ABSTRACT

This report documents a set of model output results generated by six independent modeling groups, whose collective efforts are known as the North American monsoon Model Assessment Project, or NAMAP. The simulations provide a set of comparable control runs for a single season (1990) in order to assess the ability of state-of-the-art regional and global models to simulate the seasonal evolution and diurnal cycle of atmospheric circulation, hydrometeorology, and land surface flux fields across the domain of the North American monsoon system. The motivation and model experimental setup is described, followed by a description of each of the six participating models. A subset of the full set of model output fields is shown, documenting the seasonal evolution and diurnal cycle of precipitation, surface flux and temperature, and low-level wind fields for the 1990 summer season. The report concludes with a set of potential metrics for assessing the quality of future simulations of the monsoon system.

The complete suite of model output fields is available online, in the public domain, accessible from the North American Monsoon Experiment web page hosted by University Corporation for Atmospheric Research's Joint Office for Science Support (UCAR JOSS).




1.      INTRODUCTION

This report provides a brief description of a set of numerical simulations of atmospheric variability across southwestern North America during the summer of 1990. The simulations were carried out independently by six modeling groups. The intent of this effort, called the North American monsoon Model Assessment Project or NAMAP, is to provide an indication of the state of the art of warm season climate modeling applied to the North American Monsoon System (NAMS). This initial modeling effort has been carried out in anticipation of more focused modeling activities as part of the North American Monsoon Experiment (NAME).

A complete description of NAME Science Goals is provided in a Science and Implementation Plan (NAME SWG 2003). One of the principal goals of NAME is to promote significant improvements in the capability of numerical models to simulate, and ultimately predict, warm season precipitation across the NAMS domain. A driving hypothesis of NAME science goals is that progress on modeling and understanding NAMS-related precipitation must include proper simulation of relatively small (spatial and temporal) scale climatic variability, especially the diurnal cycle across the region of maximum continental precipitation in northwestern Mexico (Higgins et al. 2003). Accordingly the NAMAP simulation protocol focuses on assessment of the diurnal cycle of precipitation and related meteorological and surface variables that serve to modulate warm season precipitation across southwestern North America.

NAME employs a multi-scale, tiered approach with focused monitoring, diagnostic and modeling activities in the heart of the monsoon region, on the regional scale, and on the continental scale, defined as Tiers 1, 2, and 3 respectively (Fig. 1). Each tier has a specific research focus aimed at improving warm season precipitation prediction, and activities related to each tier will proceed concurrently.  A primary objective of Tier 1 activities is to resolve the wind, temperature, and moisture fields at fine spatial and temporal scales around the Gulf of California and the slopes of the Sierra Madre Occidental, sufficient to develop stable monthly means during the summer.  Tier 2 focuses on broader regional-scale features over southwestern North America and the warm pool region to the southwest of Mexico.  The goal of activities in Tier 2 is an improved understanding and a more realistic simulation of intraseasonal variability of the monsoon and the low-level jet circulations that deliver moisture onto the continent from the Gulfs of California and Mexico (Fig. 1). Tier 3 focuses on aspects of the continental-scale monsoon and its variability, with the goal of improvements in understanding of spatial/temporal linkages between warm season precipitation, circulation parameters and the dominant oceanic and continental boundary forcing parameters.

The NAMAP activity has several motivations. It is intended to provide a benchmark for simulations of warm season precipitation, and the physical processes that control precipitation, in Tiers 1 and 2.  Six different modeling groups independently simulated a single warm season (1990) and archived a large set of model output described in detail in the following section.

The benchmark can be used by the entire NAME research community, as well as by the individual modeling groups as they plan sensitivity studies using their models.  In addition, examination of the cross-model variability, together with comparison to available observations, is intended to provide motivation for enhanced observations in data-poor areas during the NAME 2004 Enhanced Observing Period.

This technical note documents the performance of the six NAMAP models in simulating several of the key variables: precipitation, surface air temperature and fluxes, and low-level winds and moisture transport. Emphasis is placed on the diurnal cycles of these fields, motivated by the NAME working hypotheses regarding the importance of understanding the diurnal cycle and the seasonal evolution of the North American monsoon.  Section 2 explains the NAMAP modeling methodology and describes the individual models, and Section 3 provides brief summary descriptions of the model output. Section 4 is a brief discussion that includes some metrics as targets for future model simulations based on the NAMAP results.



2.      MODELING METHODOLOGY

From its conception, NAMAP has been designed to provide benchmark statistics for a wide range of dynamical models with quite different spatial and temporal resolutions, computational domains and physical parameterizations. Other than a common ocean surface temperature prescription, little effort has been made to constrain the specific modeling methodologies employed in the individual simulations. Because of these methodological differences, the six NAMAP simulations are not suitable for determining a "best" model and should not be interpreted for this purpose -- NAMAP is a "Model Assessment Project" and definitely not a "Model Intercomparison Project" such as AMIP (Gates et al. 1998) or CMIP (Meehl et al. 2000).

Table 1 presents a comparative summary of the model characteristics for each simulation. Individual model descriptions are outlined somewhat more completely in Section 2.2, which includes citations of more thorough documentation of the dynamical and physical characteristics of each model.

Perhaps the most fundamental distinction that should be drawn is between the four regional models and the two global models. The regional models are strongly forced by time varying atmospheric observations around the lateral boundaries of the model's computational domain. No such lateral atmospheric forcing is imposed on the global models, which are forced only by prescribed lower boundary conditions. When interpreting the results, it is important to keep in mind that the regional models are much more strongly constrained by the continual imposition of "correct" large-scale dynamical features at the lateral boundaries.

As outlined in Section 2.2 and in Table 1, the regional model simulations described herein were run with significantly different configurations. Several of these simulations are continuous simulations beginning with springtime initial conditions, while one is a succession of shorter-term simulations. Different land surface conditions and convection parameterizations are employed, and recent sensitivity studies carried out using several of these same models has demonstrated the large variability of summer precipitation associated with these different physical treatments (e.g. Gochis et al. 2002; Xu and Small 2002; Kanamitsu and Mo 2003).  Thus it should not be surprising to find a range of model simulation results commensurate with the wide variety of model characteristics employed for the NAMAP simulations.

2.1    Boundary Conditions and Archiving Protocol

The NAMAP protocols were developed to facilitate a common set of oceanic boundary conditions for the simulations. The year 1990 was chosen for simulation because precipitation was unusually intense that year, providing a vigorous monsoon season for simulation. In addition, the SWAMP 1990 field campaign occurred that year (Stensrud et al. 1995), providing some published analyses and model simulations for reference.

Lateral boundary conditions for the regional models were derived from NCEP/NCAR reanalysis fields, typically imposed four times daily along the outer periphery of the computational domain for the regional models. No such lateral boundary forcing was imposed on the global models, which are constrained only by lower boundary conditions. When interpreting the results, it is important to keep in mind that the regional models are much more strongly constrained than the global models by the continual imposition of “correct” large-scale dynamical features at the lateral boundaries.

Time-varying ocean surface temperatures are specified as a lower boundary condition using the NOAA Optimum Interpolation Sea Surface Temperature Analysis. This product, known as OI.v2, has 1°´1° spatial resolution and weekly temporal resolution. It is available with documentation online at URL:

http://www.emc.ncep.noaa.gov/research/cmb/sst_analysis/

Other boundary conditions (topography, land surface, etc.) are prescribed individually for each model. Thus the only 1990-specific boundary condition for the global models is the prescribed ocean surface temperature fields. The lack of common specification of continental lower boundary conditions potentially allows for considerable inter-model variability in land surface fluxes, and indeed the comparative analysis of the model output shows such variability.

The archiving strategy for NAMAP simulations is designed to minimize data storage and transmission requirements while preserving two of the principal climatic modes of variability of interest to NAME: the diurnal cycle and seasonal evolution of the NAMS. Thus the protocol calls for archiving monthly average maps of key surface and atmospheric variables within Tier 1 separately for each hour of the day, e.g. 24 maps valid at 0000Z, 0001Z, ..., 2300Z, each of which is a 30- or 31-day average for each month of the 1990 warm season. A set of hourly monthly mean maps was archived for each variable of interest and for each month of simulation. A smaller set of variables was archived for Tier 2 to describe the larger scale circulation. Monthly average maps of Tier 2 variables are archived with 12-hr (0000Z and 1200Z) resolution.

