Low-level jets and the summer precipitation regimes over North America

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109,, doi: /2003jd004106, 2004 Low-level jets and the summer precipitation regimes over North America Kingtse C. Mo Climate Prediction Center, NOAA/National Weather Service, National Centers for Environmental Prediction, Camp Springs, Maryland, USA Ernesto H. Berbery Department of Meteorology, Earth System Science Interdisciplinary Center, University of Maryland, College Park, Maryland, USA Received 25 August 2003; revised 31 December 2003; accepted 14 January 2004; published 30 March [1] The summer precipitation regimes over the United States and their relation to two low level jets are examined using a daily gauge-based precipitation data set, the NCEP-NCAR reanalysis and the 10-yr summer (June September) simulations based on the 50-km NCEP regional spectral model (RSM) with the initial and boundary conditions provided by the global reanalyses. The quality at regional scales of the RSM simulations is assessed through the comparison with the NCEP s operational Eta Data Assimilation System (EDAS) analyses. The RSM, as EDAS, captures the seasonal evolution of the North American monsoon rainfall and the related vertically integrated moisture fluxes, but details differ. The fluxes associated with the Great Plains low level jet (GPLLJ) and the Gulf of California low level jet (GCLLJ) depicted by the two data sets are similar, but the 50-km RSM has difficulty in capturing the vertical structure of the meridional moisture flux associated with the GCLLJ. The recurrent summer precipitation regime is a three cell pattern that consists of an inverse phase relationship between the Great Plains and the core monsoon region with an additional weak center over the southeastern United States. Over the southwestern United States, Arizona and New Mexico belong to two different rainfall regimes and have different moisture sources. The out of phase precipitation relationship is consistently related to an out of phase relationship between the two LLJs, whose variations are in turn associated with the upper level jet streams. The strong GCLLJ cases imply a weaker GPLLJ and less rainfall over the central United States and the Mississippi Valley. The strong GPLLJ cases only imply weaker meridional moisture fluxes from northern Mexico or Gulf of California to the southwestern United States and less monsoon rainfall there, but they do not have an impact on moisture fluxes along the Gulf of California. INDEX TERMS: 1854 Hydrology: Precipitation (3354); 1833 Hydrology: Hydroclimatology; 9350 Information Related to Geographic Region: North America; KEYWORDS: low level jets, summer rainfall Citation: Mo, K. C., and E. H. Berbery (2004), Low-level jets and the summer precipitation regimes over North America, J. Geophys. Res., 109,, doi: /2003jd Introduction [2] The most recurrent summer precipitation pattern over North America, from intraseasonal to interannual timescales, frequently exhibits an inverse relationship between the Great Plains and the core monsoon region over northwestern Mexico and the Southwest. Additionally, this rainfall pattern is related to a third center over the Southeast [Higgins et al., 1997; Mo et al., 1997; Higgins et al., 1998]. The out-of-phase relationship manifests itself not only on the precipitation and the dynamical fields [Barlow et al., 1998], but also in the moisture flux convergence fields Copyright 2004 by the American Geophysical Union /04/2003JD [Berbery and Fox-Rabinovitz, 2003]; therefore, it is expected that two low level jets (LLJs): the Great Plains low-level jet (GPLLJ) and the Gulf of California low-level jet (GCLLJ) could be an important part of this relationship. [3] The GPLLJ transports moisture from the Gulf of Mexico to the Great Plains and is often responsible for the region s nocturnal convective activity [Helfand and Schubert, 1995; Higgins et al., 1997]. The GCLLJ transports moisture through the Gulf of California to northwestern Mexico and the southwestern United States. The GPLLJ is stronger, and is captured by conventional observation systems and most global analyses, although in the form of intensified winds rather than with a mesoscale structure. Thus its climatology and variations are well documented [Bonner, 1968; Bonner and Paegle, 1970; Helfand and 1of18

