3576 JOURNAL OF CLIMATE VOLUME 13

Occurrence of Extreme Precipitation Events in California and Relationships with the

Madden–Julian Oscillation
CHARLES JONES
Institute for Computational Earth System Science (ICESS), University of California, Santa Barbara,
Santa Barbara, California

(Manuscript received 18 October 1999, in final form 1 February 2000)

ABSTRACT
California receives most of the annual precipitation during the boreal winter season. Additionally, large spatial
and temporal variations in the total rainfall amounts are observed. This study investigates the occurrence of
extreme precipitation events in California and the modulation by the Madden–Julian oscillation (MJO). Three
questions are investigated. 1) Are extreme precipitation events in California more likely to occur during active
MJO than inactive periods? 2) In what phase of the MJO life cycle are extreme events more likely? 3) Are
interannual variations in the frequency of extreme events in California related to interannual variations of the
MJO?
Daily totals derived from gridded hourly station data are used to define extreme precipitation events from
January 1958 to December 1996. Outgoing longwave radiation from polar orbiting satellites (1979–96) and
zonal component of the wind at 200 hPa and 850 hPa from the National Centers for Environmental Prediction–
National Center for Atmospheric Research reanalysis (1958–96) are used to describe the life cycle of the
oscillation and its interannual variability. The results indicate that the frequency of extreme events are more
common when tropical activity associated with the MJO is high, as opposed to periods of quiescent phases of
the oscillation. Second, a slight preference for a higher number of events is observed when convective anomalies
are located in the Indian Ocean. In this situation, low-level westerly and easterly wind anomalies are observed
over the Indian and western Pacific Oceans, respectively. The analysis of the interannual variability in the
amplitude of the MJO and the occurrence of extreme events over California indicates no direct and systematic
relationships with the number of extreme events.

1. Introduction occurrence and changes in California rainfall, although

not all the studies reach the same conclusions. While a
The climatology of the state of California shows that few studies indicate some modulation in the California
the bulk of the annual precipitation occurs during the rainfall by ENSO, other investigations find no clear as-
boreal winter season (Bryson and Hare 1974; Raphael sociation (Schonher and Nicholson 1989; Mitchell and
and Mills 1996; Higgins et al. 1999). Additionally, large Blier 1997). It appears that since ENSO episodes have
spatial variations in the total rainfall amounts are ob- varying characteristics in spatial structure, intensity, and
served; for example, in a typical winter season northern duration, different ENSO events may impact precipi-
California can receive 2 to 4 times more precipitation tation over the state of California differently (Null
than the southern part of the state. However and most 1993).
importantly, significant temporal variations are also ob- This study is concerned with the influence of tropical
served in rainfall extending from synoptic to intrasea- intraseasonal variations on precipitation in California.
sonal, interannual, decadal, and longer timescales. The Madden–Julian oscillation (MJO) is the main mode
On interannual timescales, the El Niño–Southern Os- of large-scale tropical intraseasonal variability with an
cillation (ENSO) is the strongest signal of large-scale important role in the climate system (Madden and Julian
variations in the ocean–atmosphere system (Horel and 1994; Meehl et al. 1996; Jones et al. 1998, 1999). On
Wallace 1981; Philander 1990). Several studies have this timescale, intraseasonal variations seem to be ex-
investigated a possible relationship between El Niño tremely important to wintertime rainfall in California as
well (Mo and Higgins 1998a). Higgins and Mo (1997)
developed a composite study to identify possible rela-
tionships between Persistent North Pacific (PNP) cir-
Corresponding author address: Dr. Charles Jones, Institute for
Computational Earth System Science (ICESS), University of Cali-
culation anomalies and tropical intraseasonal variations.
fornia, Santa Barbara, CA 93106-3060. They found that remote forcing from the Tropics seems
E-mail: cjones@icess.ucsb.edu to play a significant role in the development of PNP

q 2000 American Meteorological Society

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

15 OCTOBER 2000 JONES 3577

FIG. 2. Frequency response of a Lanczos bandpass filter with 49

weights and cutoff frequency responses of 20 and 90 days 21 .

between atmospheric circulation anomalies and hydro-

logical processes associated with California. Their re-
sults show additional evidence that the MJO modulates
California precipitation. Wet (dry) events are favored
during the phase of the oscillation associated with en-
hanced convection near 1508E (1208E) in the tropical
Pacific.
Previous studies have shown linkages between intra-
seasonal tropical–extratropical interactions and weather
variations in the western coast of North America. This
FIG. 1. Selection of the six grid boxes in the state of California has important implications for prediction purposes,
from which precipitation extremes are investigated. The dimensions
of the boxes are 28 lat. 3 2.58 long.
since very wet or very dry seasons, have profound im-
pacts with attendant serious social and economic con-
sequences. For instance, a high frequency of extreme
prior to onset. Tropical heat sources generate subtropical precipitation events can lead to saturated soil moisture
Rossby wave vorticity anomalies that are partially re- conditions and increased runoff that oftentimes is ac-
sponsible for the retraction (extension) of the Pacific jet companied by severe floods. Motivated by these rea-
stream exit region and the formation of PNP anomalies. sons, this study investigates three questions. First, are
Their results are also consistent with the study of Hig- extreme precipitation events in California more likely
gins and Schubert (1996) who analyzed simulations to occur during active MJO than inactive periods? Sec-
from a general circulation model and found that tropical ond, in what phase of the MJO life cycle are extreme
intraseasonal heat variations are phase locked with the events more likely? Third, the MJO is also known to
development of PNP anomalies. In a more recent study, exhibit large interannual amplitudes (Weickmann 1991).
Mo and Higgins (1998a,b) further investigated linkages Therefore, are interannual variations in the frequency