The description of the NAMAP simulations in this Atlas covers only a few of the monthly mean fields available for analysis. All archived output files for the NAMAP runs are freely and publicly available online from the NAME Project Office, maintained by the UCAR Joint Office for Science Support (JOSS). The complete list of fields archived at JOSS is given in Appendix 1. The main NAMAP web page hosted by JOSS contains links to the modeling protocols, brief descriptions of each model, and an online order form for obtaining output via anonymous ftp. It is not necessary to have a computing account at UCAR to obtain these data. The URL for the NAMAP web page at JOSS is:

http://www.joss.ucar.edu/name/namap/index.html

2.2    Description of Individual Participating Models

The NAMAP runs were generated using standard versions of several well-known numerical models, without special adjustments of physical parameterizations for the challenging simulation conditions of the North American monsoon domain (high and sharp topography near a coastline). A summary of the model dynamics and physics packages is, provided in Table 1 and short descriptions of the models follow. The "time series label" listed for each model corresponds to the figure legends in time series plots (Fig. 3, 7-15).

         2.2.1       Regional Spectral Model / NOAA NCEP [time series label: rsm_hj]

A hydrostatic version of the RSM (Juang and Kanamitsu 1994; Juang et al. 1997), implemented by Dr. Henry Juang of the NOAA Climate Prediction Center, was used for NAMAP. The uniqueness of the NCEP RSM is its perturbation for spectral computation. The definition of the perturbation is the difference between the value obtained from the regional model and the value obtained from the global model.  The time integration comprises linear and nonlinear computations. Linear computation is all perturbation on spectral space, including all numerical treatments such as semi-implicit, horizontal diffusion, and time-filter; and nonlinear computation is full value on physical space, including all dynamical forcing, model physics and lateral boundary relaxation. The model physics includes long-wave and short-wave radiation with aerosol, ozone and cloud interaction; a simple three layer soil model; a surface physics package including a non-local boundary layer formulation; gravity-wave drag; deep convection parameterized by a simplified Arakawa-Schubert scheme; shallow convection and large-scale precipitation; and simplified hydrologic budgets for snow depth and surface water runoff.

         2.2.2       Regional Spectral Model / SIO ECPC [time series label: rsm_mk2]

Another version of the RSM analyzed for NAMAP is implemented by Dr. Masao Kanamitsu at the Experimental Climate Prediction Center, Scripps Institution of Oceanography. This version of the RSM is the same as that used by Kanamitsu and Mo (2003).  Its convective parameterization is that of RAS.  The horizontal moisture diffusion coefficient is set to zero to prevent anomalous precipitation over steep orography.  Run 2 is a large number of 36-hour integrations starting from 0000Z each day.  The submitted fields are from 24-36 hour forecasts.  The lateral boundary conditions are taken from 12-36 hour forecasts of Tier 2, which is also a large number of 36-hour forecasts using Reanalysis-2 as lateral boundary conditions.  The nesting for Run 2 was done every hour.

Note that the principal features of model dynamics and physics are the same for the two versions of RSM analyzed for NAMAP (implemented at NCEP and ECPC). However, it will be shown that there are significant differences in the model simulations from these two implementations. As indicated in these short descriptions and in Table 1, the chief differences in the physical packages used by these models are (i) different moist convective parameterizations, (ii) different longwave radiation codes, and (iii) implementation of horizontal moisture diffusion at NCEP but not at ECPC.

Three separate NAMAP runs were carried out using the RSM/ECPC model with output archived at JOSS. Experiments 1 and 2 are the same except for a difference in the evolution of prescribed surface conditions. In Experiment 1, the surface conditions do not change in time from their prescribed 1 May 1990 initial conditions. In Experiment 2, the prescribed (from Reanalysis-2) surface conditions evolve in time. In Experiment 3, land surface conditions evolve interactively in time. This Atlas describes RSM/ECPC output derived from Experiment 2.

         2.2.3       Regional MM5 Model / University of New Mexico [time series label: mm5_lr]

The model is version 5.3 of the PSU/NCAR mesoscale model (MM5), described originally in Anthes et al. (1987), implemented by Dr. Elizabeth Ritchie at the University of New Mexico. The model is configured with two domains using a Lambert-Conformal projection. The coarse mesh is 100´100 grid points at 45 km resolution, centered at 25°N,105°W. The fine mesh is 121´151 grid points at 15 km resolution, and is embedded in the coarse mesh such that fine mesh grid point (1,1) is located at (19,34) in the coarse mesh. The coarse grid supplies boundary values to the fine mesh, which in turn feeds information back to the coarse grid over the entire fine mesh domain after the fine mesh grid integration step is completed. The two-way interaction ensures that the fine mesh structure is well represented on the coarse mesh.

There are 23 vertical sigma-p levels. The model top is 50 mb, and a radiative boundary condition is used at the top of the model. Boundary layer processes are determined from the scheme of Hong and Pan (1996), as implemented in the NCEP-MRF model. Convective processes are parameterized on both grids using the Kain-Fritsch cumulus parameterization (Kain and Fritsch 1993) and resolved convection is treated using the Goddard mixed-phase microphysical scheme (Tao et al. 1993), which includes cloud, rain, ice, snow, and graupel types. Cloud radiation is represented by a scheme that allows for longwave and shortwave interactions with explicit cloud and clear air. Over land, a 5-layer soil model (Dudhia 1996) is implemented that vertically resolves diurnal temperature variation allowing for more rapid response of surface temperature.

         2.2.4       Regional Eta Model / NOAA NCEP [time series label: eta_km]

The configuration of the NCEP mesoscale Eta model for seasonal climate modeling (including for NAMAP) is implemented by Dr. Kenneth Mitchell and Dr. Rongqian Yang of NCEP's Environmental Modeling Center.  The Eta model is a hydrostatic grid-point model that utilizes the "Eta" step-mountain vertical coordinate.  The advantage of this coordinate is that the resolution of the underlying topography can be represented at the full resolution of the native computational grid being applied, rather than having to utilize a smoothed topography common in many sigma coordinate models (the latter typically smooth the underlying terrain to remove terrain features smaller than 6-8 grid increments).  For NAMAP, the Eta model was executed in a configuration very similar to the Eta model configuration in NCEP's recently completed 24-year (1979-2002) North American Regional Reanalysis (NARR), including 45 vertical levels and 32-km horizontal resolution over a very large computational domain encompassing all of North and Central America. This NARR version of the Eta model utilizes virtually the same parameterized physics as was operational in the Eta model at NCEP as of 21 July 2001.  The overall Eta model configuration is described in Janjic (1990), including the PBL parameterization. Janjic (1994) describes the Eta model shortwave and longwave radiation parameterizations and the deep and shallow convection (known as Betts-Miller-Janjic).   Subsequently, Janjic (2000) and Janjic (2001) present later updates of the Eta model convection scheme and PBL scheme, respectively.  For the Eta model in NARR and NAMAP, Zhao and Carr (1997) describe the cloud microphysics package, while Ek et al. (2003) describe the Noah land-surface component.  The initial conditions for the Eta model land states in the NAMAP run were obtained from the NCEP/DOE Global Reanalysis 2 (Kanamitsu et al., 2002).

Eta model results shown in this Atlas are based on a simulation following the standard NAMAP protocols of initialization time (0000Z May 1 1990) and prescribed ocean surface temperature.  Results from several additional Eta runs are also archived at UCAR JOSS. One of these runs was initialized at 1200Z April 29 1990. Another simulation uses a different prescription of SST in the Gulf of California.

         2.2.5       Global Seasonal Forecast Model / NOAA NCEP [time series label: sfm_js]

The global SFM is implemented by Dr. Jae Schemm at NOAA's Climate Prediction Center. A detailed description of the SFM can be found in Kanamitsu et al. (2002).  For NAMAP, the model was run at T62 resolution (about 200 km) in the horizontal and 28 levels in the vertical.  Major physical processes in the model include the Relaxed Arakawa-Schubert convective precipitation scheme (Moorthi and Suarez 1992), longwave radiation (Chou and Suarez 1994), shortwave radiation (Chou 1992; Chou and Lee 1996), cloud-radiation interaction, non-local vertical diffusion (Hong and Pan 1996), gravity wave drag (Alpert et al. 1988), the Oregon State University land surface model (Pan and Mahrt 1987), and mean orography.

         2.2.6       Global NASA Seasonal-Interannual Prediction Project Model (NSIPP) [time series label: nsipp_pp]

The global NSIPP-1 AGCM was developed at NASA Goddard Space Flight Center as part of the NASA Seasonal-to-Interannual Prediction Project (Bacmeister et al. 2000). The dynamical core is described in Suarez and Takacs (1995). The horizontal dynamics are calculated using a 4th-order differencing scheme with explicit leapfrog time differencing. The model configuration used for NAMAP has a horizontal resolution of 0.5° latitude and 0.625° longitude. The vertical coordinate is a standard sigma-coordinate. There are 34 vertical levels with increasing resolution in the lower 2 km of the atmosphere (<200m). Vertical differencing follows Arakawa and Suarez (1983).