2 Schubert, 1995; Mo and Higgins, 1996; Higgins et al., 1997]. The GCLLJ, on the other hand, is weaker and narrower in comparison to the GPLLJ. Because of the limited wind measurements along the Gulf of California, large uncertainties exist on its structure and location. In addition, its depiction by regional and global models shows large disparities (D. S. Gutzler, H. K. Kim, and R. W. Higgins, private communication, NAMAP project ). [4] Even though the structure of the GCLLJ is still unclear, its importance to the development of convection and monsoon rainfall has been long established [Hales, 1972, 1974; Brenner, 1974]. Hales [1972] found that strong surge events occur a few days after easterly waves pass the southern part of the Gulf of California. The bridge between the surge events and the increase of moisture transport into the monsoon region is the GCLLJ [Anderson et al., 2000a, 2000b]. [5] Both LLJs and rainfall over the Great Plains and the Southwest have strong diurnal cycles. The GPLLJ maximum occurs at 0600 UTC (2300 LST) and continues with large values until 1200 UTC (0500 LST). Helfand and Schubert [1995] and Higgins et al. [1997], among many others, showed that the GPLLJ sustains nocturnal rainfall; similarly, Mo et al. [1997] examined the diurnal cycle of rainfall during wet events in the central United States and found that the maximum occurs during UTC. The diurnal cycle of the GCLLJ has large uncertainties. The diurnal variation of the meridional wind profiles at Puerto Peñasco along the northern Gulf of California coast indicates that the maximum wind in the surface to 3 km layer occurs at 0800 UTC (0100 LST) and the weakest winds occur from the late afternoon to near midnight [Douglas et al., 1998]. Over the Gulf of California at 31 N, the NCEP Eta Data Assimilation System (EDAS) analyses indicate that the maximum moisture fluxes below 925 hpa occur at 0600 UTC [Berbery, 2001]. However, the diurnal maximum may depend on the exact location of the measurement, and whether it is over land or water. Despite models inaccuracies in depicting the diurnal maximum of the GCLLJ (Gutzler et al., unpublished report, 2003, available at html), there are indications to suggest that the GPLLJ and the GCLLJ diurnal maxima may differ by a few hours. [6] The physical mechanisms responsible for the inverse relationship of rainfall are not fully understood, but an element to be taken into account is the role of the upper level tropospheric jet streams. Uccellini and Johnson [1979] and Uccellini [1980] suggested that the GPLLJ is a component of the large scale circulation associated with the upper level jet streaks. Byerle and Paegle [2003] also suggested that the GPLLJ is modulated by the orographic forcing: When ambient flow passes over the Rockies, the stronger (weaker) zonal winds have a lee side response in the form of a stronger (weaker) GPLLJ. If the two LLJs are related, then the same forcing could also modulate the GCLLJ, although in a different way. [7] Studies on summer precipitation and its association with the two LLJs are difficult because the GCLLJ in particular cannot be resolved by the coarse resolution global analyses [Schmitz and Mullen, 1996]. The understanding of the GCLLJ largely depends on diagnostic studies using limited station data [Douglas, 1995; Douglas et al., 1993], model simulations [Stensrud et al., 1995; Stensrud et al., 1997; Anderson et al., 2000a, 2000b, 2001; Kanamitsu and Mo, 2003] and regional model analyses [Berbery, 2001]. These data sets were too short to derive statistical relationships. Recently, 10-year simulations were performed using the National Centers for Environmental Prediction (NCEP) regional spectral model (RSM) for the period June September (JJAS) from 1991 to This data set, together with the NCEP - National Center of Atmospheric Research (NCEP-NCAR) reanalysis and observed precipitation data, are employed here to investigate: (1) the role played by the two LLJs in the out-of-phase precipitation regime, and (2) the linkages between LLJs and the upper level tropospheric jet streams. Data sets and procedures used are given in section 2. An intercomparison between the RSM simulations and the products from the EDAS is given in section 3 to assess the uncertainties of the RSM simulations. Composites of intense cases of precipitation and moisture fluxes are presented and discussed in section 4, while section 5 documents the corresponding upper level circulation composites. Lastly, a summary and conclusions are given in section Data and Procedures 2.1. Ten-year Simulation by the RSM [8] The primary data set for this study consists of 10 years ( ) of summertime simulations using the RSM [Juang et al., 1997; Juang and Kanamitsu, 1994]. Major recent improvements are documented in Kanamitsu and Mo [2003]. In this version, the terrain and vegetation lower boundary conditions are obtained from the United States Geological Survey and the convection scheme is the relaxed Arakawa Schubert (RAS) scheme [Moorthi and Suarez, 1992]. The model has 28 sigma levels in the vertical and the horizontal grid spacing is 50-km for a domain covering the United States and northern Mexico (18 52 N, W). The initial and boundary conditions were taken from the NCEP-DOE reanalysis II (R2 [Kanamitsu et al., 2002]). The RSM is a perturbation model: It predicts the difference from the outside model forecasts or analyses, which in this case are the R2 products. Therefore both R2 fields at the boundaries and inside the domain are used. Relaxation is used to smooth fields between the global and regional models in the shallow boundary zones [Juang et al., 1997]. [9] For each summer month (June to September) from 1991 to 2000, the model was initialized at 0000 UTC of day 1 of each month, and was integrated through the entire month. After initialization, soil conditions including soil wetness and soil temperature were predicted by the model. The spin up time for rainfall or soil moisture is about h. The dynamical quantities like temperatures and winds are less influenced by the restart. To avoid spin up problems, the composites were performed using the precipitation anomalies from the observed unified analysis. The only variables used from the RSM simulations are the vertically integrated moisture fluxes. Daily means were obtained for all variables and were used for this study. For all data sets, anomalies were defined as the departures from the 10-year JJAS climatology. To study the mesoscale features of the GCLLJ, 30-km RSM simulations for July and August from 1997 to 2000 were also performed over a smaller domain 2of18

