Abstract

This article investigates errors in forecasts of the environment near an elevated mesoscale convective system (MCS) in Iowa on 24–25 June 2015 during the Plains Elevated Convection at Night (PECAN) field campaign. The eastern flank of this MCS produced an outflow boundary (OFB) and moved southeastward along this OFB as a squall line. The western flank of the MCS remained quasi stationary approximately 100 km north of the system’s OFB and produced localized flooding. A total of 16 radiosondes were launched near the MCS’s eastern flank and 4 were launched near the MCS’s western flank.

Convective available potential energy (CAPE) increased and convective inhibition (CIN) decreased substantially in observations during the 4 h prior to the arrival of the squall line. In contrast, the model analyses and forecasts substantially underpredicted CAPE and overpredicted CIN owing to their underrepresentation of moisture. Numerical simulations that placed the MCS at varying distances too far to the northeast were analyzed. MCS displacement error was strongly correlated with models’ underrepresentation of low-level moisture and their associated overrepresentation of the vertical distance between a parcel’s initial height and its level of free convection (, which is correlated with CIN). The overpredicted in models resulted in air parcels requiring unrealistically far northeastward travel in a region of gradual meso-α-scale lift before these parcels initiated convection. These results suggest that erroneous MCS predictions by NWP models may sometimes result from poorly analyzed low-level moisture fields.

1. Introduction

Nocturnal mesoscale convective systems (MCSs) are responsible for a large percentage of warm season rainfall in the central United States (Fritsch et al. 1986), and are capable of producing extreme rainfall events and flash flooding (e.g., Moore et al. 2003; Schumacher and Johnson 2005). Whereas daytime convection frequently derives convective energy from the planetary boundary layer (PBL) (these types of systems are referred to as “surface based”), nocturnal MCSs often thrive in conjunction with a statically and convectively stable PBL, and derive the majority of their energy from layers of air above the PBL (Corfidi et al. 2008; these types of systems are referred to as “elevated”). The layer(s) of air that is the primary source of air parcels that drive deep convective overturning—often defined as the layer(s) where convective available potential energy (CAPE) 100 J kg−1 and convective inhibition (CIN) −250 J kg−1—is commonly referred to as the effective inflow layer (EIL; Thompson et al. 2007). We refer to storms within an environment where the lower bound of the EIL (as defined by the aforementioned CAPE and CIN thresholds) is above the earth’s surface as elevated. In this article, we use an observational analysis of an elevated MCS’s near environment to demonstrate how numerical weather prediction (NWP) models’ moisture errors within an MCS’s EIL influence the ability of these models to accurately predict the MCS.

Horizontal and temporal variations in CAPE and CIN influence an MCS’s intensity, what direction it will move, the likelihood of the MCS maintaining itself (e.g., Crook and Moncrieff 1988; Coniglio et al. 2007; Jirak and Cotton 2007; Trier et al. 2014a,b; Peters and Schumacher 2016), and a storm’s ability to produce hail (e.g., Brooks et al. 2003; Edwards and Thompson 1998; Jewell and Brimelow 2009), tornadoes (e.g., Smith et al. 2012; Thompson et al. 2012), and damaging winds (e.g., Mahoney et al. 2009). The presence of nonzero CIN implies that a parcel is negatively buoyant prior to reaching its level of free convection (LFC; the level after which an ascending parcel attains positive buoyancy relative to its surrounding environment). This negative buoyancy serves as a hindrance to air parcels entering deep convective updrafts. An upward-oriented perturbation pressure gradient force is required to overcome negative buoyancy and drive these air parcels to their LFCs. Convection is therefore relatively difficult to initiate and sustain in environments with large CIN values, given that a limited range of atmospheric phenomena produce lifting strong enough to overcome high CIN. In contrast, convection is more easily initiated and sustained in environments with very low CIN, given that even modest upward displacements may initiate or sustain convection (e.g., Crook and Moncrieff 1988; Trier et al. 2011; Keene and Schumacher 2013; Lane and Moncrieff 2015).

CAPE and CIN can change quickly with space and time, and these changes impact storm behavior. For instance, relatively small moisture and temperature changes at a given location may result in comparatively large changes to CAPE and CIN at that location owing to the highly nonlinear relationships between CAPE and CIN and a parcel’s temperature and moisture content at its origin level. These CAPE and CIN responses to moisture may subsequently change the depth of an MCS’s EIL (Thompson et al. 2007; Schumacher 2015b; Schumacher and Peters 2017). EIL changes have been shown to influence MCS’s location and precipitation characteristics (Schumacher 2015b; Schumacher and Peters 2017). This is most notable for elevated MCSs, given that gradual (e.g., of order 10–50 cm s−1) meso-β-scale (of order 20–200 km) to meso-α-scale (of order 200–1000 km) lift in the lower atmosphere is often the driving mechanism in the initiation and upwind propagation of elevated MCSs (e.g., Trier et al. 2010; Peters and Schumacher 2015a, 2016). Longer residence times of air parcels (and subsequently larger horizontal distances traveled) are required in regions of gradual lifting for them to reach their LFCs when CIN is large, than when CIN is comparatively small (e.g., Trier and Parsons 1993; Peters and Schumacher 2015a, 2016). Furthermore, there are frequently situations in elevated convective environments where the vertical location of the maximum atmospheric lift does not coincide with the minimum CIN in the presence of a feature. In these scenarios, flow with nonnegligible CAPE may interact with an OFB without CI occurring (e.g., Peters and Schumacher 2016). Adequate representation of the moisture and thermodynamic fields that regulate CIN in numerical weather prediction (NWP) models is therefore essential for accurately predicting the behavior and location of elevated MCSs. It is often difficult for NWP models to accurately capture the small spatial (e.g., of order 10 km) and temporal (e.g., of order 10 min) variability in these quantities, and these difficulties in predicting CIN and CAPE likely translate to errors in the NWP model’s predictions of convective behavior (e.g., Zhang et al. 2003; Melhauser and Zhang 2012; Schumacher 2015a; Peters and Schumacher 2016).

The majority of our current understanding of elevated MCS environments originates from numerical modeling experiments and the examination of coarsely gridded analysis products. High-resolution in situ observations of nocturnal MCSs and their EILs are essential to identifying potential sources of errors in numerical forecasts of elevated MCSs, and the analysis of such observations are currently lacking in the scientific literature. This literary gap exists because the necessary observations required for analyzing elevated MCS environments are difficult to obtain. For instance, regular upper-level observations of temperature, moisture, and wind are taken via radiosondes with a spatial density of 100–1000 km, and at a temporal frequency of once per 12 h per observation site. This observational frequency is insufficient to observe elevated EILs in a meaningful way, given that the spatial and temporal scales of variability associated with MCS are on the order of 10–100 km and 1–60 min, respectively. The Plains Elevated Convection at Night (PECAN; Geerts et al. 2016) experiment in 2015 aimed to provide the necessary data to fill this knowledge gap by observing MCSs in their nocturnal environment. PECAN was a multiplatform field campaign conducted during the summer months in the central United States that sought to obtain observations of the environments near, and the processes occurring within nocturnal (and potentially elevated) MCSs. During the approach and passage of MCSs, PECAN observers used the strategy of high-temporal-frequency radiosonde launches (e.g., launches every 15–30 min) at an approximately fixed location. This strategy allowed for a detailed look at how the temperature, water vapor, and horizontal wind profiles evolve, and for the observation of quasi-two-dimensional structures within MCSs via time–space conversion (Bryan and Parker 2010).

In this article we analyze high-temporal-resolution radiosonde observations of the environment near and within a warm season training line–adjoining stratiform (TL/AS)-type elevated MCS (e.g., Schumacher and Johnson 2005; Peters and Schumacher 2014) that occurred over southern Iowa, northern Missouri, and northeastern Illinois on the evening of 24 June 2015. We seek to address the following questions: (i) how do CAPE and CIN evolve in the environment near an elevated MCSs, (ii) do numerical models adequately predict the distributions of CAPE and CIN, and (iii) how might errors in CAPE and CIN in numerical models affect the ability of these models to predict the MCS? The organization of this paper is as follows: section 2a provides an overview of the evolution of the 24 June 2015 Iowa MCS, and the PECAN operations during this event, and section 2b summarizes radiosonde observations of the environment near the elevated MCS. Section 3 compares numerical modeling experiments to observations, and identifies the mechanisms for errors in the prediction of the MCS by these models. Section 4 summarizes our results, and section 5 compares our results to past research and outlines avenues for future investigation.

2. Observations of the Iowa MCS

a. Event overview

The synoptic-scale environment on the evening of 24 June 2015 was characterized by a southwesterly low-level jet (LLJ) that extended from Texas through Kansas and Missouri, and terminated along an east–west-oriented frontal boundary in southeastern Nebraska and southern Iowa (Figs. 1a,b). An MCS had moved through northeastern Nebraska and central Iowa earlier in the day (not shown), and had produced a surface cold pool that strengthened the north–south low-level temperature gradient along a preexisting synoptic front (Figs. 1a,b). A mesoscale region of low-level warm-air advection (WAA) was present along the terminus of the low-level jet in southeastern Nebraska and southern Iowa (as is typically present in the “frontal intersection zones” that foster nocturnal MCSs; Moore et al. 2003; Schumacher and Johnson 2005; Trier et al. 2010; Peters and Schumacher 2014). A broad reservoir of most-unstable CAPE (MUCAPE) 3000 J kg−1 extended from southern Iowa southward to the Gulf Coast, and westward toward Colorado (Figs. 1c,d). We must note here that the rapid refresh (RAP; Benjamin et al. 2016) model was used for this synopsis of the event, and that large errors in the RAP-analyzed thermodynamic and moisture variables relative to observations taken during PECAN are identified later in this paper. The large-scale atmospheric flow patterns and surface atmospheric fields (where high-spatial-resolution surface observations are regularly assimilated) are probably representative of the real atmosphere during this event. The local variability in fields such as CAPE and temperature above ground level should be interpreted with great caution.

Fig. 1.

(a),(c) RAP-analyzed 850-hPa horizontal temperature advection (shading, K h−1), 850-hPa temperature (red contours, °C), and 850-hPa winds (wind barbs, kt). (b),(d) RAP-analyzed MUCAPE (shading, J kg−1) and surface winds (wind barbs, kt). Analyses are valid at (left) 0000 and (right) 0300 UTC.