TABLE 1. Annual average precipitation (mm) (1 Jan 1958–31 Dec 1996) and statistics of extreme events in precipitation over California.
Values are computed over the six grid boxes. The thresholds T 5 , T10 , and T15 are 5%, 10%, and 15% values of the annual average precipitation.
These thresholds correspond approximately to 90th, 95th, and 97th percentiles, respectively. The total number of 5-day precipitation totals
exceeding the T 5 , T10 , and T15 thresholds are also indicated. These values are computed for the hydrological calendar defined as 20–24 Jul
from one year to 15–19 Jul of the succeeding year.

Box 1 Box 2 Box 3 Box 4 Box 5 Box 6

Annual average precipitation (mm) 379.26 397.25 386.05 436.00 795.37 666.13
5% Threshold (T5)(mm) 18.96 19.86 19.30 21.80 39.77 33.31
10% Threshold (T10)(mm) 37.93 39.73 38.61 43.60 79.54 66.61
15% Threshold (T15)(mm) 56.89 59.59 57.91 65.40 119.30 99.92
Number of type-I events 230 228 245 266 258 276
Number of type-II events 113 113 85 82 66 71
Number of type-III events 62 56 41 29 24 17

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

3578 JOURNAL OF CLIMATE VOLUME 13

FIG. 3. Lag composites of OLR anomalies in the 20–90-day band. Composites are computed
for periods when the amplitude of the OLR index (see text for details) is negative and less than
minus one standard deviation. Lags are from 24 to 14 pentads. The lag composites are denoted
as active MJO cases. The sample size for each composite is 181 pentads in the period 1979–96.
Dark (light) shadings indicate OLR anomalies less (greater) than 22 W m22 (12 W m22 ). Contour
interval is 2 W m22 .

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

15 OCTOBER 2000 JONES 3579

FIG. 4. Composite of OLR anomalies in the 20–90-day band. Composite is computed for periods
when the amplitude of the OLR index (see text for details) is negative but greater than 20.3
standard deviation. This case is denoted as inactive MJO cases. The sample size for each composite
is 166 pentads in the period 1979–96. Dark (light) shadings indicate OLR anomalies less (greater)
than 22 W m22 (12 W m22 ). Contour interval is 2 W m22 .

of extreme events in California related to interannual December 1996. Additional information on changes in
variations of the MJO? instrumentation, equator-crossing times, and inherent
The paper is organized as follows. We first describe biases in the OLR data can be found in Chelliah and
the datasets used in this study (section 2) and then dis- Arkin (1992) and Waliser and Zhou (1997). Since the
cuss the definition of extreme precipitation events used record of OLR only extends to 18 years, the MJO is
throughout this work (section 3). We next address the also described with variations in the zonal component
first two questions discussed above, that is, whether or of the wind at 200 hPa (U200) and 850 hPa (U850).
not extreme events are more likely when the MJO is The NCEP–National Center for Atmospheric Research
active (section 4). The relationships between interannual (NCAR) reanalysis (Kalnay et al. 1996) were used to
variations in extreme events in California and the MJO define pentads of U200 and U850 for the 39-yr record
are discussed in section 5. Our discussion and conclu- from 1–5 January 1958 to 27–31 December 1996 (total
sions are presented in section 6. of 2847 pentads).

