1. Introduction
The North American monsoon is a meteorological phenomenon that occurs during the summer months in the American Southwest. In this normally arid region, the monsoon is characterized by above-normal precipitation, relative to the annual mean. Thus, this area receives the bulk of its annual precipitation during the warm season. The strength of the monsoon varies from year to year, which yields interannual variability in the yearly precipitation totals, sometimes promoting a long-term deficit in water, or drought.
Over the last century, a variety of theoretical papers have attempted to explain the physical mechanism behind the monsoon while observational studies have documented the prominent atmospheric features associated with it (e.g., see Reed 1933; Rasmusson 1967; Hales 1972; Badan-Dangon et al. 1991; Schmitz and Mullen 1996; Higgins et al. 1997; Bordoni et al. 2004). Many of these are nicely summarized by Adams and Comrie (1997). A map illustrating the domain of the monsoon’s influence is shown in Fig. 1.
A number of studies have sought to evaluate the simulation of the North American monsoon by atmospheric models. The success of these simulations is sensitive to a variety of factors, such as horizontal resolution (Berbery and Fox-Rabinovitz 2003), boundary conditions (Yang et al. 2001) and convective and/or radiative parameterization (Xu and Small 2002; Gochis et al. 2002, 2003). Berbery and Fox-Rabinovitz (2003) showed that the National Aeronautics and Space Administration (NASA) Goddard Earth Observing System (GEOS) global climate model (GCM) was able to simulate the Gulf of California low-level jet in a high-resolution simulation, but not in a lower-resolution simulation. Yang et al. (2001) found that the National Center for Atmospheric Research (NCAR) Community Climate Model version 3 (CCM3) underestimates the zonal meanwarm-season rainfall in Arizona and New Mexico when using climatological SSTs as the lower boundary conditions. This bias was reduced when the observed SSTs from the Atmospheric Model Intercomparison Project (AMIP) were used. Some aspects of a model’s simulation of the monsoon are sensitive to the choice of convective and/or radiative parameterizations, as seen in the fifth-generation Pennsylvania State University (PSU)–NCAR Mesoscale Model (MM5) simulations by Xu and Small (2002) and Gochis et al. (2002). Testing two different convection schemes and three different radiation schemes, Xu and Small (2002) found that the combined use of a Grell convective parameterization and the rapid radiative transfer model (RRTM) results in the most realistic rainfall patterns, magnitudes, and intraseasonal variability. Gochis et al. (2002) found that among the three different cumulus parameterizations tested, the Kain–Fritsch scheme yields the modeled atmosphere most resembling observations. A continuation of this study (Gochis et al. 2003) revealed that the diurnal cycle of precipitation in the North American monsoon region is sensitive to the convective scheme used.
The simulation of the North American monsoon system by the NCAR CCM3 is the focus of the present study. We seek to quantify the differences between two versions of CCM3. In the standard version of the model, the closure for the parameterization of deep convection is based on convective available potential energy (CAPE). Zhang (2002) found that the physical assumptions on which this closure is based are not valid for midlatitude continental convection and proposed an alternative closure based on the large-scale forcing of virtual temperature. The use of the new closure showed noticeable improvement in rainfall simulation over the southern U.S. Great Plains (Zhang 2002) in the single column model of CCM3. This closure, together with a relative humidity threshold for convective triggering, leads to significant improvement in the rainfall simulation of the tropical climate and its intraseasonal variability (Zhang and Mu 2005a, b). Therefore, it is useful to evaluate such modifications in other climate regimes such as the North American monsoon region. This paper documents the differences that arise from using the two different parameterizations. In addition, it identifies the North American summertime convective regime for which each scheme is more successful. Special emphasis is placed on the model’s simulation of the diurnal cycle of precipitation in the monsoon domain. At a given location, the diurnal variation of precipitation strongly affects the hydrologic cycle and, by the interaction of radiative fluxes with clouds, the radiative budget and surface temperature. Thus, evaluating the diurnal cycle of precipitation in an atmospheric model is an important test of the parameterization for convection. For global models, such evaluations have been carried out by Randall et al. (1991), Lin et al. (2000), Chen et al. (1996), Dai et al. (1999), Dai and Trenberth (2004), Betts and Jakob (2002a, b), and Bechtold et al. (2004), and for cloud-resolving models by Chaboureau et al. (2004) and Guichard et al. (2004). For this study, simulations are carried out over a 9-yr period from 1994 through 2002, providing nine monsoon episodes.
