By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
 This paper describes orographic processes that modulate the Great Plains low-level jet (LLJ) and related hydrology of the Mississippi River basin. Mechanical flow deflection by the Rocky Mountains is diagnosed in 50 years of monthly averaged National Centers for Environmental Prediction-National Center for Atmospheric Research (NCEP-NCAR) Reanalysis fields and in a series of integrations using a primitive equation version of the Utah Global Model (UGM). Although the mountain profiles are fixed over periods of short-to medium-range climate changes, their influence on the LLJ is not stationary because evolving ambient flows produce changing LLJ responses. Ensembles of medium-range forecasts are made for the 1993 U.S. floods and for the 1988 U.S. drought. The forecasts distinguish some of the observed precipitation differences between these years, but the magnitude of the differences is underestimated. Seasonal and longer-term changes of the ambient flow occur on large scales, while the response of the LLJ occurs on smaller scales that may promote cloud generation and precipitation. Month-long simulations with monthly averaged conditions suggest that the Great Plains LLJ is a component of the large-scale circulation associated with the topography of the western United States. Orography thus provides a scale transfer mechanism that focuses global-scale features into the regional-scale responses. These are relevant to precipitation distribution and to moisture budgets of the larger individual river basins comprising the GCIP domain. A theoretical interpretation of the large-scale, mechanical influence of orography on surrounding low-level circulations is proposed.
 This investigation is related to past studies by Mo et al.  and Wang et al.  that examined extended range influences on the atmospheric portion of the hydrologic cycle over the Global Energy and Water Cycle Experiment (GEWEX) Continental-Scale International Project (GCIP) domain. Mo et al.  examined the onset and maintenance of the 1993 summer floods using Eliassen-Palm fluxes. They suggested that an anomalously strong westerly flow over the Rocky Mountains was maintained by low-pass filtered eddies over the North Pacific, and that orographic effects accelerated a cyclonic pattern in the vicinity of the Rocky Mountains. These features supported a strong southerly low-level jet (LLJ) east of the mountains and accelerated the influx of water vapor from the Gulf of Mexico toward the upper and central Mississippi River basins.
 The first goal of the present study is to examine the role of this mechanism in climatological correlations of cross-Rockies zonal flows with the hydrological cycle over North America. Section 2 uses monthly averaged NCEP/NCAR Reanalyses to demonstrate that the Great Plains LLJ during summer, related water vapor transports, and Mississippi River basin (MRB) rains tend to increase during extended periods of anomalously strong zonal flow over the western United States.
 The correlations suggest that an orographically bound, cyclonic acceleration occurs over the western United States when the summer, upper-troposphere zonal flow is relatively strong above the Rockies. Similar correlations are found between the midlatitude zonal index averaged over the full latitude circle and orographically bound cyclones in summer. Section 3 presents 13-day, ensemble simulations using the Utah Global Model (UGM) for summer 1993 and 1988, which investigate predictability over the MRB for these years of extensive zonal flow anomalies. We hypothesize that the long-term inertia of the global-scale wind anomaly may extend the period of predictability.
 Many other predictability studies for the floods of 1993 and droughts of 1988 have emphasized land-surface parameterizations. Within the the last GCIP special issue of the Journal of Geophysical Research, at least 15 articles and references therein examined the important role of land-surface processes and energy budgets upon precipitation over the continental United States. At least 8 of these investigations are modeling studies of the 1993 floods and 1988 drought, and the majority emphasize monthly to seasonal integrations. Beljaars et al.  found better forecast skill for month-long simulations resides in the representation of soil moisture and strong feedbacks between the land surface hydrology and precipitation. Viterbo and Betts  used July ensemble forecasts with the European Center for Medium-Range Weather Forecasts (ECMWF) system and found large soil water anomalies contributed to the July 1993 flooding. Replacing the July 1993 soil water with soil water from June 1988 in the ensemble forecast reduced the precipitation anomaly by 40%. Other studies have considered the important role of sea surface temperature (SST) anomalies and other remote teleconnections influencing regional and local flow anomalies (e.g., severe drought) over the Great Plains [e.g., Namias, 1983, 1991; Trenberth et al., 1988; Mo et al., 1991; Trenberth and Branstator, 1992]. Bell and Janowiak  found a strong relationship between anomalous SSTs during the warm 1992–93 ENSO and the extratropical circulations which contributed to the summer 1993 floods over the Midwest. The present study takes a similar approach to the studies listed above with emphasis on dynamical processes associated with ambient flows over the Rocky Mountains.
Section 4 examines some dynamical implications of Mo et al.'s  hypothesis that the 1993 MRB floods in part represent a mechanical, orographic response to the anomalously strong zonal circulation of this period. The single level barotropic analysis used by Mo et al.'s  spherical treatment is extended in section 4, and results suggest that the summer time flow over North America typically promotes a cyclone in the vicinity of the Rockies. This orographic cyclone may be interpreted as a subresonant response for stationary Rossby waves. A consistent interpretation of the observations (section 2) is that increased zonal flow during summer accelerates the orographic vortex as the resonant point in the response is approached.
Section 5 presents preliminary evidence that the observed winter to summer reversal of the Great Plains LLJ may also be explained as a reversal from superresonant orographic responses in winter to subresonant responses in summer. This interpretation provides an alternative, mechanical explanation to thermal influences related to surface heating that have often been used to explain observed winter-summer climatological reversals above the Rockies, as well as anomalies from the climatology. Section 6 summarizes conclusions.