A simple K-scheme is used in the boundary layer, in which turbulent diffusivities for heat and momentum are calculated based on Monin-Obukhov similarity theory (Louis et al. 1982).  The AGCM uses the relaxed Arakawa-Schubert (RAS) scheme to parameterize convection (Moorthi and Suarez 1992). The RAS scheme effectively acts as a parameterization of both deep and shallow convection in this model. The parameterization of solar and infrared radiative heating used in the model is described in Chou and Suarez (1994, 1999). The land surface model is the Mosaic model developed by Koster and Suarez (1992, 1996).



3.      DESCRIPTION OF MODEL OUTPUT PLOTS

3.1    Tier 1 fields of precipitation and low-level wind

Maps of total monthly precipitation across southwest North America for June, July, and August 1990 are shown in Figs. 2.0-2.6 . The first set of maps in Fig. 2.0 is derived from the land-only Unified Raingauge Dataset (URD) (Higgins et al. 2000b). Model-simulated precipitation fields analogous to the observed fields are shown in the six figure panels numbered 2.1-2.6. The spatial areas denoted "CORE" and "AZNM" in each of these maps defines a continental area used as an index for seasonal and diurnal time series.

The set of observed precipitation maps in Fig. 2.0 describes a relatively typical monthly monsoonal evolution, although it should be noted that 1990 was one of the wettest year on record in the CORE domain (Gutzler 2004) so the amplitudes of precipitation are unusually high. Monsoon onset occurs during June in most of western Mexico south of about 30°N, while the extension of the monsoonal domain north of 30°N remains very dry. Rainfall intensifies strongly and rapidly in July, with maximum rainfall rates (in excess of 1 cm/d in July 1990) along the western slopes of the Sierra Madre Occidental. The monsoon continues in August, with rainfall diminishing somewhat relative to July throughout northwest Mexico, but continuing at about the same lower rate as in July (generally between 5 and 10 cm/month) in Arizona and New Mexico.

Although care has been taken to incorporate as much data as possible into the URD observations, it should be kept in mind that these data are based exclusively on gauges, which tend to be located preferentially in valleys, and not on steep slopes or mountaintops where precipitation rates are higher due to orographic effects. In addition the 1°´1° gridding process tends to smooth out smaller areas of sharp gradients in precipitation. Therefore the observations are likely to be biased low in the areas of maximum precipitation on the slopes of steep terrain. An alternative analysis of summer 1990 data (Stensrud et al. 1995) suggested that total monthly precipitation in July may have exceeded 50 cm in a narrow zone along the western mountain slopes between 24°N and 28°N, where the coarser gridded URD observations show monthly precipitation only half that amount.

All of the four regional models reproduce the July seasonal precipitation maximum across northwest Mexico (Figs. 2.1-2.4), which is more heavily and tightly concentrated in the high topography of the "CORE" monsoon region (112°-106°W, 24°-30°N) in the regional models than in the observations (Fig. 2.0).  Models 2 and 3 (RSM/ECPC and MM5) exhibit the most intense precipitation. Both of these models have more intense and more widespread precipitation than the RSM/NCEP and Eta in Mexico in June; MM5 and (to a lesser extent) Eta both generate more precipitation during this month in New Mexico but this is not evident in the observations. In July all models exhibit a sharp increase in continental precipitation. In the Eta model significant precipitation is restricted to the continent whereas the other three models show precipitation spreading across the southern Gulf of California. Eta and RSM/NCEP produce no significant precipitation west of 112°W, in extreme northwest Mexico or western Arizona, in any of the summer months.

The seasonal evolution of precipitation in the two global models (Fig. 2.5, 2.6) is significantly different.  SFM, with the coarsest resolution, naturally shows a much smoother spatial distribution than the other models. Both global models show increases in precipitation across northern Mexico and the southwestern U.S. from July to August as precipitation spreads northward throughout the summer. Like Eta and RSM/NCEP, the global models generate no significant precipitation west of 112°W.

Monthly and diurnal time series of several surface variables, spatially averaged over the "CORE" and "AZNM" subregions in Fig. 2, are presented in Fig. 3 and Figs. 7-15. Gutzler (2004) showed that interannual variability of North American monsoon continental precipitation (as represented by URD observations) reaches its maximum in the CORE subregion, and demonstrated that precipitation anomalies in this region tend to persist from the early phase of the monsoon in June into the later monsoon months. The AZNM box has been used to generate an index of monsoonal activity on the U.S. side of the border in analyses by Higgins et al. (1997, etc.).

Time series of the month-to-month evolution of precipitation from May-September 1990 in the CORE and AZNM subregion, for both the observed data and the six models, are shown in Fig. 3.  The observations show a seasonal maximum in July in both regions.  All models simulate a seasonal maximum in July or August in both regions, and all models generate more precipitation in the CORE subregion relative to the AZNM subregion. The four regional models (which are provided observed lateral boundary conditions outside Tier 2) all reproduce the observed July maximum in the CORE subregion quite distinctly. MM5 and RSM/ECPC generate the most precipitation, consistently more than the URD observations; Eta generates considerably less precipitation than the observations; and RSM/NCEP follows the seasonal evolution of observed CORE precipitation quite closely. The same hierarchy of model precipitation rates is observed in the AZNM subregion, although RSM/NCEP produces a seasonal maximum in August rather than July.

Both global models tend to generate seasonal cycles of precipitation that peak somewhat later in the monsoon season than the observations. Both NSIPP and SFM produce peak precipitation in August in the CORE subregion; in AZNM the NSIPP seasonal peak is in September. The rather narrow spread of the envelope of SFM runs suggests that the delayed monsoon is common to all the ensemble members and is not just an artifact of initial conditions.

Time series of total monthly precipitation in the CORE and AZNM subregions are shown in Fig. 3. These time series cover two additional months not illustrated in Fig. 2, May (except for MM5) and September. The CORE time series confirm that all the regional models, like the observations, generate a clear seasonal precipitation maximum in July, whereas the seasonal maximum in both global models occurs in August. Precipitation is markedly less in May and September than in the three summer months in all models plus the observations, confirming the reproduction of the summer season monsoonal maximum. In AZNM the global models, and the RSM/NCEP regional model, increase precipitation from July into August while the other models (and observations) show declines from the July seasonal peak. RSM/NCEP generates a monthly precipitation time series remarkably close to the observed average (which contains considerable uncertainty). MM5 and RSM/ECPC consistently generate more precipitation than the observations indicate, while Eta and SFM generate less than observations.  NSIPP seems wetter than observations in the CORE subregion but drier than observations in AZNM.

The seasonal evolution of low level wind fields is illustrated in Fig. 4.  Each of these plots is a monthly average at 1200Z. This time of day was selected for study because it is close to the diurnal maximum in southerly flow and moisture flux in the vicinity of the Gulf of California identified in previous analyses (summarized by NAME SWG 2003) including the summer of 1990 (Stensrud et al. 1995). Subsequent plots of moisture transport will also focus on 1200Z data. When interpreting these plots it is important to keep in mind that these are monthly averages, which may underestimate the winds and moisture transport associated with transient gulf surge events (NAME SWG 2003).

The NCEP/NCAR reanalysis winds, at 2.5°´°2.5; resolution, are too coarsely defined to resolve the circulation associated with the Gulf of California (Fig. 4.0). Instead, the regime of the northwesterly winds that dominate the eastern Pacific circulation exhibit a smooth transition into southerlies associated with the Great Plains Low Level Jet, which have their maximum amplitude near 100°W. A higher resolution analysis of special data from the SWAMP experiment in July 1990 (Stensrud et al. 1995) indicate that a narrow band of southeasterly low level winds occurred over the northern half of the Gulf of California during July 1990, although that analysis did not clearly resolve whether the jet was centered over the Gulf (as depicted for a different year by Berbery 2001) or was topographically forced along the western slopes of the mountains to the east (as suggested by Anderson et al. 2001 and Fawcett et al. 2002).

Like the reanalysis data, the SFM model simulates northwesterly low-level winds extending onto the North American continent throughout the summer, consistent with the extension of the Pacific subtropical High circulation eastward across western Mexico and Arizona. The time averaged low level winds never shift to southerly along the west coast of Mexico in this simulation. The other global model, NSIPP, clearly shows a wind reversal from June to July on the western slopes of the Sierra Madre Occidental, with southerly winds extending southward to around 22°N below the mouth of the Gulf of California.

Eta winds change distinctly from June to July over the Gulf of California, from northwesterly in June to weak flow over the Gulf turning southwesterly onto the continent. Once onshore the Eta winds form a time averaged jet blowing almost straight north onto the high topography. North of 29°N these winds form a jet directed into southwestern Arizona. The Eta simulation presents the closest reproduction of the July 1990 1200Z wind field analyzed by Stensrud et al. (1995).