2. Methods
a. Data
For this model study, we used a gridded hourly precipitation rate dataset obtained from the National Weather Service (NWS) Techniques Development Laboratory. Observations are derived from first-order NWS stations and cooperative observer stations, the latter of which use the Universal or Fisher/Porter type of rain gauge. Universal gauges are accurate to the nearest hundredth of an inch while Fisher/Porter gauges are only accurate to the nearest tenth of an inch. The stations are gridded into 2.0° × 2.5° boxes over the region of 20°–60°N, 140°–60°W using a modified Cressman scheme. The data, available from 1963 to the present, is described in detail in Higgins et al. (1996). We consider this data to be the “truth” over the continental United States, though we have reason to be suspicious of it over Mexico, where precipitation gauges on the Mexican Plateau are sparse (Zou and Zheng 2004). For this reason we use measurements from the Tropical Rainfall Measuring Mission (TRMM) satellite, as a complement for the period from 1998 through 2002. Specifically, this study uses the instantaneous precipitation rates derived from the combination of the TRMM Microwave Imager (TMI) and precipitation radar (PR). These data are available in the 3G68 dataset, wherein rain rates have been averaged over 0.5° latitude × 0.5° longitude boxes between 38°S and 38°N. Satellite sampling error is mitigated by averaging over long space and time scales, which has been done successfully with previous CCM3–TRMM comparisons (Collier et al. 2004; Collier and Bowman 2004).
CAPE is calculated using temperature, specific humidity, and pressure fields from the National Centers for Environmental Prediction (NCEP)–NCAR reanalysis project (Kalnay et al. 1996), and the same algorithm as that in CCM3. In this form, CAPE is the vertical integral of the difference between the virtual temperature of the parcel and that of the environment. Because the reanalysis is in part model-based, we deferred evaluation of the precipitation fields to the above-described gauge and satellite data.
b. CCM3 simulations
CCM3 is a three-dimensional global spectral model at T42 horizontal resolution (approximately 2.8° latitude × 2.8° longitude) and 18 vertical levels (Kiehl et al. 1998). For the lower boundary conditions, we used sea surface temperatures provided to us by the Program for Climate Model Diagnosis and Intercomparison (PCMDI) at Lawrence Livermore National Laboratory (Taylor et al. 2000).
For this study, we performed two runs of CCM3, which are identical except for the following. In a control run (denoted CONTR), cloud-base mass flux is determined by CAPE, which is assumed to be consumed by convection exponentially at a given rate. From an analysis of upper-air soundings at the Southern Great Plains (SGP) site of the Atmospheric Radiation Measurement (ARM) program, Zhang (2002) found that the net change in ambient virtual temperature resulting from convective processes is approximately balanced by the net change resulting from large-scale forcing. Based on this finding, he proposed a new closure for the Zhang and McFarlane (1995) convection scheme, in which cloud-base mass flux is expressed as a function of the rate of ambient virtual temperature change resulting from large-scale processes. In an experimental run (denoted EXP), the deep convection scheme uses this closure. Additionally, a relative humidity threshold of 80% is imposed on convective parcels rising from the boundary layer. The existence of positive CAPE, the positive generation of CAPE by large-scale processes, and sufficiently moist air at the parcel-originating level (as given by the relative humidity threshold) are necessary conditions for convection in the new scheme. In the original scheme, the only necessary condition for convection is the existence of positive CAPE. For a more detailed description of the original parameterization, we refer the reader to Zhang and McFarlane (1995). For a detailed description of the new closure, we refer the reader to Zhang (2002).
When evaluating a GCM, it is important to isolate a model’s response to external variability (variations in the SST boundary condition) from its own internal variability (fluctuations resulting from weather). The inadequacy of a single simulation has been proven by Barnett (1995), among others. For this study, we simulate 9 yr (1994–2002), and choose to treat each year of the simulation as being sufficiently independent from every other year of the simulation, because one summertime monsoon likely has little dependence on the summertime monsoon of the previous year. We save the precipitation rates, both convective and large scale, as hourly averages.
For validating the development and evolution of the North American monsoon, we will focus primarily on monthly mean quantities, as available from the simulations, surface gauge data, and TRMM satellite measurements. All observed data are averaged onto the CCM3 horizontal grid before comparisons are made. Additionally, monthly means are averaged over all 9 yr of the simulation period for a “climatological” average.
3. Results
For both the control and experimental runs, nine monsoon episodes were simulated. We first evaluate monsoon onset and evolution in the daily and monthly mean precipitation fields. Then, we turn our attention to the diurnal cycle of precipitation over the North American monsoon region.
a. Onset and evolution
Nine-year mean monthly mean precipitation rates, as shown in Fig. 2, illustrate how the model’s precipitation field tracks the data through the development of the North American monsoon in the American Southwest.
