1. Introduction and background
The northeastern quadrant of Colorado offers a somewhat unique meteorological setting in which cold-season and warm-season processes may operate simultaneously during late spring and early summer. The synchronous timing between the melting of mountain snowpack and the onset of prolific thunderstorm activity on the downstream lee plains allows for a potential relationship between the two processes.
In the weeks prior to the summer solstice, vigorous sensible heating of the elevated terrain in the Rockies causes the Western Plateau to transition from a cold source in winter to a warm source during spring (Tang and Reiter 1984). During this time, the more efficient elevated heating along Colorado’s Continental Divide begins to exert an influence upon mesoscale dynamics. The specific timing of such a transition is, in part, dictated by the amount of snow stored in the mountains, which, in itself, is prone to appreciable interannual variability (Cayan 1996).
The late-spring storage of snow in the mountains represents an effective heat sink for the atmosphere. This may manifest itself in a few ways. Snow cover, especially if fresh, reflects a large portion of the incoming solar radiation. In the highest of elevations along the Continental Divide, the albedo of late spring snowpack is approximately 60% (Cline 1997). At elevations below the tree line (about 3500 m), the majority of the snow resides beneath the shielding canopies of predominantly pine forest. In this context, the relatively cold skin temperature of the snow provides for a stabilizing influence on the surface layer. Substantial decreases in turbulent fluxes occur in situations where the aerodynamic resistance is increased through stabilization. This condition is favored when potentially warmer air resides above a forest canopy harboring snow cover. Snow-induced stabilization of the surface layer would not occur if the temperature of the daytime free atmosphere is at or below freezing; however, in May and especially during June, afternoon temperatures just above the forest canopy are most often several degrees above freezing. It is therefore a safe conclusion that, in the absence of strong cold advection, late-spring days with ample solar insolation tend to stabilize the first 50 m of mountain atmosphere.
The processes involved in the melting of the snowpack require energy. A significant portion of solar radiation is expended in supplying the energy needed first to raise the temperature of the snowpack to 0°C, and then in providing for the latent energies required in melting and evaporative processes. A portion of the energy delegated to the melting process would otherwise be used in warming the surface, which would tend to increase the sensible heating. Even after the snow has melted, near-saturated soils and wetlands may persist on the time scale of weeks providing for further taxing on sensible heating. Under such conditions, a larger portion of the solar energy is required for evaporating residual melted water and driving photosynthesis.
The effects of snowpack memory processes upon the partitioning of surface energy have been cited on both regional and continental scales to be of sufficient magnitude to alter Indian monsoon rainfall (Bamzai and Shukla 1999; Vernekar et al. 1995; Dey et al. 1985), Chinese rainfall (Yang and Xu 1994), New Mexico monsoon–related precipitation (Gutzler 2000; Gutzler and Preston 1997), and severe thunderstorm activity in the southwestern United States (Lo and Clark 2002). In all of these studies, a decrease in sensible heating, and/or reduced Bowen ratio, during the first half of the warm season was cited to be the dominant mechanism in reducing the vigor of the aforementioned monsoons and redistribution of Chinese rainfall.
One consequence of heating Colorado’s elevated terrain is a diurnal oscillation of the low-level wind field. The daytime rotation of wind vectors toward higher terrain is felt not only in the immediate vicinity of the Front Range, but also a few hundred kilometers eastward onto the Great Plains (Reiter and Tang 1984). The thermally forced, mesoscale transition from nocturnal drainage flow to upslope flow occurs rather abruptly during midmorning. By early afternoon, the near-surface, easterly upslope flow and the prevailing west to southwesterly ridge-top winds typically collide. The resulting leeside convergence zone and associated vertical velocity usually aligns itself a few kilometers east of the Continental Divide where the first convective showers typically develop by early afternoon.
Numerical modeling results of the mountain–plains, pressure–density solenoid agree with observed tendencies for mountain showers to propagate downslope into the Front Range (Wolyn and Mckee 1994). Upon the eastward descent of the Front Range, the mountain showers may merge/interact with preexisting convective complexes that are forced independently by terrain-induced convergence zones and/or colliding low-level boundaries (Wilson et al. 1992; Abbs and Pielke 1986; Crook et al. 1990). These convective complexes are commonly collocated along the I-25 corridor and are especially prolific between the Palmer Divide and Cheyenne Ridge during midafternoon. Such storms may strengthen upon moving eastward if they encounter a sufficiently moist and unstable boundary layer or if dynamic forcing is present. However, in the absence of appreciable dynamic forcing, moisture advection, or strong horizontal moisture gradients, moist convective clusters exhibit a tendency to weaken as they propagate eastward toward the lee plains of northeastern Colorado and extreme southwestern Nebraska. The inhibiting factor often takes the form of an increasingly strong low-level inversion downstream to the east.
