1. Introduction
A mesoscale convective system (MCS) is a long-lived, multicellular structure composed of convective and stratiform clouds (Houze 1989). MCSs contribute greater than 70% of the rainfall in the Tropics but make up a much smaller fraction of the cloud cover (Mohr et al. 1999). As with other phenomena that occur in the Tropics, one might expect differences between an El Niño year and a La Niña year to show up as a difference in the number and distribution of MCSs.
Anomalous precipitation patterns during the negative and positive phases of the Southern Oscillation index (SOI) are well documented. The Walker circulation, an atmospheric pattern that is revealed by the removal of the mean zonal circulation, consists of rising motion over the equatorial western Pacific and sinking motion over the equatorial central Pacific. During an El Niño event the Walker circulation reverses, causing convection and therefore rainfall over the central Pacific (an area usually characterized by subsidence) and causing rainfall in the western Pacific to be suppressed (Rasmusson and Wallace 1983; Ropelewski and Halpert 1987).
Other anomalous precipitation characteristics of an El Niño event include suppression of the Indian monsoon (Rasmusson and Carpenter 1983; Meehl 1987; Ropelewski and Halpert 1987; Kiladis and Diaz 1989) and suppression of the Australian monsoon (Ropelewski and Halpert 1987; Kiladis and Diaz 1989; Barnston and Smith 1996). Drier-than-normal conditions occur in at least some seasons over southeastern Africa and Madagascar, and wetter-than-normal conditions occur in parts of equatorial eastern Africa. This characteristic possibly is due to a longer rainy season in the wetter area and a shorter rainy season in the dry area (Ropelewski and Halpert 1987; Nicholson and Kim 1997). The link between the conditions in these two regions is considered to be a dipole by Nicholson and Kim (1997) and is explained as the shift in a convergence zone by Ropelewski and Halpert (1987). The northern part of South America and southeastern South America also set up as a dry and wet dipole, respectively, according to Kiladis and Diaz (1989) and Ropelewski and Halpert (1987). Central America and the area around Hawaii also tend to be drier than normal during El Niño (Ropelewski and Halpert 1987).
During a La Niña event the same areas tend to be affected, with anomalies that are opposite in sign and somewhat weaker in strength than those during an El Niño year (Meehl 1987; Kiladis and van Loon 1988; Ropelewski and Halpert 1989; Kiladis and Diaz 1989; Nicholson and Kim 1997). Areas that tend to be drier than normal during La Niña are the central Pacific, equatorial eastern Africa, and the northern Mexico and Gulf of Mexico region. Areas that tend to be wetter than normal during La Niña include the western Pacific, the area around Hawaii, Australia, and India during the monsoon season (Kiladis and van Loon 1988; Ropelewski and Halpert 1989).
Although El Niño often tends to be used as a shorthand term for ENSO, the two parts of ENSO—El Niño (EN) and the Southern Oscillation (SO)—are not always linked (Deser and Wallace 1987). Kane (1997) classified events as combinations of EN, SO, and warm (W) or cold sea surface temperatures in the equatorial eastern Pacific. Kane (1997) found that only unambiguous ENSOW events, EN in the early part of the year and SO in the middle of the year, were reliable in giving the widely accepted El Niño precipitation patterns in all of the regions surveyed. In the periods where El Niño conditions persisted for more than one year, the second year was less likely to give expected results. Only in the Pacific did all of the EN events in Kane’s study correlate fairly well to rainfall variations. The effects on rainfall in the Pacific, the area directly related to El Niño, are more pronounced than in those areas with an indirect link.
Even though many regions of the world are known to be affected by El Niño and La Niña, there may be even more areas of anomalous precipitation that go undocumented, because in the past, scientists have tended to rely on ground stations for their climatological precipitation data. These ground stations are not evenly distributed and many areas in the Tropics and over oceans tend to have few observations. Satellite data are especially useful for documenting areas of the world where observations have been infrequent or absent in the past.
Satellites are useful for identifying and studying precipitation systems. Maddox (1980) used IR satellite data to define a mesoscale convective complex (MCC), which represents a class of MCS with a huge, near-circular, long-lasting cold cloud shield. Since that time, several groups of scientists have used satellites to study these very large, intense, and long-lasting mesoscale systems. Velasco and Fritsch (1987) used IR satellite data and surface data to study MCCs in Central and South America. They found that MCCs tend to occur over land and at night. They also suggested that El Niño might play a role in MCC activity. Miller and Fritsch (1991) found similar characteristics in their study of the western Pacific, as did Laing and Fritsch in their studies of African (1993a), Indian (1993b), and global MCCs (1997). Laing and Fritsch (1997) also suggested the possibility of a relationship between El Niño and MCCs.
Although all of the previously mentioned studies used IR satellite data, microwave channels such as the 85-GHz channel are particularly useful for studying mesoscale systems. The 85-GHz brightness temperatures respond to the precipitation-sized ice particles usually found well within the cloud system rather than to the small ice crystals near the cloud top. Stronger convection tends to produce more and larger ice particles, and therefore brightness temperature decreases as the depth of the storm containing sufficiently large ice particles increases. This behavior means that brightness temperature can be used as a proxy for strength of convection and, less directly, to make estimates of rainfall (Wu and Weinmann 1984; Mohr and Zipser 1996b).
