1. Introduction
The North American monsoon (NAM) is the regional-scale atmospheric circulation system (Stensrud et al. 1997) responsible for the dramatic increase in precipitation during the summer in northwestern Mexico and the southwest United States (Grantz et al. 2007). The NAM typically starts in mid-June in the core monsoon in northwestern Mexico (Turrent and Cavazos 2009) and in early July in Arizona–New Mexico (AZNM) and finishes around the end of September (e.g., Douglas et al. 1993; Higgins et al. 1997). It contributes 60% of the total annual precipitation in the core region and about 40% in AZNM, where it has its northern limit (Douglas et al. 1993). The core region of the NAM is centered over the Sierra Madre Occidental (SMO), a large mountain chain in northwestern Mexico, and around the SMO slopes the monsoon precipitation reaches its maximum (Douglas et al. 1993). Understanding the mechanisms that govern the timing and intensity, as well as the impacts of climate change on the NAM, is a priority for the scientific community, watershed managers, and farmers in the NAM area.
a. NAM cycle
The NAM has a well-established life cycle and has been described in many previous works (e.g., Douglas et al. 1993; Stensrud et al. 1995; Higgins et al. 1997) and so will only be described briefly here.
During the mature phase (July–August) the NAM is fully developed with heavy precipitation on the southwest coast of Mexico (Vera et al. 2006) and into Arizona and New Mexico by early July (Higgins et al. 1997). The precipitation increase corresponds to the increase in moisture flux (Douglas et al. 1993) and the formation of the southerly low-level jet over the Gulf of California (GoC) (Carleton 1986; Badan-Dangon et al. 1991; Douglas 1995) and also to displacements of the Pacific and Bermuda highs (Carleton 1986) and the formation of an upper-level anticyclone (Higgins et al. 1997). This anticyclone is associated with upper-tropospheric divergence in its vicinity and to the south with stronger easterlies (or weaker westerlies), which enhance the onset of the monsoon rainfall (Douglas et al. 1993). The decay of the NAM occurs around late September–October (Vera et al. 2006). During this phase the ridge over the western United States weakens as the monsoon high retreats southward and Mexican monsoon precipitation diminishes (Higgins et al. 1997). At the end of the monsoon season, the precipitation events are more frequently associated with synoptic-scale frontal systems rather than with localized convective instability (Vera et al. 2006). It has also been reported that, in this stage, precipitation events are heavily influenced by land-falling tropical storms in western Mexico and the U. S. Southwest (e.g., Englehart and Douglas 2001; Corbosiero et al. 2009).
However, the complex nature of the moisture source and transport mechanisms make it extremely difficult to understand (Grantz et al. 2007) and predict the variability of the NAM and the possible impacts of climate change on the region. NAM variability ranges from diurnal to decadal time scales (e.g., Carleton 1986; Vera et al. 2006; Grantz et al. 2007).
b. Models and climate change
One of the main problems in understanding the dynamics and variability of the NAM is that the observations mainly over the Mexican portion are sparse, both temporally and spatially. Hence, numerical models have become a common tool to study mesoscale phenomena in this situation. However, owing to the lack of observations it is very difficult to determine the fidelity of such models or that of the boundary conditions from general circulation models (GCMs) used to drive the regional models.
The Fourth Assessment Report (AR4) of the Intergovernmental Panel on Climate Change (IPCC) recognized that current GCMs have systematic biases, especially in the simulation of regional features in areas with complex terrain such as the NAM region (Cavazos and Marengo 2009). Consequently, understanding how changes in large-scale hydroclimatic patterns will manifest at a regional scale remains a key problem for studies on climate change (Cavazos and Marengo 2009).
The attribution of a significant anthropogenic component in current climate change and the impact of the changing global climate on human activities have produced a growing demand for climate change information to evaluate and determine mitigation and adaptation measures (Berbery and Marengo 2009). At present, climate change projections (i.e., possible future scenarios) are largely derived from GCMs (Berbery and Marengo 2009), which often do not provide reliable information at regional scales.
The Variability of the American Monsoon System (VAMOS) [a U.S. Climate Variability and Predictability (CLIVAR)–World Climate Research Program (WCRP) project for the study of variability in the American monsoon systems] Anthropogenic Climate Change (ACC) task force has identified specific issues that would lead to an increase in the credibility of regional climate change scenarios in the monsoon regions of the Americas (Cavazos and Marengo 2009). One of these issues is to evaluate the performance and uncertainties of the GCMs on the monsoon regions so as to reduce the cascade of uncertainty in regional climate prediction on the NAM. This issue of the implications of uncertainties derived from GCMs for providing reliable information on future climate change has started to be investigated in other projects, such as the Prediction of Regional scenarios and Uncertainties for Defining European Climate change risks and Effects (PRUDENCE) and Ensemble-Based Predictions of Climate Change and their Impacts (ENSEMBLES) for Europe and the North American Regional Climate Change Assessment Program (NARCCAP) for North America. To underpin these studies there is a clear requirement for detailed studies of the climate processes involved and their representation in the models.