It is of fundamental importance to assess what effect, if any, the interannual variability of residual late-spring snow mass has on the downstream moist convective processes. The lee plains region of northeastern Colorado is located close enough to the Gulf of Mexico to allow for ample transport of low-level moisture, while at the same time is elevated about 1500 m above mean sea level. These factors allow the semiarid landscape to attain high equivalent potential temperatures in the presence of relatively low wet-bulb zero heights. Such parameters are conducive to thunderstorms containing strong updrafts and large hail. This environment may also be supportive of nonsupercell tornadoes, if a low-level boundary is present, or given a favorably sheared environment, supercell tornadoes. Severe thunderstorm activity in this region typically begins in early May attaining a peak frequency in early to mid-June (Fig. 1).
A decrease in the sensible heating of the mountainous terrain tends to slow the growth rate of the planetary boundary layer (PBL). Holding synoptic-scale influences constant, it follows that a snow-covered, mountain PBL will be shallower by midafternoon than would otherwise be observed in the case of a drier, snow-free land surface. In addition, a PBL that evolves over a snow-covered landscape tends to be cooler and moister as would otherwise be the case if the sensible heating of the surface was allowed to reach its full potential (i.e., in the absence of snow cover). In any event, the mountain PBL will have a tendency to retain the characteristics of the surface, so that in a snow-covered (snow free) environment one would expect a cooler (warmer), more moist (drier), and shallower (deeper) PBL.
Typically, upon nightfall, the mountain PBL, formed during the previous day, loses convective contact with the surface and transcends into an elevated residual layer (ERL) as a radiation inversion establishes itself beneath it (Stensrud 1993). The nocturnal ERL, if wellmixed, is characterized by a near dry adiabatic lapse rate, which retains the potential temperature of the convective mixed layer of the previous afternoon. In this context, the ERL is said to be an elevated mixed layer (EML). However, in a situation where there is extensive snow cover the newly formed ERL tends not be as well-mixed and therefore exhibits a temperature profile on the stable side of dry adiabatic.
The thermodynamic effects imparted to the mountain PBL by land surface processes in the high terrain may be communicated to the downstream environment of the lee plains by advection of the ERL. The prevailing 700-mb west-southwesterly flow, typically 5–10 m s−1, transports the ERL to a position overlying northeastern Colorado by the following afternoon (16–24-h transit time). Though the initial residual layer, upon formation, may be elevated only few hundred meters relative to the mountain surfaces, after transit it may end up as high as 3 km over the downstream plains.
A portion of the mountain PBL air may become entrained into a downsloping nocturnal jet and descend into the foothills during the early morning hours (Wolyn and Mckee 1994). The subsiding air is prone to drying and may become incorporated into the evolving PBL over the immediate lee plains during midmorning. In this way, the characteristics and westerly advection of the ERL are not entirely independent from boundary layer processes evolving beneath it.
The details concerning the interaction between the ERL and the thunderstorm environment of the lee plains is not well documented. However, the near dry adiabatic lapse rates inherent to an EML is known to contribute to the volatility of severe weather environment in the southern plains states. The interested reader is referred to Lanicci and Warner (1991a, b,c) for a more detailed discussion concerning the effects of EMLs on severe thunderstorm morphology.
An observational study was undertaken in Grimaldi (2003) to assess the plausibility as to whether cold-season memory processes have the ability to affect the thermodynamic environment over the lee plains of the Colorado Rockies. The physical premise by which such a teleconnection would operate links the sensible heating of the high terrain to the characteristics of the elevated residual layer (ERL). In realizing that the evolving PBL of the lee plains grows into the ERL that originated over the mountains the previous day, one may infer that thermodynamic alterations along their interface may exhibit variability on weekly and interannual time scales.
2. Methodology
A mountain snowpack–lee plains thunderstorm relationship is twofold in that it requires knowledge of the mountain surface conditions and the thermodynamics of the downstream atmosphere. The approach of this paper is to examine, compare, and contrast the thermodynamics of the lower to middle troposphere using Denver, Colorado’s upper-air data and the severe weather record over the lee plains that occur in tandem with anomalously large and small upstream snowpack. Cumulative warm-season, streamflow discharge from two rivers that drain into the Colorado Rockies is used as a proxy for assessing the magnitude of snowpack.
After determining the 14 most extreme years, the 7 most positive and the 7 most negative were used. The seven years with the greatest runoff were classified as “high snow years,” while the lowest seven runoff years were labeled as “low snow years.” These years and associated standard deviations above and below the mean are given in Table 1. A time series of runoff data is presented in appendix A.
For these years, upper-air data from Denver’s 1200 UTC radiosonde were analyzed for the 61-day period between 1 May and 30 June. It is during this time window when snowpack memory is most profound and most likely to be a contributing factor to the increasingly active thunderstorm environment downstream.
a. Severe weather record
The thunderstorm season analyzed extends from 30 April through 1 July. The inclusion of an additional day either side of May and June was thought to be favorable since a substantial number of events happened to occur on both of these days, thereby increasing sample size. Although 1 July is soon after the midpoint of the moist convective season (see Fig. 1), snowpack memory is likely to rapidly fade as the warm season progresses. Furthermore, 1 July approximately coincides with the onset of the Arizona monsoon signaling the evolution of an altered precipitation regime across the central and southwestern United States (Higgins and Shi 2001; Higgins et al. 1998; Mock 1996).