Mohr and Zipser (1996b, hereinafter MZ96b) identified and mapped MCSs in the Tropics for January, April, July, and October of 1993 using the 85-GHz channel of the Special Sensor Microwave Imager (SSM/I) on board the Defense Meteorological Satellite Program (DMSP) F-11 satellite. MZ96b found that land areas tended to have smaller and more intense MCSs. The median size of an MCS was found to be 4500 km2. The greatest number of large MCSs was found over North America and the North Atlantic, subtropical South America, and the South Pacific. The median PCT (polarization-corrected temperature, after Spencer et al. 1989) was found to be 210 K. Tropical Africa had the greatest number of MCSs with a minimum PCT below 120 K. The most intense MCSs tended to be found over land or very near the coasts. This finding is supported by the studies of LeMone and Zipser (1980), Zipser and LeMone (1980), Jorgensen and LeMone (1989), and Lucas et al. (1994), which show that tropical oceanic updrafts tend to be much weaker than those over land. There was no statistically significant relationship between size and intensity of updrafts. MZ96b also found a strong diurnal character to the data, with oceanic MCSs being more frequent in sunrise satellite passes and land MCSs being more frequent in sunset passes.
Rainfall amounts are difficult to obtain for many tropical and oceanic regions. Therefore many algorithms have been developed for rainfall estimation using satellite data. Heymsfield and Fulton (1988), Spencer et al. (1989), Negri et al. (1993), Berg and Avery (1995), and Ferraro (1997) are among those who have used SSM/I data to make estimates of rainfall. For the 85-GHz channel, Spencer et al. (1989) suggest using a threshold of 255 K, thought to correspond to light-to-moderate rain (1–3 mm h−1). The area outlined as an MCS (250 K over an area of at least 2000 km2) in MZ96b was chosen to be close to the Spencer et al. (1989) threshold.
Another method of estimating rainfall is a combined analysis used by Xie and Arkin (1997). In their Climate Prediction Center Merged Analysis of Precipitation (CMAP), Xie and Arkin use a combination of data sources to produce an estimate of rainfall on a 2.5° by 2.5° grid for the entire globe. Their data sources are gauges, five kinds of satellite data [Geostationary Operational Environmental Satellite precipitation index, outgoing longwave radiation (OLR)-based precipitation index, SSM/I scattering, SSM/I emissions, and Microwave Sounding Unit], and numerical modeling output. Although there are problems with this method at higher latitudes, they believe that the accuracy is fairly good over land and over the tropical and subtropical oceans.
Although there have been assertions in the literature of a connection between MCCs (the very largest of the MCSs) and El Niño, there has not been a general study directly addressing the changes in mesoscale systems related to the interannual oscillation. This work will address not only changes in the number of MCSs but also changes in their size and intensity characteristics. MCSs are primary contributors to tropical rainfall. For that reason it is important to know how they change between El Niño and La Niña. Do the expected rainfall excesses or deficiencies associated with El Niño show up in the MCSs? In areas known to have enhanced precipitation, are there more MCSs observed or are those that are observed just larger or stronger?
This study compares and contrasts tropical precipitation patterns for an El Niño year and a La Niña year. This comparison is accomplished by extending the MZ96b algorithm for mapping MCSs, which already has been applied to an El Niño year, to a La Niña year. This study makes regional comparisons of various characteristics of MCSs, which are identified and classified using the microwave frequency of the SSM/I instruments on the DMSP F-11 and DMSP F-13 satellites for the years 1992/93 (an El Niño period) and 1995/96 (a La Niña period), respectively. Differences between the years are compared to documented patterns of variability. Several different characteristics of MCSs that could contribute to interannual variability in rainfall in each region are examined. These characteristics include the overall number, size, and intensity of the MCSs. This study also compares the patterns of MCSs with CMAP estimates to ascertain whether patterns of interannual variability in MCSs are comparable to observed rainfall differences. Section 2 describes the data and methods used in this study. Section 3 compares the MCSs with expected El Niño and La Niña patterns and discusses the characteristic of MCSs. Section 4 discusses the role that the characteristics of MCSs play in the interannual changes. Section 5 is a comparison of the independent CMAP estimate and the MCSs.
2. Data and methods
Two time periods are chosen for comparison in this study. The La Niña period for this study is May 1995– April 1996. This period was chosen as the first available La Niña conditions after the period of the MZ96b study. That MZ96b time period is November 1992–October 1993, so chosen because it was the first full year of available data for the DMSP F-11 SSM/I. That 1993 time period had weak El Niño conditions. In this paper, for convenience, these two time periods will be referred to as 1995 and 1993, respectively.
Table 1 shows SOI and the Pacific sea surface temperature (SST) anomalies for 1993 and 1995 from the National Oceanic and Atmospheric Administration Climate Analysis Center’s Climate Diagnostic Bulletin (Kousky 1993a,b; 1996a,b). Up to the time chosen for this study, the 1990s have been characterized by persistent El Niño conditions. Although the 1995 La Niña conditions are weak, they are the first La Niña conditions of the 1990s.