Motivated by the current climate-change research needs in the American monsoon regions, in this study we investigate the realism of the NAM system simulated by a regional climate model, the Hadley Centre Regional Model version 3P (HadRM3P), under wet and dry conditions. We analyze the moisture sources of the NAM using two different datasets of boundary conditions to drive the model; this will allow us to assess the ability/uncertainty of the regional model to reproduce rainfall under climate change conditions in the NAM region.
In the next section, the HadRM3P model and the main features of the datasets that we used are briefly described. The comparison between the observed seasonal cycle of precipitation and the three experiments carried out with HadRM3P is discussed in the first section of the results. In the second section, the impact of the moisture of the Gulf of Mexico over the northern portion of the NAM is analyzed. An overview of the implications of the findings of this work for studying climate change over the NAM is given in the last section of the results, followed by the conclusions of the study.
2. Data and methods
a. HadRM3P
HadRM3P is the model incorporated in the Providing Regional Climates for Impacts Studies (PRECIS) regional climate modeling system, developed and freely distributed by the Hadley Centre. PRECIS includes a third-generation regional climate model (RCM), a comprehensive set of boundary conditions to support modeling studies on climate variability and climate change, and software to display and process the data produced by the RCM. Its great advantage is that it can be easily set up and run on a single PC (Jones et al. 2004).
HadRM3P is an atmosphere and land surface model, based on the atmospheric component of the global coupled climate model version 3 (HadCM3) (Gordon et al. 2000). The atmospheric component uses the hydrostatic primitive equations. The model equations are solved in spherical polar coordinates, and the latitude–longitude grid is rotated so that the equator lies inside the region of interest (Jones et al. 2004). For this study HadRM3P was run with a horizontal resolution of 0.44° × 0.44° (i.e., a minimum resolution of 50 km at the equator of the rotated grid, Jones et al. 1995). In the vertical direction, the model has 19 levels, the lowest at around 50 m and the highest at 0.5 hPa (Simmons and Burridge 1981). There are terrain-following σ coordinates for the bottom four levels, pressure coordinates for the top three levels, and hybrid coordinates in the middle.
b. Experimental setup
The spatial domain (Fig. 1) for this work was defined from 8° to 46°N, 131° to 80°W. The extension of the domain was chosen to be large enough to include the inherent dynamics of the region, such as the interactions of the Pacific Ocean, the Gulf of California, and Gulf of Mexico at low and upper levels within the NAM region and the impacts of these basins on the variability of the NAM.
HadRM3P’s orography (m), contours each 500 m with contours greater than 1500 m shaded. The boxes show the location of the CORE and AZNM regions used in the precipitation analysis.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
The two regions shown in Fig. 1, denoted as CORE and AZNM, are the main focus of this work. These areas have been defined in previous works (e.g., Douglas et al. 1993; Higgins and Shi 2000) based on the amount of rainfall observed during the NAM season. The core area of the monsoon (hereafter CORE) is located between 24° and 30°N, 112° and 106°W. The other area, Arizona–Mew Mexico (AZNM) is the northern edge of the monsoon located between 32° and 35°N, 112.5° and 107.5°W.
To evaluate the skill of HadRM3P in reproducing the precipitation over the NAM region, five simulations of 10 years were carried out for the 1 December 1988–31 December 1998 period. In part, this period was chosen to include some of the years since 1995 when Sonora, a Mexican state that lies within the core region of the monsoon, has suffered from drought and, thus, is a threat to local water supplies (Hallack-Alegria and Watkins 2007). This allows us to assess the skill of HadRM3P in reproducing drought and normal precipitation phases of NAM variability and hence its ability to predict these.
The lateral boundary conditions to drive HadRM3P in the first experiment [hereafter HadRM3P–National Centers for Environmental Prediction (NCEP)] and for the ensemble experiment were derived from the NCEP reanalysis. This global reanalysis is provided by the NOAA/Office of Oceanic and Atmospheric Research (OAR)/Earth System Research Laboratory (ESRL)/Physical Sciences Division (PSD), Boulder, Colorado, from their Web site at http://www.cdc.noaa.gov/. This reanalysis has a resolution of T62 (approximately 210 km) and 28 levels in the vertical.
For the second experiment [hereafter HadRM3P–the 40-yr European Centre for Medium-Range Weather Forecasts Re-Analysis (ERA-40)] ERA-40 was used as the lateral boundary conditions to drive HadRM3P. The ERA-40 project is a global atmospheric analysis of many conventional observations and satellite data streams for the period September 1957–August 2002. This dataset is provided by the ECMWF (http://data.ecmwf.int/data/) and has a resolution of T159 (approximately 125 km near the equator) and 60 levels in the vertical.
The monthly observed precipitation data were taken from the University of Delaware (UDel_AirT_Precip_data) provided by the NOAA/OAR/ESRL/PSD, Boulder, Colorado, from their Web site at http://www.cdc.noaa.gov. This dataset is a compilation of observations interpolated onto a 0.5° × 0.5° grid. This data has worldwide coverage.