The severe thunderstorm dataset was taken from the archive of the Storm Prediction Center. The reports consist of observations collected by the National Weather Service of convective downdraft winds that either caused damage or were measured in excess of 50 kt, hail with diameters greater than ¾ in., and tornadoes. Documented locations, types, and times of severe weather events that occurred between 30 April and 1 July over the 27-yr period were noted for the lee plains area bounded by −105.5°W, −102.0°W, 42.0°N, and 38.5°N. This domain was further subdivided by a bisecting meridian at −104.0°W. In this way, two spatial subdomains were created and classified as the plains region and foothills region (Fig. 3).
The 27-season archive of severe weather reports was subjected to a filtering process in an effort to reduce the tendency for a discrete event to be overreported in the vicinity of a densely populated region (i.e., Denver). The process was similar to the one used in the severe weather climatology produced by Kelly et al. (1985). Multiple severe weather reports of identical category (i.e., hail, tornado, wind) were consolidated into one if they occurred within 30 min and 15 km of one another, retaining either the most severe report or, in cases with equal severity, the earlier report.
Use of the severe weather archive is not without limitation. A population bias is likely to exist spatially across the domain, with lower reporting efficiencies over sparsely populated, rural areas of the plains. The bias also affects the secular trend since large increases in population density along the foothills, advances in communication, and greater severe weather awareness occurred during the course of the observation record. Kelly et al. (1985) discusses these and other limitations of the dataset.
b. Upper-air data
Upper-air data were collected from the 1200 UTC [6 a.m. mountain daylight time (MDT)] Denver radiosonde dataset, for the time period 1 May to 30 June, over the 27 yr in question (1972–95, 1957, 1965, 1964, 1966). Less than 5% of the days were affected by missing data. In these limited cases, the data were either extrapolated in time, making use of the adjacent 1200 UTC soundings, or more often extrapolated in space, making use of the same 1200 UTC radiosonde profile, which sometimes contained data immediately above and below the missing parameter for the pressure level in question.
The thermodynamic parameters analyzed include 500-mb heights, 700–500-mb mean lapse rate, 700-mb relative humidity, 700-mb potential temperature and surface, and 700-mb equivalent potential temperatures. The relationship between these parameters and streamflow discharge was examined in three ways. First, the parameters were linearly correlated to the entire 27-yr streamflow record for each gauging station. The resulting correlation coefficients generated p-value thresholds in order to assess significance. Second, the 7-yr subsets were compared and statistically contrasted. The subsets were subjected to the parametric Student’s t test and the nonparametric Mann–Whitney rank sum test. Appendix B further clarifies the statistical tests. Finally, selected thermodynamic parameters corresponding to high and low snow years were subjected to relative frequency analysis to assess what range/threshold of values tended to occur most/least often over the time period in question.
3. Results
a. Thermodynamic parameters
Table 2 displays linear correlation coefficients and associated levels of significance between cumulative streamflow discharge and selected mean thermodynamic and severe weather parameters for the entire 27-season set. The analysis reveals that the cumulative streamflow discharge is inversely correlated to 500-mb heights, 700–500-mb layer thickness, 700-mb temperature, and 700–500-mb lapse rates. The combination of mean reductions in thickness and more stable lapse rates for the 700–500-mb layer, inherent to seasons having large streamflow discharge, suggests that the 700-mb level is subjected to more cooling than the 500-mb level under such conditions. There is also a direct correlation between cumulative discharge and 700-mb relative humidity, so that during periods of high discharge the 700-mb level tends to be moistened (and/or cooled) relative to periods of low discharge. Despite the suggested influence of land surface processes, it is important to realize that synoptic-scale weather features also influence the aforementioned parameters.
Equivalent potential temperatures θe were calculated for the 700-mb level and at the surface. KDEN has a mean station pressure of about 840 mb. The difference between the surface and 700-mb θe value is one proxy lower-tropospheric stability. The linear correlation coefficients of θe differences to cumulative discharge of the Big Thompson index (closer to the domain) is −0.44, which is significant at the 0.05 level, while the correlation to Halfmoon creek is −0.26, which is not significant at the 0.10 level.
b. Interclass comparisons
Mean May and June 700-mb temperatures at KDEN (∼1400 m above ground level) have a tendency to be warmer during low snow years. Average 700-mb temperatures are 1.3°C warmer in the low snow years compared to the high snow years (Table 3). The contrasted means among the 7-yr subsets are significant at the 0.01 level. Furthermore, there is a shift in the relative frequency spectrum toward warmer potential temperatures for the low snow case (Fig. 4). The linear correlation coefficients are −0.59 for Big Thompson River and −0.47 for Halfmoon Creek. These coefficients are significant at the 0.01 and 0.05 levels, respectively, thus offering further support for an inverse relationship between 700-mb temperature and antecedent snow condition.
