year, with climatological month, and with troup Southern Oscillation index. Moderate to strong events (those with wind measure greater than 2 × 106 m) show distinct seasonality for some WWE types (see section 6), with off-equatorial events tending to favor local summer and fall seasons, and type C events tending to favor boreal winter, while no seasonality is apparent in either the type NE, W, or E events. These seasonal distributions are consistent with those found by Harrison and Geise (1991) and by Hartten (1996). Correlation of annual distribution with SOI is less simply summarized, in part because the SOI was predominantly negative during our period of study—only mid-1988 through mid-1989 and late 1995 had SOI persistently positive. Overall, we find 90% statistically significant negative correlation between SOI and type C, NE, and SE events at zero lag and 90% significant negative correlation between SOI and type NW and W events with the events leading the SOI by a year. The type C event is the only event whose correlation with troup SOI is significant to the 99% level.

The relationships between the existence, preferred location, and intensity of WWEs and the large-scale environment of the atmosphere remain to be uncovered. Because the ISO is prominent in tropical convection and free atmosphere zonal wind anomalies (Madden and Julian 1972, 1994; Rui and Wang 1990), and because WWEs are often associated with enhanced convection (Kiladis et al. 1994; Meehl et al. 1996), a relationship between WWEs and the ISO (particularly the convectively active phase of the ISO) has been suggested (Lau et al. 1989; Sui and Lau 1992). In particular, the two most intense periods of surface westerly wind variability during the TOGA COARE IOP occurred in association with the convectively active phase of the ISO. Lacking a convenient ISO index for our analysis period, and not being expert in the phenomenology of the ISO, we leave further exploration of this relationship to others.

WWEs are an unusual mode of tropical atmospheric variability. According to this analysis, the off-equatorial events are almost always associated with tropical cyclonic systems (not to be confused with named tropical cyclones), but the period in which substantial westerly wind anomalies exist over regions of any significant extent is not characterized by strong translation of the cyclonic system. In this sense, it would seem that these Pacific systems are rather different from their tropical Atlantic counterparts, which usually propagate over large distances once they form. We found nothing similar to WWEs in the ECMWF analyses over the tropical Atlantic or tropical Indian Oceans. In four of our off- equatorial composite WWEs (types NW, N, S, and SE), we find moderate (2–4 m s−1) cross-equatorial inflow, as well as moderate (2–4 m s−1) inflow from the west during the days surrounding the center day (see section 4a); Hartten (1996) also has identified cross-equatorial circulations into her sort of westerly wind bursts north of the equator, as well as inflow from the west. We are not able to identify a consistent source of alternate- hemispheric flow; the tropical extratropical connection remains unclear.

The near-equatorial WWEs are sometimes, but not always, associated with cyclonic circulations on either (or both) hemisphere. In many instances, events of these types appear to be simple downgradient pressure flows, with meridional scale determined by the atmospheric first radius of deformation. The near-equatorial events in our analysis seem also to have a midlatitude connection, with cross-equatorial inflows similar to those described by Love (1985), Chu (1988), and Chu and Frederick (1990). We find moderate to strong (>2 m s−1) meridional inflows generating poleward of 20° on the days preceding and on the center day of the event (see section 4a).

The near-equatorial events are the primary events to force the ocean east of their location, and it is most likely these events that help to cause warm water to be advected eastward and downward by Kelvin wave type surges during ENSO events. This type of behavior has been modeled (Schopf and Harrison 1983; Harrison and Schopf 1984; Geise and Harrison 1991) and observed (McPhaden et al. 1992, Delcroix et al. 1993) in the equatorial Pacific. Because the SST changes associated with the remote forcing of modest amplitude WWEs is modest—typically 0.5°C over a couple of months, according to Geise and Harrison (1991)—it is not simple to observe clearly in an ocean full of variability on many frequencies. The roles of WWEs in the genesis and maintenance of ENSO-type conditions, warm equatorial anomalies in the central and eastern Pacific, (Rasmusson and Carpenter 1982; Harrison and Larkin 1996, 1997) remains unclear. There is little dispute that both the frequency and intensity of WWEs increases in the C region when the SOI is lowest [see section 6a and Harrison and Geise (1991)]. But our dataset does not contain enough realizations of ENSO for us to be able to make any analysis of the existence and intensity of WWEs before ENSO warming first occurs. The full deployment of the TOGA–TAO array substantially improves our ability to track surface and subsurface thermal changes that follow the appearance of WWEs.

Acknowledgments

This work was carried out with support from NOAA’s Pacific Marine Environmental Laboratory, ERL Headquarters, and Office of Global Programs (support to the Stanley P. Hayes Center at the University of Washington). DEH wishes to thank Gene Rasmusson for his interest and encouragement to study what we now call WWEs, more than 15 years ago. The help of the TMAP group at PMEL is also gratefully acknowledged.

REFERENCES

  • Bickel, P. J., and K. A. Doksum, 1977: Mathematical Statistics: Basic Ideas and Selected Topics. Holden-Day, 492 pp.