The NCEP North American Regional Reanalysis (NARR) dataset was used to evaluate some of HadRM3P outputs. The NARR project is an extension of the NCEP Global Reanalysis, run over the North American region. The NARR model uses the very high-resolution grid-point NCEP Eta Model (32 km/45 layer) together with the Regional Data Assimilation System (RDAS), which, importantly, assimilates precipitation along with other variables (Mesinger et al. 2006). This reanalysis, owing to its high spatial resolution and that it includes assimilation of a wider range of observations, has been used to investigate the NAM and to evaluate other regional models (i.e., Mo et al. 2005; Xu et al. 2004; Cielsielski and Johnson 2008). In previous studies (Mo et al. 2005, 2007), NARR was shown to reproduce realistically the winter atmospheric transport and precipitation. During the summer season, the basic features and evolution of the NAM, seen in NARR, compare favorably with observations. The potential of NARR to be used as a proxy for the precipitation, mainly in the AZNM region, will be demonstrated in the next section.
In an attempt to investigate more thoroughly HadRM3P precipitation, an ensemble of four members was made. The only difference between each member of the ensemble is the starting time of the simulation. The first member started in December 1988 and finished in December 1998, the second member started one year before (December 1987) and finished in December 1998, the third member started on December 1986, and the last member started on December 1985. Although it did not introduce any sort of perturbation in the boundary conditions, it has been observed that regional models are sensitive to variations in the initial conditions (ICs) because of their internal variability (Giorgi and Bi 2000). For example, varying the initialization date of each member of the ensemble even by one day can produce a change in the model simulation (Vidale et al. 2003). If each member of the ensemble is started on a slightly different synoptic state the precipitation simulated by the model can vary from member to member (e.g., Giorgi and Bi 2000; Vidale et al. 2003). The mean of the ensemble was calculated using just the 10 years that all members have in common (i.e., January 1989–December 1998).
3. Results
a. Annual precipitation
Figure 2 shows the annual cycle of precipitation (mm day−1) in the CORE area based on the four datasets used in this study. Both HadRM3P–NCEP and NARR underestimate the rainfall peaks in July and August. NARR underestimates the peak of August by almost 40%, whereas the differences between HadRM3P–NCEP and the observations in August are less pronounced, around 30%. HadRM3P–NCEP (HadRM3P–ERA-40) produces almost four times more than (20% more than) the observed precipitation in October. NARR in general underestimates the rain of the whole July–September (JAS) season and that of December.
Annual cycle of precipitation (mm day−1) (average of the 10 years simulated, 1989–98) in the CORE monsoon area: University of Delaware monthly observations (broken line, star marker), HadRM3P–NCEP (triangle marker), HadRM3P–ERA-40 (circle marker), and NARR (square marker).
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
In the AZNM area NARR matches almost perfectly the observed precipitation, except in February and May when NARR slightly overestimates the precipitation (Fig. 3). That NARR performed so well for the precipitation in this region can be explained by the fact that over the United States, in general, there are more observations available than in the Mexican portion (where the CORE region lies). HadRM3P–NCEP poorly represents the precipitation in the AZNM missing completely the NAM season [June–September (JJAS)]. HadRM3P–ERA-40 not only considerably improves the correlation with the observations (see Table 2) but it captures in a more realistic way the seasonal cycle than HadRM3P–NCEP, mainly the rain in the winter. The contribution of the winter rain is relevant to the AZNM region since it represents 30% of the total annual precipitation. In general, HadRM3P–ERA-40 has higher correlation with observations than HaDRM3P–NCEP (see Tables 1 and 2), mainly in the AZNM region, although the rms error tends to be slightly larger in the HadRM3P–ERA-40 simulation.
As in Fig. 2 but for the AZNM region.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
Correlation (Corr) and rms error (Rmse) between the simulations and the observations (University of Delaware) in the CORE area. Second and third columns are the value between HadRM3P driven by NCEP, fourth and fifth columns HadRM3P driven by ERA-40, and the last two columns NARR.
Analyzing the monthly and daily time series (not shown here) it was found that HadRM3P, with ERA-40 and NCEP, appropriately captures the interannual variability and reproduces the substantial decrease in precipitation observed from 1990 to 1996 in the CORE area.
To assess how much of the discrepancy between HadRM3P–NCEP and the observations was due to pure noise, a four-member ensemble in which the initial conditions were modified, as described in the data and methods section, was carried out. The mean of the ensemble has the same correlation with the observations as that of a single integration at daily, monthly, and annual time scales (the first two are not shown here) in both of the analyzed regions (CORE and AZNM). As shown in Fig. 4, the seasonal cycle of precipitation derived from the mean ensemble underestimates the monsoon rainfall in July–August and overestimates the winter precipitation in AZNM. There is almost no variability between the four members of the ensemble. The largest differences and the largest values of standard error occur in October, whereas in the winter months [December–February (DJF)] the differences are almost zero in both regions. Also, in the CORE region the largest difference between the members of the ensemble occurs in October (Fig. 5).