2. Interannual Variability Over North America
2.1. The Great Plains Low-Level Jet
 All diagnostics, unless noted otherwise, are based on a 50-year record (1951–2000) of monthly averaged circulations from the NCEP/NCAR Reanalysis [Kalnay et al., 1996; Kistler et al., 2001]. The assimilation system is based on NCEP's 1995 operational global spectral model with a T62 resolution on 28 vertical sigma levels (∼210 km horizontal resolution). Primary fields used here from the Reanalysis are the rotational wind and specific humidity. Kalnay et al.  described the rotational wind as one of the “type A” variables, influenced most strongly by assimilated observations. Specific humidity, used to compute vertically integrated moisture flux, is a “type B” variable, considered somewhat less reliable since both the observational data and the assimilation model affect this quantity. Surface latent heat flux, used for model initialization in section 3, is a “type C” variable, derived only from model data.
Figure 1a shows the 50-year climatology of 850 mb wind vectors and eddy heights over North America for June–July–August (JJA). The eddies are defined by subtracting the zonal average. Most of the Rocky Mountain region lies above the 850 mb level, so Reanalysis data shown there have been artificially interpolated below the surface of the Earth and the 850 mb pattern is poorly defined. A significant feature east of the Rockies is the presence of a north-to-south-oriented, southerly LLJ, which extends from the western Gulf of Mexico and Mexico, northward into the Great Plains to about 45°N. The LLJ is a persistent summer time feature over the Great Plains, and it is absent in the 850 mb winter eddy height and wind fields (not shown). Low-level northerlies off the coast of California turn cyclonically west of Baja California near 25°N.
 We first investigate how ambient flows over the Rocky Mountains may relate to the Great Plains LLJ. A 50-year time series of the area-averaged, 200 mb zonal flow is calculated over the Rockies during summer (JJA) from 120°W–100°W, 30°N–50°N (the box is outlined in Figure 4). The weakest and strongest cases of summer, ambient flow over the Rockies are then selected when the area-averaged zonal flow is above or below the mean by an amount equaling plus or minus 1.2 times the standard deviation from the mean. These cases explain about 12% of the variance, assuming a Gaussian distribution. Six cases in the 50-year record meet the criteria for strong events and six meet the criteria for weak events.
Figure 1b shows the eddy heights and winds at 850 mb for summers with the strongest 200 mb flow (“u200 high”), and Figure 1c during summers of the weakest 200 mb flow over the region (“u200 low”). u200 high values range from about 18 to 20 m/s, while low cases range from about 11.5 to 13.5 m/s. The selected years are 1951, 1958, 1978, 1980, 1992 and 1993 for u200 high; and 1955, 1961, 1970, 1971, 1983 and 1988 for u200 low. The difference field between u200 high and low cases is presented in Figure 1d. The 850 mb southerlies are almost twice as strong over Texas in the u200 high cases compared to the u200 low cases. The anticylonic circulation near the Gulf of Alaska is more pronounced during u200 high cases, and lower geopotential heights are centered over the northern United States and Canada in the difference field (Figure 1d).
Figure 2 shows the vertically integrated, meridional moisture flux for the u200 high and low composites. The highest moisture flux contour is 4 units over Texas for the u200 high cases, while it is 2.5 units for u200 low cases. The composite wind difference (Figure 1d) and the moisture flux differences in Figure 2 have similarities to the 850 mb, meridional wind differences between the “flood look-alike” (FLA) and “anti-flood look-alike” (AFLA) cases that Mo et al.  identified using ECMWF pentad data and synoptic-scale features (see their Figure 13).
Figure 3 shows 50-year (1951–2000) temporal correlations of the local, meridional, 850 mb wind in a box located near the climatological region of the Great Plains LLJ, with gauge precipitation over land. The 850 mb meridional wind is area-averaged over the indicated box, extending from 20°N to 40°N, and the precipitation is taken from the Global Precipitation Reconstruction Analysis over Land (PREC/L) produced by Chen et al. . This uses interpolated gauge observations over land from the Global Historical Climatology Network (GHCN), Version 2. Correlation coefficients of 0.3 and higher are statistically significant at the 99% level for 50 degrees of freedom. Low-level southerlies in the boxed region are strongly correlated with precipitation over the northern United States and in the Great Plains and Midwest, with the largest correlation coefficient of 0.7 over Iowa. Highest correlation coefficients are close to the exit region of the climatological LLJ depicted in Figure 1a.
Paegle  briefly summarizes some of the dynamical factors influencing LLJs over the gently sloping topography of the Great Plains, including (1) diurnal, buoyancy-forced oscillations which depend on heating [Holton, 1967]; (2) diurnal, frictional effects [Blackadar, 1957]; and (3) Wexler's  extension of western boundary dynamics in oceanography as applied to the Great Plains LLJ. Mechanisms (1) and (2) produce nocturnal enhancement of the Great Plains LLJ [e.g., Bonner and Paegle, 1970]. Wexler's  theory suggests that the blocking effect of orography influences the strength of the LLJ in association with episodic fluctuations in low-level easterlies within the subtropics that are deflected poleward by the east slope of the North American cordillera.