3 (20 40 N, W) using the same model and procedures as in the 50-km RSM Complementary Data Sets [10] Eta model products are employed to assess the quality of the RSM simulations, in particular for those regional aspects that are not captured by the coarser resolution global reanalyses. The domain of the Eta model covers all North and Central America, and extensive portions of the nearby oceans. The performance of the Eta model over the core monsoon area during was discussed by Berbery [2001]. Given the operational nature of this model, the analyses have been subject to changes during the analysis period in physics, resolution, and data assimilation procedures, which have somewhat affected the homogeneity of the data set. For example, the grid spacing started at 48 km in 1995 and increased to 32 km in 1998; it was increased again to 22 km and 50 levels in the vertical in From 27 November 2001 to the present, the operational Eta model has 12 km grid spacing and 60 vertical levels. Major changes to the assimilation/forecast system and the potential effects on precipitation are discussed in Berbery et al. [2003] and are also listed at [11] Still, these regional analyses are more advantageous than the global reanalyses or monthly RSM simulations to represent mesoscale phenomena like the LLJs. The global reanalyses do not have a realistic representation of the regional circulations associated with the North American monsoon, and least of the GCLLJ. In fact, the Gulf of California is practically missing in the global model s geography (Berbery [2001], Berbery and Fox-Rabinovitz [2003], and other authors). EDAS, on the other hand, resolves the GCLLJ and reproduces realistically other circulation features in the region. [12] The higher resolution EDAS also implies a better use of satellite data, including sea surface temperatures and radiances. Since June 1998, EDAS has had a fully continuous cycling of all atmospheric states and land states, such that the EDAS is no longer restarted every 12-hours from the global data assimilation system (GDAS) [Berbery et al., 2003]. EDAS also ingests GOES satellite-based 3-layer precipitable water estimates, which should improve the moisture fluxes. The Eta model precipitation is taken from the h forecasts. Unfortunately, daily data were not available for the present study, as only monthly means were archived. [13] The precipitation data set used is the unified United States/Mexico daily precipitation data set [Higgins et al., 2000]. The data were interpolated from daily rain gauge observations to a 1 1 grid. In cases when the study involves larger scale features, the NCEP-NCAR reanalyses at grid spacing were used as a complement. The large scale circulation anomalies were derived from the daily mean global gridded analyses from the NCEP-NCAR reanalyses (R1 [Kalnay et al., 1996]) Composites [14] In this study, the daily means of vertically integrated moisture fluxes (qflux:{qu},{qv}) were obtained from the 10-year 50-km RSM JJAS simulations. The precipitation anomalies were obtained from the observed unified analysis to avoid the spin up problem. Composites of precipitation, qflux and circulation anomalies were constructed from daily LLJ and precipitation indices for the summer period from 1991 to 2000 defined below: [15] Two indices were defined to measure the strength of the {qv} over the Great Plains and over the Southwest: [16] 1. A GPLLJ index is defined as the daily mean {qv} averaged over N, W. It measures the {qv} over the Great Plains. [17] 2. A GCLLJ index is defined as the daily mean {qv} averaged over N, W, which measures the {qv} from the northern Gulf of California to the Southwest. [18] The {qv} s were obtained from the 50-km RSM simulations. A strong positive (negative) GCLLJ index means that moisture transport from the Gulf to the Southwest is strong (weak). This index is not strictly a measure of the GCLLJ but rather of the meridional moisture flux {qv} at the location of its core as identified in Douglas [1995], and Berbery [2001]. This index does not measure the {qv} along the Gulf of California nor the intensity or frequency of surge events. [19] The jet index anomaly is defined as the departure from the given index computed from the 10-yr mean over JJAS. We then computed the standard deviation for each index. Positive (negative) LLJ events were selected based on the threshold method, that is, when the LLJ index anomaly for a given day is greater than 0.8 (less than 0.8) times the standard deviation, that day is classified as a positive (negative) event. [20] In addition to the LLJ indices, four precipitation indices are defined to identify wet and dry periods over different areas: [21] 1. P(SGP): P averaged over the Southern Great Plains (32 38 N, W). [22] 2. P(NGP): P averaged over the Northern Great Plains (38 48 N, W). [23] 3. P(AZ): P averaged over Arizona (32 36 N, W). [24] 4. P(NM): P averaged over New Mexico (32 36 N, W). [25] Precipitation events were chosen based on rainfall distribution from 1991 to When a daily precipitation index is in the top 85% (lowest 15%) of daily rainfall accumulation, that day is counted as a positive (negative) precipitation event. Composites based on different indices are used to examine the robustness of the summer rainfall regimes. [26] The statistical significance of the composite of precipitation difference between positive and negative events was assessed by the Monte Carlo method. We randomly composited equal number of positive and negative events from the same data set, and repeated the process 500 times. We then counted the number of times that the composite based on any given index is larger than the composite of random selected maps. If only 5% or less of the randomly selected cases (random probability) have values greater than the composite based on the index, then the composite based on the index is considered to be statistically significant at the 5% level. The statistical significance for the vertically integrated moisture fluxes and other circulation composites was determined by assuming a normal distribution. Then the student t-test was performed to determine the statistical significance. 3of18

4 Figure 1. Mean summer precipitation for the period derived from the unified precipitation data set for (a) June, (b) July, (c) August and (d) September. Contour interval is 1 mm day 1 and the zero contours are omitted. Areas where values are larger than 1 (4) mm day 1 are shaded light (dark). [27] The two LLJs have slightly different diurnal maxima and it is difficult to adjust this difference, therefore composites shown here are based on the daily indices. Then, the question may arise whether the LLJ indices based on daily mean {qv} are representative of the actual strength of the LLJs. The RSM outputs were archived 4 times daily for July September each year. We calculated the LLJ indices using the daily, 0600 UTC and 1200 UTC vertically integrated moisture fluxes. We then did the correlation between the GPLLJ (GCLLJ) index at 0600 UTC and the daily index, and the correlation resulted above 0.97 (0.95). This suggests that the differences between daily {qv} and 0600 UTC {qv} are largely in the mean values and not in the anomalies, and do not affect our results. We also repeated the composites for {qv} at 0600 UTC based on the observed precipitation indices. The composites of {qv} at 0600 UTC indicate that the relationships between precipitation and the LLJs are still valid for both daily and 0600 UTC data. 3. Intercomparison Between the RSM Simulations and the EDAS [28] Figure 1 presents the mean monthly evolution of observed precipitation during the warm season from June to September. The evolution shows the following basic features: (1) an increase of precipitation over western Mexico, representing the monsoon onset from June to July; (2) an increase of rainfall over the Southwest as the season progresses from June August. At the same time, precipitation decreases over the central United States; and (3) a decrease of monsoon rainfall over northern Mexico in September. The onset of the monsoon over Mexico occurs in July, while the maximum precipitation over Arizona-New Mexico occurs in August. [29] The monthly mean RSM outputs are compared with the observed precipitation and the EDAS outputs for the common period (Figures 2 and 3). The agreement among them shows the reliability of the simulations and the disagreement indicates the uncertainties. Figure 2 shows that the overall aspects of the precipitation fields in both models are similar to those from observations. The RSM precipitation in June is stronger than both the Eta and the observed precipitation over the northeastern United States. The onset of the monsoon over northwestern Mexico occurs earlier in the RSM precipitation during June than in the Eta forecasts in July; its withdrawal during September is also captured by both models, although the Eta precipitation seems to last longer. The June to September decay of precipitation over the central United States is also captured by both models. The RSM systematically simulates more rainfall over the Rockies and the Eta model is overall too 4of18

5 Figure 2. Same as Figure 1 but (a d) for the RSM simulations and (e h) for the Eta model forecasts. 5of18

6 Figure 3. Vertically integrated meridional moisture flux averaged for the period for (a) June, (b) July, (c) August and (d) September from the RSM simulations. Contour interval is 20 kg (ms) 1 and the zero contours are omitted. Values larger than 40 (60) kg (ms) 1 are shaded light (dark). (e) (h) same as Figures 3a 3d, but from the ETA model outputs. 6of18