Fig. 1.

(a),(c) RAP-analyzed 850-hPa horizontal temperature advection (shading, K h−1), 850-hPa temperature (red contours, °C), and 850-hPa winds (wind barbs, kt). (b),(d) RAP-analyzed MUCAPE (shading, J kg−1) and surface winds (wind barbs, kt). Analyses are valid at (left) 0000 and (right) 0300 UTC.

The 24 June 2015 Iowa MCS (hereafter simply “the MCS”) originated from a cluster of supercell thunderstorms that developed in west central Iowa at approximately 2300 UTC along the east–west-oriented synoptic front and preexisting OFB from earlier convection (Figs. 2a,b). A strong horizontal gradient in RAP-analyzed surface-based CAPE (SBCAPE) was present along the southwestern flank of these storms, which suggests that they formed along, or slightly northeastward, of the surface OFB (Fig. 2a). By 0300 UTC (not shown), the supercells had grown into a southeastward-moving squall line (Figs. 2c and 3c,d). At the same time, new convection had initiated along the western flank of the MCS in a northwest–southeast-oriented convective line and located over the surface cold pool (Figs. 2c,d). This simultaneous development of an upstream convective line that is decoupled from the surface OFB is known as rearward off-boundary development (Keene and Schumacher 2013; Peters and Schumacher 2014, 2015a,b, 2016). The convective line on the eastern flank of the system continued southeastward and entered northern Illinois by 0600 UTC (Figs. 3c,d). Meanwhile, new convective cells continuously developed along the western flank of the system (this process is called back-building) through 0730 UTC, and repeatedly moved in a direction parallel to the convective line over the same geographic regions (Figs. 2c,d). Both the elevated southeastward-moving flank of the MCS, and the region between the southwestern OFB and the quasi-stationary western flank of the MCS were sampled by radiosondes during PECAN operations.

Fig. 2.

Radar reflectivity (shading, dBZ) from KDMX at the lowest scanned level. The horizontal paths of radiosondes launched from the MP4 location are shown as black lines originating from a black circle. Red dashed contours are RAP-analyzed surface based CAPE (SBCAPE, J kg−1) in all panels. Surface boundaries were manually analyzed.

Fig. 2.

Radar reflectivity (shading, dBZ) from KDMX at the lowest scanned level. The horizontal paths of radiosondes launched from the MP4 location are shown as black lines originating from a black circle. Red dashed contours are RAP-analyzed surface based CAPE (SBCAPE, J kg−1) in all panels. Surface boundaries were manually analyzed.

Fig. 3.

Radar reflectivity (shading, dBZ) from KDVN at the lowest scanned level, valid at (a) 0130, (b) 0430, (c) 0600, and (d) 0700 UTC. The horizontal paths of radiosonde launches from the MUT location are shown as black lines originating from a black circle (the MUT location). Red dashed contours are RAP-analyzed SBCAPE (J kg−1) in all panels.

Fig. 3.

Radar reflectivity (shading, dBZ) from KDVN at the lowest scanned level, valid at (a) 0130, (b) 0430, (c) 0600, and (d) 0700 UTC. The horizontal paths of radiosonde launches from the MUT location are shown as black lines originating from a black circle (the MUT location). Red dashed contours are RAP-analyzed SBCAPE (J kg−1) in all panels.

b. Summary of PECAN observations

The mobile upsonde teams (MUTs, Ziegler et al. 2016)1 were situated in the southeastern corner of Iowa (Figs. 3a–d). A total of 16 radiosondes were launched by the MUTs between 2300 UTC 24 June and 0700 UTC 25 June 2015 (the horizontal paths of these sondes are shown in Fig. 3) at roughly 15-min intervals, with the eastern flank of the convective line having passed over the MUT location during the 0445–0500 UTC time frame (Fig. 3b). Balloons traveled east-southeastward, with total displacements from the launch site of 20–30 km by the time they reached the 100–200-hPa range. The teams relocated 10 km south of their initial positions between 0400 and 0430 UTC (Fig. 3b). A low-level temperature inversion was evident at approximately 900 hPa at 0030 UTC, and MUCAPE was associated with parcels lifted from the top of the inversion (Fig. 4a). A dramatic cooling and moistening of the atmospheric profile between 850 and 700 hPa occurred between 0030 and 0443 UTC (Figs. 4a,b).

Fig. 4.

Skew T–logp diagrams of vertical profiles of radiosonde observed virtual temperature (red dashed lines), temperature (red lines), dewpoint (green lines), and the lifted parcel path of the parcel with the maximum equivalent potential temperature in the column (dashed black lines) observed at (a) 0031 and (b) 0444 UTC, obtained by MUTs. Profiles observed by the MP4 are at (c) 0002 and (d) 0430 UTC.

Fig. 4.

Skew T–logp diagrams of vertical profiles of radiosonde observed virtual temperature (red dashed lines), temperature (red lines), dewpoint (green lines), and the lifted parcel path of the parcel with the maximum equivalent potential temperature in the column (dashed black lines) observed at (a) 0031 and (b) 0444 UTC, obtained by MUTs. Profiles observed by the MP4 are at (c) 0002 and (d) 0430 UTC.

To better understand the aforementioned atmospheric cooling and destabilization, we linearly interpolated temperature, relative humidity, pressure, and wind data from the soundings onto a time–height grid with a vertical grid spacing of 100 m; a temporal grid spacing of 15 min; start and end times of 0000 and 0700 UTC, respectively; and a height range of 0–15 000 m. Data gaps aloft that resulted from loss of contact with radiosondes prior to them reaching the tropopause were filled via temporal interpolation using data from temporally adjacent launches. Finally, quantities such as mixing ratio, CAPE, and CIN were computed from the interpolated analysis, rather than directly from radiosondes. In constructing this analysis, we assumed that each radiosonde profile represented an instantaneous snapshot of an atmospheric column above the launch site (i.e., we ignored radiosonde drift). Most of our analysis (with the exception of CAPE) concentrates on quantities within the lowest 3 km of the atmosphere. Typical horizontal radiosonde displacements within the lowest 3 km of the atmosphere were on the order of 5 km or less, which is comparable to the horizontal grid spacings of the models and analysis datasets that are analyzed later in this article. We do not expect that these horizontal radiosonde displacements will appreciably affect the conclusions drawn from the comparison between radiosonde profiles and vertical profiles taken from analysis datasets and simulations. Radiosonde errors in temperature are expected to be on the order of 0.3 K and 3% RH (which equates to approximately 0.5 g kg−1 of water vapor) over the temperature, height, and times of day considered in this article (Vaisala 2013).

Cooling of 2–3 K occurred between 1.25 and 3.5 km above ground level (AGL) during the 0030–0430 UTC time frame (Fig. 5a). In contrast with the aforementioned cooling, the 0.25–1 km AGL layer warmed by 3–4 K between 0300 and 0430 UTC (Fig. 5a). Abrupt cooling of 3–4 K below 2 km during the 0430 and 0530 UTC time frame signified the passage of the squall line (Fig. 5a), and an associated weak surface cold pool (evident as cooling below 0.5 km, Fig. 5a). Water vapor mixing ratio increased by 2–4 g kg−1 in the 0.5–1.5 km AGL layer during the 0200 and 0445 UTC time frame (Fig. 5b). These temperature and moisture changes in the 0.25–1.5-km layer corresponded to a dramatic increase in CAPE of nearly 2500 J kg−1 during the 0030–0445 UTC time frame (Fig. 5c), and an increases in the depth of the region of CIN 10 J kg−1. SBCAPE remained near or below 500 J kg−1 through the 0000 to 0700 UTC time frame, which was nearly an order of magnitude smaller than CAPE above 0.25 km.

Fig. 5.

Comparison of time (x axis) vs height (y axis) diagrams of various atmospheric fields from (left) MUT radiosonde launches (OBS) and (right) the same location in the RAP analysis. The potential temperature change (shading, K) from (a) 0030 UTC (the 0000 UTC sounding had spurious moisture data), and from (d) 0000 UTC, and potential temperature (gray contours, K). (b),(e) Relative humidity (shading, %) and water vapor mixing ratio (dark gray contours, g kg−1). (c),(f) CAPE as a function of lifted parcel level (shading, J kg−1), CIN (magenta contour, −10 J kg−1), and potential temperature (gray contours, K). (g)–(i) Vertical profiles of (red line), (green line), and (blue line), respectively (J kg−1). All quantities in (g)–(i) are averaged over the 0000–0700 UTC time frame.

Fig. 5.

Comparison of time (x axis) vs height (y axis) diagrams of various atmospheric fields from (left) MUT radiosonde launches (OBS) and (right) the same location in the RAP analysis. The potential temperature change (shading, K) from (a) 0030 UTC (the 0000 UTC sounding had spurious moisture data), and from (d) 0000 UTC, and potential temperature (gray contours, K). (b),(e) Relative humidity (shading, %) and water vapor mixing ratio (dark gray contours, g kg−1). (c),(f) CAPE as a function of lifted parcel level (shading, J kg−1), CIN (magenta contour, −10 J kg−1), and potential temperature (gray contours, K). (g)–(i) Vertical profiles of (red line), (green line), and (blue line), respectively (J kg−1). All quantities in (g)–(i) are averaged over the 0000–0700 UTC time frame.

One of the primary goals of this study is to evaluate the performance of NWP models in diagnosing/predicting atmospheric processes within the MCS environment. The RAP model analysis,2 which serves as a “best guess” for the atmospheric state at a given time, is an ideal starting point for comparing observations to NWP models (since it assimilates surface and radar observations on an hourly basis; Zhu et al. 2013; Pan et al. 2014). The time–height evolution of the atmosphere from the RAP analysis at the MUT location featured markedly different thermodynamic (Fig. 5d), moisture (Fig. 5e), CAPE, and CIN evolutions (Fig. 5f) from the radiosonde observed analysis. The difference in moist static energy (MSE) between observations and the RAP analysis is given by . The individual contributions to by temperature and moisture contributions are and , respectively. Temporally averaged profiles of and reveal that the RAP underanalyzed thermal energy by 500–1000 J kg−1 in the 0.5–1.5 km AGL layer (Fig. 5g) and moist energy by 2500–3000 J kg−1 in the 0.25–2.25 km AGL layer (Fig. 5h). The RAP analysis subsequently underpredicted temporally averaged CAPE values by 500–1000 J kg−1 in the 0.5–1.5 km AGL layer (Fig. 5i), and underpredicted maximum CAPE values by as much as 2500 J kg−1.