Lower 500-mb heights (<565 dm) are more frequently observed during high snow years. The low snow years dominate the higher height range between 575 and 585 dm (Fig. 5). Frequencies are remarkably similar for the highest of heights (>585 dm), perhaps negating a land surface influence on heights in late June, when a warmer troposphere is more common. However, linear correlations to cumulative discharge are in agreement with the interclass comparison as a whole, exhibiting inverse correlation coefficients of −0.60 and −0.48, both which are significant at the 0.01 level (Table 2).
The 700–500-mb lapse rates tend to be more unstable during the low snow years compared to high snow years. Low snow years have lapse rates that are on average 0.2°C km−1 steeper than mean lapse rates of the high snow years’ mean (Table 3). This is supported by correlation coefficients of −0.49 and −0.47, significant at the 0.01 and 0.05 levels for Big Thompson and Halfmoon Creek, respectively (Table 2). The relative frequency spectrum offers further support, as low snow years are shown to have more frequent values exceeding 7.5°C km−1 than high snow years, while the high snow years feature more days in the 6.5°–7.5°C range (Fig. 6).
Drier conditions at 700 mb tend to be more common to low snow years. Interclass comparison indicates that the mean relative humidity is 3.7% lower for the low snow years compared to the high snow years. The correlation coefficients associated with 700-mb relative humidity are 0.51 and 0.50 for Big Thompson and Halfmoon Creek, respectively, both of which are significant at the 0.01 level (Table 2). The relative frequency analysis renders further support for a drier 700-mb level accompanying low snow years. Days exhibiting 700-mb relative humidity between 20% and 40% occur with greater frequency for low snow years compared to high snow years, which tended to yield more occurrences in the 50%–70% range (Fig. 7).
The difference between equivalent potential temperatures at 700 mb and the surface suggest increased stability under low snow conditions. The mean amplitude of this difference for low snow years is 0.8°C larger that for the high snow class. The statistical independence between snow classes are significant at the 0.05 level (Table 3); however, when all 27 yr are considered, the correlation to streamflow discharge is not overly impressive, exhibiting significance at the 0.05 level for Big Thompson River and significance below the 0.1 level for Halfmoon Creek (Table 2). The relative frequency analysis shows a lower frequency of moderate cap strength (2°–6°C) for the low snow class versus the high snow class. Strongly capped situations (8°–14°C) are slightly more common to the low snow class (Fig. 8).
c. Severe weather reports
Hail and tornadoes constitute the overwhelming majority of severe weather observed in this region during May and June (see Fig. 1). In general, more severe weather occurred in the foothills domain compared to the plains domain. This tendency may be partially attributed to the greater population density in the foothills domain. High snow years tended to be associated with increased severe weather occurrence. An exception to this occurred in the plains domain, during low snow years, when there were a comparable number of tornado days and a greater number of total tornadic reports compared to high snow years (Table 4).
Severe weather reports occurred on average over an hour later on the plains than in the foothills domain, in agreement with the radar climatology of Karr and Wooten (1976). This time lag was most notable in low snow years. Low snow years are associated with a delay in the onset of severe weather for both domains. The delay was least noted for tornadic reports in the foothills domain. For hail events, however, the first report of the day occurred, on average, 51 min later in the foothills domain and 45 min later in the plains domain. A one-tailed Student’s t test generates the corresponding p values of 0.048 and 0.128, respectively, for the foothills and plains domains. However, when the times of all severe hail reports are considered for each domain, the associated p values are exceedingly small (order 10−5).
The correlation between the timing of severe weather occurrences and streamflow discharge considering all 27 yr were marginally significant at best. The strongest correlation was found for the foothills region, significant at the 0.05 and 0.10 levels for Big Thompson and Halfmoon Creek, respectively (see Table 2). For the plains region correlation coefficient failed to generate p values, which were significant at the 0.1 level for either gauging station. The lack of linear correlation is indicative of a scenario in which the suggested diurnal delay in the onset of severe weather is most prevalent during times of extreme snow anomalies. This is to say that during years when more typical snow conditions are observed, the aforementioned relationship breaks down as other factors may influence the diurnal timing of severe weather onset.
4. Discussion
Operating under the assumption that water from the snow-melting process is the dominant signal in the warm-season cumulative streamflow discharge, we are able to make the following assertion. Statistical evidence strongly suggests that an anomalously large late-spring snowpack leads to cooler 700-mb May and June temperatures at mesoscale distances downstream compared to anomalously deficient snowpacks. These observations agree with the a priori expectation that the memory process of a large snowpack exerts a net cooling effect on the mountain boundary layer by taxing the surface turbulent energy fluxes. The proposed physical mechanism targets snow-induced increases in surface layer stability (and subsequent aerodynamic resistance) as well increases in albedo above the tree line and in exposed elevated areas.