7 Figure 4. Vertical cross section of meridional moisture flux at 30 N averaged for the period for (a) July, (b) August, and (c) September from the RSM simulations. Contour interval is 10 g kg 1 ms 1 and zero contours are omitted. Contours 5 and 5 g kg 1 ms 1 are added, (d) (f ) same as Figures 4a 4c, but from the Eta model outputs. dry over western Arizona. The large precipitation over southern Mexico in the Eta forecasts was discussed by Berbery [2001] and attributed to the convective parameterization scheme [Manikin et al., 1998]. [30] The mean meridional moisture flux from the RSM is smaller in magnitude than that estimated from EDAS (Figure 3). Both models show large moisture flux associated with the two LLJs, but the most noticable aspect is the signature of the moisture flux associated with the GPLLJ extending from Texas to the Great Plains. The RSM tends to show two maxima for the GPLLJ: one over southern Texas and the second over Oklahoma/Arkansas. These two maxima were also reported by Uccellini [1980] and Bonner [1968] as areas that a great number of LLJ events occurs. On the other hand, the Eta analyses show mostly an elongated maximum with only hints of the two centers. [31] Both models capture the northward moisture flux maximum extending from the Gulf of California to Arizona/ California, which represents the extension of the GCLLJ into the Southwest. During August and September, both models show another maximum located south of the Gulf centered at 20 N, 108 W, where the easterly waves are most 7of18 likely to develop. Both models show that the fluxes north of the Gulf of California increase from June to July and reach a maximum in August and September, while the moisture flux over the central United States is strongest in June and diminishes as the season progresses. This is consistent with the precipitation evolution: Rainfall over the central United States decreases from June to July after the monsoon s onset over the core monsoon/southwestern U. S. [Barlow et al., 1998]. [32] The cross sections of the meridional moisture flux at 30 N for July to September are presented in Figure 4. Both models show that the moisture flux associated with the GPLLJ is strongest in June (not shown) and July and extends to 700 hpa with a maximum near 925 hpa along the eastern slopes of the Rockies. However, the intensity at the core of the GPLLJ in the RSM simulations is 30% weaker than that of the EDAS. However, more importantly, the RSM cannot resolve the vertical structure of the GCLLJ, which is apparent in the EDAS. During July, the RSM captures the positive moisture flow at 850 hpa centered around 114 W, but the magnitudes are much weaker than those of the Eta model. During August, the RSM only

8 difficulty to fully capture the vertical structure of the GCLLJ. On the other hand, there are large uncertainties in the GCLLJ due to the lack of observations. Figure 5. Same as Figure 4 but for (a) July and (b) August from the 30-km resolution RSM. shows small northward moisture flux near the surface, while in the Eta model it extends from the surface to 900 hpa. The reason for this discrepancy seems to be related to the resolution, which apparently is marginal for resolving that jet. This is suggested by the 30-km RSM simulations (Figure 5), where the GCLLJ is now evident in the July and August cross sections at 30N. [33] The difference in vertical structure in the two RSM runs might have an impact on the composites based on the GCLLJ index. For this reason, we compare the mean {qv} over the western domain for the 30-km RSM and the 50-km RSM, averaged for July and August from (Figure 6). Both simulations show positive (northward) meridional moisture fluxes from the Gulf of California to the Southwest, but the {qv} is stronger for the 30-km RSM. The {qv} differences between the 30-km and the 50-km RSM simulations appear largely in the mean, the anomalies are closer. An example is given in Figure 6c, which shows the GCLLJ index anomaly for July August 1997 from both the 30-km RSM (open circles) and the 50-km RSM (dark circles). The correlation between the daily fluctuations of the two GCLLJ indices for July and August is The 30-km RSM domain cuts off a part of the Gulf of Mexico and the eastern portion of the GPLLJ, thus the GCLLJ index from the 50-km is used to study the relationships between two LLJs. Overall, large scale features related to the vertically integrated moisture fluxes depicted by the RSM and ETA are similar, despite the 50-km RSM s 4. Composites Based on Precipitation Anomalies [34] In this section, the relationships between precipitation and the two LLJs are discussed based on composites keyed to the indices defined in section 2. To further enhance the signal, Figures 7 12 present the difference between positive and negative composites Precipitation [35] The observed precipitation composites based on the P(NGP), P(SGP) and P(AZ) indices (Figures 7a 7c) show basically a similar three-cell pattern with an inverse relationship between the Great Plains and the Southwest and the third center over the Southeast, although the details among them differ. For example, the center over the southeastern U. S. can be shifted or weak. However, when the P(NM) index is used (Figure 7d), the spatial structure of the composite pattern changes: In addition to the positive center in New Mexico, there is a very weak negative center over the Ohio Valley, but its structure bears little similarity to the other composites. This is consistent with the results of Gutzler [2000], who suggests that Arizona and New Mexico belong to two different climate regimes. In the next section, we will show that they also are associated with different moisture transports Moisture Fluxes and LLJs [36] While the inverse relationship between precipitation anomalies from the Great Plains and the core monsoon/ southwestern U. S. is present in the precipitation data, the magnitudes of rainfall anomalies and the location of rainfall maxima depend on the detailed structure of the two LLJs. [37] Figure 8a shows that the increase (decrease) of rainfall over the northern Great Plains is associated with a strengthening (weakening) of the GPLLJ near the jet maximum over Oklahoma/Kansas; similarly, the increase (decrease) of rainfall over the southern Great Plains is associated with the strengthening (weakening) of the GPLLJ near the Texas coast (Figure 8c). [38] When the Great Plains is wet and the GPLLJ is strong (Figures 8a and 8c), there are no moisture fluxes extending from the Gulf of California to Arizona, which is consistent with the reduced precipitation over Arizona when the Great Plains precipitation is intense. Notice that there is still moisture transport into the Southwest from the North Pacific. When the Great Plains is dry (Figures 8b and 8d), both branches of moisture fluxes from the North Pacific and the Gulf of California are enhanced with a strong anticyclone centered over the Southwest. [39] The composite of large precipitation events over Arizona (Figure 9a) suggests an increase of the northward moisture flux from the Gulf of California, and a decrease of the moisture flux from the Gulf of Mexico into Texas in comparison with climatology (Figure 3). The opposite is true for the dry Arizona composite (Figure 9b); therefore, this index, as well as indices based on the Great Plains precipitation, identifies similar linkages between precipitation and moisture fluxes. Both cases show the inverse 8of18

