Perhaps the most regular and predictable weather pattern in North America is the North American (NA) or Mexican monsoon. Occurring in summer, it delivers about 35% and 45% of Arizona's and New Mexico's annual precipitation, respectively, and about 60% of northern Mexico's. While recent studies have linked strong NA monsoons to summer drought in the U.S. Midwest, the sequence of events that produce the NA monsoon remain unclear.
This empirical study builds on the findings of many other studies that implicate the Gulf of California [(GOC) or simply the gulf] as the dominant moisture source for the monsoon. It examines six monsoon seasons in detail, and quantitatively relates GOC sea surface temperatures (SST) to the timing, amount, and regional extent of monsoon rainfall.
This six season study is based on satellite measurements of rainfall [using the Special Sensor Microwave Imager (SSM/I)] and GOC SST at high spatial and temporal resolution. Key findings include the following. 1) Monsoon rainfall did not occur prior to the onset of GOC SSTs exceeding 26°C, and the incremental advance of SSTs > 26°C up the mainland coast of Mexico appears necessary for the northward advance of the monsoon. 2) For the period June–August, 75% of the rainfall in the Arizona–New Mexico region (AZNM) occurred after northern GOC SSTs exceeded 29°C, with relatively heavy rains typically beginning 0–7 days after this exceedance. 3) For a given year, SSTs in the southern and central GOC reached 29.5°C during a similar time frame, but such warming was delayed in the northern GOC. This warming delay coincided with a rainfall delay for AZNM relative to regions farther south. 4) Based on the 17 yr of available SST data, 14 of those years exhibited the following behavior: When northern gulf SSTs were relatively high for some period during the first half of July, rainfall during June–August in Arizona was relatively high. Otherwise, June–August Arizona rainfall was normal or below normal. 5) Anomalously wet July–September periods in Arizona do not correspond to anomalously wet periods in New Mexico, based on data from 1950 to the present. The wettest Arizona seasons, about 1.1 standard deviations wetter than normal, were strongly related to summer drought in the Midwest, being about 0.8 standard deviations drier than normal. This was not true for the wettest New Mexico years (Midwest rainfall was near normal), but these years exhibited dry conditions in the interior Northwest, with standard deviations being about 0.6–0.9 drier than normal. Collectively, this research suggests that the cause of these two wet monsoon modes may be related to SSTs in the northern gulf, which appear to affect Arizona more than New Mexico rainfall.
During the period July–September, northwestern Mexico receives about 60% of its total annual rainfall, while Arizona and New Mexico receive about 35% and 45% of their annual rainfall, respectively (Higgins et al. 1999). This precipitation pattern is often referred to as the Mexican or North American (NA) monsoon. The monsoon was believed by some to result from moist air advected from the Gulf of Mexico and the Caribbean, due to the change in midtropospheric winds at the onset of the monsoon, from westerly to easterly (Bryson and Lowry 1955; Sellers and Hill 1974). Others (Reitan 1957; Hales 1974; Brenner 1974; Carleton 1986; Douglas et al. 1993; Douglas 1995; Stensrud et al. 1995) have argued and supplied persuasive evidence that the monsoonal moisture source is the Gulf of California [(GOC) or the gulf] and/or the tropical east Pacific. In their review, Adams and Comrie (1997) state that “there is general agreement that the bulk of monsoon moisture is advected at low levels from the eastern tropical Pacific Ocean and the Gulf of California.” In recent work (Mitchell and Brown 1996; Brown and Mitchell 1997; Mitchell et al. 1999), it was suggested that the northward evolution of the 29°C SST isotherm into the GOC may govern the northward propagation of convection and rainfall. Moreover, the modeling study of Stensrud et al. (1995) revealed that for the monsoon to be successfully simulated, exceptionally warm water (29.5°C) in the GOC must be present. The warming of GOC SSTs may be partially due to a coastal warm current during spring/summer (Collins et al. 1997; Castro et al. 1994).
Recent work (Higgins et al. 1997; Higgins et al. 1998) suggests the Mexican monsoon exerts a strong influence on summer precipitation patterns over the contiguous United States, with strong monsoons well correlated with summer droughts in the Midwest, and somewhat correlated with relatively wet summers in the southeastern United States. If indeed the NA monsoon does depend on SSTs in the GOC, a relatively small body of water, then it is not surprising that current-generation global and regional climate models have difficulty in predicting the monsoon rainfall amounts, as well as NA summer precipitation in general.
The objective of this study was to determine whether the onset and subsequent evolution of monsoon rainfall was consistent with a postulated dependence on evolving GOC SSTs. This entails 1) determining whether monsoon rainfall is predicated by a threshold SST, where rainfall occurs only after local SSTs exceed this value, and 2) determining relationships between GOC SST increases and subsequent rainfall amounts. The next section addresses the seasonal evolution of the monsoon region, including SSTs, convection, the GOC circulation, and wind fields. Section 3 describes the methodology and the results of this study. Oceanic processes related to GOC SSTs are discussed in section 4. Conclusions are given in section 5.
2. Regional context
a. Seasonal evolution of SST and OLR fields
Outgoing longwave radiation (OLR) measured by satellites can be used as a proxy for deep convection in the Tropics, and in the subtropics during summer, since this produces extensive cirrus anvil clouds that emit at relatively cold temperatures. Thus, the lowest OLR values over such regions generally indicate convective activity and cirrus anvil coverage (e.g., Zhang 1993). The OLR data used here are available on a daily basis from twice daily National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) soundings (e.g., Gruber and Krueger 1984). The grid size is 2.5° × 2.5° for the period 1974 to the present. The data provide a reasonable indicator of the extent of convection. The SST data used are based upon an optimum interpolation analysis and adjusted for biases using the method of Reynolds (1988) and Reynolds and Marsico (1993). This analysis is produced weekly on a 1° grid using in situ and satellite-derived SSTs (hence blended data). However, the SST data were not used for the main results in section 3, which have higher spatial resolution.