  • Chen, S. S., R. A. Houze Jr., and B. E. Mapes, 1996: Multiscale variability of deep convection in relation to large-scale circulation in TOGA COARE. J. Atmos. Sci.,53, 1380–1409.

  • Chu, P. S., 1988: Extratropical forcing and the burst of equatorial westerlies in the western Pacific: A synoptic study. J. Meteor. Soc. Japan,66, 4549–4564.

  • ——, and J. Frederick, 1990: Westerly wind bursts and surface heat fluxes in the equatorial western Pacific in May 1982. J. Meteor. Soc. Japan,68, 5523–5536.

  • Delcroix, T., G. Eldin, M. McPhaden, and A. Morliere, 1993: Effects of westerly wind bursts upon the western equatorial Pacific Ocean, February–April 1991. J. Geophys. Res.,98, 16379–16385.

  • ECMWF, 1989: The description of the ECMWF/WCRP level III—A global atmospheric data archive. Tech. Attachment, 72 pp. [Available from ECMWF, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Efron, B., and R. Tibshirani, 1991: Statistical data analysis in the computer age. Science,253, 390–395.

  • Eldin, G., T. Delcroix, C. Hénin, K. Richards, Y. Du Penhoat, J. Picaut, and P. Rual, 1994: Large-scale current and thermohaline structures along 156°E during the COARE intensive observation period. Geophys. Res. Lett.,21, 2681–2684.

  • Geise, B. S., and D. E. Harrison, 1990: Aspects of the Kelvin wave response to episodic wind forcing. J. Geophys. Res.,95(C5), 7289–7312.

  • ——, and ——, 1991: Eastern equatorial Pacific response to three composite westerly wind types. J. Geophys. Res.,96(Suppl.), 3239–3248.

  • Gill, A. E., 1982: Atmosphere-Ocean Dynamics. Academic Press, 662 pp.

  • Harrison, D. E., 1984: On the appearance of sustained equatorial westerlies during the 1982 Pacific warm event. Science,225, 1099–1102.

  • ——, and P. S. Schopf, 1984: Kelvin-wave-induced anomalous advection and the onset of surface warming in El Niño events. Mon. Wea. Rev.,112, 923–933.

  • ——, and B. S. Geise, 1988: Remote westerly wind forcing of the eastern equatorial Pacific: Some model results. Geophys. Res. Lett.,15, 804–807.

  • ——, and D. S. Luther, 1990: Surface winds from tropical Pacific islands—Climatological statistics. J. Climate,3, 2251–2271.

  • ——, and B. S. Geise, 1991: Episodes of surface westerly winds as observed from islands in the western tropical Pacific. J. Geophys. Res.,96, 3221–3237.

  • ——, and N. Larkin, 1996: The COADS sea level pressure signal: A near-global El Niño composite and time series view, 1946–93. J. Climate,9, 3025–3055.

  • ——, and ——, 1997: The ENSO surface temperature and wind signal: A near-global composite and time-series view, 1946–1995. Rev. Geophys., in press..

  • Hartten, L. M., 1996: Synoptic settings of westerly wind bursts. J. Geophys. Res.,101(D12), 16997–17019.

  • Joint Typhoon Warning Center, 1994a: Western North Pacific typhoons—1994. Mar. Wea. Log,38(1), 34–40.

  • ——, 1994b: Western North Pacific typhoons—1993. Mar. Wea. Log,38(4), 16–23.

  • Keen, R. A., 1982: The role of cross-equatorial cyclone pairs in the Southern Oscillation. Mon. Wea. Rev.,110, 1405–1416.

  • Kiladis, G. N., G. A. Meehl, and K. M. Weickmann, 1994: Large- scale circulation associated with westerly wind bursts and deep convection over the western equatorial Pacific. J. Geophys. Res.,99(D9), 18527–18544.

  • Kindle, J. C., and P. A. Phoebus, 1995: The ocean response to operational wind bursts during the 1991–1992 El Niño. J. Geophys. Res.,100(C3), 4803–4920.

  • Large, W. G., and S. Pond, 1981: Open ocean momentum flux measurments in moderate to strong winds. J. Phys. Oceanogr.,11, 324–336.

  • Lau, K. H., and N. C. Lau, 1992: The energetics and propagation dynamics of tropical summertime synoptic-scale disturbances. Mon. Wea. Rev.,120, 2523–2539.

  • Lau, K. M., L. Peng, C. H. Sui, and T. Nakazawa, 1989: Dynamics of super cloud clusters, westerly wind bursts, 30–60-day oscillations and ENSO: A unified view. J. Meteor. Soc. Japan,67, 2205–2219.

  • Lin, X., and R. H. Johnson, 1996: Kinematic and thermodynamic characteristics of the flow over the western Pacific warm pool during TOGA COARE. J. Atmos. Sci.,53, 5695–5715.

  • Livezey, R. E., and W. Y. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev.,111, 47–59.

  • Love, G., 1985: Cross-equatorial influence of winter hemisphere subtropical cold surges. Mon. Wea. Rev.,113, 1487–1498.

  • Lukas, R., P. J. Webster, M. Ji, and A. Leetma, 1995: The large-scale context for the TOGA Coupled Ocean-Atmosphere Response Experiment. Meteor. Atmos. Phys.,56, 3–16.