In this figure, the monthly means show that the CONTR run of CCM3 has a substantial wet bias in the western Great Plains, along longitude 105°W, during the summer months. Here, the model simulates the highest precipitation rates in June and July and relatively low rates in May, August, and September. In contrast, the observations show relatively less variation in precipitation over the western Great Plains and more variation in Arizona and New Mexico, where precipitation is observed to be at its maximum in the months of July and August. Over these states, the CONTR run is biased dry, but the EXP simulation compares favorably with the observations, showing increased precipitation in New Mexico in June over May and an increased amount in Arizona in July over June. The EXP simulation agrees well with the observations in August and September over these states as well, with less mean rainfall relative to the previous months. As for the model’s wet bias over the western Great Plains, note that it has been substantially reduced in the EXP run.
For a comparison of the monthly mean precipitation rates over Mexico, we refer to the 5-yr monthly means derived from the TRMM composite (TMI and TRMM PR) hourly rainfall, because gauges on the Mexican Plateau are few and far between (Fig. 3). These contour maps show that, in the CONTR run, May and June monthly means compare quite well with those of the TRMM-derived means. In the EXP run, the June mean is slightly low relative to the observations, particularly near the intersection of 105°W and 30°N over northern Chihuahua. For the months of July and August, both runs show a dry bias over this area. The largest bias apparent in the CONTR simulation is in the position and magnitude of the high rainfall belt extending from the southeast to northwest across western Mexico in July and August. Its location is too far west and the magnitudes of the monthly mean rainfall rates within it are too high relative to the satellite data. Neither of the biases is removed in the EXP run. There is some suggestion that the high rainfall rates here are orographically produced, because of the proximity of the Sierra Madre Occidental (see Fig. 1). We cannot expect a model at this resolution to adequately resolve such orographical effects.
For a more detailed comparison of precipitation over the southwestern United States and northwestern Mexico, we isolate regions of the “core” monsoon with the gauge (over the United States) and TRMM (over Mexico) data. In general, we look for grid boxes that exhibit large summertime departures from the annual mean rainfall (greater than 100%) and that exhibit large month-to-month increases in rainfall (also greater than 100%). These measures represent the times of monsoon peak and onset, respectively. The months of the greatest summertime departures and greatest month-to-month increases are shown in Fig. 4.
Observe that the grid boxes with the greatest summertime positive departures from the annual mean precipitation rate (i.e., greater than 100%) lie in a belt from west-central Mexico into southern Arizona and New Mexico (Fig. 4a). Toward the north, greatest deviations from the annual mean precipitation occur in August. In the far south, these departures generally occur in August or September. In between, the departures occur in June and July, where deviations over 200% of the annual mean can be found. This central region, bounded by 25° and 30°N, 110° and 100°W, may be considered to be in the core of monsoon influence.
In Arizona, the greatest rise in precipitation occurs from June to July (Fig. 4b). In New Mexico, this rise occurs from April to May. To the south, the largest increases occur from April to May or from May to June. Using the month of the greatest deviation from the annual mean (time of monsoon peak) and the pair of months that experience the greatest relative difference (time of monsoon onset), we group grid boxes with similar times of monsoon peak and onset into the regions defined by the dashed lines in Fig. 5. Region 1 is characterized by an early summer onset and a late-summer peak; region 2 is characterized by a late-spring onset and a late-summer peak; region 3 shows an early summer onset and a midsummer peak; and, finally, region 4 exhibits an early summer onset and a late-summer peak. We recognize that regions 3 and 4 are small. However, because of the distinctly different characteristics in the monsoon onset and peak, we consider them separately.
We now evaluate the model’s skill in simulating the onset and evolution of the monsoon in each of these regions with monthly mean precipitation rates as averaged over the region’s grid boxes. These time series are displayed in Fig. 6.
The monthly means from the observations show quite a large interannual spread for the northernmost regions 1 and 2 (Arizona and New Mexico, respectively). However, precipitation for region 1 is characterized by a sharp rise from June to July, a leveling off for the months of July and August, followed by a general decrease from August into October. However, in at least 2 of the 9 yr (1998 and 2000), the monsoon ends abruptly in September, followed by a rise in precipitation from September to October. These two years represent outliers for region 2 as well, where the monsoon appears less pronounced. Given the spread among the ensemble members, the experimental version of the model better simulates the timing of the monsoon in region 1 and better simulates its strength in region 2, where timing is more difficult to ascertain from the observations.