Both the SOI and the SST data in Table 1 demonstrate that these weak El Niño and weak La Niña conditions are persistent through most of the 12-month periods chosen for analysis. If taken 3 months at a time, each and every period shows that 1993 has a lower SOI than 1995 does by about one hecto Pascal. If SST is taken 3 months at a time, each seasonal average for each area (Niño-1 + 2, Niño-3, and Niño-4) shows warmer SSTs in 1993 than in 1995, with positive anomalies in 1993 and negative anomalies in 1995. The difference between the two years is 0.6°–0.8°C in each area for both December–February and June–August. Of course, these anomalies are far weaker than those for the major events of 1983, 1997, and 1998. It will be shown that many of the anomalies in the number and areal coverage of MCSs are in the expected sense; that is, it is assumed that they should be in the same sense as OLR anomalies, because both are proxies for precipitation. We will emphasize those regions in which anomalies associated with the ENSO cycle have been best documented: the equatorial Pacific Ocean and the “Maritime Continent” of Indonesia and surrounding oceanic regions.
As previously stated, MZ96b use the first full year of F-11 SSM/I data. Although F-11 SSM/I data are available for 1995, they have many missing periods and are unacceptable for this study. Therefore the F-13 SSM/I is used for the 1995 data. The F-11 and F-13 satellites are very similar. Both are sun-synchronous, low-altitude, polar-orbiting satellites with local equatorial crossing times of 0504/1704 and 0536/1736 LT, respectively. Both satellites carry an SSM/I instrument. The 85-GHz channel of the SSM/I is used for both MZ96b and this study. This channel has a resolution of 13 km × 15 km and a swath width of about 1400 km.
SSM/I samples each location on the equator once per day. Near 30°, a location is sampled either zero or two times per day. The sampling pattern shifts by a few degrees of longitude each day; therefore, an area is not sampled each day. This shift also means that the satellite revisit time varies between 12 h and 3 days. The average revisit time is fairly constant in the Tropics and is about 24 h (Berg and Avery 1995).
The average return time of 24 h means that most of the storms identified should be independent events. Velasco and Fritsch (1987) found that most MCCs they studied lasted between 8 and 17 h. The vast majority of MCSs in this study are not nearly large enough to be classified as an MCC. We assume that, on average, size and intensity of MCSs are related. Therefore, even at 35° where SSM/I return times are less than one day, most of the identified MCSs are independent events.
There are some data losses in both the 1993 and 1995 years, but they are mostly small and random. In 1993, there are no missing days of data and no day that has less than 20 orbits (Mohr 1999, personal communication). In 1995, there were 2 days of missing data and 11 days with less than 20 orbits. A perfect day should have about 28 orbits—14 ascending and 14 descending. The two missing days and two of the bad data days were early in May of 1995, when the F-13 SSM/I was first coming online. Although these data should not be a problem when yearly totals are calculated, their existence should be noted when any comparison of a shorter time period using this month is performed. The other bad days of data are scattered throughout the year, with no other month having more than three bad days.
The MZ96b criteria for identification of an MCS are as follows. There must be an area greater than or equal to 2000 km2 with a PCT less than or equal to 250 K. Within this area, there must be a pixel with minimum brightness temperature less than or equal to 225 K.
Snow-covered areas cause depression of PCTs in much the same manner as does precipitation-sized ice (Ferraro et al. 1994). This effect could cause large snow-covered areas to be mistaken as MCSs. To prevent this mistake from happening, areas likely to be snow covered were removed using the MZ96b snow screen.
Each month of data is mapped by plotting the centroid of the MCSs. An example of this is in Fig. 1, which shows maps for January of 1993 and 1996, respectively. In this map, the four classes of PCT (Mohr and Zipser 1996a) are plotted using progressively larger plus signs for more intense MCSs (colder minimum PCT). Area classes are plotted in much the same manner (not shown). Temperature and area maps both have been produced for all 12 months of the year and as of the time of writing are available on the Internet at http://www.met.tamu.edu/research/tcrp/wetnet.html.
January is a typical month in that most areas look fairly similar between the two years. One difference that stands out (as it does in the maps of all of the months) is the large dry slot in 1995, which runs along the equator in the central Pacific from 150°E to 150°W and is a typical La Niña feature. Also, as expected, the Maritime Continent has more MCSs in the La Niña year of 1995 than in 1993. The South Pacific convergence zone is characteristically farther east and north during the El Niño year of 1993. There are other differences between the two Januarys that may or may not be related to the ENSO cycle, such as for the Gulf of Mexico and interior Australia regions, both of which have more MCSs in 1993.
The area from 35°N to 35°S is broken up into 18 boxes for all 12 months of 1993 and 1995. Figure 2 shows a map of the boundaries for each region. The approximate area of each region is shown in Table 2. Only MCSs wholly within a box are included. This criterion causes about 8% of the MCSs to be excluded from the sample. Larger MCSs are more likely to cross a boundary than are smaller MCSs. When every MCS is forced into a box, it is noted that the exclusion rate is nearly the same from month to month in the regions. Therefore the MCSs in the boxes are a representative sample. Unless otherwise noted, all totals and statistics in this paper use this method of assigning MCSs.