  • Luther, D. S., D. E. Harrison, and R. A. Knox, 1983: Zonal winds in the central equatorial Pacific and El Niño. Science,222, 327–330.

  • Madden, R. A., and P. R. Julian, 1972: Description of global-scale circulation cells in the tropics with a 40–50-day period. J. Atmos. Sci.,29, 1109–1123.

  • ——, and ——, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev.,122, 814–837.

  • Mangum, L. J., H. P. Freitag, and M. J. McPhaden, 1994: TOGA- TAO array sampling schemes and sensor evaluations. Proc. Oceans ’94 OSATES,II, 402–406.

  • McBride, J. L., N. E. Davidson, K. Puri, and G. C. Tyrell, 1995: The flow during TOGA COARE as diagnosed by the BMRC tropical analysis and prediction system. Mon. Wea. Rev.,123, 717–736.

  • McPhaden, M. J., 1993: TOGA-TAO and the 1991–93, El Niño–Southern Oscillation Event. Oceanography,6(2), 36–44.

  • ——, F. Bahr, Y. Du Penhoat, E. Firing, S. P. Hayes, P. P. Niiler, P. L. Richardson, and J. M. Toole, 1992: The response of the western equatorial Pacific Ocean to westerly wind bursts during November 1989 to January 1990. J. Geophys. Res.,97, 14289–14303.

  • Meehl, G. A., G. N. Kiladis, K. M. Weickmann, M. Wheeler, D. S. Gutzler, and G. P. Compo, 1996: Modulation of equatorial subseasonal convective episodes by tropical-extratropical interaction in the Indian and Pacific Ocean regions. J. Geophys. Res.,101(D10), 15033–15049.

  • Rasmusson, E. M., and T. H. Carpenter, 1982: Variations in tropical sea surface temperature and surface wind fields associated with the Southern Oscillation/El Niño. Mon. Wea. Rev.,110, 354–384.

  • Rui, H., and B. Wang, 1990: Development characteristics and dynamic structure of tropical intraseasonal convection anomalies. J. Atmos. Sci.,47, 357–379.

  • Shopf, P. S., and D. E. Harrison, 1983: On equatorial Kelvin waves and El Niño, I, Influence of initial states on wave-induced current and warming. J. Phys. Oceanogr.,13, 936–948.

  • Spiegel, M. R., 1994: Theory and Problems of Statistics. Schaum’s Outline Series, McGraw-Hill, 504 pp.

  • Sui, C. H., and K. M. Lau, 1992: Multiscale phenomena in the tropical atmosphere over the western Pacific. Mon. Wea. Rev.,120, 407–430.

  • Tsutsui, J., and A. Kasahara, 1996: Simulated tropical cyclones using the National Center for Atmospheric Research community climate model. J. Geophys. Res.,101(D10), 15013–15032.

  • Vecchi, G. A., and D. E. Harrison, 1997: Westerly wind events in the tropical Pacific, 1986–1995; An atlas. NOAA/PMEL Tech. Memo. ERL PMEL-109, 222 pp. [Available from NTIS, 5285 Port Royal Rd., Springfield, VA 22161.].

  • Webster, P. J., and R. Lukas, 1992: TOGA COARE: The Coupled Ocean–Atmosphere Response Experiment. Bull. Amer. Meteor. Soc.,73, 1377–1416.

  • Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer, 1987: A Comprehensive Ocean–Atmosphere Data Set. Bull. Amer. Meteor. Soc.,68, 1239–1250.

  • Zwillinger, D., Ed., 1996: CRC—Standard Mathematical Tables and Formulae. 30th ed. CRC Press, 812 pp.

APPENDIX

Statistical Significance Methods

We used various methods to compute the significance of our statistically derived quantities. In this appendix, we describe the methods used to derive these significance statistics. For the computation of the significance of the composite anomaly winds and the significance of the correlation coefficients between SOI and number of events, we used standard statistical tests that assume our populations are normally distributed. To test the significance of the monthly distribution of the WWEs, we used a Monte Carlo method.

Composite wind anomaly significance test

We use a Students t-test to estimate the statistical significance of our composite results, so we require the standard deviation of the zonal and meridional wind anomalies that make up each composite. In particular, we needed to know these standard deviations for every 12-h period during the 19 days of the composite. These standard deviations were computed as follows:
i1520-0442-10-12-3131-eqa1
where σ is the standard deviation vector; U, u, x, y, tn, {τi,}, and N are the same as in section 3a; and we define a vector to the power of α to be
i1520-0442-10-12-3131-eqa2
We performed a Student’s t-test (Bickel and Docksun, 1977, 210–215) on each component of our composite vector wind anomaly field, to determine whether the true mean was distinguishable from zero, to the 99% confidence limit. We use a double-sided significance test, since we are interested in wind components that are statistically significant, regardless of the sign. We set the number of degrees of freedom in our test to be the number of individual events that went into generating the composite minus 1, using previous notation: N − 1. According to the Student’s t-test, our averages were significant to the 99% level if the following inequality was satisfied:
i1520-0442-10-12-3131-eqa3
where U, σ, x, y, tn, and N are as before; êj is the unit vector in the zonal (j = 1) or meridional (j = 2) direction; t99%,N−1 is the double-sided Student’s t-coefficient at the 99% significance level for N degrees of freedom; and ξ·η indicates the vector dot product operation. The values for t99%,N−1 are tabulated in most statistics or mathematical table handbooks, for example, Speigel (1994) and Zwillinger (1996).

