Post on 08-Aug-2018
transcript
8/22/2019 Mon. Wea. Rev. 2007 Douglas
1/21
Synoptic Variability of Rainfall and Cloudiness along the Coasts of Northern Peru and
Ecuador during the 1997/98 El Nio Event
MICHAEL W. DOUGLAS
National Severe Storms Laboratory, Norman, Oklahoma
JOHN MEJIA
Cooperative Institute for Mesoscale Meteorological Studies, University of Oklahoma, Norman, Oklahoma
NORMA ORDINOLA
Universidad de Piura, Piura, Peru
JOSHUA BOUSTEADNational Weather Service, Omaha, Nebraska
(Manuscript received 6 March 2007, in final form 3 June 2008)
ABSTRACT
This paper describes the meteorological conditions associated with large fluctuations in rainfall over the
coastal regions of northern Peru and Ecuador during the 1997/98 El Nio event. Using data from a network
of routine rain gauges and special gauges established just prior to the onset of heavy rains, it is shown that
large variations in the daily rainfall on quasi-weekly time scales occurred during the period JanuaryApril
1998. These rainfall fluctuations were approximately in phase along the coast from near the equator to
7S. The daily rainfall data was averaged to develop a subset of wet and dry days, and then these dates
were used as the basis for compositing. Special pilot balloon observations were composited with respect to
the wet and dry days, showing that westerly and northerly wind anomalies are associated with wet spells.Composites of the National Centers for Environmental PredictionNational Center for Atmospheric Re-
search (NCEPNCAR) reanalysis and outgoing longwave radiation (OLR) data support a modest asso-
ciation of anomalous westerly wind events with enhanced rainfall.
The relationship observed between westerly zonal wind anomalies and rainfall west of the Andes during
1998 suggested using the NCEP reanalysis to develop composites based on westerly wind events observed
during other years. Zonal wind anomalies at 700 hPa were used as the primary criterion for stratifying wet
and dry days, despite reservations about the association between rainfall and zonal wind. Compositing
Geostationary Operational Environmental Satellite (GOES) and OLR data for 220 west wind anomaly
events from the months of JanuaryApril for the years 19902005 showed that they are associated with
enhanced cloudiness that propagates eastward at 10 m s1. The composites using NCEP reanalyses show
the evolution of the wind field associated with the wet days and suggest a link between extratropical wave
passages across North America and anomalous westerly wind events off the coast of Ecuador and northern
Peru.
1. Introduction
The El Nio phenomenon has been the subject of
many research studies, numerous reviews (e.g., Enfield
1989), many books (e.g., Philander 1990, 2004) and ar-
ticles in the popular literature. In large part this has
been a consequence of the relatively recent, and very
large, El Nio events of 1982/83 and 1997/98. Associ-
ated with major El Nio events is an extreme enhance-
ment in the rainfall over the coastal regions of northern
Peru and Ecuador, and this has been known for some
time as one of the most dramatic climatic anomalies
found anywhere on earth (Trewartha 1962).
Corresponding author address: Dr. Michael W. Douglas, Na-
tional Severe Storms Laboratory, 120 David L. Boren Blvd., Nor-
man, OK 73072.
E-mail: michael.douglas@noaa.gov
116 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
DOI: 10.1175/2008MWR2191.1
2009 American Meteorological Society
8/22/2019 Mon. Wea. Rev. 2007 Douglas
2/21
a. Previous studies of daily rainfall variability
Although many studies have involved planetary-scale
aspects of the El Nio phenomenon, especially since
1983, comparatively few studies have focused on the
regional or local aspects of the rainfall along the coast
of northwestern South America. Even fewer studies
have considered the synoptic, or high-frequency vari-
ability of rainfall during these events, despite the ob-
servation of large daily rainfall fluctuations that occur
over northern Peru. Horel and Cornejo-Garrido (1986)
described the evolution of the 1982/83 El Nio event
over the eastern Pacific, with special emphasis on the
Peruvian coast. Using outgoing longwave radiation
(OLR) and surface wind data they described the dif-
ferences between the previous year and the El Nio
year, showing that weaker surface winds were present
during 1983. They also noted that daily rainfall data,
averaged over 17 stations, showed that the main heavyrainfall events were episodic and separated by dry pe-
riods. OLR data also showed this, as well as a large
diurnal variation in convection, with an offshore early
morning maximum similar to many other tropical re-
gions. Goldberg et al. (1987) described the peculiar me-
soscale features of the rainfall around Piura, Peru (see
Fig. 1 for geography) during the 1983 rains and their
relation to the topography of northern Peru. Their
study focused on the daily rainfall data from 66 stations
and examination of geostationary imagery and loops.
They forcefully made the point that, within the envelop
of the overall El Nio seasonal enhancement of rainfall,the especially destructive rain events responsible for
the most significant flooding and mudflows were epi-
sodic and quasi-periodic. They also speculated on the
possible forcing for these heavy rain events, and evalu-
ated the hypothesis that disturbances over the Amazon
basin might be a possible inducement for coastal con-
vective events. Bendix (2000), working principally with
satellite imagery and cloud track winds, described char-
acteristics of the cloudiness and its diurnal variation
over the region of Ecuador and northern Peru during
the 1991/92 El Nio. The role of landsea breeze cir-
culations was stressed in controlling the daily cloudi-
ness.
A recent study involving daily rainfall over northern
Peru is that of Takahashi (2004), where he examined
two recent years (1998 and 2002) when elevated sea
surface temperatures favored deep convection over
northern Peru. The unique feature of this study was the
analysis of boundary layer wind profiler observations
from Piura, which allowed for a detailed depiction of
the diurnal cycle of the winds up to the midtroposphere.