RSM/ECPC simulates very weak time-averaged winds over the Gulf of California, with a southerly component of the 925 hPa wind field evident only to the east of the Sierra Madre Occidental range. MM5, which generates high precipitation amounts comparable to RSM/ECPC, simulates a stronger southeasterly jet in both July and August along the coastal plain and western slopes of the Sierra Madre Occidental. RSM/NCEP exhibits southeasterly flow along the coastal plain north of 28°N in July and August. Both RSM simulations have strong low-level southerly winds to the east of the NAMS domain forming the Great Plains Low Level Jet. 

3.2    Longitude-height cross-sections of meridional moisture transport

To examine lower tropospheric moisture transports in the Tier 1 domain, longitude-height cross-sections of specific humidity and meridional wind were examined. The first set of these to be shown illustrate lower tropospheric conditions (1000-700 hPa) at 27°N from 115°W to 105°W, extending across the Gulf of California onto the high topography of the Sierra Madre Occidental near the middle of the Tier 1 domain (see Fig. 1). Previous analyses and model simulations (e.g. Stensrud et al. 1995) have shown that an integral part of the NAMS is a southerly low level jet extending up the Gulf of California toward Arizona, although the position of this jet relative to the Gulf and the topography has not been definitively pinned down. The Tier 1 cross-sections are followed by a second set of larger-scale cross-sections across Tier 2 at 30°N, from 120°W to 90°W. The purpose of these plots is to illustrate the joint seasonal evolution of low-level jets into southwestern North America (the NAMS domain) and into the central United States, as illustrated schematically in Fig. 1. The latter circulation feature is typically called the Great Plains Low Level Jet. 

The monthly evolution of meridional wind and specific humidity in Tier 1 from the NCEP/NCAR reanalysis during summer 1990 is shown in Fig. 5.0. At 27°N the Gulf of California extends zonally between 112°W and 110°W, a relatively low elevation coastal plain extends eastward to about 108°W leading to a sharp elevation gradient upward to the western slopes of the Sierra Madre Occidental. With its coarse 2.5° resolution the reanalysis cannot properly resolve the low level jet here, although a very small closed contour of southerly wind speed at 850 hPa is observed in August at the 112.5°W gridpoint. 110°W marks the easternmost longitude resolved in the reanalysis at 27°N below 700 hPa before the high topography begins. The observations also show a substantial increase in humidity from June to July.

Model simulations of this seasonal evolution are shown in Figs. 5.1-5.6. We did not have access to topography files for the models, and some models automatically interpolate atmospheric output downward to sea level through the topography.  Like the reanalysis, the SFM model is too coarse to resolve the distribution of water and land surfaces and the sharp topographic gradients along 27°N. Northerly winds, representing the eastern extension of the Pacific subtropical High, extend eastward to 108°W in each month.

The NSIPP model develops a very distinct low-level jet near 925 hPa between 109°W and 110°W in July and August. The jet is clearly displaced eastward from the near-surface humidity maximum near 110°-111°W that marks the Gulf of California and its high local evaporation. Note that from June to July NSIPP humidity increases sharply, and the southerly jet clearly emerges, but NSIPP produces very little precipitation in the CORE or AZNM region in July.

All the regional models (Figs. 5.1-5.4) clearly exhibit a near-surface humidity maximum near 110°W, extending westward toward 112°W, associated with the Gulf of California and its high local evaporation.  Furthermore each of the regional models develops a monthly-mean 1200Z low-level southerly wind maximum in the vicinity of the Gulf during July. The Eta model's southerly maximum is right over the maximum humidity; Eta's jet is the strongest (> 3 m/s) with maximum wind speed at the surface coincident with a shallow humidity maximum. MM5 and RSM/NCEP generate a weaker time-averaged 1200Z jet with an elevated wind speed maximum at 925 hPa amidst a deeper layer of high humidity than Eta exhibits. RSM/ECPC generates a southerly wind structure somewhat intermediate between the jets of the other models: like Eta the maximum wind speed occurs at the surface, located at the same longitude (109°W) as RSM/NCEP within a similar deep high-humidity layer.  Both RSM models eliminate the time-average 1200Z southerly winds over the Gulf in August, whereas, MM5 and Eta exhibit little change from July to August.

The seasonal evolution of larger scale moisture transport variability is considered in the cross-sections along 30°N, shown in Fig. 6. As in the Tier 1 cross-sections discussed previously, these results are monthly averages at 1200Z, a time when nocturnal low-level jet amplitudes have been shown to be large (Higgins et al. 1997).  The NCEP/NCAR reanalysis data for June shows a southerly low-level jet core exceeding 11 m/s at 850 hPa centered at 100°W (directed onto the North American continent over eastern Texas), embedded in moist lower tropospheric air with humidity exceeding 9 gm/kg (Fig. 5.0). In June the humidity west of the continental divide is low and near-surface meridional winds are northerly west of 112°W. The topographic divide at 107°W is the axis of a middle tropospheric ridge, with southerlies to the west and northerlies to the east above 650 hPa.

In July the observed high near-surface humidity values extend westward over the topographic divide and the mean near-surface winds have switched to southerly on the western flanks of the topography. The Great Plains low level jet at 100°W is 30% weaker than in June, and this jet continues to weaken in August while, on the western slopes, a modest closed contour of southerly wind speed develops.

Model simulations of this joint seasonal evolution are shown in Figs. 6.1-6.6. All the models generate a GPLLJ centered near 850 hPa and 100°W in all three months. The SFM simulation broadly reproduces the seasonal evolution described in the reanalysis, including the diminution of the low level jet at 100°W in successive summer months and the spread of humidity values exceeding 6 gm/kg in July to the west of the topographic divide at 107°W. Thus SFM reproduces the correct seasonal evolution of the Great Plains low-level jet without resolving the weaker Gulf of California jet associated with the North American monsoon.

NSIPP, the other global model, exhibits a stronger GPLLJ in August than in July. This jet then weakens in August, concurrent with an increase in humidity west of 110°W. Effectively the transition from July to August simulated by NSIPP is similar to the observed transition from June to July, i.e. the correct seasonal evolution occurs across the entire cross-section but delayed by a month relative to the observations.

MM5 and RSM/ECPC, the regional models that produce the most precipitation overall, exhibit somewhat different seasonal changes in low level winds and humidity across this section. In July and August MM5 clearly shows a diminished GPLLJ and a well-defined LLJ west of the topographic divide but the simulation of humidity seems deficient (probably too moist throughout the summer). Low-level humidity increases west of the divide in RSM/ECPC in a more realistic way from June to July, but time-averaged 1200Z winds remain weak and the GPLLJ near 100°W does not change much.

3.3    Monthly and diurnal variability of precipitation, temperature and fluxes in the CORE monsoon region

The diurnal cycle of total precipitation in the CORE subregion (Fig. 2) simulated by each model is shown in Fig. 7. The three panels of this and subsequent plots show diurnal time series averaged separately for the months of June, July, and August 1990. June is the monsoon onset month in this region of Northwest Mexico, and all models simulate a pronounced increase in total precipitation from June to July (Figs. 2,3).  Data are available hourly for four of the models (RSM/ECPC, MM5, Eta, and NSIPP) and every three hours for RSM/NCEP and SFM. 

Figure 7 shows that precipitation in most models exhibits similar diurnal phasing from month to month, but model to model differences are significant.  RSM/ECPC and MM5 generate the most precipitation in June and July; in both months MM5 produces more nocturnal precipitation between 0600Z and 1200Z. In contrast the NSIPP global model produces almost no nocturnal precipitation but sharpest and earliest (2100Z) afternoon diurnal maximum. This diurnal peak increases in amplitude with unchanged phase from June through August. RSM/NCEP produces a distinctly different diurnal cycle than RSM/ECPC, with the diurnal maximum in RSM/NCEP later in the day, despite the use of the same convection parameterization in the two models. The Eta model exhibits perhaps the most difference in diurnal cycle from month to month: the observed decrease in total precipitation from 9 cm to 7 cm between July and August (Fig. 3) occurs entirely as the result of diminished nocturnal precipitation, considering that the late afternoon (0000Z-0003Z) diurnal peak has the same amplitude and phase in the two months.

Much smaller differences between models are observed in the phasing of the diurnal cycles of surface air temperature ( Fig. 8), latent flux ( Fig. 9), and sensible flux ( Fig. 10) in the CORE subregion, although the amplitudes of the diurnal cycles vary considerably. All models show maximum temperature within an hour of 2100Z in all three summer months. The NSIPP model consistently generates the largest diurnal cycle amplitude. MM5, a relatively very wet model, consistently generates the smallest diurnal temperature amplitude. SFM is the warmest model at night; the other five models all generate very consistent nighttime temperatures, except for NSIPP in June which is quite cold relative to the other models.