2. Data 3. Extreme precipitation events in California

The precipitation used in this study is the daily totals In order to investigate possible linkages between ex-
derived from gridded hourly station data, which were treme precipitation events and the activity of the MJO,
kindly provided by Higgins et al. (1996). The daily we have selected six grid boxes over the state of Cal-
totals are displayed on 28 latitude by 2.58 longitude grid ifornia (Fig. 1). This selection focuses in the central and
covering the continental United States. The data record southern parts of the state. Although more grid boxes
analyzed here extends from 1 January 1958 to 31 De- could have been chosen, we limit our analysis to this
cember 1996. Only 54 days out of 14 246 days are miss- selection since each grid box is analyzed separately and
ing. Based on the daily precipitation, 5-day nonover- therefore this makes the display and discussion of the
lapping totals were constructed for the period above. results more manageable. In addition, we are particu-
This resulted in time series with 2847 data points, of larly interested in extreme events along the coastal re-
which eight 5-day totals were missing. The 5-day totals gions in the southern part of the state, since severe floods
are used because this study focuses on extreme events often accompany these extreme events.
that might be related to low-frequency tropical intra- Different definitions of what characterizes an extreme
seasonal variations. event in precipitation have been proposed in the liter-
The life cycle of the MJO is characterized in this ature (Dole and Gordon 1983; Mo and Higgins 1998a;
study in terms of variations in tropical convection and Cayan et al. 1999). Extreme events in this study are
zonal component of the wind, specifically those that defined as follows. For each of the six grid boxes, we
occur at the intraseasonal timescale. To describe vari- computed the annual mean precipitation from 1958 to
ations in tropical convection, we use outgoing longwave 1996. An extreme event of type I occurs if the 5-day
radiation (OLR) data, which has been frequently used total precipitation amount exceeds 5% of the annual
as a proxy for large-scale tropical convective activity mean. Similarly, we define extreme events of type II
(e.g., Lau and Chan 1986; Waliser et al. 1993; Jones et and type III if the 5-day total precipitation exceeds 10%
al. 1998). Pentads of OLR (5-day nonoverlapping means and 15% of the annual mean, respectively. The extreme
with 73 pentads per year; total of 1314 pentads) were events defined as above correspond approximately to
obtained from the National Centers for Environmental the 90th, 95th, and 97.5th percentiles, respectively. Ta-
Prediction (NCEP) from 1–5 January 1979 to 27–31 ble 1 summarizes the annual mean precipitation (mm),

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

3580 JOURNAL OF CLIMATE VOLUME 13

severe) also allows assessing the sensitivity of the sta-

tistical results discussed next.

4. Extreme events in California and the

Madden–Julian oscillation
The relationships between extreme events in precip-
itation and the MJO are now examined in detail. To
obtain intraseasonal anomalies and characterize the
MJO, a Lanczos bandpass filter with 49 weights and
cutoff frequency responses of 0.5 at 20 and 90 days21
was applied to the OLR, U200 and U850 pentads (Du-
chon 1979; Jones et al. 1998). The frequency response
of this filter is shown in Fig. 2. This bandwidth is chosen
to specifically resolve tropical intraseasonal variations
associated with the MJO. In the remainder of this sec-
tion, we investigate the hypothesis that extreme events
in California are more frequent during active phases of
the MJO.

a. Occurrences related to outgoing long wave

radiation anomalies
The life cycle of the MJO is first examined with re-
spect to variations in the OLR field. We have computed
an empirical orthogonal function (EOF) analysis of the
OLR anomalies (20–90 days) for the period 1–5 May
1979 through 3–7 September 1996 (total of 1266 pen-
tads). Several previous studies have shown that the first
two eigenvectors are separated from the remaining ones
and together they describe the propagating behavior of
the oscillation (Ferranti et al. 1990; Jones and Weare
1996). Based on the first two time coefficients (or prin-
cipal components) derived from the EOF analysis of
OLR anomalies (OLRA), we constructed an index based
on the linear combination given by
OLRA(t)
5 PC1(t) 1 [PC2(t 1 2) 1 PC2(t 1 3)]/2. (1)
FIG. 5. Total number of extreme events in each of the six grid The linear combination involves a time lag for PC2,
boxes shown in Fig. 1. The number of occurrences is counted for since PC1 leads PC2 by 2 to 3 pentads. Similar index
each lag of the OLR composites illustrated in Fig. 3. Horizontal axes
indicate the time lag in pentads. The number of occurrences is counted using anomalies of U at 850 hPa has been proposed by
for extreme events of (a) type I, (b) type II, and (c) type III. Maloney and Hartmann (1998) in their composite study
of the life cycle of the MJO. They further demonstrated
that the index for U850 effectively captures the large-
scale structure of the MJO. The results are insensitive
5%, 10%, and 15% thresholds and the total number of whether one uses PC1 and PC2 individually or with the
extreme events in each grid box for the entire data re- linear combination above.
cord. Note that the annual mean precipitation increases In order to test the hypothesis that extreme events are
from Box 1 to Box 6. Type-I extreme events are com- more frequent during active phases of the MJO, we
mon in all six grid boxes and range from 230 to 276 in further selected periods according to the amplitude of
the period 1958–96. In contrast, extreme events of type the OLRA index (1). Two cases were selected. First,
II and type III are more frequent in southern California situations of strong convective activity anomalies re-
than in the central region. This is somewhat expected, lated to the MJO were chosen when the amplitude of
since the wet and dry seasons are much more accen- the OLRA index is negative and less than minus one
tuated in the southern part of the state. The definition standard deviation. We refer to this case simply as the
of three types of extreme events (weak, moderate, and active MJO period. A second case was selected, referred

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

15 OCTOBER 2000 JONES 3581

TABLE 2. Distribution of the number of 5%, 10%, and 15% extreme events for cases of strong convective anomalies associated with the
MJO (OLRA index at lag 22), (sample size n1 5 181 pentads) and no organized convection due to the MJO (sample size n 2 5 166 pentads).
The value of the Z statistic for each case and box is also indicated. The Z values greater than Z 95 5 1.64 indicate that the difference between
the two proportions is significant at 95% confidence level.

