5.1 THE IMPACT OF PARAMETERIZED SHALLOW ON PRE-DEEP-CONVECTIVE SOUNDING STRUCTURES IN THE ETA MODEL John S. Kain*,1,2, Michael E. Baldwin1,2,3, Steven J. Weiss3, and Michael P. Kay4,5

1Cooperative Institute for Mesoscale Meteorological Studies 2University of Oklahoma/NOAA Research - National Severe Storms Laboratory 3NOAA/NWS/Storm Prediction Center 4Cooperative Institute for Research in Environmental Sciences (CIRES) 5University of Colorado/NOAA Research – Forecast Systems Laboratory

1. INTRODUCTION considerably if forecasters learned to recognize when the convective scheme has been active in the model and Convective parameterization is a necessary compo- how it has modified thermodynamic profiles. nent of mesoscale and larger scale models. An impor- Considering the important role that model soundings tant function of convective schemes is to generate have come to play in the forecast preparation process in unstable model environments before sat- and the enduring prominence of the Eta Model, a uration occurs at individual grid points. This helps mod- detailed examination of the characteristic structures els to predict the timing of convective initiation more associated with the BMJ scheme is warranted. These accurately, an obvious benefit for forecasters. However, structures are quite distinctive and the trained eye can generation of precipitation is not the most important role often recognize the “signature” of BMJ activity quite of parameterized convection. The more significant func- readily. The purpose of this paper is to provide guidance tion is to modify convective instability and redistribute for forecasters in recognizing the characteristic impact of moisture in model soundings. By stabilizing vertical - the BMJ scheme on model soundings, providing a set of umns before saturation occurs over a deep layer, con- skills that will allow them to make more insightful judg- vective parameterizations act to prevent potentially ments about model predictions. explosive and unrealistic growth of small-scale distur- This preprint is a condensed version of a full article bances, also known as “numerical point storms” (Lilly that has been accepted for publication in Weather and 1960; Rosenthal 1979; Molinari and Dudek 1986; Giorgi Forecasting and is also available on the web (Baldwin et 1991). A properly formulated convective parameteriza- al. 2002). The full article describes the algorithm used tion can suppress these unrealistic features and play an by the BMJ scheme in some detail and provides several important role in generating accurate quantitative precip- examples of the impact of the BMJ scheme on model- itation forecasts (QPF). sounding evolution. In this paper, we show one of these The operational convective parameterization in the examples and provide a discussion and summary. At the Eta Model (Black 1994) is the Betts-Miller-Janjic scheme conference, several recent cases will be highlighted as (Betts 1986; Betts and Miller 1986; Janjic 1994; hereafter we continue to monitor the impact of the BMJ scheme. BMJ). Verification of QPF from this model has been favorable ever since it was introduced (Mesinger 1996), 2. A BRIEF OVERVIEW OF THE BMJ SCHEME and the BMJ scheme deserves a lot of credit for this level of performance, particularly for warm season fore- The BMJ parameterization is a convective adjust- casts. However, certain aspects of this scheme, while ment scheme, meaning that it determines “reference” designed to produce the best possible QPF, may gener- profiles of and dewpoint towards which it ate artificial structures in vertical profiles of temperature nudges the model soundings at individual grid points. and , i.e., model-forecast soundings (Manikin et The first step in the scheme is to locate the most unsta- θ al. 2000). ble (highest e) model parcel within the lowest ~200 mb This can be problematic when these soundings are above the ground. It “lifts” this parcel to its LCL (lifting used to forecast certain elements of the weather. For condensation level), which it defines as base. example, forecasters frequently examine model sound- From there, the parcel is lifted moist adiabatically until ings to evaluate the potential for convective activity. the (EL) is reached. Cloud top is then Forecast soundings from the Eta Model are useful for defined as the highest model level at which the parcel is predicting convection, but the BMJ scheme can mask still buoyant, typically just below the EL. If the parcel is important details of the vertical structure and affect cal- not buoyant at any level, convection is not activated and culations of (CIN), Convective the scheme moves on to the next grid column. If the Available Potential Energy (CAPE), and other parame- “cloud” is less than 200 mb deep, the scheme attempts ters used in the forecast preparation process (Hart et al. to initiate shallow (non-precipitating) convection. Other- 1998). The utility of the soundings could be enhanced wise, it checks to see if deep (precipitating) convection can be activated. As discussed in Baldwin et al. (2002), the scheme often reverts to shallow convection even * Corresponding author address: Jack Kain, NSSL, 1313 Hal- ley Circle, Norman, OK 73069 e-mail: [email protected] when initial estimates of cloud depth are greater than 200 mb. This often occurs when the convectively unsta- front. This front extended along the Appalachians into ble layer is relatively dry. In the case discussed below, the southeastern U.S. The front was advancing slowly the scheme begins by parameterizing the effects of shal- eastward over the Mid-Atlantic States, into a region of low convection, but transitions to deep convection as the moderate instability over the Carolinas and Virginia. The environment moistens. upper level forcing associated with this front was weak. The model initial (1200 UTC) condition at Greens- 3. A CASE OF SHALLOW CONVECTION TRANSI- boro, North Carolina (GSO) showed good agreement TIONING TO DEEP CONVECTION with observations (Fig. 1a). The sounding was nearly saturated above about 300 mb, relatively dry through On 24 April 2001, a vigorous upper level short wave mid levels, and quite moist below 800 mb. The BMJ trough was lifting northeastward across the Great Lakes scheme determined that no CAPE existed for parcels toward the northeastern states. An associated deep sur- rooted in the lowest 200 mb, so no convection was acti- face low over Quebec was also moving to the northeast. vated at this time. CAPE continued to be absent through Trailing southwestward from this low was a surface cold 1400 UTC, but solar heating led to the development of