By comparing wet and dry day composites of the pro-
filer data, Takahashi showed that wet days had a stron-
ger westerly wind component and that this was concen-
trated in the late afternoon to early evening hours. The2002 results showed evidence of a deep positive zonal
wind anomaly during wet days that extended to 5-km
altitude. Takahashi also composited the National Cen-
ters for Environmental PredictionNational Center for
Atmospheric Research (NCEPNCAR) reanalyses for
the wet and dry days for both years, and showed that
the westerly wind anomalies extended over a 3500-km
zonal extent. The small sample size during 2002 (8 wet
days) limited the significance of the results compared
with those from 1997/98.
b. Overview of the present study
A natural step toward the possible prediction of days
with high and low rainfall during El Nio events (or
even other years) is to determine whether a large-scale
signal is present in meteorological fields that can be
resolved by either direct meteorological measurements
(such as pilot balloon or radiosonde observations) or
from routinely available analyses or satellite measure-
ments. This paper seeks to advance this goal, with its
main objective being to describe the relationship be-
tween the rainfall variability on synoptic time scales
FIG. 1. The special sounding sites established for the 1997/98 El
Nio event.
JANUARY 2009 D O U G L A S E T A L . 117
8/22/2019 Mon. Wea. Rev. 2007 Douglas
3/21
over the coastal region of northern Peru and Ecuador
and larger-scale meteorological fields. Special experi-
ment data collected during the 1997/98 El Nio event
(described in section 2), served as the initial motivation
for our work. Although a preliminary analysis of the
special observations was carried out and informally re-
ported (Douglas et al. 1999), the availability of addi-tional data sources, including daily rainfall observa-
tions from the Ecuadorean National Meteorological
Service [i.e., the Instituto Nacional de Meteorologia
e Hidrologia (INAMHI)], along with 10-km Geosta-
tionary Operational Environmental Satellite (GOES)
imagery for the period, motivated expansion and ex-
tension of our original effort. Concerns about the gen-
erality of the 1998 results led to an expansion of the
work, and this paper now presents results from our
investigation of the synoptic variability of cloudiness
over the period 19902005. To this end we have used
OLR and NCEPNCAR reanalyses to describe thesevariations.
2. Data used in this study
The Pan American Climate Studies Sounding Net-
work (PACS-SONET) project (Douglas and Fernandez
1997; Douglas and Murillo 2008) established 12 pilot
balloon stations from southern Mexico to northern
Peru in early 1997, which were to make twice-daily ob-
servations for a 6-month period. The projects initial
focus was to describe atmospheric circulation variationsassociated with wet and dry spells over Central
America (Pea and Douglas 2002). A special adapta-
tion of the projects activities for the 1997/98 El Nio
event provided additional measurements that have
been used in this study, and are described below.
a. The special El Nio pilot balloon network
The PACS-SONET project had established pilot bal-
loon stations in May 1997 at Piura, Peru, and in Ecua-
dor at Guayaquil, Esmeraldas, and San Cristobal (in
the Galapagos Islands). Support was requested in late
1997 to establish two additional pilot balloon sites in
Ecuador and five additional sites in Peru (Fig. 1). Ob-
server training was carried out between late November
1997 and early January 1998. Together, all of the sta-
tions attempted to make twice-daily observations
through May 1998. Unfortunately, major logistical
problems associated with gas cylinder transportation
occurred after the start of heavy rains. Many road seg-
ments were washed out in northern Peru, making the
delivery of the gas cylinders from the south difficult or
impossible. The frequent occurrence of low cloudiness
also prevented tracking many of the balloons to high
levels. Despite these unfavorable conditions, some 1467
pilot balloon observations were made from 1 December
1997 to 31 May 1998, with March 1998 being the most
densely sampled, with 391 observations. (The complete
dataset is available from the PACS-SONET Web site:http://www.nssl.noaa.gov/projects/pacs).
b. Rain gauge data
The rainfall data used in this study included available
Peruvian Servicio Nacional de Meteorologa e Hidrologa
(SENAMHI) and Ecuadorean (INAMHI) Meteoro-
logical Service observations, other regional network
observations, and special PACS-SONET rain gauge ob-
servations. The routine SENAMHI and INAMHI rain
gauges, with a 20-cm aperture, were read manually. The
PACS-SONET rain gauges, of which about 100 wereinstalled prior to the onset of heavy rains, were of a
wedge-shaped design (Tru-Chek brand) with an 6
cm by 6 cm rectangular opening. These gauges were
compared with the SENAMHI gauges at a few sites
and the differences between these gauges and the
SENAMHI were found to be small. For this study, any
systematic differences between the rain gauges have
been ignored.
The rain gauges were read each morning at between
0700 and 0800 LT (12001300 UTC). Local observa-
tions and satellite imagery suggested that the rainfall
over land had a strong diurnal cycle, with most rainoccurring from early afternoon to late evening. The
time of the rain gauge observation was a minimum in
the rainfall, allowing clear separation of the rainfall
events, with the morning rainfall measurement being
assigned to the previous day.
c. NCEPNCAR reanalyses and satellite imagery
To complement, and to compare with, the pilot bal-
loon observations, we also have used in this study the
NCEPNCAR reanalyses (Kalnay et al. 1996; hereafter
NCEP reanalyses). The GOES infrared imagery,
available at 3-hourly intervals, was obtained from the
National Oceanic and Atmospheric Administration
(NOAA) National Climatic Data Center (NCDC)
International Satellite Cloud Climatology Project
(ISCCP; Knapp 2004). We have used these data, with
10-km resolution, for the period 19982005. In addition,
we have also used daily-averaged OLR data (Liebmann
and Smith 1996) available at 2.5 resolution for com-
parison with the NCEP reanalyses and with the GOES
imagery.
118 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
4/21
3. Methodology
a. Identifying wet and dry days
The basic methodology of our work was that of a
compositing study, whereby we hoped to distinguish
the differences between the conditions associated with
very wet days and those associated with relatively dry
days. To develop such composites required a procedure
to distinguish wet days from dry days. Since tropical
convective rainfall can be highly variable in space, we
averaged over the rain gauge network, as done by both
Horel and Cornejo-Garrido (1986) and Goldberg et al.