 The present study, by contrast, investigates the strength of the ambient, upper troposphere westerlies impinging upon the Rocky Mountains and their possible association with the strength of the Great Plains LLJ for medium-range and longer timescales. Other studies have examined the enhancement of LLJs over the Great Plains and other parts of the United States independent of orography and for shorter timescales. Uccellini and Johnson  and Uccellini , for example, found that for synoptic scales, low-level, ageostrophic winds associated with jet streak exit regions may be associated with an enhancement of LLJs, particularly over the Great Plains and eastern North America. The next subsection examines ambient zonal flows in the upper troposphere and their correlations with wind and moisture fields over the MRB.
2.2. Correlations With Ambient, Upper Troposphere Flows
Figure 4a displays 50-year (1951–2000) temporal correlations of the local 700 mb wind with the 200 mb zonal flow for the June–July–August (JJA) average. The 200 mb zonal flow is area-averaged over the indicated box centered over the Rockies and extends from 30°N to 50°N. The eastward/northward component of the vectors denotes the magnitude of the correlation coefficient between the area-averaged, upper tropospheric wind in the box against the local zonal/meridional flow component at 700 mb. We emphasize features at 700 mb over the western United States, where artificial interpolation below the surface of the Earth is not required over most of the region.
 An increase in upper tropospheric westerlies in the box during summer (Figure 4a) results in a cyclonic circulation in the correlation vectors at 700 mb over the Rocky Mountain region. The cyclonic tendency over the mountains may therefore be related to an increase in the strength of the westerlies over the mountains. Southwesterly and southerly correlation vectors near the western Gulf of Mexico and Texas apparently converge over the Mississippi River basin (MRB), near northern Missouri. The 700 mb level is generally above the mean height of the Great Plains LLJ. In order to test whether the correlations of ambient zonal flow over the Rockies may relate to the LLJ, similar correlations with vertically integrated moisture flux and with winds at 850 mb are discussed below.
Figure 4b shows time correlations of the local, vertically integrated moisture flux with ambient, 200 mb zonal flow within the boxed region. Only vectors which are statistically significant at the 99% level (0.3) are plotted, and the statistically significant, meridional coefficients are contoured. Over the western United States, the pattern is similar to the correlations with 700 mb wind (Figure 4a). The correlation vectors and meridional contours show evidence that moisture travels from the Gulf of Mexico into the central United States during periods of increased 200 mb zonal flow over the mountains. Figure 5 is similar to Figure 4b but for a correlation of 200 mb zonal flow over the box with the 850 mb wind at all locations. Correlation vectors with statistical significance are absent over most of the boxed region, where most of the surface of the Earth lies below 850 mb. East of the Rocky Mountains, the correlation vectors are at a maximum (0.5) over Texas, with meridional correlation coefficients from 0.3 to 0.4 extending northeastward. The correlation vectors, particularly over the western Gulf of Mexico and Texas, are oriented similarly to the mean LLJ vectors at 850 mb (Figure 1a). Increased 200 mb zonal flow over the Rockies correlates with increased tropical southerlies (Figure 5) and moisture transport (Figure 4b) into the MRB.
 We next explore whether global, upper tropospheric, ambient flows within the full latitude belt encompassing the central Rocky Mountains may also influence the low level circulation over and to the east of the orography. The purpose is to test whether much larger scales and lower wave numbers within the latitude belt of the Rockies may also elicit a similar response as found for zonal flows within the boxed region (e.g., Figures 4 and 5). The area-averaged 200 mb zonal flow around the entire globe within the latitude belt from 30°N to 50°N is correlated with the local 700 mb wind in Figure 6a (the latitude belt is outlined). Upper tropospheric westerlies correlate with cyclonic turning in the 700 mb wind over the Great basin. The correlation vectors over the western and central United States are reminiscent of Figure 4a, though the magnitudes of the vectors are smaller.
 The large inertia associated with the zonally averaged wind and its interaction with orography may contribute to enhanced predictability over the GCIP region during summer. The seasonal and longer-term changes of the ambient flow occurring on large scales may relate to fluctuations in the Great Plains LLJ, whose response occurs in summer on the smaller scales of cloud generation and precipitation. Helfand and Schubert  showed the important role of the LLJ for vapor transport. Using a simulated moisture budget and the Goddard Earth Observing System (GEOS-1) atmospheric general circulation model, they found that the Great Plains LLJ transports about 30% of all the moisture entering the continental United States.
Figure 6b depicts correlations of the 850 mb wind at all locations with 200 mb winds in the outlined latitude belt around the globe. Only vectors and contours of the meridional component which are statistically significant at the 95% level (0.23) are depicted. Positive correlation coefficients extend from the western Gulf of Mexico into Texas and the Midwest. The signal is similar to, but weaker than, correlations where only the 200 mb zonal flow over the Rocky Mountain region is considered (Figure 5). Figure 7 displays time correlations similar to Figure 6b, but of vertically integrated moisture flux correlated with area-averaged, 200 mb zonal flow. The data are from a higher resolution, 29-year subset of the 50-year NCEP/NCAR Reanalysis. The subinterval is from 1968 to 1996 and contains 28 levels in the vertical with 7 levels below 850 mb (for details, see Mo et al. ). Only two levels are available below 850 mb for calculating the vertically integrated moisture flux in Figure 4b. The moisture flux data from which the correlations were computed are also based on daily, rather than monthly averaged values, thereby reflecting the effect of transients. Correlation coefficients that equal 0.31 are significant at the 95% level for this subset. Figure 7 shows a cyclonic tendency in the correlation vectors over the central Rocky Mountains. Southerly and southwesterly correlation coefficients characterize most of the Great Plains, suggesting such flow may be favored in the presence of strong, 200 mb westerly flow around the latitude circle. The correlations of ambient 200 mb flow with vertically integrated moisture flux (Figures 4b and 7) suggest that an increase in moisture transport into the MRB via the LLJ occurs during periods of enhanced, upper troposphere westerlies over the Rocky Mountains.