A fewer number of years (or ensemble members) are used for a regional evaluation in regions 3 and 4. Here, we rely upon the TRMM-derived rainfall rates, which are only available as far back as 1998. Yet, for these two regions, the monsoonal signal is clear. In region 3, rainfall rates ramp up most dramatically from May to June and ramp down from July to September. The EXP simulation shows an overall weakened monsoon here, relative to the CONTR simulation; however, it produces a more realistic monsoon peak time and decay, beginning in July instead of in June as in CCM3 (CONTR). The monsoon in the southernmost region 4 is reasonably well simulated by the EXP run. In monsoon strength, it improves upon the CONTR run. In monsoon timing, it reduces a negative phase bias, shifting the peak to July instead of June. Note that TRMM indicates a peak in August for 4 of the 5 yr.
While the core monsoon region lies in the southwestern United States and northern Mexico, the monsoon circulation is a continental-scale climate system. Therefore, to put the evaluation of the North American monsoon simulation in a broader context, we include other regions of North America in the comparison between the observations and model simulations.
The upper-tropospheric moisture content directly reflects the vertical transport of moisture by convection in active convection regimes such as the North American monsoon system. Using the NCAR CAM2 in a forecast mode and comparing against the 1997 summer intensive observing period data at the ARM SGP site, Williamson et al. (2005) showed that a significant moist bias develops within the first day in the upper troposphere after the model initialization. This bias was attributed to the convection parameterization. In Fig. 7, we show the specific humidity and wind distribution at 300 mb during the monsoon months for the two simulations. While both simulations capture the observed circulation pattern well, the moisture distribution is quite different. Compared to the reanalysis, the CONTR run is too moist during July through September over the west coast of Mexico, the Gulf of California, the Gulf of Mexico, as well as over the U.S. Midwest. The moist biases are largely removed in the EXP run, yielding excellent agreement with the reanalysis.
b. The diurnal cycle
Because much of the earth’s weather varies diurnally, the diurnal cycle is an important mode of variability for atmospheric models to simulate correctly. In this section we extend our validation of the CCM3 North American monsoon simulation to include a validation of its diurnal cycle of precipitation during the monsoon months of June, July, and August. Because of the large sampling errors associated with using the TRMM data for a single-season diurnal cycle evaluation, we rely solely on the gauge-based precipitation rates for all of North America. However, it should be noted that over northern Mexico diurnal cycles estimated from the TRMM data are qualitatively similar to those estimated from the gauge-based rain rates, when averaged over large regions. For analyzing the diurnal cycle here, we represent the phases and amplitudes of the least squares–fitted seasonal mean hourly precipitation rates with arrows in 24-h clocks. Each arrow is plotted on top of its geographic location on a map. The convention for the clock is defined such that an arrow pointing due north represents a phase of 2400 LST, one pointing due east represents a phase of 0600 LST, one pointing due south represents a phase of 1200 LST, and one pointing due west represents a phase of 1800 LST. All arrows are plotted with respect to local time. The length of each arrow represents the amplitude of the harmonic relative to the daily mean precipitation rate. Harmonics that are not statistically significant are omitted from the plots. We have found that the diurnal harmonic explains most of the diurnal variation in precipitation over North America during the summer months, and so we present a comparison of the sum of the diurnal and semidiurnal harmonics (or, the net effect) in Fig. 8.
From the figure, note that over much of the North American monsoon domain (outlined by the box), summertime precipitation is observed to peak between roughly 1200 and 1800 LST.1 In the CONTR run, for most grid boxes of the southwestern United States and extreme northern Mexico, the model leads the observations in the phase of the diurnal cycle. The phase bias is consistent with those found in several previous diurnal cycle evaluations, which have shown the tendency for global models to initiate continental precipitation prematurely, often before 1200 LST, where precipitation is observed to begin after noon or in the evening (Betts et al. 1998, 1999; Betts and Jakob 2002a; Dai and Trenberth 2004; Bechtold et al. 2004). The EXP run significantly reduces or eliminates such phase biases in southern Arizona, extreme northern Mexico, and western New Mexico. However, over the remainder of northwest Mexico, EXP diurnal cycles are out of phase with those of the observations by between 9 and 12 h, a dramatic degradation of the CONTR simulation in which simulated phases lead observed phases by at most 6 h. For a broader perspective, compare the diurnal cycles simulated over this region to those over the entire United States. The large EXP phase biases are not confined to northern Mexico. They can be seen over the southeastern and northeastern United States, as well. Meanwhile, phase biases over the U.S. Great Plains, specifically locations north of 35°N, and between 100° and 95°W, are comparatively smaller than they are in the CONTR run. It is difficult to ascertain why the modified scheme is more successful in some areas than in others, though Liang et al. (2004) found that the success of various convection schemes in the simulation of the diurnal cycle also varies geographically in a regional model.