Comparison of the seasonal cycle of precipitation (mm day−1) between observations (University of Delaware, solid line) and the HadRM3P–NCEP four-member ensemble mean (dots with error bars) for the AZNM region. (top) The error bar represents the standard deviation and (bottom) the standard error. For this experiment HadRM3P was driven by NCEP and each member of the ensemble has a different start date.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
As in Fig. 4 but for the CORE region.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
The ensemble experiment shows that HadRM3P is a stable model, which is constrained to stay close to the driving conditions (Jones et al. 1995), and that the finescale internal variability of the model has little impact on the precipitation variables of interest in this study. Since the mean of the ensemble has, in the CORE and AZNM area, the same values of correlation as the single-run experiment in each region, we can conclude that a single-run experiment is representative for the analysis carried out in this study.
b. Which large-scale drivers impact the NAM precipitation?
The summer moisture transport over Mexico and the United States is modulated by two meridional low-level jets (LLJ) (Mo et al. 2005) from the Great Plains (GPLLJ) and from the Gulf of California (GCLLJ) and a zonal one from the Caribbean area (CALLJ) (Amador 1998). The CALLJ transports moisture from the tropical Caribbean basin to the Gulf of Mexico (GoM) and some of this moisture is then transported by the GPLLJ to the southeast of the United States (Mo et al. 2005). The GCLLJ transports moisture from the eastern Pacific, along the Gulf of California (GoC) to the core monsoon in Mexico and to the southwest of the United States (Mo et al. 2005).
Mean August specific humidity and wind vector at 850 mb: (a) HadRM3P–NCEP, (b) NARR, and (c) NARR minus HadRM3P–NCEP. The squares indicate the position of the CORE and AZNM areas. The moisture interval is 1 g kg−1, and the standard vector is 10 m s−1. In (a),(b), contours larger than 8 g kg−1 are shaded, and in (c) difference values between −3 and 3 g kg−1 are shaded. A mask to remove the moisture and wind vectors corresponding to altitude lower than 1500 m (~850 mb) was applied.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
In previous studies (Mo et al. 2005, 2007), the authors compared NARR against observations and found that, in general, over all of the domain covered by the reanalysis (all of North America), the winter atmospheric transport and precipitation are realistic in NARR. During the summer season, the basic features and evolution of the NAM in NARR compare favorably with observations. The GPLLJ depicted by NARR is in agreement with the observations; however, NARR systematically overestimates the meridional southerly GCLLJ. NARR also overestimates the heat fluxes at the head of the Gulf of California. To minimize these errors, more soundings around the gulf area were added to the data assimilation process during the 2004 North American Monsoon Experiment (NAME) campaign (Mo et al. 2007). However, the inclusion of more soundings did not produce a significant improvement in the fluxes simulation over the GoC. The authors concluded that the NARR may not be suitable for studies of the GCLLJ and its relation with the monsoon precipitation (Mo et al. 2005).
1) HadRM3P–NCEP
Even though the NAM occurs during the summer season (JJAS) here we restrict the analysis to August (averaged over the 10 years simulated: 1989–98) since in this month not only is the NAM fully developed (e.g., Vera et al. 2006), but also the differences between HadRM3-NCEP precipitation and the observations are largest over the two study regions. In contrast, NARR matches almost perfectly the observed monsoon precipitation in AZNM but underestimates it in the CORE. Also, in this month there are clear differences between moisture and wind vectors from NARR and HadRM3P–NCEP.
In August (Fig. 6a), the GoM has in average 20% more moisture in NARR than in HadRM3P–NCEP, but the CORE monsoon region and the eastern tropical Pacific are more humid in the HadRM3P–NCEP than in the NARR (Fig. 6b). However, NARR shows a moisture tongue and an intense jet from the GoC (as described in Mo et al. 2005, 2007) to the CORE monsoon that is not seen in HadRM3P–NCEP. The GPLLJ in NARR advects moisture from the GoM to the Great Plains and toward the west of the AZNM region where the winds, owing to the presence of orography, start to curve. As the altitude increases (700 mb, Fig. 7b) the GPLLJ advects moisture to the CORE region and to the AZN, and the GPPLJ then curves to form an anticyclone circulation, centered over Texas.
Mean August specific humidity and wind vector at 700 mb: (a) HadRM3P–NCEP, (b) NARR, and (c) NARR minus HadRM3P–NCEP. Squares indicate the position of the CORE and AZNM areas. The moisture interval is 1 g kg−1, and the standard vector is 10 m s−1. In (a),(b), contours larger than 5 g kg−1 are shaded. In (c), difference values between −3 and 3 g kg−1 are shaded.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
In contrast, the GPLLJ in HadRM3P–NCEP is more intense but narrower than in NARR, advecting moisture from the GoM only toward the north and northeast of the United States (Figs. 6b,c). In HadRM3P–NCEP there is almost no influence from the eastern tropical Pacific nor from the GPLLJ over the AZNM region, but there is an evident intrusion of a dry air mass from the subtropical Pacific dominating the AZNM region, which explains why in the HadRM3P–NCEP experiment the AZNM region has 50% less moisture than in NARR (Fig. 6). The influence of the Pacific in the AZNM region is more evident at 700 mb (Fig. 7a), where the narrower and stronger GPLLJ does not reach AZNM and the anticyclone is located farther west (over the west coast of the United States). Over the SMO and CORE regions there are easterly winds (from the GoM), and this region has more moisture in HadRM3P–NCEP than in NARR (Fig. 7c).