The evolution of SSTs and tropical convection inferred from OLR are shown in Fig. 1 for the monsoon and surrounding regions. Monthly averaged OLR and blended SST data, from 1974 to 1993, were used to generate the 20-yr mean values. Due to the 1° resolution of SST data, the accuracy of SSTs in the GOC (especially its upper half) is degraded, but the spatial and temporal behavior of the SST data should be realistic enough to recognize the general behavior of the system. We regard the 240 W m−2 OLR isoline as an estimate of the monsoon's extent, since OLR < 240 W m−2 generally appears associated with deep convection at these latitudes (Zhang 1993). Clear skies dominate the warm pool region in March and April, allowing solar insolation to expand the warm pool. By May, a 29°C SST region (colored red) extends off southern Mexico from 109° to 90°W, and south to 7°N latitude. In June, convection in the intertropical convergence zone (ITCZ) further intensifies, possibly cooling SSTs slightly in this region (Ramanathan and Collins 1991). Also, relatively warm water begins advancing up the mainland coast of Mexico. This continues into July, with the 29°C isotherm now extended into the GOC. Also in July, the 240 W m−2 OLR isoline makes a dramatic advance northwestward, indicating the onset of the monsoon. The question can be asked: Does the advance of the 29°C isotherm play a role in initiating the monsoon? Moreover, is the poleward advance of warmer SSTs into the GOC a precondition for the poleward advance of OLR, or tropical convection? We have found that the poleward propagation of SSTs ≥ 29°C is as regular an event as the monsoon singularity, occurring to varying extents each year in July. As discussed below, this phenomenon appears to be driven by both solar insolation and a coastal current.
b. Coastal current and winds
Considerable evidence (Collins et al. 1997; Castro et al. 1994; Ripa 1997; Beier 1997) indicates that there is a poleward transport of tropical warm water along the mainland coast of Mexico up into the GOC during late spring and early summer. On the Baja Peninsula side, the flow is southerly and out of the gulf. The current is strongest in May (average flows up to 35 cm s−1), with highest velocities within 100 m of the surface. A current speed of 25 cm s−1 would transport water 670 km (416 miles) up the coast over one month, which, in combination with solar insolation, is plausible for explaining the observed northward advance of warmer water in May–June in Fig. 1. The GOC axis is about 1100 km long.
Douglas (1995) observed a persistent low-level jet flowing parallel to the GOC's axis, with maximum winds up to 15 m s−1 about 300 m above the surface. This may facilitate the “gulf surges” of southerly wind and associated heavy summer rains in southern Arizona (Hales 1972; Brenner 1974; Stensrud et al. 1997; Fuller and Stensrud 2000). Based on 11 yr of data for July, Douglas et al. (1993) demonstrate that low-level flow at 900 mb is southeasterly, emerging from south of 16°N, out of the eastern Pacific warm pool, with streamlines approximately parallel to the coast of mainland Mexico. These studies show that low-level monsoon moisture transport is from the southeast, with the GOC and tropical eastern Pacific as potential moisture sources. However, a strong “sea-breeze” effect is superimposed on this mean flow during afternoons (e.g., Douglas 1995; Douglas and Li 1996), drawing moisture inland over the deserts and the western slopes of the Sierra Madre Occidental. This results in a strong diurnal cycle of vigorous convection and rainfall, most active during late afternoon–early evening (Gourley et al. 1998; Douglas and Li 1996).
In Candela et al. (1984), based on 18 aircraft flights over the GOC from 14 July to 4 August 1983, horizontal winds at the surface over the GOC were 5.8 ± 2.4 m s−1 on average, most often from the southeast. GOC SSTs in our study exhibited little evidence of upwelling, since there was essentially no cross-gulf SST gradient. Such gradients would be expected if upwelling from winds roughly parallel up the GOC axis was producing cooler SSTs on the west side of the GOC. Moreover, observed (Ripa 1997) and modeled (Beier 1997) GOC circulation patterns are believed to be due to factors other than upwelling.
3. Empirical relations between SSTs and monsoon rainfall
The potential importance of the GOC as the low-level moisture source for monsoon rainfall suggests that changes in the SSTs of the gulf waters may be an important factor to consider in the interannual variability of the NA monsoon. To determine whether or not the SSTs may be important, an empirical study of six monsoon seasons was conducted to explore the relationship between the values of SST in the GOC region and the amounts of rainfall observed over the adjacent land regions.
Since GOC SSTs generally peak in August, the period of greatest interest here is from the first onset of monsoon rainfall to just after the highest SSTs are reached, referred to herein as the warming phase. Data from six years, 1993–97, and 1999, were examined to determine if there was any relationship between the values of SST and monsoon rainfall amounts. The 1999 analysis represents a special focus on northern GOC SSTs and their relation to Arizona–New Mexico rainfall, as gulf regions farther south were often obscured by clouds. Daily resolution SST data were used, allowing us to test findings from the 1993–97 analysis at higher resolution, and to demonstrate the utility of these findings in forecasting flood conditions.
a. Satellite data
The data consisted of satellite multichannel SST (MCSST) data at 18-km spatial and weekly mean temporal resolution, along with Special Sensor Microwave Imager (SSM/I) 5-day cumulative (pentad) precipitation data having a spatial resolution of 0.25° × 0.25°. The SST data used were derived from the NOAA AVHRR, which measures fluxes of emitted and reflected radiance from five channels in the visible, near-infrared, and thermal infrared (see McClain et al. 1985). The AVHRR instruments supplying the data were carried aboard the NOAA-11 and NOAA-14 satellites, which had a northbound equator crossing time of 1340 local standard time (LST). This data product is referred to as JPL PO.DAAC product 016, with SSTs having a global bias and rms difference of −0.1° and 0.5°C when compared with in situ buoy data. The noninterpolated data were used here, where only grid boxes corresponding to cloud-free regions are given (those obscured by clouds displayed no data). Conditions were frequently clear over the GOC from May to August during 1993–97 (i.e., most grid boxes displayed data), but on five occasions, mean weekly data were missing due to persistent and extensive cloud cover just south of the gulf's entrance. In such cases, the weekly SST value took the average of the two adjacent weeks, and a dashed line was used for the SST histogram (see appendix A). Otherwise, at least about 30% of the grid boxes displayed data within a given GOC region. When data density was near 30%, the active grids were dispersed fairly evenly over the region, allowing for realistic appraisals of SST. Overall, about 70% of the weekly images had an active grid coverage of ≳90%, based on the full region of study.