The spread of simulated diurnal cycles of latent heat flux in the CORE subregion (Fig. 9) is broadly consistent with the temperature results. MM5 and RSM/ECPC generate afternoon latent flux exceeding 200 W/m2 in June and exceeding 300 W/m2 in July and August. The diurnal peaks in MM5 and RSM/ECPC latent flux occur between 1800Z and 2100Z, leading the corresponding peaks in precipitation generated in those models by about 3 hr. This lag time seems slightly longer, 4-6 hr, in the Eta model. In contrast, there is almost no lag in time between the latent heat and precipitation peaks in the NSIPP model, in which precipitation increases sharply during the afternoon hours, then seems to dry out very quickly following sharp 2100Z peaks in both precipitation and latent flux.

Sensible heat flux reaches its peak diurnal value very consistently between 1800Z and 2100Z in each model throughout the summer (Fig. 10). Not surprisingly, the models with largest latent flux tend to generate the smallest sensible fluxes (and conversely so). NSIPP and Eta generate the largest diurnal amplitudes of sensible flux, with diurnal peaks exceeding 250 W/m2, and these models also generate the largest diurnal cycles of temperature. 

As would be expected from inspection of Figs. 9 and 10, the models exhibit a large range of Bowen ratios (B = sensible flux/latent flux) in the CORE region; values of this ratio, averaged over the full diurnal cycle for each month, are shown in Fig. 11a.  In the MM5 and RSM/ECPC simulations B<1 throughout the summer, even in June (the driest month, before monsoon onset). At the other end of the range of values exhibited by the models, B>1 for Eta throughout the summer, even after monsoon onset. The global models, NSIPP and SFM, exhibit the smallest seasonal range of B, decreasing steadily from June through August. The RSM/NCEP model evolves from very dry (B>3 in June) to very wet after monsoon onset (B<0.5 in July and August). 

3.4    Monthly and diurnal variability of precipitation, temperature and fluxes in the Arizona-New Mexico region

Diurnal cycle plots of precipitation and surface fluxes for the AZNM subregion in Figs. 12-15 are plotted using the same ordinate scale as for the CORE region to facilitate comparison. The observations have hourly resolution within the United States (Higgins et al. 1996) so model-observations comparisons are included for precipitation ( Fig. 12). Obviously precipitation rates are much smaller in AZNM compared to the CORE subregion (Figs. 3, 7). Monsoon onset occurs in July in this northern reach of the NAMS domain. Observations indicate that a small amount of rainfall occurred during this month with a diurnal peak at 2300Z. RSM/ECPC generates about the same total precipitation at a later time of day, peaking at 0400Z. MM5 generates much more precipitation than the observations indicate, with peak rainfall between 2100Z and 0600Z but significant precipitation spread over the entire diurnal cycle.

Precipitation increases sharply in July in the observations and in each of the regional models. The observations suggest that the diurnal peak in precipitation occurs between 2300Z and 0300Z, and diurnal peaks within this time window are reproduced by RSM/ECPC, MM5 and Eta (and probably by RSM/NCEP although its three-hour resolution makes this comparison difficult). Like the observations, RSM/ECPC and Eta exhibit a very distinct diurnal maximum, with precipitation falling to a very small average rate (or zero) by 1200Z. MM5 produces precipitation right through the night, so that its total monthly rainfall in July exceeds RSM/ECPC (Fig. 3) despite MM5's smaller diurnal peak rainfall rate.

This statement remains true in August. RSM/NCEP also generates nocturnal rainfall in August; the July-to-August increase in total precipitation exhibited by RSM/NCEP is principally associated with larger nocturnal precipitation rates, not an increase in the average precipitation rate during the late afternoon convective peak. The global models also exhibit an increase in precipitation from July to August, but due principally to increases in precipitation rates between 1800Z and 2400Z. The observations also show an increase in peak precipitation at 2400Z in August but total precipitation is reduced from July. Inspection of Fig. 2.0 suggests that the overall decline is associated with less rainfall in the southwestern area of the AZNM subregion, whereas precipitation along the steep topography of the Mogollon Rim (where a sharp diurnal peak in orographic thunderstorm activity would be expected) actually increases.

Many of the principal features in the diurnal cycle of surface air temperature plots ( Fig. 13) are very similar to the corresponding plots for the CORE subregion (Fig. 9). The phasing of the diurnal cycle is very consistent across the models.  The NSIPP model generates by far the largest diurnal cycle amplitude, exhibits very high and unrealistic daytime high temperatures even after the onset of significant precipitation in August. The end of the pre-monsoon dry season seems to affect nighttime low temperatures in the NSIPP model, which increase considerably from June to August. As in the CORE subregion, MM5 exhibits the smallest diurnal cycle amplitude of temperature. Eta and RSM/ECPC are notable for their month-to-month consistency in temperatures throughout the three-month period.

Latent heat fluxes in the AZNM subregion ( Fig. 14) vary dramatically from model to model and month to month. In June RSM/ECPC generates the larger fluxes than MM5, peaking at 2000Z, although MM5 generates more precipitation. Latent fluxes jump sharply from June to July in all of the regional models (though the increases are quite different in each model) but much smaller increases are seen in the global models.  The declines in precipitation from July to August in the MM5, Eta and RSM/ECPC models are not matched by corresponding declines in latent flux. NSIPP exhibits a jump in both latent flux and precipitation from July to August. 

SFM, NSIPP, RSM/NCEP and Eta exhibit huge late afternoon (1800Z-2400Z) sensible fluxes throughout the summer in AZNM ( Fig. 15), although only NSIPP shows correspondingly huge daytime temperatures (Fig. 13).  The combination of very high sensible flux with very small latent flux in SFM, NSIPP and Eta models leads to Bowen Ratios that are extremely high (Fig. 11b; note different scale of y-axis in Figs. 11a and 11b). In June these models, plus the RSM/NCEP model, exhibit monthly average values of B exceeding 10.  Each of these models exhibits a pronounced drop in B after monsoon onset occurs (the timing of which is model-dependent). The two regional models that exhibit more daytime precipitation and lower daytime temperatures, MM5 and RSM/ECPC, show much smaller sensible fluxes that decrease from June values after monsoon onset in AZNM. Bowen ratios for RSM/ECPC are nearly the same in the CORE and AZNM subregions, whereas the other models all tend to exhibit considerably higher values of B in AZNM.



4.      DISCUSSION

This analysis of the NAMAP output fields necessarily addresses only a very limited number of issues illustrated by a small subset of the full NAMAP output archive (Appendix 1). Many other fields are available for analysis and could be used for other benchmark studies. The scope of this analysis has deliberately been limited to a few selected fields to motivate such studies and to provide feedback to the NAME community in advance of the NAME 2004 field campaign. 

The archiving protocol for NAMAP acts to limit some of the issues that could be addressed with more complete output files. In particular, the NAMAP protocol is designed for analysis only of time-averaged (monthly mean) fields, thereby filtering out the structure of sub-monthly transient modes of variability such as Gulf surge events. Other potential limitations of NAMAP include the possibility that the prescribed weekly SST fields are too coarse to capture some essential aspects of the temperature field, as suggested by Gao et al. (2003). Also, the analysis here includes only limited observational validation. Future NAME-related modeling studies will address these and many other issues, e.g. examination of the sensitivity of the NAMS onset and seasonal evolution to different SST forcing.

Despite these limitations, the NAMAP project clearly points to several key modeling issues that should receive a high priority for future NAME research. Time-averaged precipitation is a key variable for the monsoon circulation and for NAME. This analysis has illustrated that current models are capable of simulating the evolution of a summer season precipitation maximum near the observed continental core of the North American monsoon ("Tier 1" as defined for NAME, as shown in Fig. 1). There are, however, important differences in the monthly evolution and diurnal cycle of precipitation generated by the models. These differences motivate observational targets for the NAME field campaign in summer 2004, and associated performance metrics for assessing improvements in model simulation after the field campaign. 

The two global models assessed in this study both show significant delays in monsoon onset (defined in terms of precipitation) compared to observations-- on the order of a full month, although the monthly mean resolution of the NAMAP output makes this time lag impossible to quantify precisely. Onset date can be reasonably captured with the existing observational precipitation data set so there is little doubt that the delayed onset is a genuine deficiency in the global model simulations.