The depth of CAPE J kg−1 nearly doubled between 0030 and 0330–0430 UTC (Fig. 6a). The speed of northwestward flow decreased with height above 250 m AGL at all times prior to the arrival of the squall line at the MUT location (Fig. 6b). As the squall line was oriented from southwest to northeast and moved toward the southeast, the system’s southeastern flank was propagating in the direction of the low-level wind shear. Air parcels ahead of downshear-propagating squall lines often experience upward displacements along the system’s outflow that substantially exceed the depth of the outflow itself (e.g., Rotunno et al. 1988; Weisman and Rotunno 2004; Bryan and Rotunno 2014),3 owing to upward air parcel accelerations due to both effective buoyancy and dynamic pressure forcing being present there (e.g., Parker and Johnson 2004; Bryan and Rotunno 2014). In many cases, this strong lifting along downshear-propagating outflow generates a persistent line of moist updrafts along this outflow flank (e.g., Moncrieff and Liu 1999). We use through the remainder of the study as a measure of convective inhibition (where is an air parcel’s source level). This quantity is formally defined as the distance a parcel must be lifted to reach its LFC and has units of distance (Davenport and Parker 2015; Peters and Schumacher 2016). It can be compared to the vertical distance a parcel is actually displaced to determine whether a parcel did or did not reach its theoretical LFC. Values of were on the order of 2 km at 0030 UTC (Fig. 6c). By 0330 and 0430 UTC, had been reduced to 500 m within the EIL (Fig. 6c). A lower bound for the actual vertical displacement of air parcels may be estimated by assuming that parcels at 0430 UTC underwent adiabatic ascent to their locations at 0530 UTC. We therefore estimate the vertical displacement of an air parcel from 0430 to 0530 UTC as the vertical displacement of isentropes over this time interval (this quantity is denoted as ). At 0330 and 0430 UTC, within the 0.7 km to 1.5 km AGL layer (Fig. 6c). Below 0.7 km and above 1.75 km AGL, on the other hand, (Fig. 6c). This suggests that parcels below 0.7 km and above 1.5 km AGL did not reach their LFCs, and were outside the EIL to the squall line.

Fig. 6.

(a) Vertical profiles from gridded MUT sounding data of CAPE at 0030 UTC (red line), 0330 UTC (blue line), and 0430 UTC (green line) (J kg−1). (b) As in (a), but for vertical profiles of southeasterly wind speed (m s−1). (c) As in (a),(b), but for vertical profiles of ; is shown as a thin gray line.

Fig. 6.

(a) Vertical profiles from gridded MUT sounding data of CAPE at 0030 UTC (red line), 0330 UTC (blue line), and 0430 UTC (green line) (J kg−1). (b) As in (a), but for vertical profiles of southeasterly wind speed (m s−1). (c) As in (a),(b), but for vertical profiles of ; is shown as a thin gray line.

Mobile PECAN Integrated Sounding Array 4 (MP4; UCAR/NCAR Earth Observing Laboratory 2016) was situated in south-central Iowa, and approximately 100 km southwest of the initial grouping of supercells at 0000 UTC. Radiosondes were launched from this location at 0002 UTC slightly to the north of a preexisting surface OFB (Fig. 2a), at 0130 UTC several minutes prior to the arrival of the grouping of supercells and their associated OFB (Fig. 2b), at 0430 UTC (Fig. 2c), and at 0600 UTC (Fig. 2d), with the last two radiosondes having been launched between the training convective line on the western side of the MCS and the southwestern flank of the MCS’s OFB. A skew T–logp diagram of the 0002 UTC launch (Fig. 4c) shows a temperature inversion near 900 hPa, and shows that the maximum CAPE was associated with parcels within the layer just above the inversion (Figs. 3a,b and 7a; though the surface CAPE here is substantially larger than that at the MUT location at a similar time). By 0430 UTC (Fig. 4d) the inversion height had increased and the inversion had moistened as a result of the passage of a new OFB, with the most unstable CAPE once again associated with parcels that originated above the inversion (Fig. 7a). Neither the 0002 UTC, nor 0130 UTC sondes reached the tropopause, and we therefore temporally extrapolated upper-level temperature onto these soundings from adjacent launches at 0430 and 0602 UTC in order to complete CAPE computations (we assume that temporal variations aloft occur gradually between radiosonde launches).

Fig. 7.

(a)–(c) As in Fig. 6, but for MP4 launches at 0002 UTC (red lines), 0130 UTC (magenta lines), 0430 UTC (blue lines), and 0603 UTC (green lines). Here is plotted in (c) for parcels lifted from 0002 to 0130 UTC locations (light gray line), 0430 UTC locations (gray line), and 0603 UTC locations (dark gray line); and (b) shows southwesterly wind speed.

Fig. 7.

(a)–(c) As in Fig. 6, but for MP4 launches at 0002 UTC (red lines), 0130 UTC (magenta lines), 0430 UTC (blue lines), and 0603 UTC (green lines). Here is plotted in (c) for parcels lifted from 0002 to 0130 UTC locations (light gray line), 0430 UTC locations (gray line), and 0603 UTC locations (dark gray line); and (b) shows southwesterly wind speed.

We estimated actual vertical displacements of parcels from their 0002 UTC positions as , in a similar manner to what was done for the eastern MCS flank in section 2a. Parcels from the 0002 and 0130 UTC profiles were convectively inhibited, and required upward of 500–1000 m of lift to reach their LFCs (Fig. 7c). Values of at 0430 UTC were less than 400 m (Fig. 7c), suggesting that parcels had not reached their LFCs as they were lifted along the OFB. In contrast with the southeastern MCS flank, the wind speed toward the boundary increased with height near the system’s southwestern flank (Fig. 7b, e.g., Moncrieff and Liu 1999). This wind shear orientation results in a downward-oriented dynamic pressure acceleration along the OFB (e.g., Parker and Johnson 2004), and often results in vertical air parcel displacements that are comparable with the depth of the outflow. These differing wind shear orientations, and the resultant differences in lifting magnitudes along the boundaries, likely explain why CI occurred on the southeastern outflow flank, but not the southwestern outflow flank (e.g., Trier et al. 2010; Peters and Schumacher 2015a, 2016).

3. Comparison to numerical simulations

In this section we examine how the large errors in CAPE, CIN, temperature, and moisture fields in NWP models may have impacted their forecasts of the 24 June 2015 Iowa MCS. We examine 13 numerical simulations of the event that produced notable errors in the placement and evolution of the MCS, and whose simulated evolutions of the MCS were quite different from each other. Each simulation was originally run for purposes other than the specific scientific problems addressed in this study, but together they have motivated our interest in the common MCS forecast errors occurring in a range of plausible convection-allowing forecasts. Given the diversity of purposes for which they were designed, there are seemingly arbitrary differences in the modeling configurations among the simulations. Our focus here is on the relationship of the varying model fields to the forecast MCSs, not on the model configurations themselves.

The first simulation of the MCS was by the real-time Colorado State University WRF-ARW Model, version 3.4.1 (Klemp et al. 2007; Skamarock et al. 2008; Skamarock and Klemp 2008; hereafter the CSUWRF; Schumacher 2015a). This model configuration used the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) as initial and lateral boundary conditions (ICs and LBCs), featured a horizontal grid spacing of 4 km, lateral boundary conditions updated every 3 h, MYJ boundary layer physics (Janjić 1994), Morrison two-moment microphysics (Morrison et al. 2005), the Noah land surface model (Mitchell et al. 2004), RRTM for GCMs (RRTMG) longwave and shortwave radiation physics (Mlawer et al. 1997), and did not implement a cumulus parameterization scheme (the domain bounds of this run are shown in Fig. 8a). The model was run for 48 h with 23–24 h of model spinup having occurred prior to the initiation of the Iowa MCS (the MCS occurred between 0000 and 1200 UTC 25 June). The CSUWRF was one of many convection-allowing models that were used during PECAN as guidance for mission planning (Geerts et al. 2016). The second version 3.4.1 WRF simulation was configured with the RAP analysis as ICs and LBCs in an attempt to produce a simulation that closely resembled the observed MCS evolution [this simulation is hereafter referred to as the downscaled RAP (DSRAP)]. The DSRAP featured an outer domain with a 15-km grid spacing, an inner domain with a 3-km grid spacing (Fig. 8a), a one-way feedback from the outer domain to the inner domain, and was run from 0000 UTC 24 June to 1200 UTC 25 June 2015 with lateral boundaries updated every hour. The third WRF simulation was configured with the North American Mesoscale Forecast System (NAM) analysis at 0000 UTC 24 June 2015 as ICs, and the subsequent 6-hourly NAM analyses as LBCs [this simulation is hereafter referred to as the downscaled NAM (DSNAM)]. This simulation featured three nests with 15-, 3-, and 1-km (we analyze the 1-km nest here) grid spacings, respectively (Fig. 8a), and two-way feedback across the lateral boundaries between nests. Additional details on the DSRAP, CSUWRF, and DSNAM model configuration are available in Table 1. All three of these simulations produced MCSs that resembled the observed one, and associated swaths of precipitation that were displaced approximately 100–200 km northeastward of stage IV (ST4) analyzed precipitation (Fig. 8b) in the CSUWRF and DSRAP simulations, and approximately 50–100 km northeastward of the stage IV precipitation in the DSNAM. Finally, we analyzed a 10-member ensemble of forecasts with similar configurations to the CSUWRF (e.g., 4-km horizontal grid spacing, Morrison two-moment microphysics, Noah land surface model, see Table 1), but with a different member of the Global Ensemble Forecast System (GEFS/R; Hamill et al. 2013) used as ICs and LBCs to each ensemble member [see Nielsen (2016) and Nielsen and Schumacher (2016) for more information on this model configuration; we refer to these runs as ENS].

Fig. 8.