To gain confidence in the use of the weekly mean MCSST data, we compared it with daily SST data from the Geostationary Operational Environmental Satellite (GOES-10) imager. The GOES imager has channels at 0.52–0.72, 3.78–4.03, 10.2–11.2, and 11.5–12.5 μm. For a description of the GOES SST algorithm and capabilities, see Wu et al. (1999). While weekly SSTs are from NOAA polar-orbiting satellites, making one ascending (daytime) and one descending (nighttime) measurement per day, the geostationary GOES satellites take images every 15 min, with a spatial resolution of 4 km (compared to 1.1 km from polar-orbiting satellites). This study used hourly GOES data for the 1999 season. This allows coverage of the SST diurnal cycle, which can be up to 3 K over calmer seas (Wu et al. 1999). Figure 2 shows a comparison between GOES-10 SSTs observed every 3 h, and weekly mean MCSSTs (horizontal bars), for the northern GOC. Blank periods were obscured by clouds. Northern gulf images containing less than 40 pixels were not used. Note that diurnal variations can reach up to 4°C. MCSSTs were about 2°C higher than mean GOES values, and MCSSTs may be too high. This problem has recently been acknowledged at the Jet Propulsion Laboratory (JPL) and may be due to thermal contamination from land masses (J. Valdez 1999, personal communication). Therefore, a −2°C offset has been applied to the MCSSTs in Fig. 2. Over the SST range experienced in this study (24°–32°C), a −2°C offset appears adequate for correcting the MCSSTs. The GOES-derived SSTs have a number of advantages over SSTs from polar-orbiting satellites, especially when daily mean SSTs are desired and when clouds are a problem (Wu et al. 1999). The reduced cloud obstruction and the averaging of the diurnal cycle make GOES SSTs attractive for use as a practical standard. Since GOES SSTs have only been archived since 1998, we used adjusted MCSSTs for the period 1993–97.
Due to the low spatial resolution and questionable reliability of station precipitation data in northwest Mexico, satellite precipitation estimates were used to qualitatively assess the spatial distribution of rainfall over the monsoon region. The SSM/I on the polar-orbiting Defense Meteorological Satellite Program (DMSP) satellites provided these estimates. Given the localized nature of convective precipitation, satellite measurements appeared most appropriate. There were generally three DMSP satellites operational during this period, making two to six overpasses per day, which should capture the diurnal cycle of convective activity. The approximate timing for the three ascending overpasses ranged between 1735 and 2106 LST [i.e., mountain standard time (MST)], spanning a period during which monsoon convection is known to be active (Gourley et al. 1998; W. Berg 1998, personal communication), with at least one ascending pass per day over the monsoon region (15°–38°N, 100°–120°W). The SSM/I swath width is 1400 km, and the swath path is roughly parallel to the GOC axis. These rainfall estimates were only used on a relative basis.
b. Temporal and spatial evolution of SSTs and rainfall
This study divided the Gulf of California up into three regions for the evaluation of SST: northern gulf, central gulf, and southern gulf, with another “pregulf region” immediately south of the mouth of the gulf. There were also four regions for evaluating 5-day cumulative (i.e., pentad) rainfall amounts: Arizona–New Mexico (AZNM), the northern gulf, the central gulf, and the southern gulf. These regions are described in Fig. 3. Since late spring and summer winds are southerly (Douglas et al. 1993; Douglas 1995), the rainfall regions were offset one “region” north relative to the SST regions.
The temporal and spatial evolution of SSTs and rainfall for each season (1993–97) is described in appendix A, where GOC SSTs are related to mainland rainfall amounts. No significant rainfall amounts were observed in any of the land regions in Fig. 3 prior to an SST onset of 26°C in the GOC or pre-GOC region. Thus, 26°C is viewed as a threshold value above which SST increases may lead to convection and rainfall in the monsoon region. While this result is based solely on the analysis of this 5-yr dataset, the same SST–deep convection threshold of 26°C was reported by Zhang (1993) and Chaboureau et al. (1998) for tropical convective activity. Moreover, an analytical modeling study (McBride and Fraedrich 1995) has related the conditional instability of the second kind (CISK) theory to underlying SSTs, and demonstrated how convective instability may manifest through two growth modes: a slow and a fast mode. The transition from slow to fast convective growth occurs at an SST threshold of about 25.5°C for the parameters chosen in their study. Our observation that an SST > 26°C in the GOC is needed before monsoon-type rainfall commences in northwestern Mexico is consistent with these observational and theoretical studies.
c. Arizona–New Mexico rainfall dependence on northern gulf SSTs
The data in appendix A also indicate that there is a relationship between cumulative rainfall over the AZNM region and SSTs in the northern gulf region. As shown in Fig. 4, 69% of the rainfall amount in the AZNM region from June to August occurs after northern gulf SSTs exceed 29.5°C. About 80% of AZNM rainfall occurred after northern gulf SSTs exceeded 28.5°C, and 91% after SSTs exceeded 27.5°C. This analysis was based on all five seasons, and gives the normalized rainfall accumulated over the period when northern gulf SSTs were less than or equal to the indicated SST. Time is implicit in Fig. 4 from cooler to warmer SSTs. But rainfall amounts do not tell the whole story. For instance, if northern GOC SSTs reside near 30°C for relatively long periods, this will inflate the rainfall amount for SST > 29.5°C in Fig. 4. More relevant to a mechanistic understanding of the monsoon would be to divide rainfall amounts in each SST interval by the corresponding time period over which this rainfall occurred, giving rainfall rates for the AZNM region. This is done in Fig. 5. It is now evident that rainfall rate generally increases with northern GOC SST, apparently punctuated with transitions at 24°, 26°, 29°/29.5°, and 30.5°C. To know whether these transitions are physically meaningful will require more data. The main result here is that substantial rainfall rates begin when SSTs reach 26°C, and continue to generally increase with SST such that rates at SST = 31°C are 3.4 times greater than rates at SST = 26°C. These same trends have also been observed over the tropical oceans: 1) deep convection is rare when SST < 26°C, and 2) deep convection increases with SST between 26° and 30°C (Zhang 1993; Chaboureau et al. 1998). A modeling study by Stensrud et al. (1995) produced similar findings: When the high SSTs in the northern GOC (29.5°C) were replaced by SSTs characteristic of the Pacific west of the Baja Peninsula, water vapor mixing ratios that advected into Arizona from the northern gulf were insufficient to produce rainfall. The importance of low-level moisture to the development of convection over the southwestern United States has been emphasized in many studies (e.g., Bryson and Lowry 1955; Green and Sellers 1964; Adang and Gall 1989). Indeed, McCollum et al. (1995) conclude that moisture is often the most important missing ingredient for the development of thunderstorms in Arizona.