The diurnal cycle of precipitation unmistakably indicates the convective nature of monsoonal precipitation, as has been amply documented previously. It is apparent from the comparison of model precipitation output that the interaction of convection with topography is handled very differently in the various models. The results are certainly sensitive to the choice of convective parameterization employed, but other model differences also contribute to the wide range of afternoon precipitation rates simulated in these runs. More detailed diagnosis that could lead to model improvements will require improved observational data, in order to determine more confidently how close each model simulation is to reality. Simulation of intense convective precipitation in regions of extremely complicated terrain poses an exceptional challenge to dynamical models, and improvements in convection modelling are a fundamental prerequisite to enhancing predictablility of climate in the NAMS domain.

It is also clear from the model output that the late afternoon convective peak is not the only issue that models need to address. Substantial differences in nocturnal precipitation are evident, suggesting that systematic propagation of convective systems is occurring to varying degrees in the model simulations, and that different model physics and dynamical schemes are generating a transition to resolved precipitation in various ways. This issue could be addressed in more detail with other NAMAP data sets, especially with better observational data from the 2004 field campaign to establish ground truth.

The differences in model precipitation discussed above do not seem to be attributable in any obvious way to the time-averaged moisture transport simulated by the models. With the exception of SFM, which is run at a horizontal resolution too coarse to define the narrow southerly flow on the Pacific side of Mexican high topography, all the models represent some manifestation of a low level 1200Z jet structure that would transport moisture northward. It is possible that transients (such as gulf surges) represent the more important component of moisture transport, and these have not been considered in the NAMAP results. Nevertheless future studies should consider how strong a linkage exists between the seasonally evolving circulation along the Gulf of California and continental precipitation.

The six models exhibit dramatic differences in surface turbulent fluxes, especially in the afternoon hours when sensible and latent fluxes tend to be maximized.  Existing observations are not sufficient to provide firm constraints on these fluxes, although some of the monthly average afternoon surface temperatures simulated by the models are clearly unrealistic.  In order to increase confidence in the utility of the models for studying land surface feedbacks, a high priority should be placed on constraining the diurnal cycle of these flux terms. Although surface radiation terms were not described in this study, they too play a fundamental role in the surface energy budget.

Some of the pronounced inter-model differences described in this analysis cannot be reconciled with existing observations. At present, observational uncertainties in many surface variables, including precipitation and turbulent fluxes, are very large across the continental North American monsoon region. The NAME 2004 Intensive Observation Period (IOP) should yield improved observations that can help to constrain the model results. Among the recommendations we make for the field campaign, directly applicable to model uncertainties discussed in this report, are:

         *    In situ and radar estimates of precipitation should include high-quality large-scale spatial averages over at least a few selected areas. The CORE and AZNM regions highlighted in this modeling study would be good candidate areas for such estimates, considering the focus on these regions developed in several previous observational papers. Other regions could be defined for special analysis as well; the NAMAP archive allows retrospective analysis of any region within Tier 1.

         *    It is essential for the field campaign to define the structure of low level winds along the Gulf of California, both with respect to monthly averages and synoptic-scale transients. NAME 2004 should constrain the location of low level jet circulations and moisture transport pathways, and elucidate the linkage between the low level jet and precipitation. 

         *    The diurnal cycle of precipitation varies widely among the different NAMAP models. NAME 2004 radar-based and in situ precipitation measurements should quantify both the magnitude of the convective peak in precipitation rate and any nocturnal precipitation following the convection (if indeed the lower rates of precipitation following the convective peak in the NAMAP models is related to antecedent deep convection). 

         *    The NAME 2004 observations should provide ground truth for model diagnostic studies of surface fluxes and land surface feedbacks, providing the basis for more thorough assessment of land-atmosphere interactions in these models.  

As additional modeling studies are planned in coordination with NAME 2004 field activities, metrics to quantify model simulation quality and improvement should be established. We propose the following metrics as targets for model simulation.

         *    Monsoon onset: Global models, in particular, exhibit delayed monsoon onset. A plausible goal for all models would be to simulate the initiation of regular deep convection (i.e. monsoon onset) within a week of its observed initiation. It is possible that improved specification of SST, especially in the Gulf of California, could significantly affect monsoon onset simulations.

         *    Afternoon precipitation maximum:  The wide range of afternoon maxima in precipitation rates demonstrated by the models presents an obvious target for model improvements. The models should seek to reproduce the full diurnal cycle of observed precipitation over the special averaging areas called for above. A goal of matching well-constrained monthly mean observations to within 20% throughout the diurnal cycle presents a challenging goal. (Other temporal and spatial scales could be envisioned as well. For example, hydrological applications demand precipitation simulated accurately at the watershed scale.)

         *    Nocturnal precipitation: Simulate the observed extent and propagation of convective systems throughout the night, including lower precipitation rates that may be associated with nonconvective precipitation. Observations taken along the transects defined by the NAME event-logging precipitation network, enhanced by radar observations (NAME SWG, 2003) could provide the validation data required to address this issue.

         *    Surface fluxes: Reproduce the magnitude of the observed afternoon peak of latent and sensible heat fluxes to within 20% on a monthly averaged basis. The range of surface fluxes is much greater than 20% in the NAMAP runs so this goal is actually quite stringent, and implies that large flux modifications will be needed in some of the models. It will be of great interest to examine the sensitivity of monsoonal precipitation to potentially large adjustments in surface fluxes in these models.

         *    Low-level jet circulations: Metrics here involve improved observations at least as much as improved models; until the observational verification data base is improved it is difficult to quantify model improvement metrics. Models must reproduce the correct position of the Gulf of California low-level jet with respect to the Gulf and the high topography to the east (but available observations are not sufficient to determine the correct position). Relationships between low-level wind, humidity (hence moisture transport) and monsoonal precipitation, and the covariation of low-level jets associated with the Gulfs of California and Mexico must be clarified in observations as well in models.

A second round of comparative model runs (NAMAP-2) is now in the planning stage.  Tentative plans for NAMAP-2 call for another model assessment based on simulations of the 2004 summer season, when the enhanced NAME observing system will yield high-resolution observations for comparison. The metrics outlined above should be revisited in NAMAP-2.



5.      ACKNOWLEDGMENTS

The authors wish to thank the NAME Science Working Group and numerous “Friends of NAME” for scientific guidance that made this analysis possible. In particular, Andrea Hahmann (University of Arizona) provided advice and assistance with the early development of the NAMAP protocol. Technical assistance with archiving and data dissemination from UCAR JOSS was essential for this project. Special thanks are extended to Linda Cully of JOSS.

D. Gutzler's research on NAMAP has been supported through a grant from the NOAA Office of Global Programs PACS/GAPP Warm Season Precipitation Initiative. Much of the NAMAP analysis was carried out during his sabbatical visit to NOAA's Climate Prediction Center, which was supported by the University of New Mexico and the CPC.



6.      REFERENCES

Alpert, J.C., and 5 co-authors, 1988: Mountain induced gravity wave drag parameterization in the NMC medium-range model. Preprints, Eighth Conf. On Numerical Weather Prediction, Baltimore MD, Amer. Meteor. Soc., 726-733.

Anderson, B.T., J.O. Roads, S.-C. Chen, and H-M. H. Juang, 2001: Model dynamics of summertime low-level jets over northwestern Mexico. J. Geophys. Res., 106, 3401-3413.

Anthes, R.A., E.-Y. Hsie and Y.-H. Kuo, 1987: Description of the PSU/NCAR mesoscale model version 4 (MM4).  NCAR Technical Note, NCAR/tn-282+STR, 66 pp.

Arakawa, A., and M. Suarez, 1983: Vertical differencing of the primitive equations in sigma-coordinates. Mon. Wea. Rev, 111, 34-45.

Bacmeister, J., P. J. Pegion, S. D. Schubert, and M. Suarez, 2000: Atlas of seasonal means simulated by the NSIPP-1 atmospheric GCM. NASA/TM-2000-104606, 7, 194pp.

Berbery, E.H., 2001: Mesoscale moisture analysis of the North American monsoon. J. Climate, 14, 121-137.

Chou, M.-D., 1992:  A solar radiation model for use in climate studies. J. Atmos. Sci., 49, 762-772.

-----, and M.J. Suarez, 1994:  An efficient thermal infrared radiation parameterization for use in general circulation models. Technical Memorandum 3, Technical Report Series on Global Modeling and Data Assimilation, NASA Tech. Rep. TM-1994-104606, 85 pp.

-----, and K.-T. Lee, 1996:  Parameterizations for the absorption of solar radiation by water vapor and ozone. J. Atmos. Sci., 53, 1203-1208.

-----, and M.J. Suarez, 1999:  A solar radiation parameterization (CLIRAD-SW) developed at Goddard Climate and Radiation Branch for Atmospheric Studies. Technical Memorandum 15, Technical Report Series on Global Modeling and Data Assimilation, NASA Tech. Rep. TM-1999-104606.

Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model.  Preprints J. Atmos. Sci., 46, 3077-3107.