(1987), to obtain a signal more representative of larger
spatial scales. Averaging the rain gauges also produced
a product with a spatial scale comparable to the pilot
balloon station separation and the 2.5 resolution of the
NCEP reanalyses. The rain gauge data were averaged
over all stations from the PeruEcuador border (3.5S)
to just north of Trujillo (8S). Only stations west of
the Andean highest terrain were used in the averaging.
Figure 2 summarizes the essential aspects of the spa-
tially averaged daily rainfall. The daily number of Pe-
FIG. 2. (a) Number of rain gauges reporting observations each day in northern Peru. (b)
Percentage of the number of station reporting rainfall each day in northern Peru. (c) Average
daily rainfall (mm) for all rain gauges reporting on a particular day (solid circles indicate the
wet spell days, solid squares indicate the dry spell centers, and open squares for all other days).
(d) As in (c), but for an average of 27 Ecuadorean stations west of the Andes. The dashed lines
in (c) and (d) show a 30-day running mean.
JANUARY 2009 D O U G L A S E T A L . 119
8/22/2019 Mon. Wea. Rev. 2007 Douglas
5/21
ruvian rainfall reports received varied from approxi-
mately 30 in early December 1997 to just over 100 in
March and April 1998 (Fig. 2a); Ecuadorean stationswere more uniform in number throughout the period
(not shown). A plot of the percentage of stations re-
porting rain showed some evidence of synoptic variabil-
ity (Fig. 2b). However, the variability became much
more evident when the total daily rainfall (sum of rain-
fall from all stations) was divided by the number of
stations reporting (Fig. 2c). The resulting daily aver-
aged rainfall varies from more than 30 mm day1 on
wet days to less than 10 mm day1 on drier days. Days
when the rainfall over the network was clearly a maxi-
mum relative to days before or after were deemed wet
and those with a distinct minimum were considered dry.
This selection was somewhat subjective, but sufficient
to clearly distinguish wet days from dry days. Because
the rainfall increased from mid-December to February
the wet days in the early season could have less rainfall
than those later in the wet season. A total of 18 wet
events and 17 dry events were selected; this gave an
average period of6 days for the eventsthough con-
siderable variability about this value is evident.
The quasi-periodicity of the events evident from Fig.
2c and the fact that these variations reflect rainfall over
the coastal region of northern Peru suggested that the
rainfall might be modulated by larger-scale controls. A
time series for the average rainfall of the 27 stationsreporting in coastal Ecuador (Fig. 2d), though less nu-
merous and less dense than those in northern Peru,
shows a general similarity, though not a close match, to
the time series from the Peruvian stations.
FIG. 3. (a) Analysis of mean rainfall (mm day1) for JanuaryApril 1998, based on rain gauge observations over northern Peru and
southern Ecuador. Special rain gauges installed for the season are shown as open dots. Average (b) wet and (c) dry days (mm day1),
and (d) relative amplitude (%) between wet and dry spells compared to mean daily rainfall. The solid (open) circles in (d) denote
positive (negative) relative amplitude between wet and dry spells. The shaded contours represent the elevation in meters.
FIG. 4. Average daily rainfall of northern Peru (mm) vs daily
rainfall of southern Ecuador (mm). The solid line is least squares
fit to all data points.
120 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
6/21
b. Compositing with respect to wet and dry days
With the wet and dry days identified, the next step
involved calculating mean fields with respect to these
dates. The special pilot balloon data are available only
during 1998 (with some limited exceptions), and thus
the initial focus has been on this year. However, the
results motivated the study of additional years, eventu-
ally leading us to generalize the results over a 16-yr
period extending from 1990 to 2005. We have focused
only on the period JanuaryApril, since this is the
height of the wet season in the coastal regions of north-
ern Peru and Ecuador, and relatively little rain falls in
the coastal parts of northern Peru during other months.
For both 1998 and the other years in this period we
have used the NCEP reanalyses and OLR data. We
averaged the 1200, 1800, 0000, and 0600 UTC NCEP
data to produce daily-averaged analyses. This yielded
an analysis centered about 2100 UTC, which was rela-
tively close to the peak time of precipitation over the
land stations. OLR data, available daily, were similarly
used to develop composite fields. GOES infrared im-
FIG. 5. (a) The wet day mean winds, averaged over the 01 km AGL layer, for all wet days.
(b) As in (a), but for dry days. (c) As in (a), but for wet day mean minus dry day mean winds.
Because all stations were within 100 m of sea level, AGL is approximately equal to above sea
level. Full wind barbs are 1 m s1, half are 0.5 m s1; numbers next to observations are the
speed in m s1. Elevations above 1000 m are shaded.
JANUARY 2009 D O U G L A S E T A L . 121
8/22/2019 Mon. Wea. Rev. 2007 Douglas
7/21
agery, available at 3-hourly intervals, have similarly
been averaged to produce daily averages, which are
24-h intervals centered on 0000 UTC.
4. Characteristics of the wet and dry days during
1998
a. Rainfall patterns on wet and dry days
In general terms the mean precipitation during the
period JanuaryApril 1998 (Fig. 3a) is similar in spatial
structure to that shown for the 1982/83 El Nio event
(see Fig. 11b in Goldberg et al. 1987). The largest rain-
fall is close to the base of the mountains on the Pacific
slope, with much smaller amounts being reported to the
east of the continental divide and smaller quantities
along the coast.
Average rainfall during wet and dry days is shown in
Figs. 3b,c. Wet days clearly have larger rainfall amounts
than dry days, though the enhancement is less evident
in Ecuador than in northern Peru. This is more easily
seen from Fig. 3d, which shows the percentage increase
of wet day rainfall relative to dry day rainfall. The
greatest enhancement of rainfall on wet days is found
around the Piura area. This result is, in part, due to
selecting the wet days based on rainfall from the Peru
stations, which are most densely concentrated in the
region around Piura. Had we used the rainfall from
Ecuadorean stations as our criteria for wet days the
anomalies would likely have been larger in Ecuador.