Off the southwest coast of Mexico, in the eastern tropical Pacific, HadRM3P–NCEP has 50% more moisture than NARR. The excess moisture and an anomalous cyclonic circulation over this region is persistent during the JAS season.
The CORE region, including the Gulf of California, has more moisture (around 20% more) in HadRM3P–NCEP than in NARR. In general, the HadRM3P–NCEP CORE region has more moisture during the NAM season, as well as in October. The excess of moisture in October seems to enhance the occurrence of daily events of extreme precipitation (not shown here), which could explain the overestimation of HadRM3P–NCEP CORE precipitation in this month (almost double the observed precipitation; see Fig. 2).
2) HadRM3P–ERA-40
In the HadRM3P–ERA-40 the simulation of moisture over the GoM and the direction and intensity of the GPLLJ compares favorably with NARR at 850 mb (Fig. 8). Over the eastern tropical Pacific and the CORE region moisture in HadRM3P–ERA-40 is larger than in NARR, but the tongue of moisture from the CORE region that reaches AZNM in NARR is not seen in HadRM3P–ERA-40, as was the case for HadRM3P–NCEP. Although the differences in moisture over AZNM are smaller than for HadRM3P–NCEP, in HadRM3P–ERA-40 there is also an influence of dry air from the subtropical Pacific toward the AZNM region explaining why in this month the peak of precipitation is not as high as in NARR or the observations (HadRM3P–ERA-40 has around 30% less rain than observed).
Mean August specific humidity and wind vector at 850 mb (August). The squares indicate the position of the CORE and AZNM areas. The moisture interval is 1 g kg−1, and the standard vector is 10 m s−1. (a) HadRM3P–ERA-40 (contours larger than 8 g kg−1 are shaded) and (b) NARR minus HadRM3P–ERA-40 (difference values between −3 and 3 g kg−1 are shaded). A mask to remove the moisture and wind vectors corresponding to altitude lower than 1500 m (~850 mb) was applied.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
At 700 mb the GPLLJ in HadRM3P–ERA-40 (Fig. 9), as in NARR, reaches the CORE region and from there diverts toward AZNM where, due to the complex topography, an anticyclone is formed centered over Texas as in NARR. This anticyclone enhances the easterlies over Mexico, advecting moisture toward the CORE region and intensifying the monsoonal precipitation over that region. The position and strength of the anticyclone determine the intensity and onset of the NAM—if the anticyclone is shifted also, the region of divergence at upper levels and convergence at lower will be displaced (Higgins et al. 1997; Douglas et al. 1993).
Mean August specific humidity and wind vectors at 700 mb: (a) HadRM3P–ERA-40 and (b) NARR minus HadRM3P–ERA-40. The squares indicate the position of the CORE and AZNM areas. The moisture interval is 1 g kg−1, and the standard vector is 10 m s−1. In (a), contours larger than 5 g kg−1 are shaded. In (b), difference values between −3 and 3 g kg−1 are shaded.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
3) NCEP
Are the differences observed between the experiments derived from the boundary and initial conditions, or is the moisture/rainfall issue related to the inherent physics of HaDRM3P? The ensemble experiment shows that HadRM3P is a model constrained to stay close to the boundary conditions, suggesting therefore that the problem comes from the dataset used to drive the regional model. Hence, here we analyze NCEP and ERA-40 moisture and wind vectors for the same period of the simulations (1989–98).1 Based on the findings of Mo et al. (2005) and Mo et al. (2007) that NARR represents well the moisture flux over the GoM and the GPPLJ, it is assumed that NARR is the best proxy for the dynamics over the region; thus, we use this dataset also in the analysis.
The GoM in NCEP is drier than in NARR, and the NCEP GPLLJ over the GoM is stronger than in NARR. These two features are persistent during the JAS season. This drier bias is then transmitted to HaDRM3P–NCEP at lower and upper levels (Fig. 10a and Fig. 10c).
Mean August specific humidity and wind vector at 850 mb: (a) NCEP, (b) ERA-40, (c) NARR minus NCEP, and (d) NARR minus ERA-40. The squares indicate the position of the CORE and AZNM areas. The moisture interval is 1 g kg−1, and the standard vector is 10 m s−1. In (a),(b), contours larger than 8 g kg−1 are shaded. In (c),(d), difference values between −3 and 3 g kg−1 are shaded. A mask to remove the moisture and wind vectors corresponding to altitude lower than 1500 m (~850 mb) was applied.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
4) ERA-40
ERA-40 moisture values over the GoM and the intensity of the GPLLJ over the GoM are consistent with those of NARR (Fig. 10b and Fig. 10d). HadRM3P–ERA-40 is sensitive to the information on that boundary and the differences between that experiment and NARR over the GoM and U.S. East Coast are almost zero (for wind vectors and moisture).