A collateral effect of warmer air near the 700-mb level is to increase the strength of the convective inhibition. Downstream of the high terrain, the combination of warmer 700-mb temperatures and larger 700 mb to surface θe differences observed during low snow years imply that, on average, surface-based (or mixed layer) parcels will need to attain higher parcel energies or be subjected to a stronger lifting mechanism in order to reach their level of free convection. The overall effect of increasing convective inhibition should be to limit the number of moist convective clusters, thereby reducing the areal coverage of precipitation.
One may surmise that warming the 700-mb layer more strongly than the 500-mb layer should tend to steepen the 700–500-mb lapse rate. Observations are in agreement with just such a scenario. Steeper midtropospheric lapse rates subject parcels, which are able to breech their level of free convection, to greater vertical accelerations. Rapid accelerations above cloud base provide for efficient tilting of streamwise vorticity into the vertical and provide for strong vortex stretching, both of which are favorable for supercell (Davies-Jones 1984) and nonsupercell tornadoes (Wakimoto and Wilson 1989). It remains plausible that the promotion of updraft instability by steep lapse rates inherent to EMLs is contributing to the increase in the number of tornadic occurrences in the plains domain observed low snow years.
An additional pathway that may increase the likelihood of supercell-related tornadogeneisis during low snow years links the thermodynamic characteristics of an EML to the degree of downdraft instability (Johns and Doswell 1992). The presence of near-neutral lapse rates (which limit subsidence warming) and low relative humidities (which increase potential evaporation), characteristic to EMLs, increase the downward velocity and lower the temperature of a supercell’s rear-flank downdraft. A collateral effect of an intense, cool convective downdraft is to increase the baroclinic generation of vorticity along its leading edge, which is known to create and amplify low-level rotational centers (Rotunno and Klemp 1985). The maintenance and intensity of low-level rotational centers as well as their colocation to the parent midlevel mesocylone are thought to be discriminating parameters between tornadic and nontornadic supercells (Brooks et al. 1994). In this way, midlevel intrusions of dry and neutrally stratified EMLs into the moist convective environment of the lee plains, thought to be a more common occurrence during low snow years, are likely to cater to environments that support and maintain intense, low-level rotational centers, thereby increasing the likelihood of supercell tornadogenesis.
Another effect of warming the 700-mb level, given identical surface conditions, is to reduce the depth of the afternoon mixed layer. An implied increase in potential temperature jump across the stronger capping inversion has the effect of slowing the growth rate of the boundary layer, ultimately reducing the depth of the mixed layer by afternoon. One effect of confining the convective mixed layer to a smaller volume is to concentrate the moist static energy. For this reason, a thin afternoon mixed layer (PBL) was cited and observed by Mahrt (1977) to be contributing factor to hail-producing thunderstorms over the semiarid landscape of northeastern Colorado.
Most notably, in both the plains and foothills domain, there is an observed tendency for severe weather to occur later in the day. The delay in convective initiation is most likely related to the presence of a stronger convective inhibition. If indeed convective inhibition is stronger over the plains domain during low snow years, a collateral effect should be to increase storm spacing and reduce the areal coverage of convective rainfall.
Composite rainfall analysis spanning the months of May and June for the high snow and the low snow years support this speculative conclusion (Fig. 9). As expected, during low snow years, for this 2-month period, below-average rainfall is observed on the lee plains of the Colorado Rockies. Even more impressive are the profound rainfall deficits extending several hundred kilometers northeastward (downstream) into the Great Plains during low snow years. The high snow years show the opposite effect. Could a persistent series of EMLs be responsible for negative precipitation anomalies at synoptic-scale distances downstream during low snow years? Considering the near-neutral stratification of an EML allows for its thermodynamic identity to be maintained during adiabatic transit to great distances downstream, it is plausible that EMLs of Rocky Mountain origin may indeed be limiting the areal coverage of moist convective clusters over the central and northern Great Plains. Furthermore, the central and western high plains are prone to large 15-day autocorrelation to summertime rainfall (Koster et al. 2003). It follows that the effects of diminished rainfall and subsequent drying of the soil tends to lend itself toward boundary layer feedback, thereby having the potential to inhibit future convective precipitation in the region.
Snow anomalies appear to be linked to the character of severe weather events in the plains domain. Most impressive is the reduction in the number of hail days observed across both domains during low snow years. As previously stated the warmer 700-mb temperatures favored during low snow years are likely strengthening the convective inhibition, thereby reducing the likelihood that parcels can be lifted to their level of free convection in the absence of appreciable dynamic forcing.