Type-I extreme events (5%) Type-II extreme events (10%) Type-III extreme events (15%)
Tropical MJO No tropical MJO Tropical MJO No tropical MJO Tropical MJO No tropical MJO
convection (lag 22) convection convection (lag 22) convection convection (lag 22) convection
Box 1 32 17 16 10 9 5
Z 5 1.988 Z 5 1.033 Z 5 0.980
Box 2 30 16 18 12 8 7
Z 5 1.913 Z 5 0.927 Z 5 0.098
Box 3 29 16 13 9 5 4
Z 5 1.780 Z 5 0.702 Z 5 0.220
Box 4 30 20 12 9 5 5
Z 5 1.197 Z 5 0.493 Z 5 20.148
Box 5 32 22 10 5 1 3
Z 5 1.127 Z 5 1.214 Z 5 21.173
Box 6 30 22 9 9 3 5
Z 5 0.862 Z 5 20.198 Z 5 20.896

to as the inactive MJO, so that the amplitude of the of extreme events during the inactive MJO periods (Fig.
OLRA index is negative but greater than 20.3 standard 4) was also counted for each of the six grid boxes. The
deviation. A total of 181 and 166 pentads were selected discussion is presented by contrasting the occurrences
for the active and inactive MJO cases, respectively. The of types-I, -II, and -III events during situations of en-
lag composite of OLR anomalies for the active MJO hanced MJO convection at lag 22 pentads with the
cases, Fig. 3, shows the typical eastward-propagating frequency of occurrence of inactive MJO convection.
behavior of the MJO in the convection field. In contrast, Table 2 summarizes the total counts. It is interesting to
the average of all inactive cases, Fig. 4, shows, as ex- note that for types I and II (5% and 10%, respectively)
pected, very weak OLR anomalies in the tropical belt. there is usually a higher frequency of occurrence during
Thus, these choices of thresholds for active and inactive enhanced MJO convection. In contrast, the same is not
MJO ensures that the sample sizes are approximately observed for the severe type-III events.
equal as well as characterize very different situations In order to determine whether or not such differences
of tropical intraseasonal variability. Note that selecting are statistically significant, we performed the following
positive amplitudes of the OLRA index simply change test. Extreme events of types I, II, and III are considered
the phases of the lag composites. binomial variables in the sense that any given 5-day
The next step consisted in counting the total number period either exceeds the 5%, 10%, or 15% thresholds
of extreme events of types I, II, and III during active or not. Next, we computed the test statistic Z for each
MJO cases for each grid box and lag (Fig. 5). For type type of extreme event and grid box defined as
I (Fig. 5a), an average of about 23 extreme events occurs
in all grid boxes and lags. There is, however, a slight P1 2 P2
Z5 . (2)
tendency for more extreme events at lag 22 pentads ÏPQ(1/N1 1 1/N2 )
(about 30 events), which corresponds to the situation
when convective anomalies are intense over the Indian In this expression, P1 and P 2 are the estimated prob-
Ocean (cf. Fig. 3). Likewise, there are less extreme abilities of occurrence of extreme events for the active
events at lag 12 pentads when convective anomalies and inactive MJO situations, respectively. The standard
are located between 1208 and 1508E. The number of error of the difference between the two proportions is
type-II extreme events for each lag and box (Fig. 5b) given by the denominator. Here, P and Q are the com-
also shows some preference for more occurrences at mon probabilities of occurrence and nonoccurrence of
negative lags. An average of eight type-II events for all extreme events, and N1 (181 pentads) and N 2 (166 pen-
lags and boxes are observed, although the variability tads) are the sample sizes of the active and inactive MJO
among boxes and lags is quite large. As the threshold periods. The null hypothesis is H 0 : P1 # P 2 and the
increases, type-III events (Fig. 5c), the number of ex- alternative is H1 : P1 . P 2 . The null hypothesis is re-
treme cases decreases. There is still, however, some jected if the sample proportion P1 is much greater than
preference for more events at lags 22 to 11 pentads, the sample proportion P 2 (see Anderson and Finn 1996,
especially in southern California (grid boxes 1 and 2). pp. 477–481). The value of the Z statistic is also shown
The number of type-III extreme events is quite low in in Table 2 for each type of extreme event and grid box.
grid boxes 5 and 6. Values of Z greater than 1.64 indicate that the differ-
We compare now the occurrence of extreme events ences between the proportions are statistically signifi-
during active MJO with inactive periods. The number cant at 95% level. Note, however, that this statistical

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

3582 JOURNAL OF CLIMATE VOLUME 13

FIG. 6. Lag composites of 20–90-day anomalies in the zonal component of the wind at 850 hPa
(U850). Composites are computed for periods when the amplitude of the U850 index (see text for
details) is negative and less than minus one standard deviation. Lags are from 24 to 14 pentads.
The lag composites are denoted as active MJO cases. The sample size for each composite is 439
pentads in the period 1958–96. Dark (light) shadings indicate U850 anomalies less (greater) than
20.5 m s21 (10.5 m s21 ). Contour interval is 0.5 m s21 .