(a) Domain bounds for the DSRAP inner domain (DO2, red box), the CSUWRF domain (green box), and the DSNAM inner domain (D03, blue box). (b) Accumulated rainfall from 0000 to 1200 UTC 25 Jun 2015 from stage-IV precipitation analysis (shading, mm), the DSRAP simulation (red contour at 60 mm), the CSUWRF simulation (green contour at 60 mm), and the DSNAM simulation (blue contour at 60 mm). The shading and contours in (b) are also shown in (a).

Fig. 8.

(a) Domain bounds for the DSRAP inner domain (DO2, red box), the CSUWRF domain (green box), and the DSNAM inner domain (D03, blue box). (b) Accumulated rainfall from 0000 to 1200 UTC 25 Jun 2015 from stage-IV precipitation analysis (shading, mm), the DSRAP simulation (red contour at 60 mm), the CSUWRF simulation (green contour at 60 mm), and the DSNAM simulation (blue contour at 60 mm). The shading and contours in (b) are also shown in (a).

Table 1.

Summary of the DSRAP, CSUWRF, and DSNAM model configurations, where the asterisk indicates that LBCs were applied to domain 1 only and “same” indicates the same configuration between the two models. The ENS runs used an identical model configuration to the CSUWRF, but with each ENS member driven by an individual member of the 0000 UTC 24 Jun 2015 initialization of the GEFS/R (Hamill et al. 2013).

Summary of the DSRAP, CSUWRF, and DSNAM model configurations, where the asterisk indicates that LBCs were applied to domain 1 only and “same” indicates the same configuration between the two models. The ENS runs used an identical model configuration to the CSUWRF, but with each ENS member driven by an individual member of the 0000 UTC 24 Jun 2015 initialization of the GEFS/R (Hamill et al. 2013).
Summary of the DSRAP, CSUWRF, and DSNAM model configurations, where the asterisk indicates that LBCs were applied to domain 1 only and “same” indicates the same configuration between the two models. The ENS runs used an identical model configuration to the CSUWRF, but with each ENS member driven by an individual member of the 0000 UTC 24 Jun 2015 initialization of the GEFS/R (Hamill et al. 2013).

Connection between displacement errors and moisture bias

A comparison between the 0130 UTC MP4 radiosonde analyzed atmospheric profile, and profiles from the same time and location in the DSRAP, CSUWRF, and DSNAM simulations shows that all three model simulations substantially underpredicted CAPE above 0.5 km by 1500–2000 J kg−1 (Fig. 9a), and overpredict in this layer by 1–2 km (Fig. 9b). In contrast, in the lowest 0.5 km both the CSUWRF and DSNAM overpredicted CAPE and underpredicted . Furthermore, was generally larger in the DSRAP than the DSNAM and CSUWRF simulations (Fig. 9b). Whereas all models overpredicted θ through most of the lower atmosphere (Fig. 9c), they substantially underpredicted above roughly 0.5 km (Fig. 9d), which was responsible for their under- and overpredictions of CAPE and , respectively (the DSNAM overpredicted near the ground). By 0430 UTC, for parcels in the layers of highest CAPE in the CSUWRF and DSNAM were approximately 1 km, and 3–4 km in the DSRAP, whereas observed remained in the 0.5–1.0-km range (Figs. 9e,f). Once again, differences in moisture were responsible for these differing distributions and magnitudes of CAPE and (Figs. 9g,h) at 0430 UTC, with the CSUWRF and DSNAM having underpredicted moisture above 0.75 km AGL and overpredicted moisture below this level, and the DSRAP having severely underpredicted moisture everywhere above the surface.

Fig. 9.

Comparisons between atmospheric fields observed by the 0130 UTC MP4 sounding (thick black dashed lines) and analogous vertical profiles from the DSRAP (red lines), the CSUWRF (green lines), and the DSNAM (blue lines) at the same location and time: (a) CAPE (J kg−1), (b) (m), (c) potential temperature (θ, K), and (d) water vapor mixing ratio (g kg−1). (e)–(h) As in (a)–(d), but at 0430 UTC.

Fig. 9.

Comparisons between atmospheric fields observed by the 0130 UTC MP4 sounding (thick black dashed lines) and analogous vertical profiles from the DSRAP (red lines), the CSUWRF (green lines), and the DSNAM (blue lines) at the same location and time: (a) CAPE (J kg−1), (b) (m), (c) potential temperature (θ, K), and (d) water vapor mixing ratio (g kg−1). (e)–(h) As in (a)–(d), but at 0430 UTC.

We hypothesized that the aforementioned differences in between model simulations and observations contributed to errors in the simulated placement of the MCSs relative to the observed MCS. In other words, if we consider two flows with nonzero CAPE traveling horizontally through similar regions of isentropic ascent, convection will initiate first within the flow that starts with a lower LFC, and will have traveled the shortest distance horizontally by the time of convection initiation (CI).4 In the case of the Iowa MCS, the southwesterly flow in the dataset with the lowest LFC and least CIN will produce an MCS farthest to the southwest, since parcels within this flow reach their LFCs prior to parcels in the dataset with higher initial LFCs. We address this hypothesis by comparing vertical sections (e.g., distance vs height plots) along the western flank of the simulated MCSs, and by connecting MCS placement errors in all simulations to errors in their moisture prediction to the southwest of the MCS. We focus our analysis on the MP4 location, since this was the location where in situ observations of the environment were available for comparison to simulations. Moisture was underpredicted by models nearly 300 km away at the MUT site as well, which suggests that that the modeled moisture errors were not localized to the MP4 location.

Surface temperature analyses of the three model simulations at 0100 UTC show general agreement in the positioning of the preexisting surface OFB between the DSNAM and DSRAP, where it was located in southwestern Iowa and northern Missouri (Figs. 10a,e) at a similar location to the observed OFB. In contrast, the CSUWRF did not assimilate surface observations and showed the surface outflow boundary slightly farther north of the DSNAM and DSRAP (Fig. 10c; especially in eastern Iowa). The convective cells that would become an MCS developed approximately 100–200 km northeast of a preexisting OFB in the DSRAP and DSNAM (Figs. 10a,e). The initial storms in both of these models were displaced northeastward of the initial supercells in observations (cf. Figs. 3a,b).

Fig. 10.

Analysis of (a),(b) DSRAP; (c),(d) CSUWRF; and (e),(f) DSNAM fields at 0100 UTC. (left) Simulated radar reflectivity at 1 km AGL (shading, dBZ) and surface temperature (gray contours, C). (right) Cross sections along the black lines in (left) showing CAPE (shading, J kg−1), potential temperature (θ, gray contours, K), and (black contours, m). Red arrows in (b),(d),(f) point to regions where m, dark blue arrows in (a),(c),(e) point to the surface outflow boundary, and blue squares on the x axis of (b),(d),(f) indicate the location of CI in the cross section.

Fig. 10.

Analysis of (a),(b) DSRAP; (c),(d) CSUWRF; and (e),(f) DSNAM fields at 0100 UTC. (left) Simulated radar reflectivity at 1 km AGL (shading, dBZ) and surface temperature (gray contours, C). (right) Cross sections along the black lines in (left) showing CAPE (shading, J kg−1), potential temperature (θ, gray contours, K), and (black contours, m). Red arrows in (b),(d),(f) point to regions where m, dark blue arrows in (a),(c),(e) point to the surface outflow boundary, and blue squares on the x axis of (b),(d),(f) indicate the location of CI in the cross section.

Vertical sections from the DSRAP (Fig. 10b) and DSNAM (Fig. 10f) show a gradual increase in the height of the layer of nonzero CAPE with northeastward distance past the OFB, along with gradually reduced along this path. The initial storms had developed in a region where dropped below 250 m in both models. The vertical displacement of isentropes with northeastward extent was in the 1000–1500-m range for the DSRAP simulation (Fig. 10b), and 500–1000 m in the DSNAM simulation (Fig. 10f). If we assume slowly evolving thermodynamic conditions, these isentrope displacements give a crude estimation of vertical parcel displacements and suggest that isentropic ascent was slightly more intense in the DSRAP than the DSNAM. As the values near the OFB in the DSNAM were much lower than in the DSRAP, the resultant CI locations were actually quite similar between the models since parcels in the DSNAM were lifted less, but required less lift to reach their LFCs, whereas parcels in the DSRAP were lifted more, but required more lift to reach their LFCs.

Though the OFB was farther north in the CSUWRF than the other models (Fig. 10c), the initial convection in the CSUWRF developed immediately along the OFB at a location much closer to where the observed cells initiated than the other models (cf. to Figs. 3a,b). A comparison of the DSRAP (Fig. 10b) and DSNAM (Fig. 10f) isentrope patterns to the CSUWRF vertical section is difficult at 0100 UTC given the irregular isentrope distribution near the OFB associated with newly developed convection. The layer of nonzero CAPE in the CSUWRF was much deeper than the other simulations, and the farthest southwest region of m occurred nearly 2 km above the ground and roughly 20 km south of the OFB (Fig. 10d). Another deep region of was present immediately above the OFB (Fig. 10d). A more pronounced upward arch in isentropes was present along the OFB in the CSUWRF than in the mother models, implying that adiabatic lifting was more intense along the OFB in the CSUWRF (Fig. 10d). The combination of this enhanced lifting and relatively low along the OFB was the likely cause of CI directly along the OFB in the CSUWRF.

A line of intense convection along the southeastern flank of the DSNAM MCS moved sharply southeastward from the region where its convection originated, having traveled over similar locations to the observed MCS (Fig. 11e). An intense line of convection was also produced by the CSUWRF simulation Fig. 11c); however, the CSUWRF convective line traveled east-southeastward, having moved along a track northeast of the observed MCS. The DSRAP MCS was characterized by weaker and less organized simulated radar reflectivity echoes along its eastern flank when compared to both the other simulations and observations, and also tracked farther northeast of the observed MCS (Fig. 11a). How might moisture differences between the models have played a role in the differing tracks of the eastern flanks of the simulated MCSs? Differences in the column maximum between the DSNAM and CSUWRF show much smaller low-level moisture in the DSNAM over eastern Iowa and Illinois than the CSUWRF (Figs. 12a–c), where the latter model had advected large quantities of low-level moisture to the east of the MCS. A corresponding region of much smaller values was present in the CSUWRF, when compared to the DSNAM, in eastern Iowa and Illinois. Conversely, the DSNAM model had much higher low-level moisture values over southern Iowa and northern Missouri than the CSUWRF, and correspondingly smaller values there. This suggests that the respective MCSs simply propagated toward regions where their model’s moisture content was highest and values were lowest. This led to a sharp southeastward movement of the DSNAM model through southern Iowa and northeastern Missouri where DSNAM moisture was highest, whereas the CSUWRF propagated on a comparatively eastward track into northern Illinois where the CSUWRF moisture was the highest. The DSRAP, which was excluded from this particular analysis, was exceedingly dry at 0400 UTC when compared to the other models (Fig. 9h), which resulted in the DSRAP MCS track being farther northeast than the other models.

