The North American monsoon is accompanied by large-scale changes in circulation and precipitation over much of Mexico and the United States during summer. Here, the influence of the North American monsoon is analyzed in terms of midlevel changes to the thermodynamic energy equation, circulation, and precipitation associated with the monsoon onset in northwest Mexico, for the 1948–2004 period. In addition to the well-known strong increase in rising motion over the core region of the monsoon during the onset, there is also a decrease in upward motion over the northern Baja California Peninsula and into the southwest United States, directly in the path of monsoon development. This area of decreased vertical motion is linked to cold advection caused by the onset itself, as the Gill-Matsuno response to the monsoon precipitation thermodynamically interacts with the mean circulation. It is in this sense that we propose that the monsoon is self-limiting.
 The North American monsoon (NAM) is characterized by an abrupt increase in precipitation over northwestern Mexico, which then spreads northwestward into the southwestern United States [Douglas et al., 1993; Stensrud et al., 1995; Mock, 1996; Adams and Comrie, 1997; Higgins et al., 1997], and also a seasonal reversal of the winds both at lower levels [Carleton, 1986; Douglas, 1995; Bordoni et al., 2004] and upper levels [Douglas et al., 1993; Stensrud et al., 1995; Higgins et al., 1997]. The summertime circulation centered over northwest Mexico has been identified to have monsoonal characteristics, including a pronounced increase in rainfall from an extremely dry May to a rainy June over large areas of the southwestern United States and northwestern Mexico, along with a surface low pressure and a strong upper level anticyclone accompanying the onset of precipitation (see discussion in Barlow et al. , Higgins and Shi , Higgins et al. , and North American Monsoon Experiment (NAME) Science Working Group ). More than half of the annual rainfall across the southwestern U.S. and northwestern Mexico occurs during the monsoon season [Carleton et al., 1990; Douglas et al., 1993; Higgins et al., 1997; Mitchell et al., 2002; Sheppard et al., 2002]. While the NAM is much smaller than its Asian counterpart, it still has a large impact on the circulation of United States and Mexico [e.g., Okabe, 1995; Barlow et al., 1998; Higgins et al., 1999], although the dynamics of this influence are not yet well understood. Here, we examine the regional influence of the NAM over the U.S. and northwest Mexico in terms of thermodynamic analysis of observed data, focusing on the changes during the onset to emphasize the direct links to the core NAM precipitation. This allows an investigation of the thermodynamic interaction of the NAM onset circulation with the mean circulation, and the possibility that this interaction results in subsidence directly in the path of further development, providing a self-limiting mechanism.
 There have been a few studies looking at the regional dynamics of the NAM, including some consideration of orography, diabatic heating, convective environment, and land-sea thermal contrast. Rodwell and Hoskins  used a dry primitive equation model to show that the diabatic heating associated with the NAM in combination with orography induces descent over the eastern North Pacific, so the mechanical influence of orography is important in addition to the thermal forcing in terms of regional dynamics. Another dynamically oriented study, Barlow et al. , shows the existence of a tropical-type thermodynamic balance between diabatic heating and adiabatic cooling in the monsoon region, switching toward a more midlatitude balance between diabatic heating and horizontal temperature advection north of 35°. Land-sea thermal contrast plays an important role in the interannual variability of the NAM onset [Zhu et al., 2005; Turrent and Cavazos, 2009; Turrent and Cavazos, 2012]. Turrent and Cavazos  have shown a dynamic mechanism that links the land-sea thermal contrast to the intensity of precipitation at onset over northwest Mexico.
 Here we consider observational analysis of the changes during the onset in northwest Mexico to focus on regional changes directly linked to the continental monsoon precipitation. The terms of the thermodynamic energy equation are calculated at midlevels to provide information on vertical velocity, which is critical to precipitation development and which is expected to change in response to the monsoon circulation and provide a mechanism by which the monsoon affects its regional environment [Rodwell and Hoskins, 1996; Saini et al., 2011]. The thermodynamic balance is also analyzed for strong and weak advection years and for heavy and light SW U.S. precipitation years to further test the role of temperature advection in the development of the monsoon. The data sets and methodology used in this study are described in the next section, which also contains a brief description of the onset index used in the study.