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

15 OCTOBER 2000 JONES 3583

test does not account for possible serial correlation be-

tween pentads, that is, the occurrence (or nonoccur-
rence) of event in a given pentad may not necessarily
be independent of occurrence (or nonocurrence) in ad-
jacent pentads. Although there are always more type-I
and -II events during the active MJO phases, the dif-
ferences are statistically significant only for grid boxes
1, 2, and 3 and type-I event. However, as it will be
discussed next, the difference between active and in-
active MJO are considerably more significant when a
dynamical field representative of the MJO cycle is used.

b. Occurrences related to zonal wind anomalies

The life cycle of the MJO is also represented by var-
iations in the zonal component of the wind at 850 hPa.
This has the advantage of the much longer data record
based on the NCEP–NCAR reanalysis (1958–96). The
approach follows the one described by Maloney and
Hartmann (1998). An EOF analysis of U850 anomalies
(20–90 days) in the equatorial belt 108S–108N was per-
formed and the time coefficients (PC1 and PC2) asso-
ciated with the first two leading EOFs were linearly
combined to produce an index given by
U850(t)
5 PC1(t) 1 [PC2(t 1 2) 1 PC2(t 1 3)]/2. (3)
Active MJO periods were selected when the ampli-
tude of the U850 index is negative and less than minus
one standard deviation. This resulted in a sample size
of 439 pentads. Periods of inactive MJO were selected
when the U850 index is negative but greater than 20.3
standard deviation resulting in a sample size of 356
pentads. The next step consisted of computing lag com-
posites of U850 anomalies for the active MJO periods.
Consistent with the OLR variations, Fig. 6 shows the
dipole of easterly (dark shading) and westerly (light
shading) wind anomalies propagating eastward. Even
though the data records of OLR and U850 are not the
same, the composites of OLR and U850 are strongly
consistent. For example, when negative OLR anomalies FIG. 7. Total number of extreme events in each of the six grid
boxes shown in Fig. 1. The number of occurrences is counted for
(enhanced convection) are located over the Indian each lag of the U850 composites illustrated in Fig. 6. Horizontal axes
Ocean, the U850 composite indicates westerly and east- indicate the time lag in pentads. The number of occurrences is counted
erly wind anomalies near the enhanced convection. As for extreme events of (a) type I, (b) type II, and (c) type III.
in the previous case, the composite of the inactive MJO
periods simply indicates weak anomalies in the entire
equatorial belt (not shown). seems to happen for lag 11 pentad. As the threshold
Next, we counted the number of extreme events for increases, Type III (Fig. 7c), the average number of
each grid box and time lag of the composites of U830. cases is about eight events with even higher differences
Figure 7 summarizes the total counts for each type of between the southern and central parts of the state.
extreme event and grid box. The average number is The total number of counts for each grid box and type
about 56 events for type-I extreme events (Fig. 7a for of event during active and inactive MJO periods in the
all grid boxes and lags. As before, a slight preference U850 field is summarized in Table 3. The comparison
for more occurrences is observed for lag 22 pentads. is made with active MJO cases at lag 22 pentads, since
The average number of occurrences for type-II cases is there is a slight preference at that phase of the MJO life
about 20, with higher variability between grid boxes cycle. The statistical significance, as measured by the
and lags (Fig. 7b). The minimum number of occurrences Z statistic, is also shown. Readily apparent is the much

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

3584 JOURNAL OF CLIMATE VOLUME 13

TABLE 3. Distribution of the number of 5%, 10%, and 15% extreme events for cases of U850 anomalies associated with the MJO (U850
index at lag 22), (sample size n1 5 439 pentads) and no U850 anomalies due to the MJO (sample size n 2 5 356 pentads). The value of
the Z statistic for each case and box is also indicated. The Z values greater than Z 95 5 1.64 indicate that the difference between the two
proportions is significant at 95% confidence level.

Type-I extreme events (5%) Type-II extreme events (10%) Type-III extreme events (15%)
Tropical U850 No tropical Tropical U850 No tropical Tropical U850 No tropical
(lag 22) U850 (lag 22) U850 (lag 22) U850
Box 1 58 22 32 9 13 6
Z 5 3.277 Z 5 3.099 Z 5 1.220
Box 2 54 19 35 11 13 5
Z 5 3.398 Z 5 3.001 Z 5 1.530
Box 3 60 22 25 9 10 5
Z 5 3.447 Z 5 2.264 Z 5 0.940
Box 4 66 24 23 8 8 2
Z 5 3.644 Z 5 2.240 Z 5 1.661
Box 5 65 23 18 6 5 1
Z 5 3.709 Z 5 2.055 Z 5 1.460
Box 6 67 26 22 5 5 2
Z 5 3.440 Z 5 2.894 Z 5 0.909