3. Results
This study examines the differences in MCSs between 1993 and 1995. The most obvious display of change caused by the interannual oscillation would be different numbers of MCSs in the regions between the years. A second characteristic is the size of MCSs. By definition, the boundary for the MCSs in this study is the 250 K isotherm, which should also be the boundary of the area covered by light–moderate rain (Spencer et. al. 1989). A larger MCS means a larger area of rain, and therefore a larger MCS would have more rain associated with it, all other factors being equal. If the distribution of areas of MCSs is significantly different between the two years, one would see a change in the precipitation pattern. Another possible difference is intensity. The lower the 85-GHz PCT is, the more intense is the system. The 85-GHz PCTs are assumed to be related inversely to rainfall rate (Spencer et al. 1989; Adler et al. 1994). All other things being equal, a lower 85-GHz PCT in a region would mean greater rainfall in that region.
To make numeric, size, and intensity comparisons between the 1995/96 and 1992/93 years, several numbers are calculated, starting with the total number of MCSs in each of the 18 regions and the total size (accumulated area) of MCSs in each region. MCSs then are ranked by size and brightness temperature for each box. This ranking allows for comparisons of the 50th and 90th percentile levels with MZ96b.
This section addresses the following questions. Do the years 1993 and 1995 show statistically significant differences in the numbers of MCSs in the regions, and are these differences similar to the expected El Niño and La Niña pattern? Regardless of number, do the basic characteristics of MCSs change from 1993 to 1995 in the regions? These questions are addressed by looking at the number, size, and intensity distributions of MCSs in 1993 and 1995.
a. Number
Figure 3 shows the number of MCSs in each region for 1993 and 1995 (in bold). The grand total of MCSs in all boxes is 52 601 in 1993 and 50 920 in 1995, a ratio of 1.033. We cannot know whether this difference (significant at the 0.01 level) is real or instrumental. Because our focus is on differences between the two years on a regional basis, we choose to normalize all the ratios used in this paper by 1.033 unless otherwise stated.
A ratio of the total number of MCSs for each box for 1993 to 1995 is also shown in Fig. 3. The largest differences between the two years occur in the central Pacific, eastern Pacific, tropical Atlantic, and the Maritime Continent. These areas are also found to be different between the two years on the 0.01 level of the chi-square test. Thirteen of the 18 regions have statistically significant differences (0.05 level) in number of MCSs between the two years. One area that often is cited in the literature as being affected by El Niño that did not show a difference is Australia (Ropelewski and Halpert 1987; Kiladis and Diaz 1989; Barnston and Smith 1996).
In most regions, the differences between the two years are in the sense expected from past studies in the literature. The central Pacific and eastern Pacific have more MCSs during 1993 (El Niño) than during 1995, and the Maritime Continent has less MCSs (Rasmusson and Wallace 1983; Ropelewski and Halpert 1987). The South Pacific has less MCSs in 1993 than in 1995. This fact is due to a northward shift in the South Pacific convergence zone that is a characteristic of El Niño (Ropelewski and Halpert 1987).
To examine the equatorial Pacific region more directly, central Pacific and Eastern Pacific regions between 5°N and 5°S are subdivided. The subregion from 5°N to 5°S is called equatorial, and the areas from 15° to 5°S and from 5° to 15°N are grouped together and called off-equatorial. Because these regions are small, every MCS is forced into a box, by assigning an MCS that falls on a boundary to the box that contains most of its area.
The numbers of MCSs in the new regions are shown in Table 3. The ratio of the number of MCSs in the equatorial central Pacific in 1993 to the number of MCSs in the equatorial central pacific in 1995 is 3.31. This same ratio for the eastern Pacific is 2.70. The ratios in the two off-equatorial areas are 1.02 and 1.20. When tested with the chi-square test, all of these smaller regions except the off-equatorial central Pacific are significantly different. Therefore, in spite of the weakness of the 1993 and 1995 events, these differences in MCS numbers are much greater in the equatorial regions and constitute a clear signal in the expected sense of interannual variability.
b. Seasonal comparisons
Because the annual cycle of precipitation or its proxies is so strong, it can overwhelm variations from El Niño in yearly totals. Seasonal totals may show characteristics that totals for the longer time period do not. The 1993 and 1995 years are broken up into 3-month seasons, and the December–January–February (DJF) and June–July–August (JJA) seasons are compared.
1) December–January–February
Figure 4 shows the number of MCSs for 1993 and 1995 (in bold) for the DJF season and the ratio of the number of MCSs in 1993 to the number of MCS in 1995 (in italics). Eleven of the 18 areas have differences of greater than 10%, all of which are significant at the 0.05 level using the chi-square test. Especially noteworthy are the large absolute and relative differences in the central Pacific, South Pacific, and Maritime Continent. These areas show the expected El Niño and La Niña patterns (Rasmusson and Wallace 1983; Ropelewski and Halpert 1987), with more MCSs in the central Pacific and fewer MCSs in the Maritime Continent and South Pacific during El Niño. Although some other regions have large relative differences between the two years, if the absolute numbers of MCSs are small then less significance should be attributed to them.
The difference between the DJF of 1993 and the DJF of 1995 is even more impressive in the equatorial Pacific. As might be expected, Table 4 shows almost 4 times the number of MCSs in the equatorial central Pacific in 1993 versus 1995. (The number of MCSs in the equatorial eastern Pacific is far too small for comparisons to be significant.)