There are two major differences between the CONTR and the EXP simulations. One of these is in the convection scheme closure. The CONTR simulation uses a closure based on CAPE while the EXP run does not. To explain how CAPE is related to convection, we compared the diurnal variations of observed CAPE and observed precipitation for the summers of 1994–2002 for all available grid boxes in North America.2 Diurnal cycle phase differences are shown in Fig. 9, where phase differences of less than 6 h are not contoured. Such small differences are insignificant because the reanalysis-based CAPE values are only available every 6 h.
This figure suggests that summertime North American convection has two distinct regimes. In one regime (regime A), the diurnal cycle of CAPE significantly leads that of precipitation (by 6 h or more). In the other regime (regime B), the diurnal cycles of CAPE and precipitation are approximately in phase with each other (within 6 h). Compare this to Fig. 10, which shows the rainfall diurnal cycle phase bias change resulting from the parameterization modifications. For nearly all points in regime A, the modified scheme improves upon the original scheme in the diurnal cycle of precipitation, which might suggest that a CAPE-based closure is less appropriate for locations in this convective regime. To explore this possibility further, we isolated the effect of the second modification, the application of a relative humidity threshold, from that of changing the closure. The relative humidity threshold used in the EXP run may have acted to delay convection from firing until the relative humidity in the convective parcel–originating layer reaches 80%.
To investigate this possibility, we conducted a short (5 yr) run of CCM3 identical to the EXP run, except without the relative humidity threshold. This simulation, which spans 1998 through 2002, shall be denoted EXPa. Figure 11 shows the diurnal cycle phase-amplitude plot. Increases in phase biases are generally small, but occur in extreme northern Mexico and in the central United States, thus generally in regime A. However, these bias changes are relatively smaller than those that result from the combined modifications. Phase bias reductions are found mostly in regime B. For example, in northwestern Mexico and in parts of the eastern United States, the model now simulates afternoon peaks in precipitation where it simulated anomalous nighttime peaks in the EXP run. However, there is not widespread improvement, because phase biases along the Gulf of Mexico and Atlantic coasts remain large.
Based on the results of this simulation, the replacement of the CAPE-based closure by the closure based on large-scale forcing is the more important modification for improving rainfall diurnal cycles in a convective regime in which CAPE and precipitation are significantly out of phase with each other. The effect of applying the relative humidity threshold has a smaller effect on the diurnal cycle in this regime. In the opposite regime, in which CAPE and precipitation are relatively well correlated, the use of the relative humidity threshold degrades the diurnal cycle simulation over northern Mexico and over parts of the eastern United States, serving to postpone convection for too long. However, while its application results in an unrealistic diurnal cycle in parts of this regime, its removal results in a significantly undersimulated summer monsoon. This is shown in Fig. 12. Comparing this figure to Figs. 2 and 3, it is apparent that the removal of the relative humidity threshold results in negligible precipitation over the American Southwest, particularly in the states of Arizona and New Mexico. Likewise, precipitation in most of northern Mexico also is eliminated, with the exception of that associated with the high-intensity rainfall belt along the coast appearing in July and August. The eastern half of northern Mexico is far too dry during the summer months. Therefore, removing this convective trigger mechanism worsens the monsoon simulation relative to the observations, as is evident in the monthly mean precipitation rates.
4. Summary and conclusions
We have compared two simulations, each for 9 yr, of the North American summer monsoon using NCAR CCM3. A control simulation uses the CCM3 deep convection parameterization with its conventional closure based on convective available potential energy. An experimental simulation uses the same parameterization with a new closure based on the large-scale forcing of the temperature and humidity fields as well as a relative humidity threshold of 80% on parcels lifted from the boundary layer. Summertime conditions, as averaged over the 9-yr period, are compared and validated with observations from both surface gauge and TRMM satellite measurements as well as the NCEP–NCAR reanalysis.