Correlation coefficient significance test

We performed a Student’s t-test (Bickel and Docksun 1977, 220) on our sample correlation coefficients, r, to evaluate whether the true correlation coefficient, ρ, approximated r was distinguishable from zero, to a prescribed significance level, l. We use a double-sided significance test because we are interested in significant correlations regardless of the sign of the correlation. The following inequality had to be satisfied in this test:
i1520-0442-10-12-3131-eqa4
where tl,N−2 is the double-sided Student’s t-coefficient for l significance level and for N − 2 degrees of freedom. Since in our case N = 10, the threshold for 90% significance is |r| > 0.55, and the threshold for 99% significance is |r| > 0.77, for the zero lag correlation tests;while for the 1-yr lagged case N = 9, so the threshold for 90% significance is |r| > 0.58, and for 95% significance |r| > 0.67. The values for tl,N−2 are tabulated in most statistics or mathematical table handbooks, for example, Speigel (1994) and Zwillinger (1996).

Monthly distribution significance test

We performed a Monte Carlo test (Livezey and Chen 1983) on the monthly distribution data to find the probability, P, that the “on” season we observed for each event had of occurring randomly. We then took the significance level of the seasonal distribution to be 100 × (1 − P)%. The probability was computed using a Monte Carlo method, with 100000 Monte Carlo samples.

The procedure to determine the probability of Nseas events being distributed over a continuous Lseas month period on a random distribution of N events is as follows. We sample 100000 times; at each sampling we randomly distribute the N events, for the particular event type, into 12 equally likely months. We then test whether there exists a continuous period of Lseas months for which the total number of events randomly placed is equal to or greater than Nseas. We count the number of times the test turns out true in our procedure and define the probability to be P = E/B, where E is the number of positive tests and B is the number of Monte Carlo samples. The probability converged for B > 1000, but we took B = 100000 for good measure.

Event sequencing significance test

We performed a Monte Carlo test (Livezey and Chen 1983) to find the probability, P, that the sequencing pattern we observed for each event pair had of occurring randomly. We then took the significance level of the event sequencing to be 100 × (1 − P)%. The probability was computed using a Monte Carlo method, with 10000 Monte Carlo samples.

We determined the probability of distributing N1 center days of type 1, and N2 center days of type 2 on a 10-yr time axis semirandomly, so that no two center days of the same type were within 7 days of each other, and having M or more ordered pairs of events (n1, n2) within 3 days of each other in the following manner. We sample 10000 times, at each sampling step semirandomly distributing N1 and N2 events as just described. We then test to see how many ordered pairs (n1, n2) are within 3 days of each other. We count the number of times the test turns out true in our procedure and define the probability to be P = E/B, where E is the number of positive tests and B is the number of Monte Carlo samples. The probability converged for B > 1000, but we took B = 10000 for good measure.

Fig. 1.
Fig. 1.

Contours of (a) March and (b) September 10-m climatological zonal wind from ECMWF (1986–95), (c) March and (d) September 10-m climatological zonal wind from COADS (1946–93), and (e) March and (f) September difference (ECMWF–COADS) 10-m climatological zonal winds. Contour intervals are 1.0 m s−1 for (a)–(d), and 0.5 m s−1 for (e) and (f). Dashed contours indicate negative values. COADS climatology data is smoothed using a five-point triangle filter in the zonal direction and a three-point triangle filter in the meridional direction.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 2.
Fig. 2.

Same as Fig. 1, except for 10-m meridional wind.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 3.
Fig. 3.

Same as Fig. 2, except for 10-m wind divergence. Contour interval is 1 × 10−6 s−1 for all six panels.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 4.
Fig. 4.

Values of the rms difference (TAO–ECMWF) from 1986 to 1995 for surface (a) zonal winds and (b) meridional winds at the TOGA–TAO buoy locations. Units are m s−1. The time period over which TAO data were available varies with location.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 5.
Fig. 5.

Classification regions superimposed on the March (a) ECMWF climatological (1986–95) 10-m wind vectors, (b) contours of COADS climatological (1946–93) SST, and (c) contours of COADS climatological (1946–93) SLP −1000 mbar. Scale vector for the winds is 5 m s−1. Contour interval for SST is 1.0°C. Contour interval for SLP is 2.0 mbar. (See text.)

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 6.
Fig. 6.

Same as Fig. 5, except for the month of September.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 7.
Fig. 7.

Type NW composite anomaly WWE 10-m wind anomaly vector map for (a) day(−2), (b) day(0), and (c) day(2). The classifying region is indicated by the thin-lined box. The scale vector is 5 m s−1. Zonal wind anomalies statistically significant at 99% are indicated by bold vectors; meridional wind anomalies significant at 99% are indicated by shaded background. Significance is determined as described in the appendix.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 8.
Fig. 8.