2 Data and Methodology
2.1 Onset Index
 The onset of the NAM is characterized by an increase in rainfall over southern Mexico in June, which quickly progresses northward along the western slopes of the Sierra Madre Occidental in northwestern Mexico and then spreads into southwestern United States by late June and early July [Douglas et al., 1993; Stensrud et al., 1995; Higgins et al., 1999; Higgins et al., 2003; Gochis et al., 2004; NAME Science Working Group, 2004]. The spatial extent of the NAM region within northwestern Mexico was studied by Higgins et al.  through examination of precipitation patterns created from a distribution of daily precipitation stations across the region. Here we define the date of onset in northwest Mexico (26°N–30°N, 112.5°W–105°W) using their selection criteria as the first day of the 5 day period when the precipitation is at least 1.0 mm day−1 using daily gridded precipitation data over the U.S. and Mexico for a period of 57 years from 1948 to 2004 with a horizontal resolution of 1° on a regular latitude-longitude grid [Higgins et al., 1996]. The onset date for all the years was defined as the first day after May 1 when the above threshold criterion was met. Table S1 in the supporting information shows the calendar date of monsoon onset over northwestern Mexico for each year based on the precipitation index.
 The reanalysis data from the National Centers for Environmental Prediction (NCEP)-National Center for Atmospheric Research (NCAR) [Kalnay et al., 1996] are used as the primary basis for calculating the terms of the thermodynamic equation and analyzing atmospheric circulation data, with the results verified in the NASA Modern-Era Retrospective Analysis for Research and Applications (MERRA) reanalysis data [Rienecker et al., 2011]. Daily-averaged (0000 and 1200 UTC) data with a horizontal resolution of 2.5° on a regular latitude-longitude grid for the period 1948–2004 are used. Previous work suggests that analysis of the thermodynamic equation is not overly sensitive to the data set used [Saini et al., 2011]. The thermodynamic equation is calculated at 500 hPa as follows [Holton, 2004]:
where T is the temperature, V is the horizontal wind vector, Sp is the static stability parameter, ω is the pressure vertical velocity, cp is the specific heat of dry air, and J is the diabatic heating. Static stability is proportional to the vertical gradient of potential temperature, so Spω represents the vertical advection of temperature, accounting for adiabatic changes. The diabatic heating term is calculated as a residual from the thermodynamic equation as in Saini et al. .
 The onset changes are considered here as the difference between the 15 day average after the onset (day 1 to day +15) and the 15 day average before the onset (day −15 to day −1). This 31 day period centered on the onset day is long enough for the regional circulation response to develop but short enough to isolate the onset signal as distinct from the more general seasonal evolution. The results are not sensitive to the exact length of the averaging period.
3 Results and Discussion
 The changes in precipitation and upper level (200 hPa) winds associated with the monsoon onset for 1948–2004 are shown in Figure 1, where the changes are the difference between the 15 day average after the onset and the 15 day average before the onset. Increased precipitation is observed throughout the monsoon region at the time of onset with maximum increase in northwest Mexico. Coinciding with the precipitation increase at the onset is a decrease in precipitation toward north and east. This out-of-phase relationship at monsoon onset between rainfall over northwest Mexico and southwest U.S. and decrease in rainfall over central U.S. has been well documented [Higgins et al., 1997, 1998, 1999; Mo et al., 1997; Higgins and Shi, 2000; Barlow et al., 1998] although the mechanism is not clear. Associated with the monsoon onset precipitation is an upper level (200 hPa) anticyclone that is centered considerably to the north of the onset precipitation. This change in the circulation represents the northward development of the Mexican High [Wang, 2006]. Similar to the Indian monsoon, the circulation is anticyclonic at upper levels and cyclonic at lower levels [Rodwell and Hoskins, 2001; Gill, 1980; Kershaw, 1985; Cadet, 1979; Hsu et al., 1999], although the longitudinal span of the anticyclone is much less than that of its Indian counterpart.