2) June–July–August
Figure 5 shows the number of MCSs for the 1993 and 1995 (in bold) JJA season and the ratios (in italics). Once again some of the regions show the expected El Niño and La Niña differences. The central Pacific and eastern Pacific have more MCSs in 1993 (Rasmusson and Carpenter 1983; Ropelewski and Halpert 1987). The tropical Atlantic and the combined North America and North Atlantic have more MCSs in 1995. Inspection of the actual MCS distributions reveals that the (expected) enhancement of tropical cyclones in the La Niña year of 1995 contributed to this result. Although India and eastern Asia had a suppressed monsoon season in 1993 (Rasmusson and Carpenter 1983; Meehl 1987; Ropelewski and Halpert 1987; Kiladis and Diaz 1989), the MCS differences do not reflect this fact. The equatorial Indian Ocean region does show reduced MCSs in 1993.
Table 5 shows that the equatorial Pacific differences are again impressive for the JJA season. The equatorial central Pacific has over 3 times the number of MCSs in 1993 as in 1995. The equatorial eastern Pacific has too few MCSs in both years for comparisons to be significant.
The year and the seasons examined do show that the majority of the regions have a statistically significant difference in the number of MCSs between 1993 and 1995. In most of these cases, the regions show a difference in number of MCSs of greater than 10%. The changes tend to be in the manner expected for El Niño and La Nina, particularly in the central Pacific, eastern Pacific, and Maritime Continent.
c. Size and intensity
Figures 6 and 7 show the 50th and 90th percentiles of MCS area for 1993 and 1995, respectively. Both of these figures are ordered by the 50th percentile. The figures show that the median size is fairly consistent from region to region and year to year at about 4500 km2. The regions with the highest median areas in 1993, the subtropical (> 15° and < −15°) oceans, are also those regions with the highest median areas in 1995. The regions with the lowest median areas in 1993, tropical South America, tropical Africa, and the Maritime Continent, also have the smallest median areas in 1995. The 90th percentiles show greater variability than the medians.
Figures 8 and 9 show the 50th and 10th percentiles of the 85-GHz PCT for 1993 and 1995, respectively. The regions are ordered by the 50th percentile. Figure 10 shows a map of the regions with the 50th percentile value placed in each region for 1993 and 1995 (in bold).
Several PCT characteristics can be noted from Figs. 8 and 9. It is easy to see that the land regions tend to have MCSs with lower PCTs than do the oceanic regions. The subtropical oceans tend to have the warmest PCTs, with the tropical oceans falling between the land and subtropical oceans.
The median PCTs tend to be fairly close in each region between the two years (Fig. 10). It is obvious that the differences between regions remain the same between the two years, overwhelming any interannnual differences in median intensity of MCSs. A similar figure for the 10th percentile (i.e., the PCT for which just 10% of the distribution is more intense) was constructed (not shown); the region-to-region differences remain large, and the interannual differences remain very small.
In summary, the size and intensity of the MCSs tend to be fairly similar over the two years. The differences among the regions are much more pronounced than are the differences between the two years in any one region. Land and water differences are much more important than the interannual oscillation, especially with regard to intensity.
d. Discussion
The last section found that there are differences in the number of MCSs between 1993 and 1995. These differences tend to match fairly well with expectations for changes between an El Niño and a La Niña year. This section explores the characteristics of MCSs further. The following question is examined: Which characteristics of MCSs show the most change or give the most information about the interannual cycle?
The total accumulated area occupied by all MCSs in a region should be related to the number of MCSs in that region, if the frequency distribution of areas is similar between the years. This relation has been examined extensively (not shown), and the results can be stated concisely. The frequency distribution of areas is approximately the same from 1993 to 1995 in each region. The relationship between the number of MCSs in a region and the total area of MCSs in that region over a year is nearly 1:1, with linear regression explaining 87% of the variance.
Although the total area is related closely to the number of MCSs, there may be some regions that show a change in the distribution of areas. To look for area distribution shifts, the Wilcoxon rank-sum test was performed on the data. Only 5 of the 18 regions show a shift in area distribution, and the shift is not large. Few regions show an area distribution shift in either of the seasons.
Another characteristic that could show a change between the years is the intensity of the MCSs. For the year, there are only 2 of the 18 regions that show a change in the median 85-GHz PCT of greater than 3 K (Fig. 10). They are Madagascar and the south Indian region. The seasonal changes in intensity have been examined and generally are very small between the two years.
The results in this section, along with those in the last section, show that the changes between the years are due almost entirely to the number and the total area of MCSs. Also, the changes in number and total area, where substantial, are generally similar to the expected patterns for El Niño and La Niña. Land and ocean effects are much more important than the interannual oscillation in determining size and intensity of MCSs.
4. Rain estimates
Mohr et al. (1999) show that, within the limitations of their method, MCSs identified with the 85-GHz PCT of SSM/I make a significant contribution to tropical rainfall (>70%) despite the fact that they account for only about 10%–20% of their precipitating systems. The differences between 1995/96 and 1992/93 are compared now with the CMAP rainfall estimates for those years, described in Xie and Arkin (1997). Comparison with the precipitation estimates also should give insight into which characteristics of MCSs best correspond to rainfall variations.