Fig. 11.

As in Fig. 10, but valid at 0500 UTC. Cross sections in (right) are valid along the westernmost black line in (left) (the easternmost black lines show the paths of the cross sections in Fig. 13). Blue squares on the x axis of (b),(d),(f) indicate the location of the training convective line in the cross section.

Fig. 11.

As in Fig. 10, but valid at 0500 UTC. Cross sections in (right) are valid along the westernmost black line in (left) (the easternmost black lines show the paths of the cross sections in Fig. 13). Blue squares on the x axis of (b),(d),(f) indicate the location of the training convective line in the cross section.

Fig. 12.

The difference in column maximum (shading; g kg−1) above 200 m between the DSNAM and CSUWRF simulations, where warm colors indicate that the DSNAM moisture is larger than the CSUWRF above 200 m, and cool colors indicate that the CSUWRF moisture is larger than the DSNAM above 200 m. (thin contours) The difference in minimum column (, intervals of −2000, −1000, −500, 500, 1000, and 2000 m) between the DSNAM and CSUWRF, where cyan contours indicate is smaller in the DSNAM than the CSUWRF, and gray contours indicate that is smaller in the CSURWF than the DSNAM. Dark black contours are the 40-dBZ radar reflectivity contour from the DSNAM. (a) 0300, (b) 0400, and (c) 0500 UTC.

Fig. 12.

The difference in column maximum (shading; g kg−1) above 200 m between the DSNAM and CSUWRF simulations, where warm colors indicate that the DSNAM moisture is larger than the CSUWRF above 200 m, and cool colors indicate that the CSUWRF moisture is larger than the DSNAM above 200 m. (thin contours) The difference in minimum column (, intervals of −2000, −1000, −500, 500, 1000, and 2000 m) between the DSNAM and CSUWRF, where cyan contours indicate is smaller in the DSNAM than the CSUWRF, and gray contours indicate that is smaller in the CSURWF than the DSNAM. Dark black contours are the 40-dBZ radar reflectivity contour from the DSNAM. (a) 0300, (b) 0400, and (c) 0500 UTC.

Outflow boundaries associated with all three modeled MCSs moved through southeastern Iowa and northeastern Missouri (Figs. 11a,c,e)—why then was the DSNAM MCS the only MCS to produce strong convection in southeastern Iowa and northeastern Missouri? South–north-oriented vertical sections through the OFB in southeastern Iowa show relatively similar wind distributions among the models, with a strong southerly low-level jet present below 2000 m in each simulation (Figs. 13a–c). The cold pool in the DSRAP simulation (its associated OFB is evident at 40.3°N in Fig. 13a) was very shallow, with vertical isentrope displacements only evident below 500 m near the OFB. Southerly flow that encountered this OFB contained only weak CAPE ( 1000 J kg−1, Fig. 14a), prohibitively large ( 2000 m), and small moisture compared to the other two models (Figs. 13a–c and 14c). These parameters suggest that southerly flow in the DSRAP was exceedingly convectively stable and was not sufficiently lifted for the development of deep convection along the OFB, with maximum w along the OFB having been far less than 1 m s−1 (Fig. 14f). Southerly flow in the CSUWRF was moister than the DSRAP (Figs. 13a,b and 14c), was accordingly smaller (Figs. 13a,b and 14a), and CAPE was accordingly higher (Fig. 14a). However, if we use and isentropes to be tracers of parcel paths for adiabatic flow, vertical displacements of contours and isentropes were only 400–500 m from the south side of the boundary to the north side, which is far smaller than the values in this region. This precluded the development of deep convection. Southerly flow within the DSNAM was moister than both other models, had larger CAPE, and smaller ( 500–700 m; Figs. 13c and 14a). A strong convective updraft was also present directly along the OFB in the DSNAM. Though the cold pool (and associated OFB lifting) in the DSNAM was substantially stronger than in the other models (Fig. 14e), the relatively low in the DSNAM implies that even the comparatively weak lifting in the other models would have lifted the DSNAM flow to its LFC. This further evidence supports the idea that moisture, which substantially modulated , determined the track of the southeastward-moving MCS. Higher moisture values in the DSNAM allowed for the MCS to propagate southeastward along the systems OFB into southeastern Iowa and northeastern Missouri; whereas, lower moisture values in the other models precluded their southward propagation into southeastern Iowa and northeastern Missouri.

Fig. 13.

Cross sections at 0500 UTC along the easternmost black lines in Figs. 11a,c,e of (shading, g kg−1), potential temperature (θ, gray contours, K), (black contours, m), and cross-sectional parallel wind (red arrows, m s−1) from (a) the DSRAP, (b) the CSUWRF, and (c) the DSNAM. The leftmost vertical black dashed lines are , and the rightmost black lines are .

Fig. 13.

Cross sections at 0500 UTC along the easternmost black lines in Figs. 11a,c,e of (shading, g kg−1), potential temperature (θ, gray contours, K), (black contours, m), and cross-sectional parallel wind (red arrows, m s−1) from (a) the DSRAP, (b) the CSUWRF, and (c) the DSNAM. The leftmost vertical black dashed lines are , and the rightmost black lines are .

Fig. 14.

Atmospheric fields at 0500 UTC along the lines in Fig. 13 (assessed within air immediately south of the OFB) from the CSUWRF (green), the DSNAM (blue). and the DSRAP (red). (a) CAPE (solid lines, J kg−1) and (dashed lines, m), (b) CIN (J kg−1), (c) (g kg−1), (d) (K), (e) υ wind (m s−1), and (f) maximum w between the and the lines in Fig. 13 (m s−1).

Fig. 14.

Atmospheric fields at 0500 UTC along the lines in Fig. 13 (assessed within air immediately south of the OFB) from the CSUWRF (green), the DSNAM (blue). and the DSRAP (red). (a) CAPE (solid lines, J kg−1) and (dashed lines, m), (b) CIN (J kg−1), (c) (g kg−1), (d) (K), (e) υ wind (m s−1), and (f) maximum w between the and the lines in Fig. 13 (m s−1).

A comparison between atmospheric profiles from the models at the MUT location at 0130 UTC, and the observed atmospheric profile reveals that thermodynamic and moisture fields varied considerably among models at this time (Figs. 15a–d). Interestingly, the DSRAP was the most comparable to observations in terms of moisture and CAPE (Fig. 15a), whereas the DSNAM was most comparable to observations in terms of temperature (Fig. 15c). The CSUWRF substantially overpredicted low-level moisture, CAPE, and thereby underpredicted , when compared with the other models and observations (all models slightly underpredicted ; Figs. 15b,d). As the convection associated with the MCS of interest was displaced several hundred kilometers to the northwest of the MUT location at 0130 UTC, the model errors at this time did not directly affect the MCS evolution (though the excessive moisture in the CSUWRF eventually advected into northern Illinois, and played a role in the erroneously easterly track of the MCS in that model; Figs. 12a–c). By 0430 UTC, the comparisons between the MUT sounding and the model simulations are quite different from the 0130 UTC comparison. The distributions of CAPE and in the DSNAM had become the closest to observations as a consequence of the DSNAM having slightly higher (and closer to observations) moisture above 700 m than the other models (Figs. 15e,f,h). On the other hand, the DSRAP had far too little low-level moisture, consistent with the findings for the RAP analysis in Fig. 5. As the moisture in the DSNAM was the “least erroneous” relative to observations, CAPE and were also closest to observations in this model, resulting in the DSNAM MCS northeastward position errors having been smaller than the other models.

Fig. 15.

As in Fig. 9, but at the MUT location.

Fig. 15.

As in Fig. 9, but at the MUT location.

As was the case in observations, a northwest–southeast-oriented training convective line developed in the wake of the initial southeastward-moving MCSs in all three simulations (Figs. 11a,c,e). Vertical sections reveal that the vertical displacement of the 306–310-K isentropes with northeastward distance was relatively similar between the DSNAM and CSUWRF, with isentrope heights increasing by approximately 750–1000 m over the 0–200-km range in the vertical section (Figs. 11d,f). Vertical displacements of the same isentropes were closer to 1500 m over the 50–250-km range in the DSRAP simulation (Fig. 11b). The training line was positioned farthest to the southwest (and closest to the observed position) in the DSNAM simulation (Figs. 11a,b), which may be a consequence of that simulation having the highest low-level moisture and lowest subsequent low-level (Figs. 12, 13, 14a, and 9b,f). Air parcels in the DSNAM therefore reached their (comparatively lower) LFCs at a location farther to the southwest of the other models as these air parcels traveled northeastward and ascended along slanting isentropes. At the other end of the spectrum, the DSRAP training line was the farthest to the northeast among the simulations, which is consistent with this model having the smallest low-level moisture magnitudes and air parcels subsequently traveling farther to the northeast along slanting isentropes before reaching their LFCs (Figs. 9b,f).

We released trajectories from the MP4 location from the 200- to 1500-m height range in all simulations in order to assess the validity of the steady-state thermodynamic assumptions used to ascertain the vertical displacements of parcels as they traveled toward the northeast (Fig. 16). Trajectories released at 0130 UTC in the CSUWRF were intercepted by the southeastward-moving convective line and were lifted sharply along the MCS’s associated outflow, which precludes their comparison with isentrope patterns in vertical sections (Fig. 16a) and with the other models (Figs. 16b,c). Trajectories that were released at 0230 and 0330 UTC, on the other hand, showed general agreement with the isentrope displacements in vertical sections, where parcels displaced 750–1250 m over 150 km of northeastward travel, which corroborates the parcel displacements that were inferred from the distribution of isotherms in vertical sections (Figs. 16g–i).