 The changes in the terms of the thermodynamic energy equation at 500 hPa during the onset are shown in Figure 2 (the averages of the 15 days before and 15 days after the onset are shown separately in Figure S1) for NCEP-NCAR (left panels) and MERRA reanalysis (right panels). The changes in the temperature tendency term (not shown) are negligible with respect to the other terms at this contour interval. The static stability is approximately constant, so the changes in the vertical velocity term are due primarily to changes in vertical velocity (Figures 2b and 2e), which are balanced by changes in temperature advection (Figures 2a and 2d) and diabatic heating (Figures 2c and 2f). The tropical balance is largely between diabatic heating and rising motion, with horizontal temperature advection becoming an important factor around 25°N. The intense downward motion at higher latitudes (Figure 2b) is in balance with diabatic cooling in those areas with small contributions from cold advection (Figure 2c). The MERRA reanalysis data verify that, while there are some differences in small-scale structure and in the vigor of vertical velocity and diabatic heating, the overall patterns and relationships are quite similar.
 The monsoon region in Figure 2b shows vigorous rising motion over the core precipitation area, consistent with previous work [Barlow et al., 1998]. Interestingly, however, there is actually a decrease in vertical motion immediately to the northwest of the onset area, directly in the path of the monsoon development. In the context of the Indian monsoon, Rodwell and Hoskins  suggested that the response to deep heating results in forced descent to the northwest of the monsoon heating over northern Africa and the Middle East, which was confirmed in observational analysis of the Indian monsoon onset by Saini et al. . For the NAM, there is also a region of subsidence to the northwest of the monsoon, but given the much smaller scales, the subsidence is much more local. This difference in spatial scales means that the same overall dynamical mechanism has very different implications for the two monsoons: in the case of the Indian monsoon, the subsidence to the northwest is sufficiently remote not to directly affect the monsoon development, but for the smaller scale of the NAM, the subsidence is directly in the path of the monsoon development.
 To further explore the changes in temperature advection, it is divided in Figure 3 into the contribution from advection of the temperature anomalies (temperature change during onset) by the mean wind and the advection of the mean temperature by the wind anomalies (wind change during onset). The midlevel warm temperature anomalies, shown in Figure 3a, are spatially coherent with the upper level anticyclone in Figure 1, as expected from the thermal wind balance. In the middle troposphere, the temperature anomalies lie within the westerly jet, thus resulting in cold advection to its west and warm advection to its east (Figure 3b) as in Rodwell and Hoskins'  “monsoon-desert” hypothesis but at a much smaller scale. The midlevel anomalous wind (Figure 3c) also spatially follows the anomalous temperature (Figure 3a) and the upper level anticyclone (Figure 1). In a similar study of the Indian monsoon done by Saini et al. , the upper level anticyclone does not extend down to midlevels as here, indicating that the monsoons have a somewhat different vertical structure in terms of the height at which the circulation switches from anticyclonic to cyclonic. As a result, the horizontal temperature advection also has significant contribution at midlevels from the advection of mean temperature by the anomalous wind (Figure 3d), but with opposite sign as that of Figure 3b.