 The correlation coefficients displayed in the region of the mean LLJ in Figures 4b and 7 suggest increased moisture flux into the MRB, and therefore heavier precipitation, may occur when upper tropospheric westerlies increase. This connection is illustrated with the time correlation of the local summer precipitation with the ambient 200 mb zonal flow between 30°N and 50°N (Figure 8a). Correlation coefficients greater than or equal to 0.3 appear over portions of the Great basin and the MRB. The same correlation for the 200 mb zonal flow averaged over the Rocky Mountain region in Figure 8b shows an even stronger signal over the MRB, with the maximum coefficient (∼0.5) centered over Iowa. The pattern is similar to the correlation of 850 mb meridional flow over the Great Plains with precipitation (Figure 3). In addition, correlations in Figures 8a and 8b appear to have a dipole pattern, with positive values over the Great Plains and negative values over the Southwest. Increased LLJ activity coupled with precipitation over the Great Plains is sometimes associated with a suppressed North American Monsoon in the southwestern United States. Several investigators have noted this out-of-phase relationship [e.g., Higgins et. al., 1997a, 1997b; Mo et al., 1997].
 The correlations presented in this section suggest that increased cross-mountain flow may accelerate the hydrologic cycle in the MRB. We have used monthly averaged fields to diagnose the interannual signal. While this shows large-scale influences on the LLJ, it does not allow an assessment of shorter timescales, including diurnal influences.
3. Model Experiments
 We next present ensemble simulations during two extreme periods within the GCIP domain. The goal is to examine the roles of dynamics and land-surface and heating processes in the analyses and 13-day forecasts.
3.1. Model and Data
 The model is a multilevel, spectral, primitive equation version of the model described by Paegle . The UGM has been used primarily to address predictability questions [e.g., Vukicevic and Paegle, 1989; Paegle et al., 1997; Wang et al., 1999; Miguez-Macho and Paegle, 2000]. It is hydrostatic and filters acoustic waves and predicts vorticity, divergence, thermal and moisture fields on pressure-based sigma coordinates. The approach is similar to that used by operational centers, with the exception of numerical approximations of horizontal derivatives that are based on Fourier series in longitude and finite elements in latitude. A low-order turbulent kinetic energy equation is used to calculate vertical mixing coefficients. Radiative processes include cloud-radiative interactions, as described by Nicolini et al. . The present experiments retain nodal spacing 2.22 degrees in latitude and 42 waves in longitude on 20 vertical levels, situated at sigma 0.99, 0.98, 0.96, 0.94, 0.91, 0.88, 0.84, 0.78, 0.72, 0.66, 0.59, 0.53, 0.47, 0.41, 0.34, 0.28, 0.22, 0.16, 0.09 and 0.03. Physical parametrizations of precipitation (convective and stratiform) are similar to those used by the 1987 version of the NCAR Community Climate Model. The model does not have a complete land parametrization. Initial surface temperature is specified over the globe from reanalysis data (ECMWF or NCEP); and subsequent surface temperature is computed from a heat balance equation including long and shortwave radiation, turbulent transfer in the atmosphere and conduction in the soil. Surface latent heat flux is specified from the NCEP/NCAR Reanalysis. A 15-minute time step is used in the present integrations.
 Ensembles of forecasts during the summers of 1993 and 1988 are presented in section 3.2. Initial conditions are from the daily, 00 UTC values of horizontal wind, specific humidity, temperature, surface pressure and geopotential height from the NCEP/NCAR Reanalysis. Initial fields are interpolated linearly from pressure levels to model sigma levels. In addition, similar initial fields from the ECMWF Reanalysis are used, with the exception of elevation and surface pressure, which in all cases are specified from the NCEP model in order to minimize surface imbalances in the adjustment of sea level pressure to the model surface. The ECMWF Reanalysis is a state-of-the-art assimilation system similar to the NCEP/NCAR Reanalysis, using observations from all available data sources. It is a spectral version (T106 resolution, with 31 vertical hybrid levels, or 17 pressure levels) of the ECMWF operational data assimilation system and model [Gibson et al., 1997].
 Surface evaporation from the NCEP/NCAR Reanalysis is specified over the globe. Evaporation, a “Type C,” model-derived product, is taken from daily averaged values for the ensemble cases in 1988 and 1993. The 1988 initial days are 26 May, 28 May, 30 May, 1 June and 3 June. The 1993 initial days are 27 June, 29 June, 1 July, 3 July and 5 July. For a set of experiments, surface evaporation is switched between 1988 and 1993, and the monthly averaged evaporation from the alternating year is used (see Figure 12). There is no diurnal cycle in the surface evaporation, which is prescribed at every time step. The model includes a diurnal cycle in solar radiation based on the season, day of the year and hour; and the radiation scheme includes water vapor absorption, cloud scattering and absorption and clear-air scattering.