In monsoonal onset and peak intensity, the control simulation shows some noticeable biases in what we consider “core” monsoonal regions of the southwestern United States and northwestern Mexico. It is negatively biased with precipitation over Arizona and New Mexico post-onset, and this bias is reduced in the experimental run. We identified four unique monsoon regions, each characterized by a unique combination of monsoon onset and peak time. A comparison of the climatological monthly mean precipitation rates, as averaged over these regions, shows that in the control run, monsoon peak occurs 1–2 months prematurely in three of the four regions and that this bias is reduced by the experimental run. However, in one of these regions, the experimental run shows a monsoon that is anomalously weak. Both versions of the model show a westward bias in a high-intensity rainfall belt located in the vicinity of the Sierra Madre Occidental. This rainfall belt is simulated to be positioned along the Pacific coastline. We suspect the positional bias is because of the model’s inability to resolve the orographically enhanced rainfall at such a coarse resolution. In particular geographic areas, the simulated diurnal cycle of precipitation shows noticeable sensitivity to changes in the deep convective parameterization. In the monsoon region, the diurnal cycle is highly sensitive over northern Mexico, where phase biases are reduced north of about 27°N and are significantly enlarged south of this latitude upon making the two above-mentioned modifications. Increases in diurnal cycle phase biases also occur in the southeastern and northeastern United States, with simulated daily precipitation peaks shifting from afternoon to just after midnight. By contrast, improvements in the diurnal cycle can be seen over extreme northern Mexico, extreme southern Arizona, and most of New Mexico, as well as over the U.S. Great Plains. The modified scheme is much more successful than the original scheme in capturing the nighttime peaks characterizing diurnal cycles in this last region. A search for commonality between the most sensitive regions reveals that where the original scheme performs relatively poorly and the modified scheme performs relatively well, the diurnal cycle of observed CAPE leads that of precipitation by more than 6 h. Regions characterized by this convective regime include part of the North American monsoon region and the U.S. Great Plains. For the opposite regime, precipitation and CAPE are observed to be approximately in phase with each other. In this regime, the modified scheme performs less well, with large diurnal cycle phase biases. Phase biases over northern Mexico can be attributed to the application of the relative humidity threshold as a convective triggering mechanism and not to the new closure. In this region, the delay associated with the buildup of relative humidity postpones convection for too long. However, its removal results in a weakened or nonexistent monsoon. The results documented here should be used in the development of convective parameterization in global climate models. Understanding the processes that initiate and regulate convection, implicit to the regimes identified here, is necessary for the most realistic representation of summer precipitation over the continent.
Acknowledgments
This research was funded by NOAA Grant GC03-074. The authors thank Karl Taylor of the Program for Climate Model Diagnosis and Intercomparison (PCMDI) at Lawrence Livermore National Laboratory for providing the sea surface temperature boundary condition. They also appreciate the comments and suggestions of Siegfried Schubert and Myong-In Lee as well as those from the anonymous reviewers.
REFERENCES
Adams, D., and A. Comrie, 1997: The North American monsoon. Bull. Amer. Meteor. Soc., 78 , 2197–2213.
Anderson, T., 1978: The Statistical Analysis of Time Series. John Wiley & Sons, 704 pp.
Badan-Dangon, A., C. E. Dorman, M. A. Merrifield, and C. D. Winant, 1991: The lower atmosphere over the Gulf of California. J. Geophys. Res., 96 , 16877–16896.
Barnett, T. P., 1995: Monte Carlo climate forecasting. J. Climate, 8 , 1005–1022.
Bechtold, P., J-P. Chaboureau, A. Beljaars, A. Betts, M. Kohler, M. Miller, and J-L. Redelsperger, 2004: The simulation of the diurnal cycle of convective precipitation over land in a global model. Quart. J. Roy. Meteor. Soc., 130 , 3119–3137.
Berbery, E. H., and M. S. Fox-Rabinovitz, 2003: Multiscale diagnosis of the North American monsoon system using a variable resolution GCM. J. Climate, 16 , 1929–1947.
Betts, A. K., and C. Jakob, 2002a: Evaluation of the diurnal cycle of precipitation, surface thermodynamics, and surface fluxes in the ECMWF model using LBA data. J. Geophys. Res., 107 .8045, doi:10.1029/2001JD000427.
Betts, A. K., and C. Jakob, 2002b: Study of diurnal cycle of convective precipitation over Amazonia using a single column model. J. Geophys. Res., 107 .4732, doi:10.1029/2002JD002264.
Betts, A. K., P. Viterbo, and E. Wood, 1998: Surface energy and water balance for the Arkansas–Red River basin from the ECMWF reanalysis. J. Climate, 11 , 2881–2897.
Betts, A. K., J. Ball, and P. Viterbo, 1999: Basin-scale surface water and energy budgets for the Mississippi from the ECMWF reanalysis. J. Geophys. Res., 104 , 19293–19306.
Bordoni, S., P. E. Ciesielski, R. H. Johnson, B. D. McNoldy, and B. Stevens, 2004: The low-level circulation of the North American monsoon as revealed by QuickSCAT. Geophys. Res. Lett., 31 .L10109, doi:10.1029/2004GL020009.
Bowman, K. P., J. C. Collier, G. R. North, Q. Wu, E. Ha, and J. Hardin, 2005: The diurnal cycle of tropical precipitation in Tropical Rainfall Measuring Mission (TRMM) satellite and ocean buoy rain gauge data. J. Geophys. Res., 110 .D21104, doi:10.1029/2005JD005763.