Same as Fig. 7, except for type N composite anomaly WWE.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 9.
Fig. 9.

Same as Fig. 7, except for type NE composite anomaly WWE.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 10.
Fig. 10.

Same as Fig. 7, except for type W composite anomaly WWE.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 11.
Fig. 11.

Same as Fig. 7, except for type C composite anomaly WWE.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 12.
Fig. 12.

Same as Fig. 7, except for type E composite anomaly WWE.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 13.
Fig. 13.

Same as Fig. 7, except for type S composite anomaly WWE.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 14.
Fig. 14.

Same as Fig. 7, except for type SE composite anomaly WWE.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 15.
Fig. 15.

(a) Type NW, (b) type N, and (c) type NE composite WWE 10-m wind vector map, for event day(0). Scale vector is 5 m s−1. Classifying region is indicated by the box.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 16.
Fig. 16.

(a) Type W, (b) type C, and (c) type E composite WWE 10-m wind vector map, for event day(0). Scale vector is 5 m s−1. Classifying region is indicated by the box.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 17.
Fig. 17.

(a) Type S and (b) type SE composite WWE 10-m wind vector map, for event day(0). Scale vector is 5 m s−1. Classifying region is indicated by the box.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 18.
Fig. 18.

Histograms for WWE (a) duration, (b) wind measure, (c) maximum averaged anomaly, and (d) maximum point anomaly, for WWEs of all types occurring during the 1986–95 period. Bar graphs indicate the number of WWEs occurring within the labeled bins, and the line graph indicates the cumulative percentage of events at each bin. Units are days for duration, and 106 m for wind measure, and m s−1 for maximum point anomaly and maximum averaged anomaly. Quantities are as defined in section 3a.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 19.
Fig. 19.

(a) Longitude vs time and (b) time vs latitude contour plots of 10-m zonal wind anomaly for the type C composite WWE; and (c) longitude vs time and (d) time vs latitude contour plots of 10-m zonal wind anomaly for type SE composite WWE. Contour intervals are 1 m s−1 and dark contour indicates e−1 level of the zonal wind anomaly. Center day of event is day 10 on the time axis.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 20.
Fig. 20.

Vector plots of the wind fields for the composite WWE modeled wind anomaly field on (a) 9 Dec 1992, (c) 5 Jan 1993, and (e) 29 Jan 1993; and ECMWF daily wind anomaly field on (b) 9 Dec 1992, (d) 5 Jan 1993, and (f) 29 Jan 1993. These are three WWE center dates occurring in the TOGA COARE IOP that are well represented by the composite WWE model. Bold wind anomaly vectors indicate zonal wind anomaly exceeding 3 m s−1. Classifying regions are shown by thin-lined boxes.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 21.
Fig. 21.

Same as Fig. 20, except for (a) and (b) 20 Nov 1992, (c) and (d) 29 Nov 1992, and (e) and (f) 6 Feb 1993. These are the three WWE center dates occurring in the TOGA COARE IOP that are most poorly represented by the composite WWE model.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 22.
Fig. 22.

Monthly distribution of WWEs for (a) type NW, (b) type N, (c) type NE, (d) type W, (e) type C, (f) type E, (g) type S, and (h) type SE.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 23.
Fig. 23.

Same as Fig. 22, except for the monthly distribution of WWEs with wind measure >2.0 × 106 m. Wind measure is as defined in section 3a.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 24.
Fig. 24.

Plots of yearly distribution of WWEs for (a) type NW, (b) type N, (c) type NE, (d) type W, (e) type C, (f) type E, (h) type S, and (i) type SE; and (g) plot of 12-month running mean (−1) × SOI. Superimposed on all the panels is the 10-yr-(1986–1995) mean of the plotted quantity.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 25.
Fig. 25.

Scatterplot of wind measure vs center date of WWE, for all the WWEs in the period 1986–95. Wind measure is as defined in section 3a, and the units are 106 m.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Fig. 26.
Fig. 26.

Cartoon description of the mega-WWEs (MWWEs) in our record. The dates for the MWWEs appear on the top of each figure. Inside the regions involved in the MWWE is listed the wind measure for the particular WWEs. Wind measure values are bolded if greater than 2 × 106 m. Units for wind measure are 106 m.

Citation: Journal of Climate 10, 12; 10.1175/1520-0442(1997)010<3131:WWEITT>2.0.CO;2

Table 1.

Number of WWEs identified during 1986–95, in each of the classifying regions, for the complete event list and for the nonoverlapping event list (see section 3a).

Table 1.
Table 2.

Duration, maximum averaged anomaly, wind measure, and maximum point anomaly for each composite WWE. Quantities as defined in section 3a.

Table 2.
Table 3.

Table of scales for the composite WWEs according to the simple Gaussian model described in section 4b.

Table 3.
Table 4.

Table of WWEs occurring during the TOGA COARE IOP (Nov 1992–Feb 1993). Quantities are as defined in section 3a. Highlighted as bold italics are the events whose wind measure statistic is greater than or equal to 2.5 × 106 m.