The ratios of the 1993-to-1995 rainfall estimates are shown in Fig. 11. The monthly CMAP totals for each region were available at the time of writing at http://www.met.tamu.edu/research/tcrp/wetnet.html. The same El Niño and La Niña pattern of differences between the years shows up in the rainfall estimates. The Maritime Continent is drier and the central Pacific and eastern Pacific are wetter in 1993 than in 1995. The monthly CMAP data show that the central Pacific shows a larger change caused by the interannual oscillation than by the seasonal cycle, despite the weakness of these events. Both the number of MCSs and the CMAP rainfall estimates show a larger difference between the two years than between the DJF and the JJA seasons. This region is the only one for which this is the case.
The tropical Atlantic, for which the number of MCSs and the total area of MCSs is more than 20% lower in 1993 than in 1995, also is much drier in 1993 in the CMAP estimates. The largest amount of rain fell in the central Pacific in 1993 and in the Maritime Continent in 1995. In most areas, the ratios of the rainfall estimates are similar to the ratio of the numbers of MCSs shown in Fig. 3.
Figure 11 also shows the ratios of the CMAP estimates for the DJF and JJA seasons of 1993 to 1995. Like the corresponding figures showing the ratios of the numbers of MCSs for the seasons (Figs. 4 and 5), the seasonal rainfall estimates often show greater differences between the seasons than are seen in the year as a whole.
The ratios of the CMAP rainfall estimates (Fig. 11) can be compared to the ratios of the total area of MCSs (not shown). As in the case of number of MCSs, the ratios of total area match well with the ratios of CMAP rainfall estimates for the year and, to an extent, the seasons.
The CMAP rainfall estimates closely match the number and total area of MCSs. The percent of variance explained for ratios of CMAP rainfall estimates with both the number and total area are close to each other. The number of MCSs and total area of MCSs closely match each other. The percent of variance explained for number with total area is slightly higher, for the year and for both seasons, than it is for CMAP rainfall with either number or total area. This result means that the number and total area are slightly closer in relation to each other than either is to the CMAP rain estimates.
a. Weak versus strong ENSO cycles
It is of interest to compare the weak ENSO cycles of 1993 versus 1995 with the extreme events of 1983 and 1997 (El Niño) versus mid–late 1998 (La Niña). At the time of this writing we can only use the CMAP estimates for such a comparison, and we have compiled these estimates for the 18 regions for the 1979–98 period. Detailed study of these patterns is well beyond our scope here; only a few main points will be made.
For DJF, we can compare the 1993:1995 ratios (Fig. 11) with only the 1983:1995 and 1997:1995 ratios, because the 1998/99 La Niña data are not available yet. For the central Pacific, these ratios are 1.57, 1.68, and 1.65—a surprisingly small difference. For the eastern Pacific, the ratios are 1.01, 2.54, and 2.63. The implication is that the 1993 versus 1995 cycle was quite strong in the central Pacific and almost nonexistent in the eastern Pacific relative to the 1983 and 1997 events. The CMAP data indicate that the strong negative rainfall anomaly in the La Niña of 1995 in the central Pacific and the absence of a positive anomaly in 1993 in the eastern Pacific control these ratio comparisons. The ratios for DJF for the Maritime Continent are 0.71, 0.51, and 0.51. The implication is that El Niño years are drier than La Niña years, with the relative anomalies being stronger in the major events of 1983 and 1997.
For JJA we can compare the El Niño year of 1993 with 1983 and 1997 and the La Niña year of 1995 with 1998 (the strong La Niña was in place by May). First, we note that the India–eastern Asia region CMAP estimates are within 10% of one another in all 5 yr, in agreement with our MCS results for 1993 and 1995 but at variance with the oft-quoted tendency for El Niño years to have deficient monsoon rainfall. The Maritime Continent estimates also show relatively minor changes for JJA between even the extreme ENSO years. The eastern and central Pacific estimates for JJA show that 1983 and 1997 were somewhat wetter than 1993 was. Although the 1998 La Niña was drier than 1995 was in the central Pacific, it is the “weak” 1995 event that was drier than 1998 was in the eastern Pacific.
The main surprise in these CMAP estimates is the lack of a 1993 versus 1995 difference in the eastern Pacific for DJF. Our data show that DJF 1993 and 1995 also have almost no difference in the number and total area of MCSs, however. We tentatively conclude that the MCS and CMAP datasets are consistent with one another. Last, the CMAP rainfall estimates indicate that there is far from a 1:1 correspondence between measures of ENSO and rainfall anomalies. Therefore, although the 1993 and 1995 events were weak, they are a fair test of the utility of using MCS data to characterize anomalous seasons in the Tropics on a regional basis.
5. Conclusions
Rainfall changes related to El Niño are documented well for various regions of the earth, but the effects of the interannual oscillation on MCSs over the global Tropics have not been studied. This study uses the MZ96b algorithm to define and to process MCSs from the 85-GHz channel of the SSM/I instrument on the DMSP F-13 satellite for May 1995–April 1996, a La Niña period. These data were compared with the MZ96b data for the period November 1992–October 1993, an El Niño period.
Despite the fact that the years chosen do not represent strong El Niño and La Niña conditions, most of the changes in the numbers of MCSs between the two years fit with expected precipitation anomaly patterns. The central Pacific and eastern Pacific have an enhanced number of MCSs in the El Niño year. The Maritime Continent and the South Pacific have fewer MCSs. The expected reduction over India and eastern Asia was virtually absent.