9 Figure 6. Vertically integrated meridional moisture flux averaged for July August from for (a) 30-km RSM, (b) 50-km RSM simulations. Contour interval is 15 kg (ms) 1 and the zero contours are omitted. Values larger than 30 (60) kg (ms) 1 are shaded light (dark) and (c) the GCLLJ index anomaly for July and August 1997 computed based on the 30-km RSM (open circles) and the 50-km RSM (dark circles). The anomaly is defined with respect to the total mean over the period July August relationships between the GPLLJ and the GCLLJ and the association between the two LLJs and precipitation. [40] Because the LLJs have strong diurnal cycles, composites for {qv} at 0600 UTC are also given in Figure 10 for comparison. The 0600 UTC {qv} data are only available for July September. The major differences between the daily and 0600 UTC composites are in their magnitudes: the LLJs are stronger when 0600 UTC data are used. The similarity between composites of daily {qv} and {qv} at 0600 UTC indicates that the inverse relationship between the two LLJs is robust. The composites for 0600 UTC data also show the same relationships between precipitation anomalies and the relative strength of the corresponding LLJ. [41] A somewhat different situation emerges when the base point is located over New Mexico (Figures 9c and 9d). For wet New Mexico, there is no increase of the moisture fluxes from the Gulf of California to the Southwest (Figure 9d), rather the moisture transport from the southern branch of the GPLLJ strengthens slightly: there are more moisture fluxes extending from the border of Texas and the Gulf of Mexico to New Mexico and northern Mexico. The jet maximum over the border of Texas and Mexico strengthens. For dry New Mexico, the zonal transport from the southern branch of the GPLLJ is weaker. This suggests that the moisture transport from the Gulf of Mexico contributes to rainfall over New Mexico and northern Mexico, and the moisture fluxes from the Gulf of California are essential to bring rainfall to Arizona. In addition, moisture fluxes from the North Pacific contribute to rainfall over both Arizona and New Mexico. Therefore Arizona and New Mexico seem to belong to different local precipitation regimes with different moisture sources. [42] In summary, from the above composites, it can be seen that the robust summer precipitation regime over the United States exhibits a three cell pattern, which includes the out of phase relationship between precipitation anomalies over the Great Plains and the Southwest (and the core monsoon) with an additional weak center over the Southeast. The precipitation anomalies largely co-vary with the low level winds so that there is also an out of phase relationship between the two LLJs. 5. Composites Based on LLJs Intensities 5.1. Precipitation and Moisture Transport [43] In the previous section, composites were based on precipitation indices; here, composites based on two LLJ indices, as defined in section 2.2, are presented to examine 9of18

10 Figure 7. Composite difference of the observed precipitation anomalies between the positive and negative P index over (a) northern Great Plains, (b) southern Great Plains, (c) Arizona, and (d) New Mexico. Contour interval is 1 mm day 1 and zero contours are omitted. Contours 0.5 and 0.5 mm day 1 are added. Areas where positive (negative) values are statistically significant at the 5% level are shaded dark (light). the corresponding pattern of precipitation, moisture fluxes, and circulation. The composites for strong and weak LLJ events have similar patterns with opposite sign, so results are presented as their difference Composites Based on the GCLLJ Index [44] The strong (weak) GCLLJ index events are associated with large (low) precipitation over Arizona and western New Mexico, and low (high) precipitation over the Great Plains and the Mississippi valley (Figure 11a). The slightly positive precipitation anomalies over Florida are only statistically significant at the 90% level. Note that the precipitation composite based on the GCLLJ index (Figure 11a) is similar to the composite based on the precipitation index over Arizona (Figure 7c), and to that based on the precipitation index over the southern Plains with a sign reversal (Figure 7b). The composite difference of moisture fluxes (Figures 11c and 12a) shows large fluxes from the Gulf of California to Arizona with a strong anomalous anticyclone centered over the Southwest (36 N, 105 W). A strong GCLLJ index implies more moisture transport from the northern Gulf into Arizona (Figure 12a), which produces more rain, in agreement with several previous studies. The corresponding negative meridional moisture flux anomalies over the central United States are indicative of a weak GPLLJ Composites Based on the GPLLJ Index [45] The precipitation composite difference based on the GPLLJ index (Figure 11b) shows the familiar three cell pattern, and is similar to the composite based on the northern Great Plains precipitation index (Figure 7a). Note that by the GPLLJ index definition, it will produce results that are closer to the northern Great Plains precipitation index rather than to the southern Great Plains index. The composite difference of meridional moisture flux based on the GPLLJ index (Figures 11d and 12b) shows positive anomalies with a maximum located over Oklahoma and negative anomalies over the western United States. Figure 12b shows that there is a strong anomalous cyclone centered at 42 N, 108 W, which is located northeast of the anomalous anticyclone from the GCLLJ composite. Note that there are no anomalous flux along the Gulf of California (Figure 12b). [46] The moisture flux composites based on the GPLLJ index and the GCLLJ index (Figure 12) are not the same pattern with opposite phase: the composite based on the GPLLJ (Figure 12b) shows that during strong GPLLJ events, there is no meridional moisture flux along the Gulf of 10 of 18