Only four stations showed less rainfall on the wet days,and these were by relatively small amounts. All other
sites show higher rainfall during wet days. The relation-
ship between daily mean rainfall in Ecuador and in
northern Peru over the JanuaryApril period is shown
in Fig. 4. The agreement is reasonable (correlation 0.69),
showing that coherent rainfall variations, suggested by
Fig. 3d, extend from near the equator to near 8S.
b. Pilot balloon-based wind analyses
To seek an explanation for the strong synoptic vari-
ability in the rainfall series (Figs. 2c,d) we first examine
the pilot balloon data. Using the criteria mentioned in
section 3 the wet and dry day mean profiles of the
vector wind were calculated for each pilot balloon sta-
tions data (Figs. 5a,b). Differences between the wet
and dry day mean winds were then calculated to more
clearly show the differences (Fig. 5c). The observa-
tional period and total number of observations varied
with the station and it was not possible to produce a
uniform composite. Figure 5 shows the composite based
only on observations made during the period 1 January
30 April, when the observations were most complete.
Overall, the wet and dry day means show a change in
the position of the zero-meridional wind in the 01-km
layer along the coast of Ecuador, with a southward dis-
placement during wet days. The difference between the
wet and dry day means (Fig. 5c) shows most stations
with a northerly wind anomaly during wet days, except
for those in Central America, and at Cartagena and
Iquitos. The anomaly flow at most Ecuadorean and Pe-
ruvian stations is approximately parallel to the coastline.
Because there are variations in the observational pe-
riods at the different coastal pilot balloon sites and the
heights reached by different balloons varied from day
to day, we prepared a multistation mean of the zonal
and meridional winds during wet and dry days. Three
stations with the most complete records were used:
Tumbes, Piura, and Chiclayo. The zonal winds (Fig. 6a)
show a stronger westerly component during wet days,
of about 1.5 m s1, through a layer extending to above
4 km. The mean meridional wind (Fig. 6b) is about 1
FIG. 6. Mean profiles of the (a) zonal and (b) meridional wind
during wet (solid circles) days and dry (open circles) days, based
on observations from three coastal pilot balloon stations (Piura,
Tumbes, and Chiclayo) in northern Peru and NCEP reanalysis
data (squares) for a grid point (5
S, 82.5
W) offshore northernPeru. Mean profiles are obtained by including morning (AM) and
afternoon (PM) soundings.
122 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
8/21
m s1 stronger (southerly) during dry days than on wet
days below 2 km ASL. Above 2.5 km ASL, the merid-
ional wind is more northerly during wet days. The re-
sults from the pilot balloon observations are quite simi-
lar to those reported by Takahashi (2004) using profiler
data at Piura for the same period.
Previous studies, notably Bendix (2000) and Taka-
hashi (2004) have examined the diurnal cycle relation-
ship with rain events. The twice-daily pilot balloon
data, though limited by cloudiness, do permit a depic-
tion of the diurnal variation of the wind field. Figure 7a
shows the diurnal variation during both wet and dry
days. In general, the mean zonal wind is more westerly
during wet days than during dry days. The difference
between the PM and AM profiles for the wet and dry
day means (Fig. 7b) shows that the amplitude of the
diurnal cycle of the zonal wind is approximately the same
(2.5 m s1) in the lowest 1 km but the profiles diverge
above this, with the dry day profile showing little change
in the 1.53-km layer. During wet days the PM zonal wind
is stronger than the AM zonal wind by about 1 m s1 up
to 5 km. These results agree with the Piura diurnal cycle
shown by Takahashi (2004) in that the afternoon west-erly flow is much deeper during wet days.
5. Structure of wet and dry days from
NCEPNCAR reanalyses, GOES infrared
imagery, and OLR data
a. Selecting wet and dry days from NCEPNCAR
reanalysis and OLR data alone
The generation of composite wind fields associated
with wet and dry days during 1998 is relatively straight-
FIG. 8. (a) Relationship between 1998 rainfall in northern Peru
and zonal wind based on observations from three coastal pilot
balloon stations (solid dots) averaged over the 01 km AGL layerand the zonal wind based on daily NCEPNCAR reanalysis data
(open dots) at 925 hPa for a grid point (5S, 82.5W) near the
center of the Peruvian rain gauge network. Dotted (dashed) line
is least squares fit between Peruvian rainfall and pilot balloon
observations (pibals) (NCAPNCAR reanalysis). (b) Correlo-
grams for the rainfall and zonal wind data shown in (a) and also
for OLR and 700-hPa wind anomalies with daily rainfall.
FIG. 7. (a) Mean profiles of the zonal wind for morning (circles)
and afternoon (triangles) soundings during wet (solid circles) days
and dry (open circles) days, based on observations from three
coastal pilot balloon stations (Piura, Tumbes, and Chiclayo) innorthern Peru. (b) Mean profiles of zonal wind for afternoon
minus morning soundings during wet (solid circles) days and dry
(open circles) days.
JANUARY 2009 D O U G L A S E T A L . 123
8/22/2019 Mon. Wea. Rev. 2007 Douglas
9/21
forward using the selected wet and dry days. Likewise,
the evolution of these days can be determined by
choosing days prior to and after the wettest days. How-
ever, to anticipate results that we later show related to
the generality of the 1998 results, it is not straightfor-
ward to identify wet and dry days for years other than
1998. First, during nonEl Nio years there is little rain-
fall in northern Peru, making the identification of wet
and dry days difficult. Then there is the unavailability
of sufficient Ecuadorean rainfall data for other years,
making a rainfall-based index unreliable. We are left
with several nonrainfall based indices to estimate wet
and dry days.
To generate a multiyear composite of wet and dry
events we needed a quantity that could ideally be re-
lated to rainfall along the Peruvian and Ecuadorean
coast. Perhaps the most obvious index related to rain-
fall would be OLR or some quantification of GOES
FIG. 9. (a) Timelongitude Hovmller plots of NCEP reanalysis zonal wind for 700 hPa averaged between 7.5 S and 2.5N during
JanuaryApril 1998. (b) As in (a), but for 925 hPa. (c) As in (a), but for OLR. The black squares in (a)(c) are wet days determined
from Peruvian rain gauge data, red squares in (a) are dry days. The dash marks at 85 W in (a) show subjectively selected 700-hPa
anomalous westerly wind events.