higher number of extreme events during active MJO

periods than in inactive ones. The differences in oc-
currences are statistically significant at 95% level (Z .
1.64) for type-I and -II extreme events in all six grid
boxes. An important issue to consider in this analysis,
however, is how many of the extreme cases occurring
during active MJO periods were also coincident with
the warm phase of ENSO. We counted the number of
extreme events occurring from November to March for
the El Niño years of 1957/58, 1965/66, 1968/69, 1972/
73, 1982/83, 1986/87, and 1991/92. Only a small frac-
tion of the events shown in Table 3 occurred also during
warm ENSO phases. For type-I events the counts were
11, 10, 11, 13, 12, and 14 cases for grid boxes 1–6,
whereas for type-II events the counts were 4, 5, 6, 5,
4, and 5. In addition, only one event was reported during
the 1965/66 and no extreme events occurred in the 1986/
87 El Niño. These results further substantiate the hy-
pothesis that extreme events in California are not only
modulated by ENSO but also by tropical intraseasonal
activity such as the MJO.
To better illustrate the differences between active and
inactive MJO cases and summarize the main results of
Table 3, Fig. 8 shows the percentage of occurrences of
extreme events of types I and II (number of occurrences
divided by the sample size) during active MJO (lag 22
pentads) and inactive periods. While the average per-
centage of type I for all grid boxes during active MJO
periods is 14%, the average for inactive cases is only
6.3%. Likewise, for type II, the average percentages are
5.8% and 2.2% for active and inactive MJO periods,
respectively.

5. Interannual variability of the Madden–Julian

FIG. 8. (a) Percentage of type-I (5%) extreme events that occur oscillation and extreme events in California
during active and inactive MJO periods. (b) Same as in (a), but for
type-II (10%) events. Sample sizes are 439 pentads for active MJO Since the MJO is known to exhibit significant inter-
and 356 pentads inactive MJO. Percentages are expressed as occur- annual variations, we now investigate the question of
rences from the sample size. whether or not variations in the amplitude of the oscil-

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

15 OCTOBER 2000 JONES 3585

FIG. 9. Interannual variability of the MJO based on U200 index (solid line in units of m 2 s22 ,
see text for explanation). Also shown is the number of extreme events of Type II averaged over
the six grid boxes in California (vertical bars with square and number of occurrences on top of
each bar). The number of occurrences was counted from 20–24 Jul in one year to 15–19 Jul of
the succeeding year and plotted on 20–24 Jul of each year. Time extends from 1958 to 1996.

lation can be related to high frequency of extreme events of the U200 index, especially during 1980/81 and the
on a yearly basis. In this analysis, we consider two late 1980s and early 1990s. Also displayed in the figure
indexes to describe interannual variations in the MJO. is the number of occurrences of type-II extreme events
The first index is similar to the one used by Slingo et averaged over the six grid boxes in California. The num-
al. (1999) in their investigation of the long-term be- ber of occurrences is plotted on 20–24 July of each year
havior of the MJO, SST, and ENSO. The index uses the in the period 1958–96. A higher than average number
zonal component of the wind at 200 hPa (U200), since of occurrences are seen during the El Niño years of
the activity of the MJO strongly influences the atmo- 1968/69 and 1982/83. Interestingly, very few events oc-
spheric angular momentum (Madden and Julian 1994). curred during the 1986/87 El Niño. In contrast, the high
The bandpassed anomalies (20–90 days) of U200 were number of cases during 1977/78 is unrelated to the warm
first averaged from 108S to 108N. We next computed phase of ENSO. We have counted the number of extreme
the zonal mean of the equatorial U200 anomalies and events when the amplitude of the U200 index is above
raised that quantity to the second power. Then a running or below one standard deviation, but no large difference
mean of 21 pentads (equivalent to 105 days) was applied is observed between the two situations. In fact, the com-
to that time series resulting in the index that describes parison between the interannual amplitudes of the MJO
the interannual variability of the MJO. and the occurrence of extreme events in California does
In order to compare the interannual activity of the not indicate any obvious relationship. For instance, the
MJO and variations in precipitation over California, the high amplitudes during 1981/82 and 1988/89 are not
total number of each type of extreme event (I, II, and related to high numbers of extreme events.
III) was counted for each year from 1958 to 1996 and Since the U200 index described above may not nec-
for each gridbox. Because most of the precipitation in essarily capture variations that are entirely related to the
central and southern California occurs during the winter MJO (Hendon et al. 1999), a second index was con-
season, the number of occurrences was counted from structed based on the zonal components of the wind at
20–24 July in one year to 15–19 July of the succeeding 850 hPa (U850). The procedure is identical as before.
year. Anomalies (20–90 days) of U850 were first averaged
Figure 9 shows the U200 index from 1958 to 1996. from 108S to 108N. We next computed the zonal mean
Despite differences in methodology, a close correspon- of the equatorial U850 anomalies and raised it to the
dence is observed between our index and the one dis- second power. A running mean of 21 pentads was ap-
cussed by Slingo et al. (1999, see their Fig. 7). It is plied to that time series resulting in the U850 index.
interesting to note the large variations in the amplitude Figure 10 shows the U850 index from 1958 to 1996.

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

3586 JOURNAL OF CLIMATE VOLUME 13

FIG. 10. As in Fig. 9, but based on the U850 index to describe the interannual variability of the
MJO (solid line in units of m 2 s22 , see text for explanation).