The physical mechanisms by which SST increases may lead to convective rainfall are not well understood, but it stands to reason that higher moisture levels resulting from higher SSTs are an important factor. The effect of increasing GOC SSTs by 1°C on precipitable water in the marine boundary layer, and the subsequent potential impact on latent heat release at the cloud base, is explored in appendix B.
1) Daily SST pentad rainfall analysis for 1999
Unlike the 1993–97 monsoon seasons, the 1999 season was anomalously wet, especially in Arizona, resulting in a rise in state reservoir levels and the coolest summer temperatures in Phoenix in 32 yr. The main purpose of this analysis is to test at higher time resolution whether heavier rainfall follows after northern GOC SSTs attain or exceed 29°C. This occurred for each of the five seasons analyzed (see appendix A).
Daily mean GOES-10 SSTs are plotted in Fig. 6 for the northern GOC, along with 5-day cumulative rainfall amounts (dashed) for the AZNM region. The GOES SSTs < 30°C tend to be about 0.5°C cooler than the weekly MCSSTs. Hence, the threshold for the northern gulf GOES SST for supporting heavier rainfall should be about 28.5°C. This SST was exceeded on 6 July. To interpret GOES SSTs more precisely, a 6-order polynomial fit was used to relate weekly MCSSTs to corresponding mean weekly GOES SSTs for the 1999 season. A GOES SST of 28.56°C translates to a MCSST of 29.0°C, and this threshold SST is indicated in Fig. 6 by the dotted line. Missing SSTs in Fig. 6 are due to cloud cover over the GOC.
During the evening of 6 July, the first heavy monsoon rains arrived in Arizona, as indicated by National Weather Service (NWS) station data and daily SSM/I images (not shown). SSM/I and IR satellite images for 5 July indicate rainfall was light at best, with some cloud cover over the eastern half of Arizona and western New Mexico. These findings are supported by 6-hourly upper-tropospheric water vapor (UTWV) GOES images and NWS station data for Tucson reported in Berg et al. (2000), given for 4–8 July 1999. On the evening of 6 July, rainfall was heavy throughout the southern half of Arizona, with very little or no rainfall over New Mexico. Over the next two days, as indicated by GOES UTWV images (Berg et al. 2000), this moisture moved farther west, resulting in a major flash flood in Las Vegas, Nevada, beginning around 1100 LST on 8 July. This flood caused $20,000,000 in property damage, and President Clinton declared Las Vegas a disaster area on 19 July (Haro et al. 1999). The arrival of rain in southern Arizona on 6 July is consistent with the fact that the “critical” SST was exceeded on that particular day, as well as expected moisture transport times from the northern gulf, which are on the order of 9 h for 10 m s−1 winds from the south. The 850-mb National Centers for Environmental Prediction's (NCEP's) streamline analysis indicates winds were south-southwesterly on 6 July, flowing into Arizona from the northern gulf, and an Eta Model forecast for 8 July shows southerly 600-mb winds at 20 m s−1 (Haro et al. 1999). Satellite imagery of UTWV and total precipitable water (Berg et al. 2000) indicate that moisture surges travel up the GOC axis and beyond at 500 ± 100 km day−1. For reference, the GOC axis is about 1100 km long. Our results in combination with Berg et al. suggest that the heavy monsoon rains beginning on 6 July depended on two factors: 1) a gulf surge event and 2) northern GOC SSTs.
This daily analysis of part of the 1999 season reinforces the previous analysis of the 1993–97 seasons, indicating that northern GOC SSTs around 29°C or higher are a prerequisite for sustained heavy rains in Arizona. But while the lag between the northern gulf reaching 29°C (in terms of MCSSTs) and monsoon rainfall was less than a day for 1999, a considerable lag may exist in other seasons, as shown in Table 1. Table 1 gives the AZNM precipitation lag time following the day the northern gulf reached or exceeded 29°C, based on histogram midpoints. Rainfall had to be 0.8 cm pentad−1 or more to terminate a lag period. In 1995, no SSM/I data were available during a 2–7-day interval after the SST threshold. The mean lag period including 1995 was 6.7 days, while the mean lag period excluding 1995 was 3.6 days.
2) Relative delay in Arizona monsoon onset
We now focus on the observation that the warming of the northern GOC lags behind that of gulf waters to the south by usually 2–3 weeks (SST > 26°C), and discuss what this may imply in terms of the monsoon onset in AZNM. Higgins et al. (1999) show that the mean calender date of monsoon onset systematically advances from south to north, consistent with an SST onset mechanism, presuming that warmer water incrementally advances northward up the coast during spring/early summer. We have plotted the Higgins et al. results as latitude of monsoon onset versus days beginning 1 June in Fig. 7. The rainfall dataset was based on the period 1963–88 using daily rainfall samples. Since a relation between northern GOC SSTs and rainfall is more likely for Arizona than for New Mexico (due to Arizona's proximity to the northern GOC), only Arizona was considered north of 31° latitude in Fig. 7. It is seen that the mean time lag for onset between the northern GOC and Arizona is 11.5 days, over twice the time lag relative to adjacent regions farther south. But once onset occurs in southern Arizona, there is relatively little delay in the northbound propagation of the monsoon. These observations are consistent with this study in that the northern GOC warming lagged behind the gulf regions to the south, and that the primary monsoon period in Arizona occurred after the northern GOC SSTs exceeded 29°C. Hence, it is tempting to view the Higgins et al. results as an advance of warm water up the coast, producing conditions favorable for convection as SSTs climb beyond a threshold value.