Though observational evidence provides strong support to the hypothesis that surface moisture conditions in the elevated terrain have the potential to alter stratification over the downstream lee plains, the analysis is not without its limitations. There was no attempt to filter the dataset to eliminate those days where the wind direction was unfavorable and/or cap strength irrelevant to the meteorological situation, such as would occur after the passage of a strong cold front. Use of the severe weather log is subject to temporal bias in addition to spatial bias. The plains domain is less densely populated and would likely have less efficient severe weather reporting as the I-25 corridor of the foothills region. Furthermore, advances in communication, the advent of spotter networks, and greater severe weather awareness (storm spotter networks) should have the effect of increasing reporting efficiency of severe weather in both domains since the late 1980s.
One must also consider the possibility that a cut-and-dry snow–thunderstorm relationship may be contaminated by longer-time-scale synoptic patterns. Dry winters, in some cases, may be the result of persistent longwave ridging over the Rockies. This pattern may persist into and throughout the warm season and could lead to diminished spring–summer convective rainfall. The possibility of a winter global circulation anomaly persisting into the summer is cited by Gutzler (2000) for providing upper-tropospheric convergence (ridging), thereby limiting the influence of smaller-scale feedbacks that Rocky Mountain snowpack has on summertime Arizona precipitation. However, severe weather days in northeastern Colorado are frequently synoptically benign, further emphasizing that mesoscale processes contribute strongly to modulating thunderstorm activity in this region.
5. Conclusions and future work
The memory processes of anomalously large Colorado Rocky Mountain snowpack is statistically tied to cooler, May and June 700-mb temperatures over the foothills and plains. The trend toward cooler 700-mb temperatures is likely a result of reduced sensible heating of the elevated terrain where migrating ERLs originate. The less efficient sensible heating of mountain surfaces during high snow years is likely related to higher albedo above the tree line and increased surface layer stability within elevated forest canopies. It follows that a prolific snow-melting season results in the formation of ERLs, which are cooler and more stable than would otherwise be observed for drier mountain surfaces. Statistical evidence suggests that an anomalously large late-spring snowpack affects the thermodynamic environment of the downstream atmosphere by lowering the convective inhibition, which apparently results in a greater number of severe hail days and increased rainfall along the lee plains.
It is furthermore suggested that mountain-generated elevated mixed layers, as would tend to more often form following low snow years, have the potential to inhibit warm-season convective rainfall not only in the immediate lee of the Rockies but farther downstream over the central/northern Great Plains and the upper Mississippi Valley. While a stronger capping inversion can reduce the probability of parcels reaching their level of free convection during June, boundary layer feedback, which is especially strong in the western high plains during July, can amplify a dry pattern, potentially leading to a drought scenario. It is during these months when agriculture in this region is most sensitive to reductions in rainfall.
Upon extrapolating a 1°–4°C global warming scenario, the amount of spring snow stored in the Rockies is expected to decrease due to the tendency for more precipitation to fall as rain instead of snow and the earlier arrival of warm-season processes (Service 2004). It is therefore of great meteorological interest to determine to what degree snowpack memory is tied to not only to thunderstorm activity in the lee of the Rockies but also to convective rainfall patterns at synoptic scales downstream. Future work should attempt to lock down the specific mechanism that would allow such a teleconnection to exist. Research should also be directed toward larger scales, considering years with exceptional north-to-south snowpack gradients along the Continental Divide of the United States and Canada in order to evaluate the potential for a nonlocal, continental-scale, atmospheric response to moisture conditions along the entire length of the American and Canadian Rockies. If such relationships can indeed be verified, a lagged seasonal indicator of warm-season processes could be established and would prove to be of great value to agricultural, forecasting, climatic interests for the wide expanse of the downstream plains.
Acknowledgments
The majority of this research was conducted at The South Dakota School of Mines and Technology under the supervision of Andy Detwiler. The input from Steve Hodanish of the Pueblo NWS was appreciated. I would like to thank my research assistant at SUNY Oneonta, Heather Frees, for her coordinating and logistical efforts in Colorado as well as the University of Colorado personnel who granted us access to their mountain research station at Niwot Ridge.
REFERENCES
Abbs, D. J., and Pielke R. A. , 1986: Thermally forced surface flow and convergence patterns over northeast Colorado. Mon. Wea. Rev., 114 , 2281–2296.
Bamzai, A. S., and Shukla J. , 1999: Relation between Eurasian snow cover, snow depth, and the Indian summer monsoon: An observational study. J. Climate, 12 , 3117–3132.
Brooks, H. E., Doswell C. A. III, and Cooper J. , 1994: On the environments of tornadic and nontornadic mesocyclones. Wea. Forecasting, 9 , 606–618.
Cayan, D. R., 1996: Interannual climate variability and snowpack in the western United States. J. Climate, 9 , 928–948.
Cline, D. W., 1997: Effect of seasonality of snow accumulation and melt on snow surface energy exchanges at a continental alpine site. J. Appl. Meteor., 36 , 32–51.
Crook, N. A., Clark T. L. , and Moncrieff M. W. , 1990: The Denver Cyclone. Part I: Generation in low Froude number flow. J. Atmos. Sci., 47 , 2725–2742.