 Model precipitation forecasts are compared with daily averaged observations from the unified precipitation data set of the Climate Prediction Center (CPC) [Higgins et al., 1996]. The data are available on a 0.25° × 0.25° grid from 140°W to 60°W and 20°N to 60°N. Data are derived from three sources using a Cressman scheme: (1) National Climatic Data Center co-op stations, (2) the CPC data set, and (3) the daily accumulations from hourly precipitation data set.
 Averages of the vertically integrated, meridional moisture transport in the UGM are compared to those of the ECMWF Reanalysis for the 13-day ensembles in 1988 and 1993. ECMWF fluxes are computed on pressure coordinates, while UGM fluxes are computed on sigma coordinates.
3.2. Thirteen-Day Ensemble Simulations
 Ten forecasts are made during the 1993 U.S. summer floods, and another ten during the U.S. drought of 1988. July 1993 was characterized by Mo et al.  and Bell and Janowiak  as a persistent, highly amplified wave period. Mo et al.  showed strong, negative 500 mb height anomalies were maintained over the west-central United States (90°W to 120°W) and positive 500 mb height anomalies were maintained over the Eastern Pacific and Eastern United States. The flood region of the Midwest was located east of a large-scale trough axis, in an area of anomalously strong southwesterly flow [Bell and Janowiak, 1995]. By contrast, June 1988 was during the height of a severe drought in which an anomalous ridge was anchored over the north-central United States [Mo et al., 1991; Namias, 1991] with a northward displacement of the jet stream.
 The initial times for each 13-day forecast during the 1993 flood are 00 UTC 27 June, 29 June, 1 July, 3 July and 5 July 1993. Each forecast is initialized with NCEP data and again with ECMWF data, for a total of ten ensemble members (“June–July 1993” ensemble). During the 1988 drought, the forecast days are 00 UTC 26 May, 28 May, 30 May, 1 June and 3 June 1988. As in 1993, each time is initialized with NCEP data and again with ECMWF data to produce an ensemble of ten, 13-day forecasts (“May–June 1988” ensemble).
Figure 9a shows the average, observed precipitation accumulation for the five, 13-day periods during June–July 1993. More than 20 cm of rain falls over northeastern Kansas, northern Missouri and southern Iowa. A local peak of 27 cm occurs near the Kansas-Nebraska-Missouri-Iowa borders. Other maxima include 8 cm in western Montana and North Dakota, 10 cm over Wisconsin and 10 cm in Mississippi and southeastern Louisiana. The UGM ensemble for the same period is shown in Figure 9b. The model has two, 8 cm maxima in precipitation near the region of observed maximum precipitation. One is in western Kansas and another is located in west-central Illinois. The model precipitation pattern represents some of the major features across the north-central United States, but lacks precipitation west Texas, Oklahoma, and across the Gulf coast and into the southeastern United States. The forecast values are substantially less than observations. The areal extent of both observed and model precipitation partially resembles the intraseasonal signal. This is shown by comparing them to the correlation coefficients of area-averaged, upper tropospheric zonal flow with precipitation, depicted in Figures 8a and 8b.
Pan et al.  demonstrated with a mesoscale model that a major contributor to the summer 1993 floods was a continuous generation of mesoscale convective complex (MCC)-like systems. The UGM simulations presented here are on a 2.2° (latitude) by 2.8° (longitude) horizontal grid and are too coarse to adequately resolve mesoscale features. Pan et al.  also concluded that the anomalous, large-scale environment during summer 1993 created conditions favorable for MCCs. Such interactions occurred in their 30-day simulations (and verified against observations) in regions away from the simulated boundary-forcing locations, where MCCs were favored substantially by the large-scale anomalies. The large-scale circulation resolved by the UGM may have contributed to the precipitation produced in the GCIP region (Figure 9b). Wang et al.  found similar results to the present ensemble flood cases using much higher local resolution (50 km) in a global, variable resolution model for two-week simulations beginning 27 June 1993.
Figure 10 is similar to Figure 9, but for the 13-day, ensemble precipitation accumulation for May–June 1988. In contrast to 1993, the midwestern United States is relatively dry, with only 2 to 4 cm of precipitation over Iowa, Missouri and Illinois (Figure 9a). Accumulations exceed 6 cm over the Texas panhandle, and over central Texas and south Florida. The model simulation (Figure 10b) is also dry over the Mississippi basin and midwestern United States. Deficits in precipitation occur along the Gulf Coast and Florida, as in the 1993 ensemble. Figures 9 and 10 suggest the forecasts for the ensembles of 1993 and 1988 distinguish some of the observed precipitation differences between these years, although the magnitude of the differences is underestimated.