11 Figure 8. (a) Composite of the vertically integrated daily moisture flux (vector) for the wet events based on the P(Northern Plains) index. The meridional moisture flux is contoured. The contour interval is 30 kg (ms) 1. Values greater than 60 kg (ms) 1 are shaded. The unit of vector is 150 kg (ms) 1, (b) same as Figure 8a, but for dry events, (c) same as Figure 8a, but based on P(Southern Plains) index and (d) same as Figure 8c, but for dry events. Composites were based on the 50-km RSM outputs. California. The anomalous anticyclone from the GCLLJ composite and cyclone from the GPLLJ composite are located over the western United States, but they are not at the same location. The anomalous anticyclone (32 N, 85 W) over the Southeast on the GPLLJ composite is absent from the GCLLJ composite. While a stronger (weaker) GCLLJ is associated with a weaker (stronger) GPLLJ and less (more) rainfall over the Great Plains, a stronger (weaker) GPLLJ only influences the meridional transport in midlatitudes over the Southwest. There is little change in moisture flux along the Gulf of California. This suggests that the large scale flow pattern associated with the intensification of the GPLLJ has influence on {qv} from the northern Gulf to the Southwest, but has little influence on {qv} along the Gulf of California. The {qv} along the Gulf of California may be related to the easterly wave disturbances in the eastern Pacific and moisture surge events, but not related to the GPLLJ Upper Level Tropospheric Jet Stream [47] In this section, we discuss the relation between the upper level jet streams and the LLJs. To cover a larger area than that defined by the RSM domain, the circulation anomalies are taken from the NCEP-NCAR reanalysis (but still using the same low level jet index defined in section 2.4 from the RSM simulations). The composites of the 200 hpa zonal wind, which represents the tropospheric jet streams, show very different pictures for the strong/weak GPLLJ and the weak/strong GCLLJ cases (Figure 13). [48] For the strong GPLLJ case (Figure 13a), the flow is confluent between the west coast and the Rockies; strongest intensities (jet streak) extend from the Great Lakes to the east coast of Canada, where the flow is mostly diffluent; consequently, the GPLLJ is located under the diffluent region of the tropospheric jet streak. There is an anticyclone located near the entrance region of the Gulf of California and the subtropical jet is confined west of 140 W. The composite for the weak GCLLJ case (Figure 13d) also shows confluent flow upstream and diffluent flow downstream in midlatitudes and a subtropical anticyclone over southern Mexico. However, the center of the jet in midlatitudes is shifted roughly 15 degrees to the east and not as 11 of 18

12 Figure 9. (a) Composite of the vertically integrated moisture flux (vector) for the wet events based on the P(Arizona) index. The meridional moisture flux is contoured. The contour interval is 30 kg (ms) 1. Values greater than 60 kg (ms) 1 are shaded. The unit of vector is 150 kg (ms) 1, (b) same as Figure 9a, but for dry events, (c) (d) same as Figures 9a 9b, but based on P(New Mexico) index. Composites were based on the 50-km RSM outputs. strong. The center of the subtropical anticyclone is also shifted southeastward. [49] The composites of 200 hpa zonal wind for strong GCLLJ and weak GPLLJ cases both show that a stronger subtropical jet extends northeastward from the eastern Pacific to the west coast of California (Figures 13b and 13c) and merges with the subpolar jet over the Rockies. Both show a jet maximum over the Atlantic Ocean. In addition, the strong GCLLJ composite also shows a jet maximum extending from 110 W to the Great Lakes. Both composites show a ridge-trough pair over the United States. However, the center of the ridge position for the strong GCLLJ composite (Figure 13c) is located degrees to the east in comparison to the weak GPLLJ case and the anticyclone in the subtropics is also shifted northeastward. These differences can be better demonstrated by the composites of zonal wind anomalies based on the GPLLJ and the GCLLJ indices (Figures 14a and 14b). Both show a four cell pattern with two north-south dipoles opposite in sign. However, the centers based on the GCLLJ index are located southwestward in comparison to the GPLLJ composite. For strong GCLLJ cases, positive 200 hpa zonal wind anomalies located near the Baja California indicate the northeastward shift of the subtropical jet stream and positive anomalies centered at 50 N, 100 W are consistent with the northwestward shift of the jet stream. When the GPLLJ is strong (Figure 14a), there are no negative anomalies over the western coast nor the extension of the subtropical westerlies to California. Therefore these conditions are not favorable for strong {qv} over the Southwest. [50] The positive anomalies upstream of the Rockies on the GPLLJ composite map (Figure 14a) indicate stronger upper level wind consistent with recent findings of Byerle and Paegle [2003]. They suggested that the GPLLJ is a component of the large scale circulation generated by the interaction between the ambient flow and the Rockies. In this context, they find that strong (weak) upper level winds upstream of the Rockies are responsible for strong (weak) GPLLJs and more (less) rainfall over the Great Plains. [51] To investigate further, we computed a zonal wind index (U200 index) defined as the average zonal wind over the box (25 40 N, W) marked in Figure 14d, 12 of 18

13 Figure 10. Composite of the vertically integrated meridional moisture flux {qv} at 6Z for (a) the wet events based on the P(Northern Plains) index, (b) wet events based on P(Southern Plains) index and (c) wet events based on P(AZ) index. The contour interval is 30 kg (ms) 1. Values greater than 60 kg (ms) 1 are shaded, (d) (f ) Same as Figures 10a 10c, but for dry events. Composites were based on the 50-km RSM outputs for the period July September of 18