The JJA and DJF seasons also are examined for El Niño patterns. This examination highlighted differences that, in some regions, are masked in totals for the entire year. Most of the seasonal differences also fit with the expected precipitation anomaly patterns. For JJA, the El Niño year has a greater number of MCSs for the central Pacific and eastern Pacific and has a lesser number of MCSs for the Maritime Continent, North America and the North Atlantic, and the tropical Atlantic. For DJF, the El Niño year is wetter for the central Pacific and Australia and is drier for the Maritime Continent and the South Pacific.
Overall, the areas with direct ties to ENSO—central Pacific, eastern Pacific, and Maritime Continent—show the largest changes in the number and total area of MCSs. An equatorial subregion (5°N to 5°S) shows by far the greatest difference between the two years.
Although there are many differences over the Tropics related to the interannual oscillation, the properties of the MCSs themselves do not change significantly. In both years, the largest median areas of MCSs are found over the subtropical oceans, and the smallest are found over tropical South America, tropical Africa, and the Maritime Continent. In both years, the most intense MCSs are found over land and the least intense are found over the subtropical oceans.
The greatest differences between the two years are due to the change in number and total area of MCSs. The number of MCSs did have a statistically significant change in at least one of the seasons or the year for 17 of the 18 regions studied. The ratios of the total area of MCSs between the two years match closely with the ratios of the numbers of MCSs between the two years. Some regions do show minor area distribution changes and/or intensity changes between the two years, but these changes are less frequent than the change in number and total area of MCSs. In general, intensity does not change between the two years.
The number of MCSs, total area of MCSs, and CMAP rainfall estimates match well for most regions for the year and for the JJA and DJF seasons. The ratios of rainfall show a significant relationship to the number and total area of MCSs for the year as a whole and both seasons. The close match between the CMAP estimates and the number and total area of MCSs reemphasizes the importance of number and total area in the interannual changes.
Acknowledgments
This research was supported by NASA Grant NAG5-4699. The suggestions of James McGuirk and Benjamin Giese improved the earlier drafts. Additional thanks are due Dan Cecil, Chris Lucas, Steve Nesbitt, Rick Toracinta, Chris West, and Dave Wolff for their assistance and support. The constructive criticism of Robert Maddox and an anonymous reviewer substantially improved the manuscript.
REFERENCES
Adler, R. F., G. J. Huffman, and P. R. Keehn, 1994: Global tropical rain estimates from microwave-adjusted geosynchronous IR data. Remote Sens. Rev.,11, 125–152.
Barnston, A. G., and T. M. Smith, 1996: Specification and prediction of global surface temperature and precipitation from global SST using CCA. J. Climate,9, 2660–2696.
Berg, W., and S. K. Avery, 1995: Evaluation of monthly rainfall estimates derived from the Special Sensor Microwave/Imager (SSM/I) over the tropical Pacific. J. Geophys. Res.,100, 1295– 1315.
Deser, C., and J. M. Wallace, 1987: El Niño events and their relation to the Southern Oscillation: 1925–1986. J. Geophys. Res.,92, 14 189–14 196.
Ferraro, R. R., 1997: Special Sensor Microwave Imager derived global rainfall estimates for climatological applications. J. Geophys. Res.,102, 16 715–16 735.
——, N. C. Grody, and G. F. Marks, 1994: Effects of surface conditions on rain identification using the DMSP-SSM/I. Remote Sens. Rev.,11, 195–209.
Heymsfield, G. M., and R. Fulton, 1988: Comparison of high-altitude remote aircraft measurements with the radar structure of an Oklahoma thunderstorm: Implications for precipitation estimation from space. Mon. Wea. Rev.,116, 1157–1174.
Houze, R. A., Jr., 1989: Observed structure of mesoscale convective systems and implications for large-scale heating. Quart. J. Roy. Meteor. Soc.,115, 425–461.
Jorgensen, D. P., and M. A. LeMone, 1989: Vertical velocity characteristics of oceanic convection. J. Atmos. Sci.,46, 621–640.
Kane, R. P., 1997: Relationship of El Niño–Southern Oscillation and Pacific sea surface temperature with rainfall in various regions of the globe. Mon. Wea. Rev.,125, 1792–1800.
Kiladis, G. N., and H. van Loon, 1988: The Southern Oscillation. Part VII: Meteorological anomalies over the Indian and Pacific sectors associated with the extremes of the oscillation. Mon. Wea. Rev.,116, 120–135.
——, and H. F. Diaz, 1989: Global climate anomalies associated with extremes in the Southern Oscillation. J. Climate,2, 1069–1090.
Kousky, V. E., Ed., 1993a: Climate Diagnostics Bulletin. October, 1993, 79 pp.
——, 1993b: Climate Diagnostics Bulletin. November, 1993, 74 pp.
——, 1996a: Climate Diagnostics Bulletin. April, 1996, 79 pp.
——, 1996b: Climate Diagnostics Bulletin. May, 1996, 78 pp.
Laing, A. G., and J. M. Fritsch, 1993a: Mesoscale convective complexes in Africa. Mon. Wea. Rev.,121, 2254–2263.
——, and ——, 1993b: Mesoscale convective complexes over the Indian monsoon region. J. Climate,6, 911–919.