Table 4.

* PMEL Contribution Number 1811 and JISAO Contribution Number 1745.

Save
  • Bickel, P. J., and K. A. Doksum, 1977: Mathematical Statistics: Basic Ideas and Selected Topics. Holden-Day, 492 pp.

  • Chen, S. S., R. A. Houze Jr., and B. E. Mapes, 1996: Multiscale variability of deep convection in relation to large-scale circulation in TOGA COARE. J. Atmos. Sci.,53, 1380–1409.

  • Chu, P. S., 1988: Extratropical forcing and the burst of equatorial westerlies in the western Pacific: A synoptic study. J. Meteor. Soc. Japan,66, 4549–4564.

  • ——, and J. Frederick, 1990: Westerly wind bursts and surface heat fluxes in the equatorial western Pacific in May 1982. J. Meteor. Soc. Japan,68, 5523–5536.

  • Delcroix, T., G. Eldin, M. McPhaden, and A. Morliere, 1993: Effects of westerly wind bursts upon the western equatorial Pacific Ocean, February–April 1991. J. Geophys. Res.,98, 16379–16385.

  • ECMWF, 1989: The description of the ECMWF/WCRP level III—A global atmospheric data archive. Tech. Attachment, 72 pp. [Available from ECMWF, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Efron, B., and R. Tibshirani, 1991: Statistical data analysis in the computer age. Science,253, 390–395.

  • Eldin, G., T. Delcroix, C. Hénin, K. Richards, Y. Du Penhoat, J. Picaut, and P. Rual, 1994: Large-scale current and thermohaline structures along 156°E during the COARE intensive observation period. Geophys. Res. Lett.,21, 2681–2684.

  • Geise, B. S., and D. E. Harrison, 1990: Aspects of the Kelvin wave response to episodic wind forcing. J. Geophys. Res.,95(C5), 7289–7312.

  • ——, and ——, 1991: Eastern equatorial Pacific response to three composite westerly wind types. J. Geophys. Res.,96(Suppl.), 3239–3248.

  • Gill, A. E., 1982: Atmosphere-Ocean Dynamics. Academic Press, 662 pp.

  • Harrison, D. E., 1984: On the appearance of sustained equatorial westerlies during the 1982 Pacific warm event. Science,225, 1099–1102.

  • ——, and P. S. Schopf, 1984: Kelvin-wave-induced anomalous advection and the onset of surface warming in El Niño events. Mon. Wea. Rev.,112, 923–933.

  • ——, and B. S. Geise, 1988: Remote westerly wind forcing of the eastern equatorial Pacific: Some model results. Geophys. Res. Lett.,15, 804–807.

  • ——, and D. S. Luther, 1990: Surface winds from tropical Pacific islands—Climatological statistics. J. Climate,3, 2251–2271.

  • ——, and B. S. Geise, 1991: Episodes of surface westerly winds as observed from islands in the western tropical Pacific. J. Geophys. Res.,96, 3221–3237.

  • ——, and N. Larkin, 1996: The COADS sea level pressure signal: A near-global El Niño composite and time series view, 1946–93. J. Climate,9, 3025–3055.

  • ——, and ——, 1997: The ENSO surface temperature and wind signal: A near-global composite and time-series view, 1946–1995. Rev. Geophys., in press..

  • Hartten, L. M., 1996: Synoptic settings of westerly wind bursts. J. Geophys. Res.,101(D12), 16997–17019.

  • Joint Typhoon Warning Center, 1994a: Western North Pacific typhoons—1994. Mar. Wea. Log,38(1), 34–40.

  • ——, 1994b: Western North Pacific typhoons—1993. Mar. Wea. Log,38(4), 16–23.

  • Keen, R. A., 1982: The role of cross-equatorial cyclone pairs in the Southern Oscillation. Mon. Wea. Rev.,110, 1405–1416.

  • Kiladis, G. N., G. A. Meehl, and K. M. Weickmann, 1994: Large- scale circulation associated with westerly wind bursts and deep convection over the western equatorial Pacific. J. Geophys. Res.,99(D9), 18527–18544.

  • Kindle, J. C., and P. A. Phoebus, 1995: The ocean response to operational wind bursts during the 1991–1992 El Niño. J. Geophys. Res.,100(C3), 4803–4920.

  • Large, W. G., and S. Pond, 1981: Open ocean momentum flux measurments in moderate to strong winds. J. Phys. Oceanogr.,11, 324–336.

  • Lau, K. H., and N. C. Lau, 1992: The energetics and propagation dynamics of tropical summertime synoptic-scale disturbances. Mon. Wea. Rev.,120, 2523–2539.

  • Lau, K. M., L. Peng, C. H. Sui, and T. Nakazawa, 1989: Dynamics of super cloud clusters, westerly wind bursts, 30–60-day oscillations and ENSO: A unified view. J. Meteor. Soc. Japan,67, 2205–2219.

  • Lin, X., and R. H. Johnson, 1996: Kinematic and thermodynamic characteristics of the flow over the western Pacific warm pool during TOGA COARE. J. Atmos. Sci.,53, 5695–5715.