124 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
10/21
infrared imagery. However, we were initially motivated
to use zonal wind anomalies as an index, based on the
difference between wet and dry day mean zonal winds
evident from the pilot balloon data (Fig. 5). Despite the
clear difference in the mean profiles from Fig. 5, the
daily data (Fig. 8a) show considerable scatter. Compari-
son of the correlations with NCEP 925-hPa zonal winds
(roughly the same level as the pilot balloon data), 700-
hPa zonal winds, and OLR for 1998 shows that the
NCEP winds are the least correlated with rainfall (Fig.
8b), though the differences are not large.
Figure 9 shows three Hovmller diagrams for 1998.
The first two show NCEP zonal wind anomalies, while
the third displays OLR anomalies. The agreement be-
tween wet and dry days (determined from the Peruvian
rainfall data) and the zonal wind (and OLR) anomalies
is clearly imperfect. This reflects the scatter evident in
Fig. 8a. Dates of 700-hPa westerly zonal wind events,
identified subjectively, are also shown in Fig. 9. It is
clear that the wind-based Hovmller diagrams are rela-
tively insensitive to the particular level, with both 925
and 700 hPa showing similar patterns. The OLR Hov-
mller shows greater variability than those of zonal
wind, both in time, and importantly, along a given lon-
gitude. These variations make identifying synoptic-
scale variations from OLR data more difficult than
from zonal wind variations, and they appear to reflect
smaller longitudinal scales.
We evaluated different options for identifying wet
and dry days. Figure 10 shows the average wettest
day wind field at 925 hPa, based on four criteria we
used for selecting the wet and dry days. One criterion
was based on maxima and minima in OLR, two used
zonal wind anomalies at 700 hPa [one subjectively de-
termined from Hovmller examination (see the ex-
ample in Fig. 9) and one objectively determined], and
one used objectively determined zonal wind anomalies
at 925 hPa. The same domain was used for all criteria,
FIG. 10. Comparison of wet day minus dry day composite wind fields for 925 hPa, using four different criteria for
determining the wet and dry days for the months of JanuaryApril 19902005. See text for an explanation of each.
The gray shaded region indicates where the differences between westerly and easterly composites exceed the 95%significance level of the Students t test.
JANUARY 2009 D O U G L A S E T A L . 125
8/22/2019 Mon. Wea. Rev. 2007 Douglas
11/21
from 2.5N7.5S to 8090W, though the subjective
procedure used a slightly broader longitudinal band for
evaluating zonal continuity of the anomalies. The im-
portant aspect of Fig. 10 is that all composites repro-
duce certain basic features of the wind field. The north-
erly winds over the western Gulf of Mexico, the cy-
clonic circulation near the east coast of the United
States, and westerly winds over the equatorial eastern
Pacific are common to all analyses, though the intensity
and precise positions vary. Not surprisingly, the three
zonal wind-based composites tend to be more similar
and to show stronger westerly wind anomalies along the
equator than does the OLR-based composite. We
chose to use the 700-hPa zonal wind as the index for
developing our composites, with positive u-wind events
being the equivalent of wet days and negative u
anomalies being dry days. We adopt this terminology
hereafter for ease of expression. However, it should be
clearly stated that this index should be considered more
an indication of westerly wind events over the domain
FIG. 11. (a) Difference between cloud frequency for days with westerly wind anomalies and cloud frequency for
days with easterly wind anomalies for years 19982005 during the period JanuaryApril. The cloudiness was
quantified by using a temperature threshold of38C. (b) As in (a), but only for JanuaryApril 1998. (c) As in (b),
but for frequency differences between wet and dry days based on the Peruvian rainfall data. (d) As in (c), but for
the one day after wet and dry days based on the Ecuadorean rainfall data.
126 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
12/21
of interest (coastal Ecuador and northern Peru) rather
than a reliable indication of precipitation.
In summary, to determine the specific dates of
anomalous zonal wind events we restricted our com-
parison to the months of JanuaryApril (rainy season in
northern coastal Peru) for the years 19902005 and gen-
erated Hovmller diagrams of 700-hPa zonal wind
anomalies from the NCEP climatology, averaged over
the latitude band of 2.5N7.5S. These were inspected
to identify the dates of anomalous westerly wind events
and anomalous easterly wind events at 85W. A total
of 220 westerly wind events and 215 easterly wind
events were identified over the 16 yr examined, though
the selection was subjective and the events varied in
intensity and duration. An average period was 9 days
(1920 total days divided by 215 wet events), though
there was considerable variation in the period between
events. This average period is longer than that obtained
from using the 1998 rainfall data (6 days). Without
rainfall data for other years, and for more Ecuadorean
stations as well, it is not possible to determine the
source of this difference.
b. OLR- and GOES-based cloudiness for wet and
dry days
Using the 700-hPa zonal wind criterion described in
the previous section we can describe the satellite-
observed evolution of the wet spells (positive zonal
wind events) along the coast of Ecuador and Peru. To
this end we use the data discussed in section 2c.
1) WET DAY CLOUDINESS FROM GOES IMAGERY
GOES-estimated cloudiness was quantified by deter-
mining the number of times a particular pixel was
colder than 38C. This was done independently for
each 3-hourly image, which were then summed to pro-
duce daily frequency values. These values were then
averaged over all of the west wind events for the 8 yr.
Figure 11a shows that enhanced cloudiness is concen-
trated south of Costa Rica, and extends to the Colom-
bian coastline. The maximum anomaly is close to the
mean position of the ITCZ cloudiness during the boreal
winter (Wang et al. 2004). The result for 1998 is shown
in Fig. 11b. The positive cloudiness anomalies are more
FIG. 12. Hovmller diagram of the difference between OLR for days with westerly wind anomalies and OLR with easterly wind
anomalies during the period JanuaryApril (left) 19902005 and (right) 1998. The evolution is from three days prior to three days after
the westerly/easterly wind anomalies in the eastern equatorial Pacific. The latitudinal extent ranges from 10S to 10N. An eastward
propagation is evident.