For the west coast of Mexico ERA-40 has 20% more moisture than in NARR and NCEP. There are no large differences in that region between HadRP3P–ERA-40 and HadRM3P–NCEP, probably because of the presence of the cyclonic circulation that concentrates the moisture in that region and does not allow the moisture to move northward. This cyclonic circulation over the Pacific is present in ERA-40 and the two HadRM3 experiments, but NCEP and NARR do not exhibit this feature. The most noticeable differences between NCEP and ERA-40 are the amount of moisture over the GoM and the intensity and direction of the GPLLJ. These features seem to be transmitted to the regional model. When the regional model is fed with ERA-40 reanalysis, which has similar values of moisture over the GoM and similar GPLLJ (intensity and direction) to those of NARR, the precipitation over AZNM is more realistic.
It has been suggested that the influence of the GoM on the NAM only occurs at upper levels (Schmitz and Mullen 1996; Turrent and Cavazos 2009) and at lower levels its influence on the monsoon is minor, at least at the initial and medium stages (Turrent and Cavazos 2009). However, this work shows that the low-level (850–700 mb) moisture, originated from the GoM and advected by the GPLLJ, plays an important role in the occurrence and intensity of the NAM rain in the AZNM region. Although the realistic simulation of moisture from the GoM improves the simulation of rain in the AZNM, there is still a component from the eastern Pacific and the GoC that needs to be resolved not only in the HadRM3P but in NARR as well. As one of the reviewers pointed out, more detailed analysis between the model output and observations (e.g., soundings) is needed to properly assess the impacts of the moisture flux from the GoM to the CORE area as well as the impacts of the eastern Pacific. This issue will be part of another paper.
c. Climate change implications
Until now, the efforts from the regional modeling groups that work on the NAM have concentrated on developing better models and/or parameterization (e.g., Gochis et al. 2002, 2003; Xu et al. 2004), but the impact of the boundary conditions has not been analyzed for the NAM.
The findings of this work suggest that more realistic moisture and winds from the eastern tropical Pacific and the GoM are relevant for an appropriate simulation of precipitation in the AZNM region. This result raises the question of what are the implications for the prediction of possible changes in the NAM under climate change. Since the NAM region has a semiarid climate, the impacts of droughts/floods are devastating; therefore, the capacity to predict the precipitation is a priority for the governments of Mexico and the United States.
To determine the realism of GCMs used in climate change studies, we compared NARR moisture and wind vectors with the results from two of the 23 GCMs that participated in the recent Coupled Model Intercomparison Project phase 3 (CMIP3) (Covey et al. 2003) for the Fourth Assessment Report of the IPCC. Among all of the IPCC models, HadCM3 was chosen because in a recent study of global warming in Mexico (Montero-Martinez et al. 2008) this model best reproduced the current Mexican precipitation and temperature climatology. The second climate model, the anthropogenically forced medium-resolution Model for Interdisciplinary Research on Climate [MIROC(medres); K-1 Model Developers (2004)] was selected because it was found that this model captures the hydrological cycle in the western United States and reproduces well the Pacific decadal oscillation (PDO) (Barnett et al. 2008). Castro et al. (2001) found that SST anomalies over the Pacific are correlated to a high (low) PDO phase and El Niño (La Niña), a northerly (southerly) shifted monsoon ridge, and a late (early) monsoon onset and below (above) average early monsoon precipitation (Grantz et al. 2007).
In HadCM3 (Fig. 11a), the GoM is drier than in NARR; the largest differences in moisture between the reanalysis and the model are found over the CORE and AZNM regions where HadCM3 is more than 5 g kg−1 drier that NARR (Fig. 11c). The winds are poorly resolved with an unrealistic circulation over the entire domain. The CALLJ does not exist in the HadCM3 simulation or it is merged with the GPLLJ that has an unrealistic component from the west coast of Mexico.
Mean August specific moisture and wind vectors at 850 mb: (a) HadCM3, (b) MIROC, (c) NARR minus HadCM3, and (d) NARR minus MIROC. In (a),(b), contours larger than 8 g kg−1 are shaded. In (c),(d), difference values between −3 and 3 g kg−1 are shaded.
Citation: Journal of Climate 24, 11; 10.1175/2011JCLI3846.1
In the MIROC simulation, during July and August (Fig. 11b and Fig. 11d), the GPLLJ and CALLJ are in agreement with those of NARR; however, the GoM is drier (around 50% less moisture) than in the NARR. The CORE and AZNM regions are influenced by the dry and cold air masses from the North Pacific; hence, there is a lack of moisture in these months over these areas. RCMs driven by these GCMs therefore would not give realistic simulations of the current climate of the region and therefore would not offer a realistic projection of climate change of the NAM.