14 Figure 11. Composite observed P difference between positive and negative events based on the GCLLJ index. Contour interval is 0.5 mm day 1 Zero contours are omitted. Areas where positive (negative) values are statistically significant at the 5% level are shaded dark (light), (b) same as Figure 11a, but based on the GPLLJ index. Contours 0.2 mm day 1 are added, (c) same as Figure 11a, but the composite difference of daily meridional moisture flux {qv}. Contour interval is 40 kg (ms) 1 for positive anomalies and 10 kg (ms) 1 for negative anomalies, (d) same as Figure 11c, but based on the GPLLJ. The {qv}s are from the 50-km RSM outputs. where wind anomalies are positive (Figure 14a). The mean U200 index for the strong (weak) GPLLJ case is 14.5 (10.9) m s 1. Therefore strong GPLLJ is associated with a strong U200 index as expected. Composites of observed precipitation and the vertically integrated moisture flux are formed from cases when the U200 index exceeds a threshold of 0.8 times the standard deviation. The patterns of precipitation anomalies for positive and negative U200 index cases are similar, but opposite in sign so the difference is given in Figure 14c. Consistent with Byerle and Paegle [2003], the strong U200 index is associated with positive rainfall anomalies over the Great Plains and negative anomalies over the Southwest and the western slopes of the Sierra Madre Occidental (SMO). The pattern is similar to the rainfall composite based on the GPLLJ index (Figure 11b). [52] There is an anomalous cyclone located over the western United States on the composite difference of moisture fluxes based on the U200 index (Figure 14e), but the center (40 N, 110 W) is located about 5 degrees to the west in comparison with the composite based on the GPLLJ index (Figure 12b). The cyclone is associated with intensification of the GPLLJ and weaker {qv} anomalies over the Southwest. Results here are consistent with Byerle and Paegle [2003], in that the orographic effect plays a role in modulating the GPLLJ and rainfall from the Great Plains. Additionally, the anomalous cyclone is also linked to weaker {qv} over the Southwest and less monsoon rainfall there. 6. Summary and Conclusions [53] The summer precipitation regimes over the United States and their relation to the Great Plains and Gulf of California low level jets are examined using a rain gaugebased precipitation data set, and 10-yr summer (JJAS) simulations based on NCEP s RSM. These data sets are complemented by the NCEP-NCAR global reanalysis when larger scale features are studied. First, the quality of the RSM simulations is assessed with the regional analyses produced by EDAS. While the general features depicted by the two models are similar, the details differ. The RSM, is able to capture the evolution of the North American monsoon rainfall and vertically integrated moisture transport {qv} 14 of 18

15 point using the data from the North American Monsoon Experiment (NAME) and when the regional reanalysis from the 32-km Eta model become available. [54] The dominant summer precipitation regime is a three cell pattern that includes the inverse phase relationship over the Great Plains and the core monsoon/southwestern United States; the third (weak) center is found over the southeastern United States. The two LLJs exhibit an inverse relationship that is consistent with the inverse relationship in precipitation. Cases with strong GCLLJ index imply a weaker GPLLJ and less rainfall over the central United States; however, the strong GPLLJ index only implies weaker northward moisture flux transport over Arizona in midlatitudes, but not {qv} along the Gulf of California. [55] Composites also show that Arizona and New Mexico seem to belong to two different precipitation regimes and supported by different moisture transports [see also Berbery and Fox-Rabinovitz, 2003]. While the moisture transport from the Gulf of California is important for rainfall over Figure 12. (a) Composite difference of moisture flux (vector) based on the GCLLJ index. The unit vector is 250 kg (ms) 1. The letter A marks the location of the anticyclone. (b) same as Figure 12a, but for the difference based on the GPLLJ index. The letter C marks the location of the cyclone. The {qv}s are from the 50-km RSM outputs. associated with the two LLJs, but it has difficulty representing the vertical structure of the GCLLJ; instead, it produces a region of stronger low-level winds. The structure of the GCLLJ is better described in experiments at higher resolution (30 km grid spacing), which is consistent with several other studies that highlight the need for higher resolution to capture the mesoscale features of the North American monsoon. Differences in the vertical structure also influence the vertically integrated meridional moisture flux, but this influence is largely in the mean values. Overall, the mean {qv} depicted by the 30-km RSM simulation is stronger than that of the 50-km RSM. The high correlation (0.9) between the {qv} s from the two RSM simulations suggests that the GCLLJ index computed based on the 50-km RSM {qv} anomalies is an adequate index for selecting strong and weak GCLLJ events. On other hand, there are large uncertainties in depicting the {qv} fluxes over the Gulf of California, as there are little observational data to verify model simulations. We will revisit this Figure 13. Composite of the zonal wind at 200 hpa (shaded) and streamline for (a) the strong GPLLJ events, (b) weak GPLLJ events, (c) strong GCLLJ events, and (d) weak GCLLJ events. Areas that the zonal wind greater than 14 and 24 m s 1 are shaded light (dark). Composites were based on the global CDAS reanalysis. 15 of 18

16 Figure 14. (a) Composite of zonal wind anomalies at 200 hpa for the strong GPLLJ case. Contour interval 1 m s 1. Areas where positive (negative) values are statistically significant at the 5% level are shaded dark (light), (b) same as Figure 14a, but for the strong GCLLJ case, (c) composite observed P difference between positive and negative events based on the U200 index. Contour interval is 0.5 mm day 1 Zero contours are omitted and (d) the location of the box where the u200 index is based, and (e) same as Figure 14c, but for the composite difference of vertically integrated moisture flux (vector). The unit vector is 100 kg (ms) of 18