-----, 1996: A multi-layer soil temperature model for MM5.  Preprints, Sixth PSU/NCAR Mesoscale Model Users' Workshop, 22-24 July 1996, Boulder CO, 49-50.

Ek, M. B., K. E. Mitchell, Y. Lin, P. Grunmann, E. Rogers, G. Gayno, V. Koren, 2003: Implementation of the upgraded Noah land-surface model in the NCEP operational mesoscale Eta model. J. Geophys. Res. (accepted).

Fawcett, P.J., J.R. Stalker, and D.S. Gutzler, 2002: Multistage moisture transport into the interior of northern Mexico during the North American summer monsoon. Geophys. Res. Lett., doi 10.1029/2002GL015693, 05 Dec 2002.

Fels, S.B, and M.D. Schwarzkopf, 1975:  The simplified exchange approximation: A new method for radiative transfer calculations. J. Atmos. Sci., 32, 1474-1488.

Gao, X., S. Sorooshian, and J. Xu, 2003:  SST data improve modeling of North American Monsoon rainfall. EOS, 84, 457 ff.

Gates, W. L., and 15 co-authors, 1998: An overview of the results of the Atmospheric Model Intercomparison Project (AMIP I). Bull. Amer. Meteor. Soc., 73, 1962-1970.

Gochis, D., 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.

Gutzler, D.S., 2004: An index of interannual precipitation variability in the core of the North American Monsoon region. J. Climate, submitted for publication.

Higgins, R. W., J. E. Janowiak and Y. Yao, 1996:  A gridded hourly precipitation data base for the United States (1963-1993).  NCEP/Climate Prediction Center ATLAS No. 1., 47 pp.

-----, Y. Yao, and X. Wang, 1997: Influence of the North American Monsoon System on the United States summer precipitation regime. J. Climate, 10, 2600-2622.

-----, Y.Yao, E. Yarosh, J. E. Janowiak and K. C. Mo, 1997: Influence of the Great Plains low-level jet on summertime precipitation and moisture transport over the central United States.  J. Climate, 10, 481-507.

-----, Y. Chen, and A.V. Douglas, 1999:  Interannual variability of the North American warm season precipitation regime. J. Climate, 12, 653-679.

-----, A. Leetmaa, Y. Xue and A. Barnston, 2000a: Dominant factors influencing the seasonal predictability of U.S. precipitation and surface air temperature. J. Climate, 13, 3994-4017.

-----, W. Shi and E. Yarosh, 2000b: Improved United States Precipitation Quality Control System and Analysis. NCEP/Climate Prediction Center ATLAS No. 7, 40 pp.

-----, and 16 co-authors, 2003: Progress in Pan American CLIVAR Research: The North American Monsoon System. Atmósfera, 16, 29-63.

Hong, S.-Y., and H.-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium range forecast model. Mon. Wea. Rev., 124, 2322-2339.

Janjic, Z., 1990:  The step-mountain coordinate: Physical package. Mon. Wea. Rev., 118, 1429-1443.

-----, 1994:  The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927-945.

-----, 2000:  Comments on "Development and evaluation of a convection scheme for use in climate models". J. Atmos. Sci., 57, 3686.

-----, 2001:  Nonsingular implementation of the Mellor-Yamada Level 2.5 scheme in the NCEP Meso model. NOAA/NWS/NCEP Office Note #437, 61 pp. Available from the NCEP Environmental Modeling Center, NOAA Science Center, WWB room 207, 5200 Auth Road, Camp Springs MD 20746.

Juang, H.-M.H. and M. Kanamitsu, 1994: The NMC nested regional spectral model. Mon. Wea. Rev., 122, 3-26.

-----, S.-Y. Hong and M. Kanamitsu, 1997: The NCEP regional spectral model: an update. Bull. Amer. Meteor. Soc., 78, 2125-2143.

Kain, J., and J. M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain-Fritsch scheme.  The Representation of Cumulus Convection in Numerical Models, Meteor. Monograph No. 46, Amer. Meteor. Soc., Boston, 165-170.

Kanamitsu, M., and 10 co-authors, 2002: NCEP dynamical Seasonal Forecast System 2000.  Bull. Amer. Meteor. Soc., 83, 1019-1037.

-----, W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fioino, and G. L. Potter, 2002: NCEP-DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631-1643.

-----, and K. Mo, 2003: Dynamical effect of land surface processes on summer precipitation over the southwestern United States. J. Climate, 16, 496-509.

Koster, R. D., and M. J. Suarez, 1992: Modeling the land surface boundary in climate models as a composite of independent vegetation stands. J. Geophys. Res., 97, 2697-2715.

-----, and -----, 1996: Energy and water balance calculations in the Mosaic LSM. NASA Tech. Memo. 104606, 9, 194pp.

Lacis, A. A., and J. E. Hansen, 1974: A parameterization of the absorption of solar radiation in the earth's atmosphere. J. Atmos. Sci., 31, 118-133.

Louis, J.F., M. Tiedtke, and J.F. Geleyn, 1982: A short history of the operational PBL parameterization at ECMWF. Proceedings of the Workshop on Planetary Boundary Parameterization, European Centre for Medium-Range Weather Forecasts, Shinfield Park, Reading RG2 9AX, UK, pp. 59-79.

Meehl, G.A., G.J. Boer, C. Covey, M. Latif, and R.J. Stouffer, 2000: The Coupled Model Intercomparison Project (CMIP). Bull. Amer. Meteor. Soc., 81, 313--318.

Mesinger, F., and 16 co-authors, 2004: North American Regional Reanalysis, AMS Annual Meeting, Seattle WA, January 2004.

Moorthi, S., and M.J. Suarez, 1992: Relaxed Arakawa-Schubert:  A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120, 978-1002.

NAME SWG, 2003: North American Monsoon Experiment (NAME): Science and Implementation Plan. Available online at: http://www.cpc.ncep.noaa.gov/products/precip/monsoon/NAME.html

-----, 2003: NAME Modeling and Data Assimilation: A Strategic Overview. (in revision).

Pan, H.-L., and L. Mahrt 1987: Interaction between soil hydrology and boundary layer developments. Boundary-Layer Meteorology, 38, 185-202.

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.

Suarez, M.J., and L.L. Takacs, 1995: Documentation of the ARIES/GEOS dynamical core: Version 2. Technical Memorandum 5, Technical Report Series on Global Modeling and Data Assimilation, NASA Tech. Rep. TM-1995-104606.

Tao, W.-K., and J. Simpson, 1993: Goddard cumulus ensemble model. Part I: Model description. Terrestrial, Atmospheric, and Oceanic Sciences, 4, 35-72.

Xu, J.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(D23), 4727, doi:10.1029/2001JD002047.

Zhao, Q., and F. H. Carr, 1997: A prognostic cloud scheme for operational NWP models.  Mon. Wea. Rev., 125, 1931-1953.

 




Table 1

Participating Models (see Section 2 for more details)

Model No. Model abbrev. time series abbrev. Institution Contact Model Resolution Computational Domain Integration Period (1990) Time Step
A. Regional Models
1 RSM rsm_hj NCEP Dr. H. Juang 20 km / L28 18-37.5°N/100-117°W May 1-Sep 30 120s
2 RSM rsm_mk2 SIO ECPC Dr. M. Kanamitsu 20 km / L28 NAME Tier 2 May 1-Sep 30 120s
3 MM5 mm5_lr UNM Dr. E. Ritchie 15 km / L23 18-38°N/101-119°W June 1-Sep 30 40s
4 Eta eta_km NCEP Dr. K. Mitchell 32 km / L45 10-90°N/58-164°W May 1-Sep 30 90s
B. Global Models
5 SFM sfm_js NCEP Dr. J. Schemm 2.5° / L28 global May 15-Sep 30 1200s
6 NSIPP nsipp_pp NASA Dr. S. Schubert 0.5° / L34 global June 1-Sep 30 60s


Table 1 (continued)

Model Physics (see Section 2 for more details)

Model No. Model abbrev. Institution Moist Convection Radiation: SW Radiation: LW Radiation: Clouds Boundary Layer Parameterization Land Surface
A. Regional Models
1 RSM NCEP Pan & Wu (1995) Chou (1992) Chou & Suarez (1994) Slingo (1987) Hong & Pan (1996) Pan & Mahrt (1987)
2 RSM SIO ECPC Moorthi & Suarez (1992) Chou (1992) Fels & Schwarzkopf (1975) Slingo (1987) Hong & Pan (1996) Pan & Mahrt (1987)
3 MM5 UNM Kain & Fritsch (1993) Dudhia (1989) Dudhia (1989) Dudhia (1989) Hong & Pan (1996) Pan & Mahrt (1987)
4 Eta NCEP Betts-Miller-Janjic Lacis & Hansen (1974) Fels & Schwarzkopf (1975) Zhao & Carr (1997) Janjic (1994) Ek et al. (2003)
B. Global Models
5 SFM NCEP Moorthi & Suarez (1992) Chou & Lee (1996) Chou & Suarez (1994) Slingo (1987) Hong & Pan (1996) Pan & Mahrt (1987)
6 NSIPP NASA Moorthi & Suarez (1992) Chou & Suarez (1999) Chou & Suarez (1994) diagnostic Louis et al. (1982) Koster & Suarez (1996)