 To evaluate the role of surface latent heat flux on the 13-day forecast ensembles, we present the same ten cases in each summer with different surface evaporation. The June–July 1993 members are initialized with daily, 1993 variables, as before. However, the monthly averaged, reanalyzed surface evaporation from June 1988 is used for each case instead of daily 1993 values. For the May–June 1988 cases, all initial variables are daily, 1988 variables, with surface evaporation from the monthly averaged reanalysis value of July 1993. Figure 11 shows the monthly averaged, surface latent heat flux from the NCEP/NCAR Reanalysis for July 1993 (Figure 11a) and the difference field between July 1993 and June 1988 (Figure 11b). Simulated June–July 1993 precipitation (Figure 12a) is substantially reduced over the Midwest and north-central United States compared to the observations and to forecasts in Figure 9. Kansas and Missouri are excessively dry. A peak of 4 cm is located over Iowa and Illinois and a maximum of 5 cm is evident over North Dakota. The May–June 1988 simulations with July 1993 evaporation do not deviate significantly from the May–June 1988 with May–June 1988 evaporation cases (compare Figure 12b with Figure 10b), and accumulation over most of the Great Plains and Midwest is only about 2 cm.
 We next examine moisture transport in the UGM simulations and the ECMWF Reanalysis. The ECMWF Reanalysis may provide better resolution of moisture transport, since it includes an additional grid level below 700 mb (775 mb) not contained in the NCEP/NCAR Reanalysis. The UGM and ECMWF Reanalysis values at 00 and 12 UTC each day are summed for each 13-day case and the ensemble average is then computed. The June–July 1993 ensemble average, vertically integrated, meridional moisture flux is depicted in Figure 13. The ECMWF Reanalysis shows a pronounced northward flux of moisture over north-central Texas and the western Gulf of Mexico (Figure 13a). The moisture flux is associated with the Great Plains LLJ [e.g., Mo et al., 1995; Paegle et al., 1996]. The heaviest rain occurs just north of this jet core (see Figure 9a). UGM vertically integrated moisture flux is in fair agreement with the ECMWF Reanalysis (Figure 13b). The highest values of northward flux over the Great Plains are displaced westward over the Texas panhandle, just south of the modeled Kansas precipitation maximum.
Figure 13 suggests a significant, remote source of moisture for the June–July 1993 ensemble, associated with the the Great Plains LLJ. Dirmeyer and Brubaker  used a quasi-isentropic back trajectory analysis to show that the moisture source of peak flooding over the upper Mississippi and Missouri basins in July 1993 originated over the western Gulf of Mexico and the Caribbean Sea. They showed in contrast, moisture sources from the April to June 1988 drought period were mostly terrestrial in nature and moisture recycling from the surface was more prevalent.
 The meridional moisture flux for the May–June 1988 ensemble is depicted in Figure 14. The maximum, reanalyzed, southerly moisture flux over the Great Plains is about 60% smaller than that for June–July 1993 (compare Figure 14a with Figure 13a). The UGM shows similar results (Figure 14b), but moisture does not penetrate significantly northward into Nebraska or the Dakotas as shown in the ECMWF Reanalysis.
 The meridional moisture flux within present model integrations does not depend sensitively upon surface evaporation. For the June–July 1993 ensemble with June 1988 surface evaporation (Figure 15a), the average moisture flux over the Great Plains and Midwest is slightly weaker, but the pattern resembles the June–July 1993 ensemble with June–July 1993 surface evaporation (compare Figure 15a with Figure 13b). For the May–June 1988 ensemble with July 1993 surface evaporation (Figure 15b), the southerly moisture flux increases about 25–30% over Texas and western Oklahoma. The magnitude is similar to the May–June 1988 ensemble with May–June 1988 surface evaporation (Figure 14b).
 The time evolution of area-averaged, accumulated precipitation for a box encompassing much of the GCIP region (105°W to 88°W, 34°N to 50°N) is displayed in Figure 16. The model under-predicts precipitation by about 30% to 50% after the first week in the June–July 1993 ensemble (Figure 16a, solid circles). Accumulations increase steadily in time, similar to observations. For June–July 1993 with June 1988 surface evaporation, accumulations are lower (Figure 16a, open circles). The curve tends to closely follow the June–July 1993 ensemble with June–July 1993 surface evaporation to about hour 168 (day 7), and then it levels off to between 2 and 2.5 cm. For May–June 1988, the accumulations in both sets of model ensembles are very similar and in close agreement with observations until about hour 216 (Figure 16b).
 Surface evaporation appears to have little impact on model precipitation averaged over the bulk of the MRB during the first week of the June–July 1993 ensemble forecast. This suggests an important role for the large-scale circulation during this period. Surface processes and land-surface interactions appear to play a significant role in predictability beyond the first week in 1993, as supported by many previous studies using regional and global models over the GCIP domain.
 The present results, and examination of vertical motion simulations (not shown), suggest that the dynamical effects of the large-scale circulation contribute to the predictability in the 1988 ensemble more than in the 1993 ensemble. Several investigators have noted the importance of large-scale forcing on the 1988 U.S. drought. Namias , for example, noted the spring-to-summer persistence of anomalous upper-level anticyclones over the North Pacific, the central United States and the North Atlantic. Trenberth and Branstator  demonstrated the role of anomalous equatorial SSTs on the drought of 1988, and concluded local soil moisture feedbacks were secondary to the large-scale forcing.
 Comparison of the 1993 and 1998 UGM simulations (Figure 16) suggests that the large-scale circulation influences whether evaporation impacts precipitation. Evaporation tends to influence precipitation only if the large-scale conditions are favorable for precipitation [e.g., Barnston and Schikedanz, 1984], as during the summer of 1993. Pan et al.  for instance, demonstrate the important contribution of MCC-like systems to precipitation totals during the summer of 1993. Namias  finds that the large-scale circulation was not conducive to precipitation during the spring and early summer of 1988.