Also displayed is the number of occurrences of type-II b), who showed that the MJO favors wet conditions
extreme events averaged over the six grid boxes in Cal- when enhanced convection is near 1508E. The extreme
ifornia (plotted again on 20–24 July of each year). As events seem to happen more frequent, however, when
with the U200 index, these results lead us to infer that convective anomalies decay to the east of the date line.
high activity of the MJO on interannual timescales is At this phase of the oscillation, suppressed convection
not always connected with high number of extreme establishes over the western Pacific and new enhanced
events in California. convection forms again over the Indian Ocean. The third
aspect addressed in this study, investigated the inter-
annual variability in the amplitude of the MJO and the
6. Summary and conclusions
occurrence of extreme events over California. Appar-
This study investigated the important issue of extreme ently, no direct and systematic relationship between the
events in precipitation over the state of California. These two is found. This leads us to the important issue of
events, which occur primarily during the Northern how exactly the MJO interacts with ENSO. During the
Hemisphere winter season, can bring serious economic peak of the warm ENSO phase, when convective activ-
losses with great social damages. While most studies ity shifts over the central and eastern Pacific, eastward
have focused on relationships between rainfall in Cal- propagation related to the MJO is not always clear. Dur-
ifornia and the ENSO phenomenon (Cayan et al. 1999), ing these periods, the number of extreme events in Cal-
this study showed that systematic and significant rela- ifornia seems to be related to seasonal changes due to
tionships exist with tropical intraseasonal activity. The ENSO rather than intraseasonal variability associated
Madden–Julian oscillation, the most distinct mode of with the MJO.
tropical intraseasonal variability, seems to modulate the The frequency of extreme precipitation events in Cal-
frequency of occurrence of these events as well. Three ifornia is modulated by large-scale circulation patterns
aspects were investigated. First, the statistical results from intraseasonal to interannual timescales. The MJO
indicate that the frequency of extreme events are more and ENSO phenomenon are the two leading modes of
common when tropical activity associated with the MJO tropical variations modulating these occurrences. Evi-
is high, as opposed to periods of quiescent phases of dently, the occurrence of these extreme events is also
the oscillation. Second, a slight preference for a higher largely dictated by baroclinic waves in the midlatitudes
number of events is observed when convective anom- of the North Pacific that are unrelated to either MJO or
alies are located in the Indian Ocean. In this situation, ENSO. Ultimately, what determines high rainfall in any
westerly and easterly wind anomalies are observed over particular location is a combination of factors involving
the Indian and western Pacific Oceans, respectively. the large-scale, mesoscale, and local features. Recent
This result is consistent with Mo and Higgins (1998a, examples of extreme precipitation to impact the state of