To partially test this hypothesis, the data from this study were evaluated in a similar manner. However, both rainfall and SST information were plotted with respect to time and latitude to look for relationships between the timing of northern GOC warming and rainfall. Shown in Fig. 8 are mean values (and standard deviations) for the times at which GOC SSTs exceeded 26°C and were ≥29.5°C (solid curves), based on the five seasons. Each point represents one of the GOC regions, identified by the latitude of the region's center. These SSTs were chosen since a prerequisite for convection appeared to be SST > 26°C, and the monsoon is well developed when SST ≥ 29.5°C. Due to the rapid increase in rainfall amount with SST following the first one-third of June–August rainfall (e.g., Fig. 4), we also plot the mean time (and standard deviations) at which 33% of the June–August rainfall total was attained or exceeded for each land region. This time is denoted R1/3 and is indicated by the dashed curve.
All individual years exhibit the general trend of 26°C water moving up the coast over time, indicating that any time past this threshold, rainfall may commence. The tendency for the 29.5°C curve was similar for all years, with the timing of these SSTs being similar in the pre-, southern, and central GOC, but delayed in the northern GOC. There is a similarity between the 29.5°C curves and the R1/3 curves, with the delay in the R1/3 curves corresponding to AZNM rather than the northern GOC (recall that prevailing winds are southerly). Moreover, the mean delay associated with 29.5°C SSTs in the northern GOC (relative to the central GOC) is similar to the mean delay for R1/3 in AZNM, being 18 days and 14 days, respectively. Relative to the mean 29.5°C SST for all gulf regions farther south, the mean 29.5°C delay time for the northern GOC is 13.5 days, almost matching the mean R1/3 delay time. This mean R1/3 delay time of 14 days is similar to the corresponding onset delay time of 11.5 days reported in Higgins et al. (1999), as shown in Fig. 7. These findings suggest that the usual relative delay for monsoon onset in Arizona, as observed in Fig. 7, may be attributed to a relative delay in the warming of the northern GOC. Finally, the temporal proximity of the R1/3 curve to the 29.5°C curve suggests that the timing of the indicated rainfall fraction is sensitive to the attainment of SSTs near 29.5°C.
If northern GOC SSTs are a critical factor determining moisture availability in the regions downwind (flows out of the northern GOC are typically toward the north or northeast), then a natural inference is that the warming of northern GOC SSTs may influence the western extent of the monsoon, since this body of water is farthest west. The results of this section support this idea. In addition, the results from Higgins et al. (1999) also show that there is no delay in the monsoon onset in western New Mexico, relative to regions farther south. While Arizona may be affected mostly by the northern GOC, New Mexico could be more affected by SSTs farther south.
d. Arizona monsoon season strength and northern gulf SSTs
June–August precipitation for Arizona statewide is shown in Fig. 9 including the period 1983–99. This period was selected since MCSST data go back to 1983, allowing us to investigate whether the timing of the 29°C MCSST threshold is related to the strength of a monsoon season in Arizona. From Fig. 9, the wettest June–August periods in Arizona were in 1984, 1986, 1988, 1990, 1992, and 1999. The seasons 1993–97 analyzed in detail for this study are seen to be relatively weak monsoon seasons based on June–August. For the two wettest periods, 1984 and 1999, northern GOC mean weekly SSTs (MCSSTs) are plotted from June to August, as shown by the dashed and dotted horizontal bars in Fig. 10. Also plotted are mean SSTs based on the five weaker monsoon seasons, 1993–97. These are shown by the thick horizontal bars, and the corresponding range in SST values are shown by the vertical bars. It is seen that for the two wettest years, the 29°C threshold is reached about a week earlier than for any of the weaker monsoon years, during the first half of July.
This analysis was repeated for the moderately wet June–August periods of 1986, 1988, 1990, and 1992, shown in Fig. 11. For 1986 and 1992, the 29°C threshold was exceeded in early July, over a week prior to any of the weak years, and the mid-July (12–18) SST for 1990 exceeds the weak monsoon SST range for this period. Note that the horizontal bar for 1990 for this period is superimposed on the 1992 pattern in Fig. 11. An exception to this trend is the 1988 season, which exhibits SSTs comparable to or less than the SSTs associated with the weaker monsoon seasons. Hence, five of the six relatively wet monsoon years, for which we have SST data, exhibit higher SSTs earlier in the season relative to the five weaker monsoon years studied. The fact that one year, 1988, did not exhibit this behavior, indicates that other factors are also important in determining how wet a monsoon season in Arizona will be. For instance, the meridional temperature gradient and thermal wind circulation, set up by diabatic heating (Barlow et al. 1998), may play an important role.
One can also ask, for the 17-yr record considered above, were there weaker monsoon years (relative to the six wet years considered) for which northern GOC SSTs were above the weak monsoon SST range (shown in Figs. 10 and 11) during the first half of July? The answer is yes: 1989 and 1998. In summary, we can say that 14 out of 17 years exhibited the following behavior for June–August Arizona rainfall: when northern GOC SSTs for some period during the first half of July were above the SST range for the drier monsoon years 1993–97, rainfall during June–August was relatively high. When northern GOC SSTs during the first half of July were within or below this SST range, June–August rainfall was normal or below normal. These results are consistent with the findings from Higgins et al. (1999), which show that early monsoon onsets in AZNM tend to be anomalously wet monsoon seasons.
e. Arizona and New Mexico: Different factors influencing monsoon rainfall?
This section addresses the question: Are wet monsoon seasons for Arizona also wet for New Mexico? To answer this, we evaluated statewide precipitation for AZ and NM for the years 1950–99, for the period July–September. Wet AZ seasons were considered to be those exceeding 6 in. of rain, and wet NM seasons to be those attaining or exceeding 8 in. of rain (roughly one standard deviation above the mean). Using this dataset, standardized precipitation anomalies are plotted over the United States for the AZ wet years in Fig. 12 and for NM wet years in Fig. 13 (using Web software at the NOAA Climate Diagnostics Center). The color scale is in units of std dev, where std dev is relative to the 1950–95 long-term mean.