 Student's t-statistics for precipitation (not shown) were computed for the June–July 1993 and May–June 1988 UGM ensembles to study the significance of the forecasts during the two summers. The results show that a 5% confidence level for 10 degrees of freedom is maintained over the MRB after one week. During the second week, the response weakens as predictability weakens.
 Currently selected cases represent extremes in the zonal flow over the Rocky Mountains. The area-averaged 200 mb zonal wind, averaged over all ensembles, calculated over the Rockies in the region described in section 2 (120°W to 100°W, and 30°N to 50°N) is 22.4 m/s for the June–July 1993 cases. In contrast, the magnitude for the May–June 1988 cases is 12.1 m/s. As previously noted, the strength of the cross-Rockies zonal flow appears to influence the circulation east of the Rockies, including the strength of the LLJ. The next section describes the mechanical effect of orography with emphasis on longer timescales than those that characterize the two sets of ensembles presented here. This is helpful for dynamical interpretation of prior results.
4. Ambient Flow Oscillations: Subcritical and Supercritical Flows
 A possible interpretation ambient flow oscillations is found using the steady solution of the barotropic vorticity equation, summarized below. The phase speed, C, of Rossby waves in a one-level, non-divergent, barotropic fluid is:
where U is the uniform, zonal background flow, k is horizontal wave number, and β is the meridional gradient of the Coriolis parameter, assumed constant for a “beta-plane” approximation. For a quasi-stationary seasonal response, C = 0, and “supercritical” conditions exist when
“subcritical” conditions exist when
and resonance occurs when
 The strength of the response maximizes near resonance. It can be shown that the steady state response forms anticyclones over mountains and cyclones over valleys for supercritical conditions, and the opposite distribution, with cyclones over the mountains for subcritical conditions. Strong westerlies favor supercritical flows, while weak westerlies favor subcritical conditions. A more detailed analysis is provided by Byerle and Paegle .
Nogués-Paegle  and Charney and DeVore  studied an idealized case, forced by just one spectral harmonic in the surface orography. They emphasized the solution destabilization and bifurcation that occur, respectively, as the flow fluctuates about resonant values that separate supercritical and subcritical conditions. Paegle et al.  showed topographically induced ultralong waves in the Northern Hemisphere have a nearly equivalent barotropic structure in the troposphere.
 Over North America, supercritical conditions are satisfied more easily in winter when stronger zonal winds prevail across the Rocky Mountains compared to summer, when the winds weaken. A consistent interpretation of reanalyzed fields (section 2) is that increased zonal flow during summer accelerates the orographic vortex as the resonant point in the response is approached.
 One interpretation of the correlations of cross-orography zonal flows with lee-side responses of LLJs, moisture flux, and precipitation is that increasing cross-mountain flow accelerates the summer cyclone situated over the Rockies through mechanical processes relating to subcritical ambient flows approaching resonance. This interpretation is reasonable only to the extent that the dynamical processes in question actually tend to develop a cyclone over the central Rockies in the summer season. The next section demonstrates that the lower-tropospheric cyclone commonly observed above the central Rockies in summer may be at least partly explained by a subcritical, stationary, orographic Rossby wave response, and that the observed and modeled intraseasonal oscillations may therefore be partly explained by the simple dynamical processes described in the present section.
5. Seasonal Cycle Over North America
 To more clearly support or refute the possibility that the observed summer cyclone reflects the orographic mechanism, it is necessary to demonstrate that the effects are simulated in a model that retains a full and realistic spectrum of orography. The model should also use realistic ambient flow, taken from observed conditions for each season and allow reasonable vertical shear and more general processes than does a purely barotropic model. To show that the simple barotropic results are relevant to actual prediction, they should be reflected in a global, primitive equation model such as the Utah Global Model (UGM). We present evidence which suggests that mechanical effects influence the seasonal reversal around the Rocky Mountains, and hence the LLJ east of the mountains, using simulations initialized with monthly averaged variables and comparisons to reanalyses. Much of the Rocky Mountain region lies above 850 mb. We therefore focus on 700 mb (sigma level 0.72 in UGM integrations), where artificial interpolation below the surface of the Earth is not required over most of the area.
 The integrations are performed omitting radiative and latent heating. They consist of month-long simulations of monthly averaged conditions taken from January and July from the NCEP/NCAR Reanalysis, averaged from 1951–2000. Orography is represented in a spherical harmonic, wave number 42 triangular truncation. The zonally averaged portion of the simulated circulation is maintained at the reanalysis value for each month by rapid relaxation of the rotational portion of the zonally averaged flow toward the climatology for each month. To isolate the mechanical, orographic effect of the Rockies, differences are taken between runs with orography over the entire globe and runs without orography over the North American region. In the latter case, initial surface pressure over North America is set to a constant (1013 mb) over the region from 20°N to 60°N, and 130°W to 70°W.
Figure 17 shows differences in the average wind vectors (mountain minus no-mountain) for simulations with and without orography. The average is for the last 30 days of a 40 day forecast. During January (Figure 17a), the mechanical effect of orography produces a lower tropospheric anticyclone over the western United States and northerly flow east of the mountains. The orographic effect for July produces cyclonic curvature within the lower troposphere above the central Rocky Mountains (Figure 17b) and southwesterly flow east of the mountains over the southern Great Plains.