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC

15 OCTOBER 2000 JONES 3587

California have been discussed by Masutani and Leet- Centers for Environmental Prediction, Climate Prediction Center
Atlas 1, 47 pp.
maa (1999). The months of January and March of 1995 , Y. Chen, and A. V. Douglas, 1999: Interannual variability of
were characterized by active MJO variability as well as the North American warm season precipitation regime. J. Cli-
high frequency of intense storms that imposed above- mate, 12, 653–680.
average precipitation over most of California. These Horel, J. H., and J. M. Wallace, 1981: Planetary scale atmospheric
events also caused extensive property damage and loss phenomenon associated with the Southern Oscillation. Mon.
Wea. Rev., 109, 813–829.
of life. Estimates showed over $3 billion in damages Jones, C., and B. C. Weare, 1996: The role of low-level moisture
and 27 lives claimed by the flooding. convergence and ocean latent heat fluxes in the Madden and
Since the MJO has a slow evolution relative to syn- Julian Oscillation: An observational analysis using ISCCP data
optic weather systems, prediction of the oscillation has and ECMWF analyses. J. Climate, 9, 3086–3104.
, D. E. Waliser, and C. Gautier, 1998: The influence of the Mad-
the potential to significantly improve extended-range den–Julian oscillation on ocean surface heat fluxes and sea sur-
weather forecast in midlatitudes. Until the prediction of face temperature. J. Climate, 11, 1057–1072.
the oscillation with numerical weather prediction mod- , , J. E. Schemm, and W. K. Lau, 2000: Prediction skill of
els is improved (Jones et al. 2000), forecasts provided the Madden and Julian Oscillation in dynamical extended range
by statistical models (Waliser et al. 1999) can provide weather forecasts. Climate Dyn., 16, 273–289.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Re-
valuable information for local forecasters. analysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Lau, K. M., and P. H. Chan, 1986: Aspects of the 40–50 day oscil-
Acknowledgments. This study was funded by a re- lation during the northern summer as inferred from outgoing
search grant (Contract W-905) from the Water Resources longwave radiation. Mon. Wea. Rev., 114, 1354–1367.
Madden, R. A., and P. R. Julian, 1994: Observations of the 40–50-
Center, Centers for Water and Wildland Resources, Uni- day tropical oscillation—A review. Mon. Wea. Rev., 122, 814–
versity of California, Riverside. Data support from the 837.
National Center for Atmospheric Research (NCAR, Maloney, E. D., and D. L. Hartmann, 1998: Frictional moisture con-
which is sponsored by the National Science Foundation) vergence in a composite life cycle of the Madden–Julian oscil-
and the National Centers for Environmental Prediction lation. J. Climate, 11, 2387–2403.
Masutani, M., and A. Leetmaa, 1999: Dynamical mechanisms of the
(NCEP) is greatly appreciated. Dr. Wayne Higgins kind- 1995 California floods. J. Climate, 12, 3220–3236.
ly supplied the gridded precipitation data and provided Meehl, G. A., G. N. Kiladis, K. M. Weickmann, M. Wheeler, D. S.
valuable discussions. I am also thankful for the en- Gutzler, and G. P. Compo, 1996: Modulation of equatorial sub-
couragement and support from Professor Catherine Gau- seasonal convective episodes by tropical–extratropical interac-
tion in the Indian and Pacific Ocean regions. J. Geophys. Res.,
tier. 101, 15 033–15 049.
Mitchell, T. P., and W. Blier, 1997: The variability of wintertime
precipitation in the region of California. J. Climate, 10, 2261–
REFERENCES 2276.
Mo, K. C., and R. W. Higgins, 1998a: Tropical influences on Cali-
Anderson, T. W., and J. D. Finn, 1996: The New Statistical Analysis fornia precipitation. J. Climate, 11, 412–430.
of Data. Springer–Verlag, 712 pp. , and , 1998b: Tropical convection and precipitation regimes
Bryson, R. A., and F. K. Hare, 1974: Climates of North America. in the western United States. J. Climate, 11, 2404–2423.
Vol. 11, World Survey of Climatology, Elsevier, 420 pp. Null, J., 1993: Relationships between Type 1 ENSO events and Cal-
Cayan, D. R., K. T. Redmond, and L. G. Riddle, 1999: ENSO and ifornia rainfall, 1949–1991. Preprints, Eighth Conf. on Applied
hydrological extremes in the western United States. J. Climate, Climatology, Anaheim, CA, Amer. Meteor. Soc., 82–88.
12, 2881–2893. Philander, S. G., 1990: El Niño, La Niña and the Southern Oscillation.
Chelliah, M., and P. Arkin, 1992: Large-scale interannual variability Academic Press, 289 pp.
of monthly outgoing longwave radiation anomalies over the Raphael, M., and G. Mills, 1996: The role of mid-latitude cyclones
global tropics. J. Climate, 5, 371–389. in the winter precipitation of California. Prof. Geogr., 48, 251–
Dole, R. M, and N. D. Gordon, 1983: Persistent anomalies of the 262.
extratropical Northern Hemisphere wintertime circulation: Geo- Schonher, T., and S. E. Nicholson, 1989: The relationships between
graphical distribution and regional persistent characteristics. California rainfall and ENSO events. J. Climate, 2, 1258–1269.
Mon. Wea. Rev., 111, 1567–1586. Slingo, J. M., D. P. Rowell, K. R. Sperber, and F. Nortley, 1999: On
Duchon, C. E., 1979: Lanczos filter in one and two dimensions. J. the predictability of the interannual behaviour of the Madden–
Appl. Meteor., 18, 1016–1022. Julian Oscillation and its relationship with El Niño. Quart. J.
Ferranti, L., T. N. Palmer, F. Monteni, and E. Klinker, 1990: Tropical– Roy. Meteor. Soc., 125, 583–609.
extratropical interaction associated with the 30–60 day oscil- Waliser, D. E., and Zhou, W., 1997: Removing satellite equatorial
lation and its impact on medium and extended range prediction. crossing time biases from the OLR and HRC datasets. J. Climate,
J. Atmos. Sci., 47, 2177–2199. 10, 2125–2146.
Hendon, H. H., C. Zhang, and J. D. Glick, 1999: Interannual variation , N. E. Graham, and C. Gautier, 1993: Comparison of the highly
of the Madden–Julian Oscillation during austral summer. J. Cli- reflective cloud and outgoing longwave radiation datasets for
mate, 12, 2538–2550. use in estimating tropical deep convection. J. Climate, 6, 331–
Higgins, R. W., and S. D. Schubert, 1996: Simulations of persistent 353.
North Pacific circulation anomalies and interhemispheric tele- , C. Jones, J. E. Schemm, and N. E. Graham, 1999: A statistical
connections. J. Atmos. Sci., 53, 188–207. extended-range tropical forecast model based on the slow evo-
, and K. C. Mo, 1997: Persistent North Pacific circulation anom- lution of the Madden–Julian Oscillation. J. Climate, 12, 1918–
alies and the tropical intraseasonal oscillation. J. Climate, 10, 1939.
223–244. Weickmann, K. M., 1991: El Niño/Southern Oscillation and Madden–
, J. E. Janowiak, and Y.-P. Yao, 1996: A Gridded Hourly Pre- Julian (30–60 day) oscillations during 1981–1982. J. Geophys.
cipitation Data Base for the United States (1963–1993). National Res., 96, 3187–3195.

Unauthenticated | Downloaded 10/10/22 05:49 PM UTC