Davies-Jones, R., 1984: Streamwise vorticity: The origin of updraft rotation in supercell thunderstorms. J. Atmos. Sci., 41 , 2991–3006.
Dey, B., Kathuria S. N. , and Kumar O. B. , 1985: Himalayan summer snow cover and withdrawal of the Indian summer monsoon. J. Appl. Meteor., 24 , 865–868.
Grimaldi, R. T., 2003: Two candidate mechanisms potentially influencing severe thunderstorm activity in the Central United States. Ph.D. dissertation, South Dakota School of Mines and Technology.
Gutzler, D. S., 2000: Covariability of spring snowpack and summer rainfall across the southwest United States. J. Climate, 13 , 4018–4027.
Gutzler, D. S., and Preston J. W. , 1997: Evidence for a relationship between spring snow cover in North America and summer rainfall in New Mexico. Geophys. Res. Lett., 24 , 2207–2210.
Higgins, R. W., and Shi W. , 2001: Intercomparison of the principal modes of interannual and intraseasonal variability of the North American monsoon system. J. Climate, 14 , 403–417.
Higgins, R. W., Mo K. C. , and Yao Y. , 1998: Interannual variability of the U.S. summer precipitation regime with emphasis on the southwestern monsoon. J. Climate, 11 , 2582–2606.
Johns, R. H., and Doswell C. A. III, 1992: Severe local storms forecasting. Wea. Forecasting, 7 , 588–612.
Karr, T. W., and Wooten R. L. , 1976: Summer radar echo distribution around Limon, Colorado. Mon. Wea. Rev., 104 , 728–734.
Kelly, D. L., Schaefer J. T. , and Doswell C. A. III, 1985: Climatology of nontornadic severe thunderstorm events in the United States. Mon. Wea. Rev., 113 , 1997–2014.
Koster, R. D., Suarez M. J. , Higgins R. W. , and Van den Dool H. M. , 2003: Observational evidence that soil moisture variations affect precipitation. Geophys. Res. Lett., 30 .1241, doi:10.1029/2002GL016571.
Lanicci, J. M., and Warner T. T. , 1991a: A synoptic climatology of the elevated mixed-layer inversion over the southern Great Plains in spring. Part I: Structure, dynamics, and seasonal evolution. Wea. Forecasting, 6 , 181–197.
Lanicci, J. M., and Warner T. T. , 1991b: A synoptic climatology of the elevated mixed-layer inversion over the southern Great Plains in spring. Part II: The life cycle of the lid. Wea. Forecasting, 6 , 198–213.
Lanicci, J. M., and Warner T. T. , 1991c: A synoptic climatology of the elevated mixed-layer inversion over the southern Great Plains in spring. Part III: Relationship to severe-storms climatology. Wea. Forecasting, 6 , 214–226.
Lo, F., and Clark M. P. , 2002: Relationships between spring snow mass and summer precipitation in the southwestern United States associated with the North American monsoon system. J. Climate, 15 , 1378–1385.
Mahrt, L., 1977: Influence of low-level environment on severity of high-plains moist convection. Mon. Wea. Rev., 105 , 1315–1329.
Mock, C. J., 1996: Climatic controls and spatial variations of precipitation in the western United States. J. Climate, 9 , 1111–1125.
Reiter, E. R., and Tang M. , 1984: Plateau effects on diurnal circulation patterns. Mon. Wea. Rev., 112 , 638–651.
Rotunno, R., and Klemp J. , 1985: On the rotation and propagation of simulated supercell thunderstorms. J. Atmos. Sci., 42 , 271–292.
Service, R. F., 2004: As the West goes dry. Science, 303 , 1124–1127.
Stensrud, D. J., 1993: Elevated residual layers and their influence on surface boundary-layer evolution. J. Atmos. Sci., 50 , 2284–2293.
Tang, M., and Reiter E. R. , 1984: Plateau monsoons of the Northern Hemisphere: A comparison between North America and Tibet. Mon. Wea. Rev., 112 , 617–637.
Vernekar, A. D., Zhou J. , and Shukla J. , 1995: The effect of Eurasian snow cover on the Indian monsoon. J. Climate, 8 , 248–266.
Wakimoto, R. M., and Wilson J. W. , 1989: Non-supercell tornadoes. Mon. Wea. Rev., 117 , 1113–1140.
Wilson, J. W., Foote G. B. , Crook N. A. , Fankhauser J. C. , Wade C. G. , Tuttle J. D. , Mueller C. K. , and Krueger S. K. , 1992: The role of boundary-layer convergence zones and horizontal rolls in the initiation of thunderstorms: A case study. Mon. Wea. Rev., 120 , 1785–1815.
Wolyn, P. G., and Mckee T. B. , 1994: The mountain-plains circulation east of a 2-km-high north–south barrier. Mon. Wea. Rev., 122 , 1490–1508.