Fig. 16.

(a)–(c) Trajectories (black lines) releases from the MP4 location at 0130 UTC with initial heights ranging from 200–1500 m at 25-m intervals, and the estimated number of 1-min trajectory points per km2 (computed by a Gaussian filter with a radius of influence of 3 km). (d)–(f) As in (a)–(c), but at 0230 UTC. (g)–(i) As in (a)–(c), but at 0330 UTC. (left) CSUWRF, (middle) DSRAP, and (right) DSNAM.

Fig. 16.

(a)–(c) Trajectories (black lines) releases from the MP4 location at 0130 UTC with initial heights ranging from 200–1500 m at 25-m intervals, and the estimated number of 1-min trajectory points per km2 (computed by a Gaussian filter with a radius of influence of 3 km). (d)–(f) As in (a)–(c), but at 0230 UTC. (g)–(i) As in (a)–(c), but at 0330 UTC. (left) CSUWRF, (middle) DSRAP, and (right) DSNAM.

We also compared MCS positions in the ENS model members to their low-level moisture at the MP4 location. We specifically hypothesized that low-level moisture at the MP4 location was negatively correlated with northeastward MCS displacement among ENS members. All 10 of the ENS members produced an MCS in Iowa and Illinois (not shown). Much like the CSUWRF, DSRAP, and DSNAM, however, the MCS precipitation in each ENS member was displaced to the northeast of the observed MCS (Figs. 17a,b). The profiles of moisture at the MP4 location at 0100 UTC from the 10 ENS members featured an 2 g kg−1 range of below 1 km AGL (Fig. 17c), with all members slightly drier than the DSRAP, CSUWRF, and DSNAM, and substantially drier than observations. We smoothed the 0000–1200 UTC 25 June accumulated rainfall fields from each simulation (including the DSNAM, DSRAP, CSUWRF) and ST4 observations with a Gaussian filter with a radius of influence of 20 km. We then computed the northeastward displacement (in kilometers) of the maximum accumulated precipitation in a model simulation from the analogous ST4 maximum. This northeastward displacement was compared to the 0100 UTC mean 200–750-m in the ENS, DSNAM, DSRAP, and CSUWRF simulations at the MP4 location (Fig. 17d; the 200–750-m layer was where the strongest correlation was found). Values of were strongly negatively correlated with both northeastward displacement of precipitation areas in models from ST4 (R2 = 0.67, p = 0.001), and this correlation was statistically significant to the 95% confidence level (e.g., based on a Student’s t test; Fig. 17d), which affirms that that the underrepresentation of low-level moisture, and the subsequent overrepresentation of in the models was responsible for the northeastward position errors of the simulated MCSs. The DSNAM featured the highest low-level moisture content ( 16.4 g kg−1) of all the models considered, along with the smallest northeastward displacement at 100 km (Fig. 17d). The largest displacement ( 260 km) occurred with the ENS member that had the lowest moisture ( 13.9 g kg−1) of the simulations considered.

Fig. 17.

(a) 50 mm 24-h precipitation contours ending at 1200 UTC 25 Jun 2015 from the 10 ENS members (each member has a different color; the Ctrl simulation was not analyzed here). (b) 24-h ST4 precipitation contours ending at 1200 UTC (shading, mm). (c) 0100 UTC vertical profiles of water vapor mixing ratio from the 10 ENS runs at the MP4 location (gray lines, , g kg−1), from the DSRAP at 0130 UTC at the MP4 location (red line), from the CSUWRF at 0130 UTC at the MP4 location (green line), from the DSNAM at 0130 UTC at the MP4 location (blue line), and from the observed sounding by MP4 at 0130 UTC (black dashed line). (d) Northeastward displacement (km) of the 12-h precipitation maxima (valid at 1200 UTC) from the observed ST4 centroid for the 10 ENS runs (gray dots), the DSRAP (red dot), the CSUWRF (green dot), and the DSNAM (blue dot) plotted against the 200–750-m mean water vapor mixing ratio (g kg−1) at the MP4 location at 0100 UTC. The gray dashed line is a best-fit line for northeastward displacements vs water vapor mixing ratios from the ENS runs; and p values for these lines are shown in the legend.

Fig. 17.

(a) 50 mm 24-h precipitation contours ending at 1200 UTC 25 Jun 2015 from the 10 ENS members (each member has a different color; the Ctrl simulation was not analyzed here). (b) 24-h ST4 precipitation contours ending at 1200 UTC (shading, mm). (c) 0100 UTC vertical profiles of water vapor mixing ratio from the 10 ENS runs at the MP4 location (gray lines, , g kg−1), from the DSRAP at 0130 UTC at the MP4 location (red line), from the CSUWRF at 0130 UTC at the MP4 location (green line), from the DSNAM at 0130 UTC at the MP4 location (blue line), and from the observed sounding by MP4 at 0130 UTC (black dashed line). (d) Northeastward displacement (km) of the 12-h precipitation maxima (valid at 1200 UTC) from the observed ST4 centroid for the 10 ENS runs (gray dots), the DSRAP (red dot), the CSUWRF (green dot), and the DSNAM (blue dot) plotted against the 200–750-m mean water vapor mixing ratio (g kg−1) at the MP4 location at 0100 UTC. The gray dashed line is a best-fit line for northeastward displacements vs water vapor mixing ratios from the ENS runs; and p values for these lines are shown in the legend.

Given that connection between data and northeastward MCS placement shown in the simulations, and the fact that observations were less convectively inhibited than the simulations, it is reasonable to infer that the models’ collective erroneous northeastward displacement of the MCS and rainfall locations relative from observations were ultimately a result of the models’ collective underrepresentation of low-level moisture and subsequent overrepresentation of low-level .

4. Discussion

This research provides valuable observational support for a series of recent studies that have shown that the behavior of MCSs simulated in idealized environments is highly sensitive to low-level moisture (e.g., Schumacher 2015b; Schumacher and Peters 2017). Here, we show that the magnitudes of moisture differences investigated by those authors in idealized simulations are of similar magnitude to moisture errors between weather forecast models and observation. Furthermore, these moisture errors lead to MCS forecast errors that are consistent with the variability among idealized MCSs simulated by those authors. Moisture errors are quite problematic in that they may easily “slip under the radar” when model forecasts are initialized and run. Moisture is a relatively weak dynamic constraint on an atmosphere dominated by hydrostatic and gradient wind balance. If we consider the hydrostatic equation , where is perturbation hydrostatic pressure, , and all other terms retain their traditional meanings. A variation in of 4 g kg−1 over a depth of 1 km leads to a pressure variation of order 10−1 Pa. In contrast, a variation of 4 K over the same depth leads to a pressure variation of order 100 Pa. A given balanced thermodynamic and kinematic environment may therefore correspond to a wide range of moisture distributions, since this moisture variation has little impact on the pressure and wind field. If we consider, on the other hand, the impact of moisture variations on thermodynamic potential energy (e.g., MSE), the same and variations lead to changes in MSE of order and 5 × 103, respectively, meaning that has a comparatively larger impact on thermodynamic potential energy than . Since the development and behavior of deep moist convection is strongly dependent on thermodynamic potential energy, moisture errors whose impact on the large-scale atmosphere is not otherwise apparent, may have a dramatic impact on how convection behaves.

Previous numerical modeling experiments have shown that horizontal advection of moisture and temperature can lead to the reduction of CIN in environments such as the one we have analyzed here (e.g., Trier et al. 2014a,b), however, the potential for such dramatic increases in CAPE of an elevated layer at night has seldom been discussed in the literature. It is troubling that best-guess analysis datasets such as the RAP did not adequately capture this destabilization, particularly considering that short-range human forecasts of the maintenance of an MCS often make use of the RAP analyzed environment ahead of an existing MCS [e.g., the Storm Prediction Center mesoanalysis MCS maintenance parameter; Coniglio et al. (2007)]. Forecasters that are relying on incorrect data within the RAP analysis to forecast the short-term behavior of an MCS may underestimate the probability of the MCS maintaining itself, or predict its direction of motion.

An obvious factor that is left indeterminate in this study is the origin of the moisture errors in the models, and future research will aim to determine the source of moisture errors. Another logical next step is to determine whether such moisture errors are ubiquitous among the environments of elevated MCSs. Displacement biases of MCSs are common among NWP models (e.g., Yost 2012), and moisture errors are a potential candidate for the cause of such errors. Multiple MCS cases were observed during the PECAN field campaign, and these observations are an ideal initial dataset for the investigation of moisture errors in other MCS events. Do errors such as the ones explored here influence other modes of convection, such as the positioning of surface-based convection, or the placement of orographically forced precipitation? What processes within NWP models are responsible for such errors, and can these biases be corrected? Plans are ongoing for the investigation of the aforementioned research avenues.

5. Summary

This article uses observations of the environment near and within an elevated MCS that occurred on 24 June 2015 during the PECAN field campaign to identify processes that lead to forecast errors of the event by numerical weather prediction models. This MCS featured a southeastward-moving squall line on its eastern flank and a quasi-stationary training convective line on its western flank, which are frequent characteristics of the training line–adjoining stratiform (TL/AS) MCS archetype (Schumacher and Johnson 2005; Peters and Schumacher 2014). A total of 16 radiosondes were launched at a high temporal frequency prior to, during, and after the passage of the squall line on the eastern flank of the MCS, and 4 radiosondes observed the environment near the quasi-stationary western flank of the MCS.

Substantial increases in temperature and moisture between 0.25 and 1 km AGL occurred within the environment ahead of the southeast flank of the MCS prior to its arrival at the observing location. These temperature and moisture increases resulted in a dramatic increase in CAPE within this layer from a maximum of 2400 J kg−1 to a maximum of nearly 4500 J kg−1, and a corresponding decrease in CIN within this layer. CAPE was underrepresented and CIN was overrepresented by the RAP analysis of the event due to underpredictions of moisture by the RAP of nearly 4 g kg−1 within the 0.25–1 km AGL layer. The magnitude of CAPE on the western flank of the MCS was observed to be similar to that on the eastern MCS flank; however, lifting along the MCS’s southwestern OFB was insufficient for air parcels to reach their LFCs. These parcels continued to travel toward the northeast over the low-level cold pool, having entered convective updrafts approximately 100 km away from the surface OFB.