 We next compare the mechanical effect of the UGM (Figure 17) to reanalyzed, climatological features of the low-level circulation flanking the orography over western North America for summer and winter. As in section 2, the data are based on a 50-year record (1951–2000) of monthly averaged circulations from the NCEP/NCAR Reanalysis.
Figure 18 shows the seasonal composites of the eddy height and wind field for December–January–February (DJF) and June–July–August (JJA), for the years 1951 to 2000. The eddy fields are constructed by subtracting the zonal mean from the climatological average. The winter composite at 700 mb shows an anticyclone over the western United States, centered near southeastern Washington (Figure 18a). A trough becomes prominent in summer, centered just off the west coast of California (Figure 18b). The spring and autumn fields (not shown) have intermediate patterns to summer and winter. The spring configuration is more similar to summer, while the fall is more similar to winter. Paegle et al.  showed the upper troposphere exhibits ridge-trough reversals beginning in northern spring. The 850 mb eddy wind field for JJA (see Figure 1a) displays the Great Plains LLJ over the western Gulf of Mexico and Texas.
Figure 19 shows the annual march of relative vorticity area-averaged from 30°N to 50°N and 130°W to 110°W, in the vicinity of the circulation reversals shown in Figure 18. Figure 19a depicts the complete vorticity field, including the zonally averaged contribution, as well as the longitudinally varying wave portions. This clearly indicates a winter anticyclone and spring cyclone near the west coast of North America, with the strongest values of cyclonic (positive) vorticity occurring in May. Figure 19b displays the seasonal evolution of the longitudinally varying (eddy) portion of the vorticity pattern. Here, the winter to summer reversals are particularly evident. The wave field shows a peak in vorticity in early summer, and then slightly weaker values in July and August.
 The mean cyclonic circulation during summer may be partly explained by seasonally reversing heating/cooling influences on elevated plateaus. Comparison of the orographic effect of the UGM in Figure 17 with reanalyzed eddy fields in Figure 18 suggests that the mechanical effect of orography also plays a prominent role in the seasonal reversals over the Rocky Mountains and indirectly supports the relevance of this mechanism to observed intraseasonal oscillations described in section 2 and modeled events in section 3.
 The flood and drought ensembles presented in section 3 may be examples in which the orographic mechanism influences predictability over the MRB through ambient flow interaction with the Rocky Mountains. In section 5 we have discussed the potential for longer-term (e.g., intraseasonal) predictability and further understanding of the seasonal reversals based on this mechanism. Previous studies have also described anomalously strong (weak) ambient flows in the upper troposphere and their persistence over the Rocky Mountains during the summer of 1993 (1988) (e.g., Mo et al. [1995, 1991], respectively). We hypothesize that the classic theory of ambient zonal flows over topography may provide an alternative mechanism for enhanced predictability of such features.
6. Summary and Conclusions
 The present study promotes the hypothesis that the orography of western North America provides a scale transfer mechanism, focusing global-scale features into regional responses. The LLJ east of the Rocky Mountains is one component of a larger-scale, orographically bound, summer cyclone whose strength increases with increasing cross-barrier flow. This interpretation provides an alternative, mechanical explanation to thermodynamic processes related to surface heating influences that have often been used to explain the anomalous response of the 1993 floods [e.g., Giorgi et al., 1996; Paegle et al., 1996, and references therein; Beljaars et al., 1996; Dirmeyer and Brubaker, 1999; Viterbo and Betts, 1999].
 The mechanical, orographic effect also implies an alternative explanation to the pronounced seasonal reversals that characterize the western United States. Analysis of the seasonal cycle provides one context in which to investigate the relative role of surface heating and mechanical, orographic influences. The observed winter-to-summer reversal from low-level anticyclonic flow (winter) to cyclonic flow (summer) is most commonly explained in terms of seasonal reversal of surface heat balance from summer to winter, as occurs in monsoon climates, but may also be strongly affected by the scale-transfer mechanism due to orography.
 Thermal effects of sensible and latent heating have been frequently cited as potential sources of enhanced predictability of the coupled ocean-land-atmosphere system. In particular, a large number of investigations have been made of the influence of soil moisture anomalies upon model forecasts; others have considered ice and snow states and the influence of surface albedo. One intent of the present study is to re-emphasize the mechanical influence of orography as a land-surface effect. Unlike soil moisture, ice state, and albedo, orography is relatively fixed over periods of short-to medium-range climate change. The orographically induced response is nevertheless not fixed, but depends upon the ambient atmospheric state and its anomalies.
 In cases where anomalies of the atmospheric state possess relatively large scales, the orographic influence may promote a relatively predictable lee-side response, and may help to explain and possibly predict extreme events such as those observed over the GCIP domain in 1988 and 1993. In these cases there may be some basin-scale skill in anomalous precipitation prediction for one to two weeks.
 This research was partly supported by NSF grants ATM0106776, ATM0109241, and NOAA/PACS NA06GP0451 to the University of Utah. The first author's participation in this research is also supported by the USAF. We would like to thank the anonymous reviewers for many helpful comments that improved our research and the manuscript.