Table 1 (continued)

Initial and Boundary Conditions for Regional Models (see Section 2 for more details)

Model No. Model abbrev. Institution Atmospheric Initial Conditions Source: Day/Time Integration Length Lateral Boundary Conditions Source: Update Freq. Land Surface Initial Source
1 RSM NCEP Reanal2: 0000Z 1 May 1990 5 mo (uninterrupted) Reanal2:    6 hr USGS
2 RSM SIO ECPC Reanal2: 0000Z 1 May 1990 5 mo (from chain of 26-hr forecasts) Reanal2:    6 hr USGS
3 MM5 UNM Reanal1: 0000Z 1 Jun 1990 4 mo (uninterrupted) Reanal1:   12 hr USGS
4 Eta NCEP Reanal2: 0000Z 1 May 1990 5 mo (uninterrupted) Reanal2:    6 hr Reanal2





Appendix 1

Output fields archived for NAMAP analysis

Each NAMAP archive file is a concatenated set of monthly average maps for a particular variable and time of day, written in binary x-y format suitable for import into GrADS data analysis software. Thus for NAME Tier 1 high-resolution output, hourly data for a single variable and month (e.g. surface latent flux for the month of July 1990) would be written out as 24 x-y maps for 0000Z, 0100Z, ..., 2300Z concatenated into one file. Each model has a different horizontal resolution we'll need clear instructions about the x-y format and how to convert your model's horizontal coordinates into latitude and longitude.

The NAMAP protocol called for the following fields to be written out across Tier1 (with hourly resolution) and Tier 2 (with 12-hourly resolution). Due to individual modeling constraints, not all the variables are available for every model and not all the Tier 1 fields are available with hourly resolution.  Full documentation for each set of model output is provided online together with the data files at URL

http://www.joss.ucar.edu/name/namap/index.html .

A.  Tier 1 fields (hourly)

Tsfc   surface temperature           LWnet   surface net longwave radiation
qsfc   surface specific humidity           SW   surface shortwave radiation
usfc   surface zonal wind component           OLR   TOA outgoing longwave radiation
vsfc   surface meridional wind component           CAPE   convective available potential energy
SLP   sea level pressure           cldlg   resolved cloudiness
T925   925 hPa temperature           cldconv   subgrid cloudiness
q925   925 hPa specific humidity           SH   surface sensible heat flux
u925   925 hPa zonal wind component           LH   surface latent heat flux
v925   925 hPa meridional wind component           qucol   column-integrated zonal moisture flux
z925   925 hPa geopotential height           qvcol   column-integrated meridional moisture flux
T850   850 hPa temperature           PWcol  column-integrated precipitable water content
q850   850 hPa specific humidity           albedo   planetary albedo
u850   850 hPa zonal wind component           precip   precipitation
v850   850 hPa meridional wind component          
z850   850 hPa geopotential height          
T700   700 hPa temperature          
q700   700 hPa specific humidity          
u700   700 hPa zonal wind component          
v700   700 hPa meridional wind component          
z700   700 hPa geopotential height          


B.  Tier 2 fields (0000Z and 1200Z)

precip total precipitation
SLP sea level pressure

and at 9 levels (surface, 925-850-700-500-300-250-200-100 hPa):

T temperature
q specific humidity
u zonal wind component
v meridional wind component
z geopotential height



List of Plots

1.      Schematic diagram of operational domains for the North American Monsoon Experiment (NAME) as shown in the NAME Science and Implementation Plan (NAME SWG, 2003).

2.      Total monthly precipitation maps [cm], Tier 1.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

2.0     Observed precipitation over land, calculated from the 1°´ daily NCEP “Unified” Raingauge Dataset (URD) (Higgins et al. 2000b).  Boxes denoted “CORE” and “AZNM” show the spatial subregions used for time series analysis in subsequent plots.

2.1     Simulated by RSM/NCEP (model 1 in Table 1)

2.2     Simulated by RSM/ECPC (model 2 in Table 1)

2.3     Simulated by MM5 (model 3 in Table 1)

2.4     Simulated by Eta (model 4 in Table 1)

2.5     Simulated by SFM (model 5 in Table 1)

2.6     Simulated by NSIPP (model 6 in Table 1)

3.      Time series of total monthly precipitation [cm], May-Sep 1990, averaged over the (a) CORE subregion (b) AZNM subregion.

         The boxes in Figure 2 depict the CORE and AZNM subregion boundaries.

Black dashed line:                   Unified observed data.

Dark blue line/squares:            RSM/NCEP simulation.

Orange line/diamonds:             RSM/ECPC simulation.

Red line/squares:                    MM5 simulation.

Green line/Xs:                        Eta simulation.

Black solid line/squares:          arithmetic mean of the ensemble of 10 SFM simulations.

         Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.

Light blue line/circles:             NSIPP simulation.

4.      Maps of monthly averaged 1200Z wind at 925 hPa. Vectors on all plots are drawn to a constant scale, shown by the 10 m/s arrow beneath each map.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990.

4.0     Observed, NCAR/NCEP reanalysis.

4.1     Simulated by RSM/NCEP (model 1 in Table 1)

4.2     Simulated by RSM/ECPC (model 2 in Table 1)

4.3     Simulated by MM5 (model 3 in Table 1)

4.4     Simulated by Eta (model 4 in Table 1)

4.5     Simulated by SFM (model 5 in Table 1)

4.6     Simulated by NSIPP (model 6 in Table 1)

5.      Longitude-height cross-sections of specific humidity [kg/kg] and meridional wind [m/s]    along 27°N, Tier 1.

(a) Jun 1990, 1200Z          (b) Jul 1990, 1200Z          (c) Aug 1990, 1200Z.

5.0     Observed, NCAR/NCEP reanalysis.

5.1     Simulated by RSM/NCEP (model 1 in Table 1)

5.2     Simulated by RSM/ECPC (model 2 in Table 1)

5.3     Simulated by MM5 (model 3 in Table 1)

5.4     Simulated by Eta (model 4 in Table 1)

5.5     Simulated by SFM (model 5 in Table 1)

5.6     Simulated by NSIPP (model 6 in Table 1)

6.      Longitude-height cross-sections of specific humidity [kg/kg] and meridional wind [m/s]    along 30°N in Tier 2.

(a) Jun 1990, 1200Z          (b) Jul 1990, 1200Z          (c) Aug 1990, 1200Z.

6.0     Observed, NCAR/NCEP reanalysis.

6.1     Simulated by RSM/ECPC (model 2 in Table 1)

6.2     Simulated by MM5 (model 3 in Table 1)

6.3     Simulated by SFM (model 5 in Table 1)

6.4     Simulated by NSIPP (model 6 in Table 1)

7.      Diurnal cycle of total precipitation rate [mm/hr], CORE subregion.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.

8.      Diurnal cycle of surface air temperature [K], CORE subregion.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.

9.      Diurnal cycle of upward latent flux [W/m2], CORE subregion.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.

10.    Diurnal cycle of upward sensible flux [W/m2], CORE subregion.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.

11.    Time series monthly values of Bowen ratio [sensible flux/latent flux], Jun-Aug 1990, based on 24-hr averages of fluxes over the           (a) CORE subregion   (b) AZNM subregion. 

Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.

12.    Diurnal cycle of total precipitation rate [mm/hr], AZNM subregion.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

Black dashed line:   Observed data from U.S. hourly gridded data set (Higgins et al. 1996).
Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.

13.    Diurnal cycle of surface air temperature [K], AZNM subregion.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.

14.    Diurnal cycle of upward latent flux [W/m2], AZNM subregion.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.

15.    Diurnal cycle of upward sensible flux [W/m2], AZNM subregion.

(a) Jun 1990                      (b) Jul 1990                       (c) Aug 1990

Dark blue line/squares:   RSM/NCEP simulation.
Orange line/diamonds:   RSM/ECPC simulation.
Red line/squares:   MM5 simulation.
Green line/Xs:   Eta simulation.
Black solid line/squares:  arithmetic mean of the ensemble of 10 SFM simulations.
 * Yellow shading surrounding SFM average values shows ± 1s envelope of variability of the ten SFM simulations.
Light blue line/circles:  NSIPP simulation.