In Fig. 12, it is seen that in wet AZ years, wet conditions are extensive throughout the Southwest, extending into Utah, Nevada, California, and western NM, while eastern NM is slightly drier than normal. In Fig. 13, in wet NM years, the monsoon season is near normal in AZ, with anomalously wet conditions confined to NM and parts of Colorado and Texas. Hence, a strong monsoon year for NM will generally not benefit other parts of the desert Southwest.
We have repeated this analysis for dry summers as well, defining dry monsoon seasons in AZ as less than 4 in. of rainfall and dry seasons in NM as less than 6 in. of rainfall. For the driest NM years, AZ was slightly drier but near normal, and for the driest AZ years, NM was moderately dry in the west, and slightly wetter than normal in the east.
It is also of interest that the wettest summers in AZ (Fig. 12) correspond to the driest summers in the Midwest. However, the same is not true when NM is wettest (Fig. 13). This suggests that the anticorrelation between desert Southwest and Midwestern summer precipitation observed by Higgins et al. (1997, 1998) and Barlow et al. (1998) may primarily result from conditions that produce wet monsoons in AZ. This study suggests that one of these conditions may be relatively high SSTs in the northern GOC during the first half of July. Factors affecting the monsoon moisture flow into AZ and NM include the location of the subtropical high pressure ridge axis over the western United States, and topography, where the Gulf of Mexico may also be a moisture source. It is also of interest that when NM is wettest (Fig. 13), Idaho and surrounding areas are anomalously dry. The ability to forecast wet conditions in AZ and NM may also make it possible to forecast anomalously dry conditions in the Midwest and Northwest, respectively. Based on all the results here, we postulate that when the atmospheric circulation favors wet monsoon conditions in AZ, another factor determining whether heavy monsoon rains will commence in AZ and whether drought is likely in the Midwest is the northern GOC SST.
In summary, though both AZ and NM are affected by the monsoon, the means by which they are affected appears to be different. We suggest that a principal moisture source for wet AZ monsoon seasons is the northern GOC. This moisture source may be less important for wet NM seasons. Also, the monsoon begins in western NM about a week earlier on average than in western AZ (Higgins et al. 1999). If the NM monsoon moisture comes from the Pacific (which satellite imagery suggest), it apparently comes from lower regions of the GOC or the eastern Pacific off southern Mexico. In these regions, SSTs warm earlier than in the northern GOC, and may explain the earlier onset times in NM. Of course, this idea needs testing and further study.
4. Discussion: Northern gulf SSTs
Knowledge of the causes behind the delay in the warming of the northern gulf waters, relative to gulf waters farther south, may have relevance to the ultimate predictability of the NA monsoon. A potentially strong influence on northern GOC SSTs are the islands, or archipelago region, which separate the northern and central GOC (see Fig. 3). The archipelago has the impact of 1) mixing the water column and 2) presenting an obstacle to the northward flow of warm water up the GOC. Both factors could delay the warming of the northern GOC. Regarding 1), Paden et al. (1991) present findings that tidal mixing occurs over the relatively shallow sills in the island region, especially in mid- to late spring, affecting the upper 300–500 m. Tidal mixing over such depths pumps heat away from the surface, deep into the water column, hence cooling SSTs in the archipelago by 2°–4°C. The same degree of SST cooling in the archipelago was found in this study. Simpson et al. (1994) show that, in addition to tidal or vertical mixing, advection over the sills in the archipelago must occur to explain the observed temperature reductions. Paden et al. found that this pool of relatively cool water in the archipelago was advected northward over much of the northern GOC, while SSTs south of the archipelago remained warmer owing to advection of water from the southern GOC, which was not influenced by tidal mixing.
Eventually, northern GOC SSTs become as warm or warmer as gulf SSTs farther south, typically by late July to early August for the seasons studied here. As suggested by Paden et al. (1991), increased solar insolation leads to increased stratification of the water column (due to warmer SSTs near the surface), resulting in less mixing. With less mixing, SSTs in the northern GOC climb rapidly to their peak values, as high as 32°C. However, the maximum net energy flux into the GOC through the surface, mostly due to solar insolation, occurs during June (Castro et al. 1994), around 10 June (Ripa 1997). Yet the SST differences between the northern and central GOC remain, until they are minimized about one to two months after 10 June. This suggests that during the one to two months following 10 June, stratification of the water column is being inhibited by some other process. It is conjectured here that this other process is the advection of the upper layer through the archipelago. Once advection decreases, mixing and associated cooling should decrease (Simpson et al. 1994), allowing northern GOC SSTs to become comparable to or to exceed those farther south. This would produce favorable conditions for heavy rainfall in the AZNM region, assuming northern GOC SSTs were ≥29°C.
Evidence for this conjecture is found in Ripa (1997) and Beier (1997), which indicate surface current velocities through the island region peak about 28 June, with reduced advection thereafter possibly promoting increased stratification of the water column in the northern GOC. Using sea level measurements in the central GOC, Ripa (1997) estimated the annual harmonic of the geostrophic surface velocity (top 70 m), usfc. Peak values of usfc (toward the gulf's tip) occur around 28 June, while peak sea levels at Guaymas (central GOC, mainland coast) and Santa Rosalia (opposite Guaymas on Baja coast) occur 9 and 27 August, respectively. After late September, usfc reverses, with net flow toward the gulf entrance. This reversal coincides with the reversal in GOC surface currents modeled by Beier (1997), from cyclonic in summer to anticyclonic in winter. Using a two-dimensional linear two-layer model initialized with the same hydrographic observations used by Ripa (1997), Beier found the annual cycle of an internal wave, initiated in the mouth of the gulf as a baroclinic Kelvin wave by the Pacific Ocean, was primarily responsible for both sea level changes and upper layer (top 70 m) currents in the GOC, although wind stress was also important. Since the annual cycle of the internal wave and wind direction were in phase, their effects were additive. While upper-current velocities reached 70 cm s−1, current velocities of 30–40 cm s−1 were more typical. Such velocities appear sufficient to induce mixing in the relatively shallow archipelago and to cool SSTs in the northern GOC. Moreover, current velocity reductions occurring after June may also affect northern GOC SSTs by reducing both mixing over the sills and advection to the northern GOC, thus allowing northern GOC SSTs to increase by responding more directly to solar insolation.