JANUARY 2009 D O U G L A S E T A L . 127
8/22/2019 Mon. Wea. Rev. 2007 Douglas
13/21
widespread and are larger, especially south of the
equator. This is not surprising, since the higher seasurface temperatures south of the equator during 1998
favored more convection in this region than in other
years. And, where the 8-yr mean cloudiness shows the
western Amazon basin to have somewhat more fre-
quent cloudiness during wet days, the opposite is seen
in 1998.
Cloud frequencies were also computed using the wet
and dry days based on the Peruvian rainfall data. This
produced Fig. 11c, which while showing a maximum
over northern Peru, was considerably different from
the wind-based composite in other areas. In particular,
large differences are evident over southern Peru and
most of Bolivia, and the frequencies are lower over the
Pacific coast of Colombia and Ecuador. Using a rainfall
index based on only the Ecuadorean rain gauges pro-
duced still different results, with the greatest frequency
of cloudiness along the Ecuadorean coast apparent on
the day after the wet day (Fig. 11d). It may be that the
number of rain gauges is insufficient to provide a reli-
able indicator of wet events, or that the number of
events is too small during one year for stable means to
be obtained.
2) PROPAGATION OF THE CLOUDINESS FROM OLRDATA
Hovmller diagrams of the OLR wet day minus dry
day OLR values for both the 19902005 period and the
1998 period alone were constructed using the 700-hPa
zonal wind features described in section 5a (Fig. 12).
The most striking feature of each diagram is the east-
ward propagation of the OLR anomalies over the equa-
torial eastern Pacific. The 1998 anomalies appear some-
what more distinct than those of the multiyear mean,
but the propagation velocity is similar (10 longitude
day1 or 12 m s1). The wavelength estimated from
successive positive or successive negative OLR anoma-
lies in Fig. 12 is about 50006000 km.
A Hovmller diagram of the OLR wet dry day dif-
ferences, similar to Fig. 12 but based on a selection of
212 negative anomaly OLR days (wet days) and 200
positive anomaly days (dry days), is shown in Fig. 13.
The eastward propagation of the OLR patterns is less
obvious than in Fig. 12, but is still apparent in the 16-yr
composite. However, there is an obvious nearly station-
ary aspect, strongest around day zero, and some evi-
dence of a westward-propagating signal as well.
FIG. 13. As in Fig. 12, but based on an index using OLR instead of zonal wind.
128 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
14/21
The differences between the zonal wind-based and
OLR-based Hovmller diagrams are not easy to ex-
plain. It is possible that heavy, but localized, rainfall
such as might occur along the coastal strip of northern
Peru, might not be fully represented in the coarser reso-
lution (2.5) OLR data. OLR variations often are asso-
ciated with broad cirrus shields, and although these are
generally indicative of convective precipitation in the
tropics, such canopies can be extensive or relatively
small depending on the strength of upper-level winds
and the upper-tropospheric relative humidity. Synop-
tic-scale wind perturbations associated with tropical
waves, of the spatial scale that can be resolved by the
NCEP reanalyses, tend to have better time continuity
than their associated cloud fields, which are related to
variations in the moisture field, static stability, and
small variations in the vertical motion field. The data
assimilation procedure used to produce the NCEP re-
analyses also ensures a level of time continuity in the
wind field that the OLR data, interpolated from an
independent satellite dataset (Liebmann and Smith
1996), do not possess.
c. Vertical structure of the wet minus dry day wind
fields
Figure 14 shows the vertical structure of the wet mi-
nus dry day mean wind field for the period 19902005
using the 700-hPa zonal wind index described in section
FIG. 14. Difference between the mean wind field for
days with westerly wind anomalies and the mean wind
field for days with easterly wind anomalies during
JanuaryApril 19902005. Wind fields based on dailyNCEPNCAR reanalysis data for (a) 1000, (b) 850, (c)
700, (d) 500, and (e) 300 hPa. Westerly and easterly
wind anomalies are determined from the 700-hPa wind
field in the equatorial eastern Pacific (see text). Dif-
ferences significant at the 95% level are shaded.
JANUARY 2009 D O U G L A S E T A L . 129
8/22/2019 Mon. Wea. Rev. 2007 Douglas
15/21
5a. The most striking feature associated with wet days is
the strong westerly wind anomaly in the lower tropo-
sphere along the equator, being strongest from 100 to70W. This is not unexpected, since the compositing
procedure was based on zonal anomalies over this re-
gion. The westerly anomaly is largest at the 850- and
700-hPa level, becoming notably weaker and more
southwesterly by 500 hPa. The anomaly winds at 1000
hPa, while zonal on the equator, are strongly diffluent
at the coast, with northwesterly wind anomalies along
the Peruvian coast and southwesterly anomalies west of
Colombia and extending over Panama. The anomalies
along the Peruvian coast agree with the pilot balloon
observations (Fig. 5), however, major differences exist
over Central America.
Perhaps the most surprising result of the compositingprocedure is the cyclonic vortex off the east coast of
North America that tilts westward with height. Rela-
tively strong northerly flow is present over the Gulf of
Mexico at low levels. Taken together, the composite re-
sembles an extratropical cyclone off the central east coast
of the United States, with trailing cold frontal zone
extending toward Central America. At higher levels
(500 and 300 hPa) a wave train extends zonally across
the entire domain at 3040N, with one cyclonic and
two anticyclonic eddies evident at 300 hPa along 40N.
FIG. 15. As in Fig. 14, but for JanuaryApril 1998.