  • Livezey, R. E., and W. Y. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev.,111, 47–59.

  • Love, G., 1985: Cross-equatorial influence of winter hemisphere subtropical cold surges. Mon. Wea. Rev.,113, 1487–1498.

  • Lukas, R., P. J. Webster, M. Ji, and A. Leetma, 1995: The large-scale context for the TOGA Coupled Ocean-Atmosphere Response Experiment. Meteor. Atmos. Phys.,56, 3–16.

  • Luther, D. S., D. E. Harrison, and R. A. Knox, 1983: Zonal winds in the central equatorial Pacific and El Niño. Science,222, 327–330.

  • Madden, R. A., and P. R. Julian, 1972: Description of global-scale circulation cells in the tropics with a 40–50-day period. J. Atmos. Sci.,29, 1109–1123.

  • ——, and ——, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev.,122, 814–837.

  • Mangum, L. J., H. P. Freitag, and M. J. McPhaden, 1994: TOGA- TAO array sampling schemes and sensor evaluations. Proc. Oceans ’94 OSATES,II, 402–406.

  • McBride, J. L., N. E. Davidson, K. Puri, and G. C. Tyrell, 1995: The flow during TOGA COARE as diagnosed by the BMRC tropical analysis and prediction system. Mon. Wea. Rev.,123, 717–736.

  • McPhaden, M. J., 1993: TOGA-TAO and the 1991–93, El Niño–Southern Oscillation Event. Oceanography,6(2), 36–44.

  • ——, F. Bahr, Y. Du Penhoat, E. Firing, S. P. Hayes, P. P. Niiler, P. L. Richardson, and J. M. Toole, 1992: The response of the western equatorial Pacific Ocean to westerly wind bursts during November 1989 to January 1990. J. Geophys. Res.,97, 14289–14303.

  • Meehl, G. A., G. N. Kiladis, K. M. Weickmann, M. Wheeler, D. S. Gutzler, and G. P. Compo, 1996: Modulation of equatorial subseasonal convective episodes by tropical-extratropical interaction in the Indian and Pacific Ocean regions. J. Geophys. Res.,101(D10), 15033–15049.

  • Rasmusson, E. M., and T. H. Carpenter, 1982: Variations in tropical sea surface temperature and surface wind fields associated with the Southern Oscillation/El Niño. Mon. Wea. Rev.,110, 354–384.

  • Rui, H., and B. Wang, 1990: Development characteristics and dynamic structure of tropical intraseasonal convection anomalies. J. Atmos. Sci.,47, 357–379.

  • Shopf, P. S., and D. E. Harrison, 1983: On equatorial Kelvin waves and El Niño, I, Influence of initial states on wave-induced current and warming. J. Phys. Oceanogr.,13, 936–948.

  • Spiegel, M. R., 1994: Theory and Problems of Statistics. Schaum’s Outline Series, McGraw-Hill, 504 pp.

  • Sui, C. H., and K. M. Lau, 1992: Multiscale phenomena in the tropical atmosphere over the western Pacific. Mon. Wea. Rev.,120, 407–430.

  • Tsutsui, J., and A. Kasahara, 1996: Simulated tropical cyclones using the National Center for Atmospheric Research community climate model. J. Geophys. Res.,101(D10), 15013–15032.

  • Vecchi, G. A., and D. E. Harrison, 1997: Westerly wind events in the tropical Pacific, 1986–1995; An atlas. NOAA/PMEL Tech. Memo. ERL PMEL-109, 222 pp. [Available from NTIS, 5285 Port Royal Rd., Springfield, VA 22161.].

  • Webster, P. J., and R. Lukas, 1992: TOGA COARE: The Coupled Ocean–Atmosphere Response Experiment. Bull. Amer. Meteor. Soc.,73, 1377–1416.

  • Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer, 1987: A Comprehensive Ocean–Atmosphere Data Set. Bull. Amer. Meteor. Soc.,68, 1239–1250.

  • Zwillinger, D., Ed., 1996: CRC—Standard Mathematical Tables and Formulae. 30th ed. CRC Press, 812 pp.

  • Fig. 1.

    Contours of (a) March and (b) September 10-m climatological zonal wind from ECMWF (1986–95), (c) March and (d) September 10-m climatological zonal wind from COADS (1946–93), and (e) March and (f) September difference (ECMWF–COADS) 10-m climatological zonal winds. Contour intervals are 1.0 m s−1 for (a)–(d), and 0.5 m s−1 for (e) and (f). Dashed contours indicate negative values. COADS climatology data is smoothed using a five-point triangle filter in the zonal direction and a three-point triangle filter in the meridional direction.

  • Fig. 2.

    Same as Fig. 1, except for 10-m meridional wind.

  • Fig. 3.

    Same as Fig. 2, except for 10-m wind divergence. Contour interval is 1 × 10−6 s−1 for all six panels.

  • Fig. 4.

    Values of the rms difference (TAO–ECMWF) from 1986 to 1995 for surface (a) zonal winds and (b) meridional winds at the TOGA–TAO buoy locations. Units are m s−1. The time period over which TAO data were available varies with location.

  • Fig. 5.