4. Conclusions
To study the key mechanisms that control the precipitation in the North American monsoon HadRM3P, a regional atmospheric model, was configured for the NAM region and two 10-yr simulations and a 10-yr four member ensemble of simulations was performed.
Analysis of the ensemble experiment demonstrates that seasonal precipitation in HadRM3P is constrained by the driving boundary conditions. Individual members of the ensemble are highly correlated with each other, so we are confident that a single experiment is representative for the analysis carried out in this study.
The seasonal precipitation in the AZNM region in the experiment where HadRM3P was driven with ERA-40 (HadRM3P–ERA-40) has a correlation of 0.82 with the observations, which is almost double that of the experiment in which the regional model was driven with NCEP (HadRM3P–NCEP). The largest differences between both regional simulations (HadRM3P–NCEP and HadRM3P–ERA-40) occur in the low-level (850–700mb) moisture over the Golf of Mexico and in the intensity and direction of the Great Plains low-level jet. These differences in the moisture over the GoM and the GPLLJ are derived from the boundary conditions used to drive HadRM3P.
After a thorough comparison of the moisture and wind vector at upper and lower levels between reanalysis of observations and the regional models used, we conclude that the moisture from the GoM advected by the GPLLJ also plays an important role in the AZNM precipitation, therefore a realistic simulation of these features is necessary for a more accurate simulation of precipitation in the NAM.
It has also been suggested that the influence of the GoM for the AZNM precipitation takes place at middle–upper levels above 700 mb (Schmitz and Mullen 1996). Although in that work the authors define a slightly different region, covering mainly just Arizona. The regional model results analyzed here show that, if there is not enough moisture from the GoM and the GPLLJ does not reach the boundary margin of the AZNM region below the 700 mb, then a dry westerly flow dominates that area and the summer precipitation is small. In the upper levels of the GPLLJ an anticyclone is apparent that produced divergence and, if the GPLLJ is not simulated appropriately, the anticyclone is shifted and less intense. The intensity and position of the anticyclone determine the intensity and onset of the NAM. The anticyclone enhances easterlies that advect moisture toward the CORE region at 700 mb. This moisture is then transported toward the AZNM region.
These findings also have implications for studies of climate change; two of the most commonly used GCMs that simulate well the NAM precipitation (HadCM3 and MIROC) do not reproduce correctly the GPLLJ intensity nor the moisture in the GoM. This would mean that the precipitation in the region would not be simulated accurately, even within a nested RCM. Thus, this study demonstrates that the simulation of moisture and winds should be improved when investigating the applicability of climate models for studying or predicting the climate and climate change of the region.
Acknowledgments
This research was funded by the CONACyT-Mexico through a Ph.D. scholarship granted to the first author. We wish to give thanks to David Hein and other members of the PRECIS team in the Met Office Hadley Centre for their support in installing and running HadRM3P. We would also like to acknowledge Dr. Tereza Cavazos for her very helpful suggestions. We also appreciate the comments of Dr. David Gochis and the two anonymous reviewers. Richard Jones was supported by U.K. government programmes [the Joint DECC and Defra ICP (GA01101, CBC/2B/0417_Annex C5)]. This work was also supported by Microsoft Research and the EU FP6 WATCH integrated programme.
REFERENCES
Amador, J. A., 1998: A climatic feature of the tropical Americas: The trade wind easterly jet. Top. Meteor. Oceanogr., 5, 1–13.
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, 16 877–16 896.
Barnett, T. P., and Coauthors, 2008: Human-induced changes in the hydrology of the western United States. Science, 319, 1080–1083.
Berbery, H., and J. Marengo, 2009: Activities on anthropogenic climate change. VAMOS Newsletter, No. 5, CLIVAR Project Office, Southampton, United Kingdom, 16–18.
Carleton, A. M., 1986: Synoptic-dynamic character of burst and break in the southwest U.S. summer precipitation singularity. J. Climatol., 6, 605–623.
Castro, C. L., T. B. McKee, and R. A. Pilke, 2001: The relationship of the North American monsoon to tropical and North Pacific surface temperatures as revealed by observational analysis. J. Climate, 14, 4449–4473.
Cavazos, T., and J. Marengo, 2009: Activities on anthropogenic climate change. VAMOS Newsletter, No. 5, CLIVAR Project Office, Southampton, United Kingdom, 3–6.
Cielsielski, P., and R. Johnson, 2008: Diurnal cycle of surface flows during 2004 NAME and comparison to model reanalysis. J. Climate, 21, 3890–3913.
Corbosiero, K. L., M. J. Dickinson, and L. F. Bosart, 2009: The contribution of eastern North Pacific tropical cyclones to the rainfall climatology of the Southwest United States. Mon. Wea. Rev., 137, 2415–2435.
Covey, C., K. M. AchutaRao, U. Cubasch, P. Jones, S. J. Lambert, M. E. Mann, T. J. Phillips, and K. E. Taylor, 2003: An overview of results from the Coupled Model Intercomparison Project. Global Planet. Change, 37, 103–133.