Chaboureau, J-P., F. Guichard, J-L. Redelsperger, and J-P. Lafore, 2004: The role of stability and moisture in the diurnal cycle of convection over land. Quart. J. Roy. Meteor. Soc., 130 , 3105–3117.
Chen, M., R. E. Dickinson, X. Zeng, and A. N. Hahmann, 1996: Comparison of precipitation observed over the continental United States to that simulated by a climate model. J. Climate, 9 , 2233–2249.
Collier, J. C., and K. P. Bowman, 2004: Diurnal cycle of tropical precipitation in a general circulation model. J. Geophys. Res., 109 .D17105, doi:10.1029/2004JD004818.
Collier, J. C., K. P. Bowman, and G. R. North, 2004: A comparison of tropical precipitation simulated by the NCAR Community Climate Model CCM3 with that observed by the Tropical Rainfall Measuring Mission (TRMM) satellite. J. Climate, 17 , 3319–3333.
Dai, A., and K. E. Trenberth, 2004: The diurnal cycle and its depiction in the Community Climate System Model. J. Climate, 17 , 930–951.
Dai, A., F. Giorgi, and K. E. Trenberth, 1999: Observed and model-simulated diurnal cycles of precipitation over the contiguous United States. J. Geophys. Res., 104 , 6377–6402.
Gochis, D. J., W. J. Shuttleworth, and Z-L. Yang, 2002: Sensitivity of the modeled North American monsoon regional climate to convective parameterization. Mon. Wea. Rev., 130 , 1282–1298.
Gochis, D. J., W. J. Shuttleworth, and Z-L. Yang, 2003: Hydrometeorological response of the modeled North American monsoon to convective parameterization. J. Hydrometeor., 4 , 235–250.
Guichard, F., and Coauthors, 2004: Modelling the diurnal cycle of deep precipitating convection over land with cloud-resolving models and single-column models. Quart. J. Roy. Meteor. Soc., 130 , 3139–3172.
Hales Jr., J. E., 1972: Surges of maritime tropical air northward over the Gulf of California. Mon. Wea. Rev., 100 , 298–306.
Higgins, R. W., J. Janowiak, and Y. Yao, 1996: A gridded hourly precipitation data base for the United States (1963–1993). NCEP/Climate Prediction Center Atlas, No. 1, NOAA/NWS/NCEP, 47 pp.
Higgins, R. W., Y. Yao, and X. Wang, 1997: Influence of the North American monsoon system on the U.S. summer precipitation regime. J. Climate, 10 , 2600–2622.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437–471.
Kiehl, J., J. Hack, G. Bonan, B. Boville, D. Williamson, and P. Rasch, 1998: The National Center for Atmospheric Research Community Climate Model: CCM3. J. Climate, 11 , 1131–1149.
Liang, X-Z., L. Li, A. Dai, and K. E. Kunkel, 2004: Regional climate model simulations of summer precipitation diurnal cycle over the United States. Geophys. Res. Lett., 31 .L24208, doi:10.1029/2004GL021054.
Lin, X., D. A. Randall, and L. D. Fowler, 2000: Diurnal variations of the hydrologic cycle and radiative fluxes: Comparisons between observations and a GCM. J. Climate, 13 , 4159–4179.
Randall, D. A., and D. A. Dazlich, and Harshvardhan, 1991: Diurnal variability of the hydrologic cycle in a general circulation model. J. Atmos. Sci., 48 , 40–62.
Rasmusson, E. M., 1967: Atmospheric water vapor transport and the water balance of North America: Part I. Characteristics of the water vapor flux field. Mon. Wea. Rev., 95 , 403–426.
Reed, T. R., 1933: The North American high-level anticyclone. Mon. Wea. Rev., 61 , 321–325.
Schmitz, J. T., and S. L. Mullen, 1996: Water vapor transport associated with the summertime North American monsoon as depicted by ECMWF analyses. J. Climate, 9 , 1621–1634.
Taylor, K. E., D. Williamson, and F. Zwiers, 2000: The sea surface temperature and sea-ice concentration boundary conditions for AMIP II simulations. Lawrence Livermore National Laboratory Program for Climate Model Diagnosis and Intercomparison Tech. Rep. 60, 25 pp.
Williamson, D., and Coauthors, 2005: Moisture and temperature balances at the Atmospheric Radiation Measurement Southern Great Plains Site in forecasts with the Community Atmosphere Model (CAM2). J. Geophys. Res., 110 .D15S16, doi:10.1029/2004JD005109.
Xu, J., and E. E. Small, 2002: Simulating summertime rainfall variability in the North American monsoon region: The influence of convection and radiation parameterizations. J. Geophys. Res., 107 .4727, doi:10.1029/2001JD002047.