17 Arizona, moisture transport from the Gulf of Mexico is important for rainfall over New Mexico. The inverse relationship between the two LLJs is supported by the upper level tropospheric circulation. However, the moisture flux composites based on the GPLLJ index and the GCLLJ index are not the same pattern with opposite phase. When the GPLLJ is strong, the upper level zonal wind anomalies upstream of the Rockies are also strong, and there is a strong jet stream over the Great Lakes. When the GCLLJ is strong, there is an eastward extension of subtropical westerlies to the Baja California and the northwestern shift of the upper level jet stream. Consistent with Byerle and Paegle [2003], the interaction between the ambient flow and the Rockies does play a role in strengthening of the GPLLJ. The cyclonic circulation over the western United States associated with the strong upper level zonal wind also implies weaker {qv} from the northern Gulf of California and northern Mexico to the Southwest. Therefore there is more rainfall over the Great Plains and less rainfall over the Southwest and northern Mexico. In addition to the orography, the thermal effect generated by soil moisture may also contribute to the strength and the location of two LLJs [Kanamitsu and Mo, 2003; Mo and Juang, 2003]. Notice that for the strong GPLLJ cases, the strong zonal flow over the south central Rockies is accompanied by the negative zonal flow anomalies to the north (Figure 14a). This type of cyclonic flow can reflect the presence of mid-tropospheric cyclones that moved over the Rockies from the Pacific and subsequently induced the cyclonic pattern in favor of the development of the GPLLJ. More work is needed to clarify the relative role of orographic cyclogenesis, thermal forcing and other conditions on the development of the GPLLJ. [56] The reason for the GCLLJ vertical structure not being reproduced in the RSM 50-km simulation is not clear, although it is likely that a combination of factors involving the resolution and vertical coordinate are in play. For example, NCEP s operational Eta model reproduced the GCLLJ even at the time when its resolution was 48 km [Berbery, 2001]. Whether the vertical coordinate (sigma or eta) could influence or not the representation of a jet that is small and located near the steep slopes of the Sierra Madre Occidental, is still to be investigated. Moreover, the NAME Model Assessment Project (NAMAP) [Gutzler, 2000] has shown that different models have variations on the location, structure and intensity of the simulated GCLLJ. [57] Both precipitation over the Southwest and Great Plains have strong diurnal cycles. While it is clear that the diurnal maximum of the GPLLJ occurs around 0600 UTC to support the nocturnal rainfall, the diurnal maximum of the GCLLJ is not so well defined, and we expect that the enhanced observations during the 2004 NAME Campaign may clarify this issue. [58] Given the absence of observations in this area, the analysis of the diverging model results will need to be addressed during and after the NAME field campaign. The RSM, as other regional models, has limitations and these results need further assessment. We expect that the upcoming regional reanalyses being developed at NCEP will help clarify those points. [59] Acknowledgments. The comments of three anonymous reviewers are appreciated. They helped clarify several aspects of the manuscript. K. Mo was supported by the grant from the OGP/GAPP program GC02-102, and E. H. Berbery was supported by NOAA Grant NA76GP0291 (GCIP). References Anderson, B. T., J. O. Roads, S. C. Chen, and H. M. H. Juang (2000a), Regional simulation of the low level monsoon winds over the Gulf of California and southwestern United States, J. Geophys. Res., 105(D14), 17,955 17,969. Anderson, B. T., J. O. Roads, and S. C. Chen (2000b), Large scale forcing of summer time monsoon surges over the Gulf of California and southwestern United States, J. Geophys. Res., 105(D19), 24,455 24,467. Anderson, B. T., J. O. Roads, S. C. Chen, and H. M. H. Juang (2001), Model dynamics of summer time low level jets over northwestern Mexico, J. Geophys. Res., 106(D4), Barlow, M., S. Nigam, and E. H. Berbery (1998), Evolution of the North American monsoon system, J. Clim., 11, Berbery, E. H. (2001), Mesoscale moisture analysis of the North American monsoon, J. Clim., 14, Berbery, E. H., and M. S. Fox-Rabinovitz (2003), Multiscale diagnosis of the North American monsoon system using a variable resolution GCM, J. Clim., 16, Berbery, E. H., Y. Luo, K. E. Mitchell, and A. K. Betts (2003), Eta model estimated land surface processes and the hydrologic cycle of the Mississippi basin, J. Geophys. Res., 108(D22), 8852, doi: / 2002JD Bonner, W. D. (1968), Climatology of the low level jet, Mon. Weather. Rev., 96, Bonner, W. D., and J. Paegle (1970), Diurnal variations in the boundary layer winds over the south central United States in summer, Mon. Weather Rev., 98, Brenner, I. S. (1974), A surge of maritime tropical air-gulf of California to the southwestern United States, Mon. Weather Rev., 102, Byerle, L., and J. Paegle (2003), Modulation of the Great Plains low level jet and moisture transports by orography and large scale circulation, 108(D16), 8611, doi: /2002jd Douglas, M. W. (1995), The summer time low level jet over the Gulf of California, Mon. Weather Rev., 123, Douglas, M. W., R. A. Maddox, K. Howard, and S. Reyes (1993), The Mexican monsoon, J. Clim., 6, Douglas, M. W., A. Valdez-Manzanilla, and R. C. Cueto (1998), Diurnal variation and horizontal extent of the low level jet over the Northern Gulf of California, Mon. Weather Rev., 126, Gutzler, D. S. (2000), Covariability of spring snow pack and summer rainfall across the Southwest United States, J. Clim., 13, Hales, J. E. (1972), Surges of maritime tropical air northward over the Gulf of California, Mon. Weather Rev., 100, Hales, J. E. (1974), The southwestern United States summer monsoon source:gulfofmexicoorpacificocean?,j. Appl. Meteorol., 13, Helfand, H. M., and S. D. Schubert (1995), Climatology of the Great Plains low level jet and its contribution to the continental moisture budget of the United States, J. Clim., 8, Higgins, R. W., Y. Yao, E. S. 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. Clim., 10, Higgins, R. W., K. C. Mo, and Y. Yao (1998), Interannual variability of the U. S. summer precipitation regime with emphasis on the Southwest monsoon, J. Clim., 11, Higgins, R. W., W. Shi, E. Yarosh, and R. Joyce (2000), Improved United States precipitation quality control system and analysis, ATLAS 7, 47 pp., Clim. Prediction Cent. Natl. Cent. for Environ. Prediction, Camp Springs, Md. Juang, H. M. H., and M. Kanamitsu (1994), The NMC nested regional spectral model, Mon. Weather Rev., 122, Juang, H. M. H., S. Y. Hong, and M. Kanamitsu (1997), The NCEP regional spectral model: An update, Bull.Am.Meteorol.Soc., 78, Kalnay, E., et al. (1996), The NMC/NCAR CDAS/Reanalysis Project, Bull. Am. Meteorol. Soc., 77, Kanamitsu, M., and K. C. Mo (2003), Dynamical effect of the land surface processes on summer precipitation over the southwestern Untied States, J. Clim., 16, Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. Hnilo, M. Fiorino, and G. L. Patter (2002), NCEP/DOE AMIP-II reanalysis, Bull. Am. Meteorol. Soc., 82, Manikin, G. F., K. E. Mitchell, D. J. Stensrud, J. S. Kain, J. Garrity, and M. E. Baldwin (1998), Convective scheme tests on the coastal precipita- 17 of 18