 To further study the influence of temperature advection and associated changes to vertical motion on the northward progression of the monsoon, we compare the onsets during years of strong midlevel cold advection over northwest Mexico with the onsets during years of weak cold advection. To do this, we have developed an index defined based on horizontal temperature advection for the 15 day period before the onset (day −15 to day −1) at 500 hPa averaged over the monsoon region (25°N–37.5°N, 115°W–105°W). We use this index as a criterion to differentiate between strong and weak advection years and then consider the third strongest (18 years) and weakest (18 years) advection years for our analysis. The onset changes in terms of thermodynamic equation are compared between the averages of 18 strong (Figure 4, left panels) and 18 weak (Figure 4, right panels) advection years. During strong advection years (Figure 4a), there are strong decreases in vertical motion (Figure 4b) to the northwest of the monsoon, and the monsoon onset, while vigorous, does not extend beyond 36°N (Figure 4c). During weak advection years (Figure 4d), there are strong increases in vertical motion (Figure 4e) in the monsoon region and to the northwest of the monsoon, and the monsoon onset is somewhat weaker (Figure 4d) but extends considerably farther to the north (Figure 4f). We have verified that these differences in diabatic heating are representative of the actual changes in precipitation (Figure S2). This comparison of onsets during conditions of strong and weak local temperature advection confirms that the temperature advection is closely related to the northward development of the monsoon. We have also stratified the years based on strong and weak precipitation onset changes over southwest U.S. (120°W–100°W and 32°N–45°N) and studied the change in thermodynamic balance. The thermodynamic terms for the 18 strongest and weakest precipitation years in that region are shown in Figure S3, and the precipitation changes are shown in Figure S4. The “weakest” years, when the monsoon has its least northward extent, are indeed years with very strong cold advection over the northwesternmost part of Mexico and a strong area of subsidence anomalies to the northwest of the core monsoon, whereas the “strongest” years, when the monsoon has its greatest northward extent, actually have switched to warm advection to the northwest of the monsoon and uninterrupted rising motion all the way to 40°N (compare Figure S3 with Figure 2 of the paper). This is another supporting piece of evidence for our hypothesis that temperature advection is closely linked to the northern development of the monsoon.
 Note that since the temperature advection is directly linked to the monsoon itself, the strength of the onset is related to the strength of the temperature advection, so it is not possible to consider cases of strong monsoons without the influence of temperature advection because the mean wind and temperature gradient are almost never displaced so far to the north as to not result in advection. However, it would be expected that onsets when the mean wind is relatively weak would extend farther to the north, and we have verified that this is indeed the case.
 On further examining the strong and weak advection years and comparing them to early and late onset years, we found that there is a moderate relationship between strength of advection during onset and the timing of the onset. Of the 18 weak advection onset years, 10 are late and 4 are early onsets; of the 18 strong advection onset years, 8 are early and 4 are late onsets. The study on interannual variability of NAM by Higgins and Shi  showed a stronger relationship between early and wet monsoons and a weaker relationship between late and dry monsoons. This interannual variability of the NAM and the relationship between the strength and the timing of monsoon onset needs to be further explored.
 For the NAM, the thermodynamic interaction of the Gill-Matsuno response to the diabatic heating of the monsoon onset with the westerly jet results in a region of cold advection (descent) just north and west of monsoon onset and a region of warm advection (ascent) farther to the northwest of the monsoon heating. These cold and warm advection areas are a result of both the advection of the onset temperature changes by the mean wind and the advection of mean temperature by the onset wind changes. In both cases, the cold advection and associated descent just to the northwest of the core onset region are a direct response to the monsoon onset; that is, the thermodynamics response to the onset, itself, results in unfavorable conditions (midlevel subsidence) directly in the path of the monsoon development. It is in this sense that we suggest that the NAM is self-limiting—the more vigorous the precipitation, the more subsidence will be forced in the way of its northward development. This is in contrast to the Indian monsoon, where the much larger scale of the response produces forced subsidence remote from the monsoon development. To further test the proposed mechanism, we have compared onsets with weak and strong advection as well as onsets with greater and lesser development into the southwest U.S.: both perspectives show a close relationship between temperature advection and northward development of the monsoon. The current analysis has focused on the onset to consider the direct response to monsoon precipitation. An important next step will be to consider the role of thermodynamic negative feedback in seasonal variability.
 This research effort was supported by NOAA grant NA08OAR4310592 and NSF grant ATM-0621737.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this manuscript.