Yang, Z-L., D. Gochis, and W. J. Shuttleworth, 2001: Evaluation of the simulations of the North American monsoon in the NCAR CCM3. Geophys. Res. Lett., 28 , 1211–1214.
Zhang, G. J., 2002: Convective quasi-equilibrium in midlatitude continental environment and its effect on convective parameterization. J. Geophys. Res., 107 .4220, doi:10.1029/2001JD001005.
Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre general circulation model. Atmos.–Ocean, 33 , 407–446.
Zhang, G. J., and M. Mu, 2005a: Effects of modifications to the Zhang-McFarlane convection parameterization on the simulation of the tropical precipitation in the National Center for Atmospheric Research Community Climate Model, version 3. J. Geophys. Res., 110 .D09109, doi:10.1029/2004JD005617.
Zhang, G. J., and M. Mu, 2005b: Simulation of the Madden–Julian oscillation in the NCAR CCM3 using a revised Zhang–McFarlane convection parameterization scheme. J. Climate, 18 , 4046–4064.
Zou, C-Z., and W. Zheng, 2004: Simulation of diurnal patterns of summer precipitation in the North American monsoon: An assessment using TRMM. Geophys. Res. Lett., 31 .L07105, doi:10.1029/2004GL019415.
Topographical map of the North American monsoon region, with elevation in meters. Elevation data courtesy of NCAR through the U.S. Navy Global Elevation dataset.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Monthly mean precipitation rates (mm day−1) for (left) CCM3 CONTR, (middle) CCM3 EXP, and (right) surface gauges, as averaged over the years 1994–2002. For clarity, only U.S. and Mexico land precipitation rates have been contoured.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Monthly mean precipitation rates (mm day−1) for (left) CCM3 CONTR, (middle) CCM3 EXP, and (right) TRMM (TMI + TRMM PR), as averaged over the years 1998–2002.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Months of (a) greatest relative departure from the annual mean precipitation rate and (b) those of greatest relative month-to-month departures in mean precipitation rate from the gauge and TRMM data, as averaged over 1994–2002 (gauges) and 1998–2002 (TRMM). Note that only positive departures over land are represented in the figure.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Map showing the grouping into regions of grid boxes with similar times of monsoon onset and peak.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Monthly mean precipitation rates (mm day−1) for years 1994–2002 (regions 1 and 2) and 1999–2002 (regions 3 and 4). Note that regions 1 and 2 show comparisons between (bottom) CCM3 CONTR, (middle) CCM3 EXP, and (top) gauge-based rainfall rates, while regions 3 and 4 show comparisons between model simulations and TRMM-based rainfall rates. The climatological (or ensemble) mean is given by the heavy curve.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Monthly mean 300-mb specific humidity (g kg−1) (contoured) and horizontal winds (m s−1) for (left) CCM3 CONTR, (middle) CCM3 EXP, and (right) NCEP–NCAR reanalysis as averaged over the years 1994–2002.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Diurnal + semidiurnal harmonics for CCM3 and gauge precipitation rates. Harmonics from (a) CCM3 (CONTR) and the gauges, and those from (b) CCM3 (EXP) and the gauges are shown. Only harmonics significant at the 95% level are plotted. Note that a vector’s length represents the amplitude of the harmonic as a fraction of the daily mean.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Difference between the phase of the NCEP–NCAR reanalysis-based CAPE diurnal cycle and that of the gauge-based precipitation. Diurnal cycles represent averages for JJA 1994–2002.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Change in absolute diurnal cycle phase bias resulting from the convection parameterization modifications. Note that negative changes represent improvements while positive changes represent degradations.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Monthly mean precipitation rates (mm day−1) for years 1998–2002 and for the monsoon regions depicted in Fig. 5, and diurnal + semidiurnal harmonics for CCM3 (EXPa) and gauge precipitation rates. Only harmonics significant at the 95% level are plotted. As in Fig. 8, a vector’s length represents the amplitude of the harmonic as a fraction of the daily mean.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
Monthly mean precipitation rates (mm day−1) for CCM3 EXP, CCM3 EXPa, and the gauge-based and TRMM-based data as averaged over the years 1998–2002. Note that the gauge-based rainfall rates are shown for north of 30°N and TRMM-based rainfall rates are shown south of this latitude.
Citation: Journal of Climate 19, 12; 10.1175/JCLI3732.1
In agreement with the gauge data, regionally averaged TRMM data indicates an afternoon peak in precipitation over northern Mexico.
Observed CAPE refers to the quantity calculated from the temperature, specific humidity, and pressure fields of the NCEP–NCAR reanalysis for June–August (JJA) 1994–2002. Note that the calculation follows the method used for calculating CAPE in the CCM3.