130 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
16/21
Figure 15 shows the same mean fields as Fig. 14, but
for 1998. There are many similarities with the 16-yr
mean, including 1) the westerly wind anomalies along
the equator from 1000 to 700 hPa, 2) the cyclonic vortex
off the U.S. east coast, 3) the cyclonic eddy/trough
southwest of Peru over the southeastern Pacific (most
evident at 850700 hPa), 4) the pronounced trough off
the north Chilean coast at 300 hPa, and 5) the easterly
flow at 300 hPa over the equatorial eastern Pacific near
90W. There are numerous differences, especially in
higher latitudes, but the overall impression is one of
moderate agreement between the 1998 anomaly com-
posite and the 16-yr mean composite. In fact, Takahashi
(2004), using rainfall-based wet and dry days, found
NCEP reanalysis wet minus dry day wind differences at
850 and 700 hPa that were quite similar to those shown
in Fig. 15. This suggests that the synoptic-scale condi-
tions associated with wet days in 1998, despite the
strong El Nio conditions, may be generally similar to
those associated with wet days during other years. It
FIG. 16. Evolution of the difference between the mean SLP for days with westerly wind anomalies and the mean
SLP for days with easterly wind anomalies for JanuaryApril 19902005. Differences for (a) 3, (b) 2, (c) 1,
(d) 0, (e)1, and (f) 2 days about westerly/easterly wind anomalies in the eastern equatorial Pacific. The contour
interval is 0.5 hPa. Differences significant at the 95% level are shaded.
JANUARY 2009 D O U G L A S E T A L . 131
8/22/2019 Mon. Wea. Rev. 2007 Douglas
17/21
also indicates that our zonal wind-based compositing
approach yields results similar to one based directly on
the rainfall data.
d. Evolution of the wet minus dry day anomalies
A natural extension of compositing the NCEP re-
analyses to describe the wet day conditions is to com-
posite days prior to and after the wet day. This was
done from three days prior to the wet day to three days
after the wet day. For brevity, we discuss only sea level
pressure and 850- and 500-hPa wind evolutions; the lat-
ter only in the Northern Hemisphere. Also, 1998 analy-
ses are shown only for 850 hPa, since the confidence is
greater for the 16-yr means.
1) SEA LEVEL PRESSURE EVOLUTION
The main feature in the evolution of sea level pres-
sure (Fig. 16) is the strong extratropical signal over
North America that moves eastward with time. The
major tropical feature is the broad westeast gradient in
FIG. 17. Evolution of the difference between the mean 850-hPa wind field for days with westerly wind anomalies
and the mean wind field for days with easterly wind anomalies during JanuaryApril 19902005. Evolution for (a)
3, (b) 2, (c) 1, (d) 0, (e) 1, and (f) 2 days about westerly/easterly wind anomalies in the eastern equatorial
Pacific. Differences significant at the 95% level are shaded.
132 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
18/21
pressure, with positive anomalies over the Pacific and
lower values over South America during days t 2 to
t 0.
2) 850-HPA WIND FIELD EVOLUTION
The equatorial eastern Pacific westerly wind anoma-
lies (Fig. 17) seen on day t 0 are evident on both the
day before and day after, reflecting the duration of the
event and probably also the difficulty in assigning a
precise time to these events from the NCEP analyses
(Fig. 9). The largest extratropical changes are over
North America, with southerly flow over the central
United States and the Gulf of Mexico at t 3 be-
coming northerly at t 0. The amplitudes of the
anomalies decay noticeably after t 1, and day t
3 is not shown, with only small-amplitude featuresbeing present.
The 1998 wind field anomaly evolution at 850 hPa
(Fig. 18) is qualitatively similar to the 16-yr mean, but
differs in many details. Areas of agreement are the an-
ticyclonic anomaly off the U.S. east coast on days prior
to t 0 and the cyclonic vortex to its west. However,
the position of this cyclonic vortex is somewhat east of
the position in the 16-yr mean. The equatorial westerly
wind anomalies from day t 1 to day t 1 are
broadly similar in both composites. Given the small
FIG. 18. As in Fig. 17, but only for JanuaryApril 1998.
JANUARY 2009 D O U G L A S E T A L . 133
8/22/2019 Mon. Wea. Rev. 2007 Douglas
19/21
FIG. 19. Hovmller diagram of the difference between the mean wind field at 500
hPa for days with westerly wind anomalies and the mean wind field for days with
easterly wind anomalies during JanuaryApril 19902005. Evolution is from three
days prior to three days after the westerly wind anomaly in the eastern equatorial
Pacific. The latitudinal extent ranges from 10 to 45N. Differences significant at
the 95% level are shaded.
134 M O N T H L Y W E A T H E R R E V I E W VOLUME 137
8/22/2019 Mon. Wea. Rev. 2007 Douglas
20/21
sample size for 1998 perhaps not much more can be
made of the similarities.
The clear link between extratropical waves of the
Northern Hemisphere and the zonal west wind anoma-
lies along the equator is provided by a Hovmller dia-
gram of the 500-hPa wind anomalies for the 16-yr mean
fields (Fig. 19). The wave train shown in Fig. 19 main-tains its intensity until day t 0 and thereafter decays
rapidly. It is not clear why the anomaly wind field
should show a stronger correlation with days prior to
the reference day, as opposed to days after. The west-
ward propagation of the eddies is 10 m s1 (10
longitude day1 at 35N) and the wavelength is close to
5000 km, estimating from the positions of the cyclonic
and anticyclonic circulation centers on day t 0.
6. Summary
Rain gauge measurements along the coast of north-ern Peru and the Pacific coast of Ecuador during the
strong El Nio event of 1997/98 showed large variabil-
ity on synoptic (1 week) time scales. With rain gauge
observations serving to identify relatively wet and dry
days over this region, pilot balloon observations, NCEP
reanalyses, GOES infrared imagery, and OLR data
were used to describe the characteristics of these wet
and dry days. The 1998 results motivated an additional
effort to evaluate the generality of the 1998 results to a
longer period, from 1990 to 2005. The lack of rainfall
data forced a different compositing approach, and the
procedure selected used zonal wind anomalies at 700hPa. The main results of our study are as follows:
1) Wet days along the coasts of Peru and Ecuador are
associated locally with reduced southerly boundary
layer flow and stronger than normal westerly winds
extending at least 1000 km offshore and up to 4
km ASL.