Regarding factor 2), GOC circulations are strongest and cyclonic during spring/summer (Bray 1988; Merrifield and Winant 1989; Paden et al. 1991; Castro et al. 1994; Collins et al. 1997; Ripa 1997), and can be explained by means of an internal wave trapped against the coast, propagating up the mainland coast, and back down the Baja California coast (Beier 1997). The GOC circulations predicted by Beier (1997) indicate two cyclonic gyres in summer, separated by the archipelago that, in effect, blocks some of the warm water advected up the mainland coast from entering the northern GOC. Hence, horizontal heat advection is expected to contribute less to SSTs in the northern GOC than in other gulf regions (see also Figs. 8 and 12 in Ripa 1997).
In summary, a working hypothesis is proposed here as a possible aid for future research. It postulates that AZ rainfall is strongly dependent on northern GOC SSTs, and that tidal mixing and mixing from horizontal advection over the sills in the archipelago are largely responsible for the relatively cooler SSTs in the northern GOC during much of the warming phase. Partial blocking by the sills of the poleward advection of warm water may also be a factor. These factors, by producing relatively cooler northern GOC SSTs, may be responsible for the relative delay observed for the monsoon onset in AZ. The eventual increase of northern GOC SSTs may be due to a decrease in northward advection of water through the archipelago, allowing the water column to stratify and respond more directly to solar insolation. The complex interactions of these and other factors will need to be addressed in more detail via observational and modeling studies before confident predictions of northern GOC SSTs can be made, which might ultimately be used to help predict the timing and strength of the NA monsoon in AZ.
Interrelationships between SSTs in the Gulf of California (GOC) and rainfall amounts in adjacent land regions were evaluated for six monsoon seasons (June–August), with one season (1999) addressing only SSTs in the northern gulf due to persistent cloud cover farther south. This was made possible by using high temporal and spatial resolution satellite data. These results lend considerable support to the hypothesis that GOC SSTs play an important role in determining the timing, rainfall amount, and northwestward extent of the Mexican or North American (NA) monsoon. The major findings from this study are as follows.
A 20-yr mean seasonal cycle, from April to July, of SSTs and OLR was evaluated, highlighting a developing warm pool off southern Mexico and Central America in spring, when solar insolation is high. Relatively warm (e.g., 29°C) water advances far up the Mexican west coast into the GOC during July, coinciding with a similar extreme northward advance in deep convection (i.e., OLR < 240 W m−2). A five season intensive observation period revealed that convection did not occur over northwestern Mexico until corresponding GOC SSTs exceeded 26°C, implicating the advance of warm water up the coast as a precondition for monsoon onset.
For the period of June–August, 69% of the rainfall in the Arizona–New Mexico region occurred after SSTs in the northern GOC exceeded 29.5°C. Regionally averaged rainfall rates when northern GOC SSTs exceeded 30°C were about 3 times greater than rates at 26°C, with significant rainfall amounts and rates beginning at 26°C.
A relative climatological delay in the monsoon onset for Arizona, as observed by Higgins et al. (1999) via 26 yr of daily precipitation data, appears to result from a delay in the warming of northern gulf waters relative to gulf waters farther south.
For five of the six seasons studied, relatively heavy rains occurred within 0–7 days following northern gulf SSTs reaching or exceeding 29°C. For the strong monsoon year of 1999, the timing of daily northern gulf SSTs exceeding 29°C coincided with the timing of the first heavy monsoon rains in Arizona and a major flood in Las Vegas two days later.
Based on the 17 yr of available SST data, 14 of those years exhibited the following behavior: When northern gulf SSTs during the first half of July exceeded the northern gulf SST range of the relatively dry years 1993–97 (regarding AZNM), rainfall during June–August in Arizona was relatively high (one exception out of six). Otherwise, June–August Arizona rainfall was normal or below normal.
Anomalously wet July–September periods in Arizona do not correspond to anomalously wet periods in New Mexico, based on 1950–99. The wettest Arizona seasons, about one standard deviation wetter than normal, were strongly related to summer drought in the Midwest, being about 0.8 standard deviations drier than normal. This was not true for the wettest New Mexico seasons, but these periods exhibited dry conditions in the interior northwest, with standard deviations being about 0.6–0.9 drier than normal. The cause of these two wet monsoon modes may be related to SSTs in the northern gulf, which appear to affect Arizona more than New Mexico rainfall.
In conclusion, the factors governing the timing, intensity, and extent of the NA monsoon may be oceanographic in nature as much as they are atmospheric, and greater “cross fertilization” between the fields of oceanography and the atmospheric sciences will likely provide the key to optimizing progress toward a mechanistic understanding of the NA monsoon.
The authors gratefully acknowledge the Desert Research Institute in Reno, Nevada, for providing funds for this research effort. We also wish to thank our past and present division directors Dr. Peter Barber and Dr. Kent Hoekman for their constant support of this research. Additional support was provided by the Environmental Sciences Division, U.S. Department of Energy, Atmospheric Radiation and Measurement (ARM) Program. The GOES SST processing was courtesy of Dr. Xiangqian Wu at the Cooperative Institute for Meteorological Satellite Studies (CIMSS), University of Wisconsin—Madison. Fruitful conversations with Dr. Steven Chai are gratefully acknowledged.
Detailed SST–Rainfall Analysis
The results for each year are given in Figs. A1–A5. Mean weekly SST variations are shown for each ocean region in the left column, with land regions likely to be affected by SST changes shown in the right column (due to moisture from southerly or onshore flow). Rainfall amounts per pentad (in cm) are shown for these land regions during the SST warming phase [i.e., the first rainfall jump above “background” (≲1.0 cm) to the time about 10 days after SSTs level off].