    Classification regions superimposed on the March (a) ECMWF climatological (1986–95) 10-m wind vectors, (b) contours of COADS climatological (1946–93) SST, and (c) contours of COADS climatological (1946–93) SLP −1000 mbar. Scale vector for the winds is 5 m s−1. Contour interval for SST is 1.0°C. Contour interval for SLP is 2.0 mbar. (See text.)

  • Fig. 6.

    Same as Fig. 5, except for the month of September.

  • Fig. 7.

    Type NW composite anomaly WWE 10-m wind anomaly vector map for (a) day(−2), (b) day(0), and (c) day(2). The classifying region is indicated by the thin-lined box. The scale vector is 5 m s−1. Zonal wind anomalies statistically significant at 99% are indicated by bold vectors; meridional wind anomalies significant at 99% are indicated by shaded background. Significance is determined as described in the appendix.

  • Fig. 8.

    Same as Fig. 7, except for type N composite anomaly WWE.

  • Fig. 9.

    Same as Fig. 7, except for type NE composite anomaly WWE.

  • Fig. 10.

    Same as Fig. 7, except for type W composite anomaly WWE.

  • Fig. 11.

    Same as Fig. 7, except for type C composite anomaly WWE.

  • Fig. 12.

    Same as Fig. 7, except for type E composite anomaly WWE.

  • Fig. 13.

    Same as Fig. 7, except for type S composite anomaly WWE.

  • Fig. 14.

    Same as Fig. 7, except for type SE composite anomaly WWE.

  • Fig. 15.

    (a) Type NW, (b) type N, and (c) type NE composite WWE 10-m wind vector map, for event day(0). Scale vector is 5 m s−1. Classifying region is indicated by the box.

  • Fig. 16.

    (a) Type W, (b) type C, and (c) type E composite WWE 10-m wind vector map, for event day(0). Scale vector is 5 m s−1. Classifying region is indicated by the box.

  • Fig. 17.

    (a) Type S and (b) type SE composite WWE 10-m wind vector map, for event day(0). Scale vector is 5 m s−1. Classifying region is indicated by the box.

  • Fig. 18.

    Histograms for WWE (a) duration, (b) wind measure, (c) maximum averaged anomaly, and (d) maximum point anomaly, for WWEs of all types occurring during the 1986–95 period. Bar graphs indicate the number of WWEs occurring within the labeled bins, and the line graph indicates the cumulative percentage of events at each bin. Units are days for duration, and 106 m for wind measure, and m s−1 for maximum point anomaly and maximum averaged anomaly. Quantities are as defined in section 3a.

  • Fig. 19.

    (a) Longitude vs time and (b) time vs latitude contour plots of 10-m zonal wind anomaly for the type C composite WWE; and (c) longitude vs time and (d) time vs latitude contour plots of 10-m zonal wind anomaly for type SE composite WWE. Contour intervals are 1 m s−1 and dark contour indicates e−1 level of the zonal wind anomaly. Center day of event is day 10 on the time axis.

  • Fig. 20.

    Vector plots of the wind fields for the composite WWE modeled wind anomaly field on (a) 9 Dec 1992, (c) 5 Jan 1993, and (e) 29 Jan 1993; and ECMWF daily wind anomaly field on (b) 9 Dec 1992, (d) 5 Jan 1993, and (f) 29 Jan 1993. These are three WWE center dates occurring in the TOGA COARE IOP that are well represented by the composite WWE model. Bold wind anomaly vectors indicate zonal wind anomaly exceeding 3 m s−1. Classifying regions are shown by thin-lined boxes.

  • Fig. 21.

    Same as Fig. 20, except for (a) and (b) 20 Nov 1992, (c) and (d) 29 Nov 1992, and (e) and (f) 6 Feb 1993. These are the three WWE center dates occurring in the TOGA COARE IOP that are most poorly represented by the composite WWE model.

  • Fig. 22.

    Monthly distribution of WWEs for (a) type NW, (b) type N, (c) type NE, (d) type W, (e) type C, (f) type E, (g) type S, and (h) type SE.

  • Fig. 23.

    Same as Fig. 22, except for the monthly distribution of WWEs with wind measure >2.0 × 106 m. Wind measure is as defined in section 3a.

  • Fig. 24.

    Plots of yearly distribution of WWEs for (a) type NW, (b) type N, (c) type NE, (d) type W, (e) type C, (f) type E, (h) type S, and (i) type SE; and (g) plot of 12-month running mean (−1) × SOI. Superimposed on all the panels is the 10-yr-(1986–1995) mean of the plotted quantity.

  • Fig. 25.

    Scatterplot of wind measure vs center date of WWE, for all the WWEs in the period 1986–95. Wind measure is as defined in section 3a, and the units are 106 m.

  • Fig. 26.

    Cartoon description of the mega-WWEs (MWWEs) in our record. The dates for the MWWEs appear on the top of each figure. Inside the regions involved in the MWWE is listed the wind measure for the particular WWEs. Wind measure values are bolded if greater than 2 × 106 m. Units for wind measure are 106 m.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 2018 947 37
PDF Downloads 650 192 24