2) Wet days are associated with enhanced cloudiness
over the eastern Pacific. The 1998 El Nio event
departed from this mean pattern with more cloudi-
ness south of the equator.
3) Multiyear composites of NCEP reanalysis and OLRdata suggest that extratropical waves crossing North
America are associated with the near-equatorial en-
hanced cloudiness and positive zonal wind anoma-
lies. Both the extratropical waves and the OLR pat-
terns propagate eastward at 10 of longitude
day1.
Limitations of this study stem from the lack of mul-
tiyear daily rain gauge data from coastal Ecuador and
northern Peru that could be used to improve the deter-
mination of wet and dry days. In this regard, Ecuador-
ean stations receive rainfall every year, whereas Peru-
vian stations may not, so the detection of synoptic
variations in the rainfall data should be easier at Ecua-
dorean stations. Wind soundings from this region have
been uncommon and sporadic, but it may be possible to
develop daily indices from a mix of wind profiler, ra-
diosonde, and pilot balloon observations that havebeen made in the region during the past two decades. In
addition, it may be possible to relate surface wind vari-
ability over the ocean to coastal rainfall variations using
satellite scatterometer data.
Given the apparent relationship between the North-
ern Hemisphere synoptic-scale extratropical waves,
which may be predictable to a week or more, and wind
and rainfall variations along the Ecuadorean and north
Peruvian coasts, this could be an area of fruitful re-
search for the meteorological services of the region.
Acknowledgments. The present work was supportedby the NOAAs Office of Global Programs, during the
early stages of the PACS-SONET project. PACS-
SONET funding was provided by the NOAA/Office of
Oceanic and Atmospheric Research under NOAA
University of Oklahoma Cooperative Agreement
NA17RJ1227, U.S. Department of Commerce. The
various program managers are thanked for their sup-
port of the 1998 special measurements. This study was
initially started during a visit by one of the authors
(NO) to NSSL and while one author (JB) was partici-
pating in a Research for Undergraduates activity sup-
ported by the National Science Foundation underGrant ATM-9820587. Antonio Rodriguez of INAMHI
graciously provided the INAMHI rainfall data from Ec-
uador. Ken Knapp of NESDIS made the GOES imag-
ery available for this work. Many observers made the
observations used in this study and many others as-
sisted in different aspects of the field work. Special
thanks are due to the reviewers (Ken Takahashi and
two anonymous reviewers) of this manuscript for en-
couraging us to explore in more detail some ideas that
led to the expansion of this papers scope.
REFERENCES
Bendix, J., 2000: Precipitation dynamics in Ecuador and northern
Peru during the 1991/92 El Nio: A remote sensing perspec-
tive. Int. J. Remote Sens., 21 (3), 533548, doi:10.1080/
014311600210731.
Douglas, M. W., and W. Fernandez, 1997: Strengthening the me-
teorological sounding network over the tropical eastern Pa-
cific Ocean and the Intertropical Americas. WMO Bull., 46,
348351.
, and J. Murillo, 2008: The Pan American Climate Studies
Sounding Network (PACS-SONET). Bull. Amer. Meteor.
Soc., 89, 17091725.
JANUARY 2009 D O U G L A S E T A L . 135
8/22/2019 Mon. Wea. Rev. 2007 Douglas
21/21
, M. Pea, J. L. Santos, N. Ordinola, and L. Flores, 1999: El
Nio 19971998 as seen from PACS-SONET observations.
Preprints, Second Hayes Symp. on Seasonal-to-Interannual
Climate Variability1997/1998 ENSO Cycle, Dallas, TX,
Amer. Meteor. Soc., 4346.
Enfield, D. B., 1989: El Nio, past and present. Rev. Geophys., 27,
159187.
Goldberg, R. A., M. G. Tisnado, and R. A. Scofield, 1987: Char-acteristics of extreme rainfall events in northwestern Peru
during the 19821983 El Nio period. J. Geophys. Res., 92
(C13), 14 22514 241.
Horel, J. D., and A. G. Cornejo-Garrido, 1986: Convection along
the coast of northern Peru during 1983: Spatial and temporal
variation of clouds and rainfall. Mon. Wea. Rev., 114, 2091
2105.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-
analysis Project. Bull. Amer. Meteor. Soc., 77, 437471.
Knapp, K. R., 2004: ISCCP B1 Data at NCDC: A new climate
resource. Preprints, 13th Satellite Meteorology and Oceanog-
raphy Conf., Norfolk, VA, Amer. Meteor. Soc., P6.5. [Avail-
able online at http://ams.confex.com/ams/pdfpapers/77999.
pdf.]
Liebmann, B., and C. A. Smith, 1996: Description of a complete
(interpolated) outgoing longwave radiation dataset. Bull.
Amer. Meteor. Soc., 77, 12751277.
Pea, M., and M. W. Douglas, 2002: Characteristics of wet and dry
spells over the Pacific side of Central America during the
rainy season. Mon. Wea. Rev., 130, 30543073.
Philander, S. G., 1990: El Nio, La Nia, and the Southern Oscil-
lation. Academic Press, 286 pp.
, 2004: Our Affair with El Nio: How We Transformed an
Enchanting Peruvian Current into a Global Climate Hazard.
Princeton University Press, 288 pp.
Takahashi, K., 2004: The atmospheric circulation associated with
extreme rainfall events in Piura, Peru, during the 19971998
and 2002 El Nio events. Ann. Geophys., 22, 39173926.
Trewartha, G. T., 1962: The Earths Problem Climates. University
of Wisconsin Press, 334 pp.
Wang, C., S.-P. Xie, and J. A. Carton, 2004: A global survey of
ocean-atmosphere and climate variability. Earth Climate: The
OceanAtmosphere Interaction, Geophys. Monogr., Vol. 147,
Amer. Geophys. Union, 119.
136 M O N T H L Y W E A T H E R R E V I E W VOLUME 137