Numerical simulations of the MCS that produced errors in the position and evolution of the MCS were analyzed and compared to observations to elucidate the source of the errors in these models. We determined that underpredictions of low-level atmospheric moisture (which resulted in overrepresentation of and underrepresentation of CAPE) were primarily responsible for model errors. The distance parcels needed to be lifted to reach their LFCs that regulated the residence time required for parcels within the region of meso-α-scale lift to achieve CI, whereby parcels with higher initial LFCs required longer residence times (and thus greater vertical lifting) than parcels with lower initial LFCs. In a regime characterized by southwesterly low-level flow, the length of residence times of parcels within meso-α-scale lifting corresponded to the distance of northeastward travel of air prior to CI, and low therefore equated to a small northeastward MCS position error, and high equated to a large northeastward MCS position error (Fig. 18). This conclusion is further supported by an analysis of a 10-member ensemble of convection-allowing simulations of the event, where low-level moisture southwest of the MCS was strongly negatively correlated with northeastward errors in the MCS placement. Variations in low-level moisture along the southeastern flank of the MCS were also connected to the strength of convection along the systems eastward-moving flank in simulations. High values of low-level moisture (and correspondingly low ) resulted in a strong eastward-moving squall line with convection closely following the eastern outflow boundary, whereas comparatively low values of low-level moisture (and correspondingly high ) resulted in weaker convection that lagged behind the eastern outflow boundary.

Fig. 18.

A modified version of the diagram shown in Fig. 22 of Trier and Parsons (1993) that shows how differences in parcel LFC heights lead to differing positions of the MCS among model simulations and observations. Dashed gray and orange lines are streamlines (also isentropes for nearly steady conditions), and the dashed blue line is the LFC height at the left side of the diagram for the parcels traveling along the dashed orange line. (a) Parcels along the orange dashed line are lifted to their LFCs near a preexisting surface OFB, as in what occurred in observations and the DSNAM. (b) Parcels with a higher initial LFC height than those in (a) are lifted to their LFC at some point beyond the surface OFB, as in the CSUWRF. (c) Parcels with higher initial LFC heights than (a) and (b) are lifted to their LFC farther to the right than both OBS and the CSUWRF, as in the DSRAP.

Fig. 18.

A modified version of the diagram shown in Fig. 22 of Trier and Parsons (1993) that shows how differences in parcel LFC heights lead to differing positions of the MCS among model simulations and observations. Dashed gray and orange lines are streamlines (also isentropes for nearly steady conditions), and the dashed blue line is the LFC height at the left side of the diagram for the parcels traveling along the dashed orange line. (a) Parcels along the orange dashed line are lifted to their LFCs near a preexisting surface OFB, as in what occurred in observations and the DSNAM. (b) Parcels with a higher initial LFC height than those in (a) are lifted to their LFC at some point beyond the surface OFB, as in the CSUWRF. (c) Parcels with higher initial LFC heights than (a) and (b) are lifted to their LFC farther to the right than both OBS and the CSUWRF, as in the DSRAP.

Acknowledgments

The writing and publishing of this article was supported by NSF Awards AGS-PRF 1524435 and AGS-1359727. The mobile upsonde systems that collected the radiosonde data used in this article were supported by the NSF Award AGS-1359726. We thank the NOAA/National Severe Storms Laboratory (NSSL) for its role in providing and configuring the two NSSL mobile sounding vehicles and the NOAA/Office of Atmospheric Research (OAR) grant, which provided all mobile sounding expendables used by the CSU and two NSSL sounding vehicles during PECAN. Special thanks go out to Mike Coniglio, Conrad Ziegler, Clark Evans, Gregory Herman, Stan Trier, and two anonymous reviewers for helpful conversations and feedback. High-performance computing resources from Yellowstone (ark:/85065/d7wd3xhc) and the stage-IV analysis were provided by the National Center for Atmospheric Research’s (NCAR) Computational and Information Systems Laboratory (NSSL), which is sponsored by the National Science Foundation. Finally, thanks to the NSSL, North Carolina State University, and Colorado State University mobile sounding teams for collecting the data used in this study.