Yang, S., and Xu L. , 1994: Linkage between Eurasian winter snow cover and regional Chinese summer rainfall. Int. J. Climatol., 14 , 739–750.
APPENDIX A
Streamflow Data
Streamflow data are used to indicate the amount of snow stored in the Front Range of Colorado during the previous cold season. The data cover a total of 27 melt seasons. The 27 yr include 1972 to 1995 with the addition of the anomalously high streamflow years of 1957 and 1965 and the anomalously low streamflow years of 1964 and 1966. The streamflow data were furnished by two U.S. Geological Survey (USGS) gauging stations: the Big Thompson River (northern Front Range) and Halfmoon Creek (southwest of the Front Range) (Fig. A1). Both datasets are 100% complete.
The five lowest and highest snow years that occurred between 1972 and 1995 were determined by creating a combined weighted streamflow index, calculated as follows: 60% of the normalized seasonal discharge (cubic feet) from the Big Thompson River (given in units of standard deviations above or below the sample mean of the 23 consecutive years) added to 40% of the normalized seasonal discharge (same normalized representation) from Halfmoon Creek. This is the magnitude of the normalized index presented in Table 1. The greater weighting of the Big Thompson River discharge reflects the fact that the drainage basin of the Big Thompson River is closer to the warm-season severe weather domain than Halfmoon Creek (see Fig. 2).
APPENDIX B
Statistical Tests
Student’s t test
The Student’s t statistic tests the probability that two sample populations x and y have significantly different means. Note that x and y may be of different lengths. The default assumption is that the data are drawn from populations with the same true variance. This type of test is often referred to as the t-means test.
Mann–Whitney Rank-Sum Test
The Mann–Whitney U (or Z) statistic, unlike the Student’s t statistic, does not assume the distribution of data to be normally distributed. In this case, the U (or Z) statistic tests the hypothesis that two sample populations x and y have the same medians against the hypothesis that they differ. Samples x and y may be of different lengths. This type of test is often referred to as the “Wilcoxon Rank-Sum Test” or the “Mann–Whitney U-Test.”
The test first involves calculation of a test statistic named U that is calculated as follows:
Observations are arranged into a single ranked series regardless of which sample they are in.
Add up the ranks in sample x. The sum of ranks in sample y may be shown to equal N(N + 1)/2, where N is the number of observations.
- The U statistic is then given bywhere Nx and Ny are the two sample sizes, and Wx and Wy are the sum of the ranks in samples x and y, respectively. For large samples (>20), a Z statistic may be generated that closely follows a normal distribution. Here Z is defined as follows:
Time series analysis for severe weather report frequency within the domain bounded by 42.0° to 38.5°N, −102.0° to −105.5°W based on 20 seasons’ worth of data (1976–95). Dashed line represents hail report frequency. Data are subject to a 5-day moving weighted average with weights (0.1, 0.25, 0.3, 0.25, 0.1) respective to days (−2, −1, 0, +1, +2); 30 June represents the 181st day of a non–leap year.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Shaded relief map of Colorado showing locations of both USGS gauging stations.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Map of the high plains showing the division between the foothills and the plains domains. Thick solid line denotes the Continental Divide. Interstates and densely populated areas (i.e., Denver) are shown.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Frequency analysis of 700-mb potential temperature according to the low snow subset (LS) and the high snow subset (HS). Sample size is 61 days × 7 seasons per subset.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Same as in Fig. 4, but for 500-mb heights.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Same as in Fig. 4, but for 700–500-mb lapse rate.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Same as in Fig. 4, but for 700-mb relative humidity.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Same as in Fig. 4, but for equivalent potential temperature difference.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Composite May and June mean precipitation anomalies for (top) the high snow case and (bottom) the low snow case. From NOAA/CIRES Climate Diagnostics Center.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
Fig. A1. USGS streamflow data for (left) the Big Thompson River, 7492 ft above sea level, and (right) Halfmoon Creek, 9830 ft above sea level, for (top) 1992–95 and (bottom) the entire time series.
Citation: Journal of Hydrometeorology 9, 1; 10.1175/2007JHM826.1
For the 14 yr chosen for statistical analysis, indices are based on the seven highest/lowest ranking combined weighted streamflow index, calculated by the sum of 60% of the Big Thompson standardized index and 40% of the Halfmoon Creek standardized index.
Linear correlation coefficients and associated levels of significance between seasonal (May–June) sample means of given upper-air parameters (1200 UTC KDEN) and diurnal timing of severe weather occurrence to cumulative (1 Apr to 31 Jul) streamflow. The p values are based on a two-tailed t test (N = 27). (NS = not significant at 0.1 level)
Intercomparison of parameters between the seven seasons with the largest seasonal runoff (high snow) and the seven seasons with the lowest runoff (low snow).
Spatial analysis of storm report characteristics illustrating the differences between high snow years and low snow years. Severe weather days and total number of discrete events (in parentheses) are indicated.