To aid data interpretation, the threshold SST value of 26°C is shown by the dashed line in the SST plots in Figs. A1–A5. SST increases are labeled chronologically as 1, 2, 3, etc., followed by A, B, C, or D, corresponding to the pre-, southern, central, or northern gulf regions, respectively. Rainfall peaks possibly associated with specific SST increases are labeled accordingly. Since SST jumps ≤0.4°C appeared unrelated to rainfall increases, only SST jumps >0.4°C were labeled and related to rainfall increases. For the 1993–95 seasons, any rainfall increase occurring within 5–17 days after such an SST increase was associated with that SST increase, and was labeled accordingly. The lag period for the 1996–97 seasons was 0–15 days, based on histogram midpoints. These lag periods likely depend on wind and other conditions, and may be a function of the moisture surge events known as gulf surges (e.g., Hales 1972; Stensrud et al. 1997) or other phenomena that periodically transport boundary layer moisture poleward. Given prevailing wind directions, rainfall increases were never attributed to an SST increase north of the rainfall region.
For all five years, there is a general tendency for rainfall to increase after an SST increase, provided that SST > 26°C. Moreover, the rainfall amount tends to be related to the product of the magnitude of the SST jump times the difference (SST − 26°C). This is most apparent in the two southernmost regions for SST and for rainfall. Albeit subjective, rainfall peaks were related to SST increases to facilitate this trend analysis.
Except for the 1996 season, no tropical depressions were evident north of about 20° latitude (based on SSM/I and GOES infrared satellite images) during the warming phase, and hence, were not responsible for rainfall during those periods. Three tropical depressions were identified after the warming phase, and are labeled “TD” in Figs. A2, A3, and A4.
Possible examples of northward-propagating rainfall events are shown by peak 1A in Fig. A1; peak 3A/3B in Fig. A3; the TD peak in Fig. A3; three peaks associated with 3B, 4B, 5B/6B in Fig. A4; and three peaks associated with 1A, 6B, and 5A/7B in Fig. A5. The data are consistent with other observations that indicate gulf surge events and associated convection often originate in the pre–gulf region or eastern Pacific (Stensrud et al. 1997; Fuller and Stensrud 2000), as far south as 10°N latitude (Berg et al. 2000).
Particularly interesting is the rainfall period associated with 1A in Fig. A5. Although significant monsoon rainfall in any of the four regions of Fig. 3 does not normally start until June or July, monsoonlike rains occurred from about 7 to 25 May in all four regions in 1997. This event coincided with the timing of GOC SSTs exceeding 26°C, exhibiting an increase of about 2°C. There was no rainfall indicated for these regions during the previous month. Historical GOES satellite imagery shows strong convection in these four regions with upper-level winds apparently from the southwest or south, and indicated this rainy period, lasting at least 13 days, was not due to any organized weather system. Another example of early rainfall occurred in 1994 in association with SST increase 1A (exceeding 26°C), with rain beginning around 22 May in AZNM. Inspection of historical GOES satellite images reveals that a cutoff low, with little associated cloud cover, was drawing moisture off the southern gulf and possibly the pre–gulf regions, circulating it into AZNM, where it then organized into convective complexes.
Some of these northward-propagating convective complexes may be due to gulf surge events, which often appear to be triggered by the passage of easterly waves off western Mexico (Fuller and Stensrud 2000). These surges appear to dramatically deepen the boundary layer, transporting moisture up the GOC into Arizona, and even as far north as Lake Tahoe (Berg et al. 2000). A deep boundary layer will allow GOC SSTs to increase humidities throughout this layer. The periodicity of gulf surges, being about every 10 days on average, but quite variable (Fuller and Stensrud 2000), may at least partially account for the lag periods apparent in the data here. However, while a gulf surge may last 3–4 days, the rainy period described above (beginning about 7 May 1997) lasted 13 days or more. Gulf surge events, and apparently other processes, may work in conjunction with GOC SST increases to free up boundary layer moisture for convection over land.
It is noteworthy that the heaviest periods of rainfall in AZNM occur after SSTs in the northern gulf attain about 29°C or higher, as discussed in section 3c. This was observed for all five seasons.
Sensitivity of Latent Heat Release to GOC SST Changes
It may be the sensitivity of convection to relatively high moisture levels that best explain Figs. 4–5 and other results in this paper. This is illustrated in Fig. B1, which describes results from an idealized situation. We assumed an SST jump of 1°C took place under a boundary layer 300 m deep. The relative humidity (RH) of the boundary layer air was assumed to be 80% relative to the saturation vapor density at equilibrium with the SST. Using the Clausius–Clapeyron equation, the increase in precipitable water (ΔPW) in the boundary layer for a 1°C SST jump was calculated over a range of initial SST values, ranging from 21° to 31°C. The SSTs in Fig. B1 are the initial SSTs prior to a 1°C increase. The PW increase (ΔPW) are given by the long-dashed curve. The total boundary layer PW (based on initial SSTs) is described by the short-dashed curve. The boundary layer air is assumed to arrive at cloud base undiluted. By equating the change in latent heat via the SST-induced vapor density change with the sensible heat increase of the air, the change in temperature near cloud base due to latent heat release can be estimated:
where ρυs = water saturation vapor density, 2 and 1 denote after and prior to SST jump, Lυ = latent heat of vaporization, ρa = air density at cloud base (assumed 700-mb pressure; 10°C), and cpa = specific heat of dry air at constant pressure. Here ΔT is shown by the solid curve in Fig. B1. Note that ΔT results only from ΔPW (i.e., a given SST jump), and not the cumulative increase in total PW. Hence, the fractional increase of ΔPW and ΔT with increasing SST are identical.
Figure B1 suggests that the latent heat released due to a 1°C SST jump may be significant to convection by raising cloud-base temperature up to several degrees. It is possible that a delicate balance exists in regards to factors determining convection, such that a modest change in ΔT could produce conditions more favorable for convection. Such reasoning appears consistent with the concept of a critical SST for rapid convection growth, as found from observations (Zhang 1993; Chaboureau et al. 1998) and theory (McBride and Fraedrich 1995). The ΔT values in Fig. B1 are only rough estimates, with entrainment and smaller SST jumps producing lower estimates. A more realistic and comprehensive treatment of this issue is desirable but is beyond the scope of this study.
Corresponding author address: Dr. David L. Mitchell, Atmospheric Sciences Division, Desert Research Institute, 2215 Raggio Parkway, Reno, NV 89512-1095. Email: firstname.lastname@example.org