REFERENCES

REFERENCES
Benjamin
,
S. G.
, and Coauthors
,
2016
:
A North American hourly assimilation and model forecast cycle: The Rapid Refresh
.
Mon. Wea. Rev.
,
144
,
1669
1694
, doi:.
Brooks
,
H. E.
,
J. W.
Lee
, and
J. P.
Craven
,
2003
:
The spatial distribution of severe thunderstorm and tornado environments from global reanalysis data
.
Atmos. Res.
,
67–68
,
73
94
, doi:.
Bryan
,
G. H.
, and
M. D.
Parker
,
2010
:
Observations of a squall line and its near environment using high-frequency rawinsonde launches during VORTEX2
.
Mon. Wea. Rev.
,
138
,
4076
4097
, doi:.
Bryan
,
G. H.
, and
R.
Rotunno
,
2014
:
Gravity currents in confined channels with environmental shear
.
J. Atmos. Sci.
,
71
,
1121
1142
, doi:.
Chen
,
F.
, and
J.
Dudhia
,
2001
:
Coupling an advanced land surface-hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity
.
Mon. Wea. Rev.
,
129
,
569
585
, doi:.
Coniglio
,
M. C.
,
H. E.
Brooks
,
S. J.
Weiss
, and
S. F.
Corfidi
,
2007
:
Forecasting the maintenance of quasi-linear mesoscale convective systems
.
Wea. Forecasting
,
22
,
556
570
, doi:.
Corfidi
,
S. F.
,
S. J.
Corfidi
, and
D. M.
Schultz
,
2008
:
Elevated convection and castellanus: Ambiguities, significance, and questions
.
Wea. Forecasting
,
23
,
1280
1303
, doi:.
Crook
,
N. A.
, and
M. W.
Moncrieff
,
1988
:
The effect of large-scale convergence on the generation and maintenance of deep moist convection
.
J. Atmos. Sci.
,
45
,
3606
3624
, doi:.
Davenport
,
C. E.
, and
M. D.
Parker
,
2015
:
Observations of the 9 June 2009 dissipating supercell from VORTEX2
.
Wea. Forecasting
,
30
,
368
388
, doi:.
Edwards
,
R.
, and
R. L.
Thompson
,
1998
:
Nationwide comparisons of hail size with WSR-88D vertically integrated liquid water and derived thermodynamic sounding data
.
Wea. Forecasting
,
13
,
277
285
, doi:.
Fritsch
,
J. M.
,
R. J.
Kane
, and
C. R.
Chelius
,
1986
:
The contribution of mesoscale convective weather systems to the warm-season precipitation in the United States
.
J. Climate Appl. Meteor.
,
25
,
1333
1345
, doi:.
Geerts
,
B.
, and Coauthors
,
2016
:
The 2015 Plains Elevated Convection at Night (PECAN) field project
.
Bull. Amer. Meteor. Soc.
,
98
,
767
786
, doi:.
Grell
,
G. A.
,
1993
:
Prognostic evaluations of the assumptions used by cumulus parameterizations
.
Mon. Wea. Rev.
,
121
,
764
787
, doi:.
Hamill
,
T. M.
,
G. T.
Bates
,
J. S.
Whitaker
,
D. R.
Murray
,
M.
Fiorino
,
T. J.
Galarneau
,
Y.
Zhu
, and
W.
Lapenta
,
2013
:
NOAA’s second-generation global medium-range ensemble reforecast dataset
.
Bull. Amer. Meteor. Soc.
,
94
,
1553
1565
, doi:.
Hong
,
S.-Y.
,
Y.
Noh
, and
J.
Dudhia
,
2006
:
A new vertical diffusion package with an explicit treatment of entrainment processes
.
Mon. Wea. Rev.
,
134
,
2318
2341
, doi:.
Iacono
,
M. J.
,
J. S.
Delamere
,
E. J.
Mlawer
,
M. W.
Shephard
,
S. A.
Clough
, and
W. D.
Collins
,
2008
:
Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models
.
J. Geophys. Res.
,
113
,
D13103
, doi:.
Janjić
,
Z. I.
,
1994
:
The step-mountain eta coordinate model: Further developments of the convection, viscous sublayer, and turbulence closure schemes
.
Mon. Wea. Rev.
,
122
,
927
945
, doi:.
Jewell
,
R.
, and
J.
Brimelow
,
2009
:
Evaluation of Alberta hail growth model using severe hail proximity soundings from the United States
.
Wea. Forecasting
,
24
,
1592
1609
, doi:.
Jirak
,
I. L.
, and
W. R.
Cotton
,
2007
:
Observational analysis of the predictability of mesoscale convective systems
.
Wea. Forecasting
,
22
,
813
838
, doi:.
Keene
,
K. M.
, and
R. S.
Schumacher
,
2013
:
The bow and arrow mesoscale convective structure
.
Mon. Wea. Rev.
,
141
,
1648
1672
, doi:.
Klemp
,
J. P.
,
W. C.
Skamarock
, and
J.
Dudhia
,
2007
:
Conservative split-explicit time integration methods for the compressible nonhydrostatic equations
.
Mon. Wea. Rev.
,
135
,
2897
2913
, doi:.
Lane
,
T. P.
, and
M. W.
Moncrieff
,
2015
:
Long-lived mesoscale systems in a low-convective inhibition environment. Part I: Upshear propagation
.
J. Atmos. Sci.
,
72
,
4297
4318
, doi:.
Mahoney
,
K. M.
,
G. L.
Lackmann
, and
M. D.
Parker
,
2009
:
The role of momentum transport in the motion of a quasi-idealized mesoscale convective system
.
Mon. Wea. Rev.
,
137
,
3316
3338
, doi:.
Melhauser
,
C.
, and
F.
Zhang
,
2012
:
Practical and intrinsic predictability of severe and convective weather at the mesoscales
.
J. Atmos. Sci.
,
69
,
3350
3371
, doi:.
Mitchell
,
K. E.
, and Coauthors
,
2004
:
The multi-institution North American Land Data Assimilation System (NLDAS): Utilizing multiple GCIP products and partners in a continental distributed hydrological modeling system
.
J. Geophys. Res.
,
109
,
D07S90
, doi:.
Mlawer
,
E. J.
,
S. J.
Taubman
,
P. D.
Brown
,
M. J.
Iacono
, and
S. A.
Clough
,
1997
:
Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave
.
J. Geophys. Res.
,
102
,
16 663
16 682
, doi:.
Moncrieff
,
M. W.
, and
C.
Liu
,
1999
:
Convection initiation by density currents: Role of convergence, shear, and dynamical organization
.
Mon. Wea. Rev.
,
127
,
2455
2464
, doi:.
Moore
,
J. T.
,
F. H.
Glass
,
C. E.
Graves
,
S. M.
Rochette
, and
M. J.
Singer
,
2003
:
The environment of warm-season elevated thunderstorms associated with heavy rainfall over the central United States
.
Wea. Forecasting
,
18
,
861
878
, doi:.
Morrison
,
H.
,
J. A.
Curry
, and
V. I.
Khvorostyanov
,
2005
:
A new double-moment microphysics parameterization for application in cloud and climate models. Part I: Description
.
J. Atmos. Sci.
,
62
,
1665
1677
, doi:.
Morrison
,
H.
,
G.
Thompson
, and
V.
Tatarskii
,
2009
:
Impact of cloud microphysics on the development of trailing stratiform precipitation in a simulated squall line: Comparison of one- and two-moment schemes
.
Mon. Wea. Rev.
,
137
,
991
1007
, doi:.
Nielsen
,
E. R.
,
2016
: Using convection-allowing ensembles to understand the predictability of extreme rainfall. M.S. thesis, Department of Atmospheric Sciences, Colorado State University, Fort Collins, CO, 144 pp.
Nielsen
,
E. R.
, and
R. S.
Schumacher
,
2016
:
Using convection-allowing ensembles to understand the predictability of an extreme rainfall event
.
Mon. Wea. Rev.
,
144
,
3651
3676
, doi:.
Pan
,
Y.
,
K.
Zhu
,
M.
Xue
,
X.
Wang
,
M.
Hu
,
S. G.
Benjamin
,
S. S.
Weygandt
, and
J. S.
Whitaker
,
2014
:
A GIS-based coupled EnSRF-En3DVar hybrid data assimilation system for the operational Rapid Refresh model: Tests at a reduced resolution
.
Mon. Wea. Rev.
,
142
,
3756
3780
, doi:.
Parker
,
M. D.
, and
R. H.
Johnson
,
2004
:
Structures and dynamics of quasi-2D mesoscale convective systems
.
J. Atmos. Sci.
,
61
,
545
567
, doi:.
Peters
,
J. M.
, and
R. S.
Schumacher
,
2014
:
Objective categorization of heavy-rain-producing MCS synoptic types by rotated principal component analysis
.
Mon. Wea. Rev.
,
142
,
1716
1737
, doi:.
Peters
,
J. M.
, and
R. S.
Schumacher
,
2015a
:
Mechanisms for organization and echo training in a flash-flood-producing mesoscale convective system
.
Mon. Wea. Rev.
,
143
,
1058
1085
, doi:.
Peters
,
J. M.
, and
R. S.
Schumacher
,
2015b
:
The simulated structure and evolution of a quasi-idealized warm-season convective system with a training convective line
.
J. Atmos. Sci.
,
72
,
1987
2010
, doi:.
Peters
,
J. M.
, and
R. S.
Schumacher
,
2016
:
Dynamics governing a simulated mesoscale convective system with a training convective line
.
J. Atmos. Sci.
,
73
,
2643
2664
, doi:.
Rotunno
,
R.
,
J. B.
Klemp
, and
M. L.
Weisman
,
1988
:
A theory for strong, long-lived squall lines
.
J. Atmos. Sci.
,
45
,
463
485
, doi:.
Schumacher
,
R. S.
,
2015a
:
Resolution dependence of initiation and upscale growth of deep convection in convection-allowing forecasts of the 31 May–1 June 2013 supercell and MCS
.
Mon. Wea. Rev.
,
143
,
4331
4354
, doi:.
Schumacher
,
R. S.
,
2015b
:
Sensitivity of precipitation accumulation in elevated convective systems to small changes in low-level moisture
.
J. Atmos. Sci.
,
72
,
2507
2524
, doi:.
Schumacher
,
R. S.
, and
R. H.
Johnson
,
2005
:
Organization and environmental properties of extreme-rain-producing mesoscale convective systems
.
Mon. Wea. Rev.
,
133
,
961
976
, doi:.
Schumacher
,
R. S.
, and
J. M.
Peters
,
2017
:
Near-surface thermodynamic sensitivities in simulated extreme-rain-producing mesoscale convective systems
.
Mon. Wea. Rev.
,
145
,
2177
2200
, doi:.
Skamarock
,
W. C.
, and
J. B.
Klemp
,
2008
:
A time-split nonhydrostatic atmospheric model for weather research and forecasting applications
.
J. Comput. Phys.
,
227
,
3465
3485
, doi:.
Skamarock
,
W. C.
, and Coauthors
,
2008
: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., doi:.
Smith
,
B. T.
,
R. L.
Thompson
,
J. S.
Grams
,
C.
Broyles
, and
H. E.
Brooks
,
2012
:
Convective modes for significant severe thunderstorms in the contiguous United States. Part I: Storm classification and climatology
.
Wea. Forecasting
,
27
,
1114
1135
, doi:.
Thompson
,
G.
,
P. R.
Field
,
R. M.
Rasmussen
, and
W. D.
Hall
,
2008
:
Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization
.
Mon. Wea. Rev.
,
136
,
5095
5115
, doi:.
Thompson
,
R. L.
,
C. M.
Mead
, and
R.
Edwards
,
2007
:
Effective storm-relative helicity and bulk shear in supercell thunderstorm environments
.
Wea. Forecasting
,
22
,
102
115
, doi:.
Thompson
,
R. L.
,
B. T.
Smith
,
J. S.
Grams
,
A. R.
Dean
, and
C.
Broyles
,
2012
:
Convective modes for significant severe thunderstorms in the contiguous United States. Part II: Supercell and QLCS tornado environments
.
Wea. Forecasting
,
27
,
1136
1154
, doi:.
Trier
,
S. B.
, and
D. B.
Parsons
,
1993
:
Evolution of environmental conditions preceding the development of a nocturnal mesoscale convective complex
.
Mon. Wea. Rev.
,
121
,
1078
1098
, doi:.
Trier
,
S. B.
,
C. A.
Davis
, and
D. A.
Ahijevych
,
2010
:
Environmental controls on the simulated diurnal cycle of warm-season precipitation in the continental United States
.
J. Atmos. Sci.
,
67
,
1066
1090
, doi:.
Trier
,
S. B.
,
J. H.
Marsham
,
C. A.
Davis
, and
D. A.
Ahijevych
,
2011
:
Numerical simulations of the postsunrise reorganization of a nocturnal mesoscale convective system during 13 June IHOP_2002
.
J. Atmos. Sci.
,
68
,
2988
3011
, doi:.
Trier
,
S. B.
,
C. A.
Davis
,
D. A.
Ahijevych
, and
K. W.
Manning
,
2014a
:
Use of the parcel buoyancy minimum (Bmin) to diagnose simulated thermodynamic destabilization. Part I: Methodology and case studies of MCS initiation environments
.
Mon. Wea. Rev.
,
142
,
945
966
, doi:.
Trier
,
S. B.
,
C. A.
Davis
,
D. A.
Ahijevych
, and
K. W.
Manning
,
2014b
:
Use of the parcel buoyancy minimum (Bmin) to diagnose simulated thermodynamic destabilization. Part II: Composite analysis of mature MCS environments
.
Mon. Wea. Rev.
,
142
,
967
990
, doi:.
UCAR/NCAR Earth Observing Laboratory
,
2016
: MP4 NCAR/EOL QC soundings, version 2.0. UCAR/NCAR Earth Observing Laboratory, accessed 1 July 2016, doi:.
Vaisala
,
2013
: Vaisala radiosonde RS92 performance in the WMO intercomparison of high quality radiosonde systems. Vaisala, 17 pp. [Available online at http://www.vaisala.com/Vaisala%20Documents/White%20Papers/WEA-MET-WMO-Test-White_Paper-B211129EN-D-LOW.pdf.]
Weisman
,
M. L.
, and
R.
Rotunno
,
2004
:
“A theory of strong, long-lived squall lines” revisited
.
J. Atmos. Sci.
,
61
,
361
382
, doi:.
Yost
,
C.
,
2012
: Investigation into a displacement bias in numerical weather prediction models’ forecasts of mesoscale convective systems. M.S. thesis, Department of Atmospheric Science, Colorado State University, 103 pp.
Zhang
,
F.
,
C.
Snyder
, and
R.
Rotunno
,
2003
:
Effects of moist convection on mesoscale predictability
.
J. Atmos. Sci.
,
60
,
1173
1185
, doi:.
Zhu
,
K.
,
Y.
Pan
,
M.
Xue
,
X.
Wang
,
J. S.
Whitaker
,
S. G.
Benjamin
,
S. S.
Weygandt
, and
M.
Hu
,
2013
:
A regional GIS-based ensemble Kalman filter data assimilation system for the Rapid Refresh configuration: Testing at reduced resolution
.
Mon. Wea. Rev.
,
141
,
4118
4139
, doi:.
Ziegler
,
C.
,
M. C.
Coniglio
,
M.
Parker
, and
R.
Schumacher
,
2016
: CSU/NCSU/NSSL MGAUS radiosonde data, version 2.0. UCAR/NCAR Earth Observing Laboratory, accessed 1 July 2016, doi:.

Footnotes

This article is included in the Plains Elevated Convection At Night (PECAN) Special Collection.

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

1

The MUTs consisted of three vehicles equipped to launch Vaisala radiosondes.

2

This analysis did not assimilate PECAN observations.

3

In this context, we are referring to the general scenario of outflow propagating downshear, rather than the specific scenario discussed by those authors, where the ratio of vertically integrated buoyancy to vertical wind shear is unity.

4

In this context, CI refers to the onset of a particular air parcel’s deep convective overturning, and does not imply the absence of prior convection.