——, and ——, 1997: The global population of mesoscale convective complexes. Quart. J. Roy. Meteor. Soc.,123, 389–405.
LeMone, M. A., and E. J. Zipser, 1980: Cumulonimbus vertical velocity events in GATE. Part I: Diameter, intensity, and mass flux. J. Atmos. Sci.,37, 2444–2457.
Lucas, C., E. J. Zipser, and M. A. LeMone, 1994: Vertical velocity in oceanic convection off tropical Australia. J. Atmos. Sci.,51, 3183–3193.
Maddox, R. A., 1980: Mesoscale convective complexes. Bull. Amer. Meteor. Soc.,61, 1374–1387.
Meehl, G. A., 1987: The annual cycle and interannual variability in the tropical Pacific and Indian Ocean regions. Mon. Wea. Rev.,115, 27–50.
Miller, D., and J. M. Fritsch, 1991: Mesoscale convective complexes in the western Pacific region. Mon. Wea. Rev.,119, 2978–2992.
Mohr, K. I., and E. J. Zipser, 1996a: Defining mesoscale convective systems by their 85-GHz ice-scattering signatures. Bull. Amer. Meteor. Soc.,77, 1179–1189.
——, and ——, 1996b: Mesoscale convective systems defined by their 85-GHz ice scattering signature: Size and intensity comparisons over tropical oceans and continents. Mon. Wea. Rev.,124, 2417–2437.
——, J. S. Famiglietti, and E. J. Zipser, 1999: The contribution to tropical rainfall with respect to convective system type, size, and intensity estimated from the 85-GHz ice-scattering signature. J. Appl. Meteor.,38, 596–606.
Negri, A. J., R. F. Adler, R. A. Maddox, K. W. Howard, and P. R. Keehn, 1993: A regional rainfall climatology over Mexico and the southwest United States derived from passive microwave and geosynchronous infrared data. J. Climate,6, 2144–2160.
Nicholson, S., and J. Kim, 1997: The relationship of the El Niño– Southern Oscillation to African rainfall. Int. J. Climatol.,17, 117–135.
Rasmusson, E. M., and T. H. Carpenter, 1983: The relationship between eastern equatorial Pacific sea surface temperatures and rainfall over India and Sri Lanka. Mon. Wea. Rev.,111, 517– 528.
——, and J. M. Wallace, 1983: Meteorological aspects of the El Niño– Southern Oscillation. Science,222, 1195–1202.
Ropelewski, C. F., and M. S. Halpert, 1987: Global and regional scale precipitation patterns associated with the El Niño–Southern Oscillation. Mon. Wea. Rev.,115, 1606–1626.
——, and ——, 1989: Precipitation patterns associated with the high index phase of the Southern Oscillation. J. Climate,2, 268–284.
Spencer, R. W., H. M. Goodman, and R. E. Hood, 1989: Precipitation retrieval over land and ocean with the SSM/I: Identification and characteristics of the scattering signal. J. Atmos. Oceanic Technol.,6, 254–273.
Velasco, I., and J. M. Fritsch, 1987: Mesoscale convective complexes in the Americas. J. Geophys. Res.,92, 9591–9613.
Wu, R., and J. A. Weinman, 1984: Microwave radiances from precipitating clouds containing aspherical ice, combined phase, and liquid hydrometeors. J. Geophys. Res.,89, 7170–7178.
Xie, P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc.,78, 2539– 2558.
Zipser, E. J., and M. A. LeMone, 1980: Cumulonimbus vertical velocity events in GATE. Part II: Synthesis and model core structure. J. Atmos. Sci.,37, 2458–2469.
(top) Jan 1993 (Mohr and Zipser 1996a) and (bottom) Jan 1996 MCSs by PCT class
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
Map of regions. The east and west boundaries of the South American regions are at the coastlines. Note that the North, central, and South Pacific regions straddle the 180° meridian and extend to 155°E
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
No. of MCSs for 1993 and 1995 (bold). Also given is the ratio of number of MCSs for 1993 to 1995 (italics)
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
No. of MCSs in DJF only for 1993 and 1995 (bold). Also given is the ratio of number of MCSs for DJF for 1993:1995 (italics)
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
No. of MCSs in JJA only for 1993 and 1995 (bold). Also given is the ratio of number of MCSs for JJA for 1993:1995 (italics)
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
50th and 90th percentile of area of MCSs in 1993
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
50th and 90th percentile of area of MCSs in 1995
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
50th and 90th percentile of minimum PCT in 1993
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
50th and 90th percentile of minimum PCT in 1995
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
50th percentiles of minimum PCT for 1993 and 1995 (bold)
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
Ratio of CMAP rain estimates for 1993 to 1995. Boldface indicates full year, top set of numbers corresponds to DJF, and italics indicate JJA
Citation: Journal of Climate 13, 18; 10.1175/1520-0442(2000)013<3314:ACOTMC>2.0.CO;2
SOI (hPa) and Pacific SST anomalies (°C) for Niño-1 and -2 (0°–10°S, 90°–80°W), Niño-3 (5°N–5°S, 150°–90°W), and Niño-4 (5°N–5°S, 160°E–150°W)
Areas (approximate) of regions
No. and ratio of MCSs for subdivided regions
No. and ratio of MCSs for subdivided regions in DJF
No. and ratio of MCSs for subdivided regions in JJA