Douglas, M. W., 1995: The summertime low-level jet over the Gulf of California. Mon. Wea. Rev., 123, 2334–2347.
Douglas, M. W., R. Maddox, K. Howard, and S. Reyes, 1993: The Mexican monsoon. J. Climate, 6, 1665–1677.
Englehart, P. J., and A. V. Douglas, 2001: The role of eastern North Pacific tropical storms in the rainfall climatology of western Mexico. Int. J. Climatol., 21, 1357–1370.
Giorgi, F., and X. Bi, 2000: A study of internal variability of a regional climate model. J. Geophys. Res., 105, 29 503–29 521.
Gochis, D. J., W. J. Shuttleworth, and Z. 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. Yang, 2003: Hydrometeorological response of the modeled North American monsoon to convective parameterization. J. Hydrometeor., 4, 235–250.
Gordon, C., C. Cooper, C. A. Senior, H. Banks, J. M. Gregory, T. C. Johns, J. F. B. Mitchell, and R. A. Wood, 2000: The simulation of SST, sea ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. Climate Dyn., 16, 147–168.
Grantz, K., B. Rajagoopalan, M. Clark, and E. Zagona, 2007: Seasonal shifts in the North American monsoon. J. Climate, 20, 1923–1935.
Hallack-Alegria, M., and D. W. Watkins Jr., 2007: Annual and warm season drought intensity–duration–frequency analysis for Sonora, Mexico. J. Climate, 20, 1897–1909.
Higgins, R. W., and W. Shi, 2000: Dominant factors responsible for interannual variability of the summer monsoon in the southwestern United States. J. Climate, 13, 759–776.
Higgins, R. W., Y. Yao, and X. L. Wang, 1997: Influence of the North American Monsoon System on the U.S. summer precipitation regime. J. Climate, 10, 2600–2622.
Jones, R. G., J. M. Murphy, and M. Noguer, 1995: Simulation of climate change over Europe using a nested regional-climate model. I: Assessment of control climate, including sensitivity to location of lateral boundaries. Quart. J. Roy. Meteor. Soc., 121, 1413–1449.
Jones, R. G., M. Noguer, D. Hassel, D. Hudson, S. Wilson, G. Jenkins, and J. Mitchell, 2004: Generating high resolution climate change scenarios using HadRM3P. Met Office Hadley Centre Rep., 40 pp.
K-1 Model Developers, 2004: K-1 Coupled Model (MIROC) description. H. Hasumi and S. Emori, Eds., Center for Climate System Research, University of Tokyo K-1 Tech. Rep. 1, 40 pp.
Mesinger, F., and Coauthors, 2006: North American regional reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360.
Mo, K., M. Chelliah, M. L. Carrera, R. W. Higgins, and W. Ebisuzaki, 2005: Atmospheric moisture transport over the United States and Mexico as evaluated in the NCEP regional reanalysis. J. Hydrometeor., 6, 710–728.
Mo, K., E. Rogers, W. Ebisuzaki, W. Higgins, J. Woollen, and M. L. Carrera, 2007: Influence of the North American Monsoon Experiment (NAME) 2004 enhanced soundings on NCEP operational analyses. J. Climate, 20, 1821–1842.
Montero-Martinez, M. J., N. Pavon-Gonzalez, and J. Martinez-Jimenez, 2008: Future climate trends of precipitation and surface temperature in Mexico under global warming. Geophysical Research Abstracts, Fall Meeting 2008, Abstract GC51A-0660.
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.
Simmons, A. J., and D. M. Burridge, 1981: An energy and angular-momentum conserving vertical finite difference scheme and hybrid vertical coordinates. Mon. Wea. Rev., 109, 758–766.
Stensrud, D. J., R. L. Gall, S. L. Mullen, and K. W. Howard, 1995: Model climatology of the Mexican Monsoon. J. Climate, 8, 1775–1794.
Stensrud, D. J., R. L. Gall, and M. K. Nordquist, 1997: Surges over the Gulf of California during the Mexican monsoon. Mon. Wea. Rev., 125, 417–437.
Turrent, C., and T. Cavazos, 2009: Role of the land–sea thermal contrast in the interannual modulation of the North American Monsoon. Geophys. Res. Lett., 36, L02808, doi:10.1029/2008GL036299.
Vera, C., and Coauthors, 2006: A unified view of the American monsoon systems. J. Climate, 19, 4977–5000.
Vidale, P. L., D. Lüthi, C. Frei, S. Seneviratne, and C. Schär, 2003: Predictability and uncertainty in a regional climate model. J. Geophys. Res., 108, 4586, doi:10.1029/2002JD002810.
Xu, J., X. Gao, J. Shuttleworth, S. Sorooshian, and E. Small, 2004: Model climatology of the North American monsoon onset period during 1980–2001. J. Climate, 17, 3892–3906.
From the variables ingested to the HadRM3P model through the boundaries just the specific moisture and winds change dramatically in both simulations. SSTs and surface pressure were analyzed as well but there were not any remarkable differences between the simulations.