a d

e

f

b

c

Fig. 11. A sequence of Eta model forecast soundings valid from 1200 UTC to 2200 UTC 24 April 2001 over Greensboro, NC (KGSO). a) Model initial condition (thick dark curves) overlaid on the observed sounding (thick light curves); b) BMJ shallow- convection reference profiles (thick dark curves) overlaid on the model 3 h forecast, valid 1500 UTC; c) as in b), but for the 5 h forecast, valid 1700 UTC; d) as in b), but for the 6 h forecast, valid 1800 UTC; e) as in b), but for the 8 h forecast, valid 2000 UTC; f) as in b), but for the 10 h forecast, valid 2200 UTC. some instability during the following hour. profile is subsaturated with a dewpoint depression that During this hour, the scheme activated shallow con- varies linearly from about 3o C at cloud base, to ~7 o C at vective feedbacks, nudging the environment toward the the freezing level, and back to ~4 o C at cloud top. reference profiles shown in Fig. 1b. Note the distinctive character of the shallow convective reference profiles. 4. SUMMARY AND DISCUSSION The temperature profile appears as a nearly straight line on a skew-T/log P diagram, while the dewpoint profile All convective parameterizations contain arbitrary has a convex shape, tailing off to sharply lower dew- parameter settings and have characteristic behaviors points at higher levels (i.e., at the top of the shallow that are sometimes inconsistent with reality. So, this cloud layer, as determined by the BMJ scheme). Even study is not intended to single out the BMJ scheme as though parameterized shallow convection had been somehow inferior or inadequate. On the contrary, this active for less than an hour by 1500 UTC, the scheme scheme has been critically important to the success of had clearly placed its footprint on the sounding. Specifi- the Eta Model running at the National Centers for Envi- cally, the scheme had significantly cooled and moistened ronmental Prediction (NCEP) and its enduring status as the layer near 750 mb, introduced a monotonic decrease the primary 1-2 day operational forecast model in the in temperature from about 920 to 740 mb and imposed a United States. Other convective parameterizations have smooth convex moisture profile over the same layer (Fig. been tested in the Eta Model but none has produced 1b). At this time the scheme was acting over a depth of consistently higher QPF verification scores than the BMJ 200 mb, which is the maximum allowable depth the shal- scheme. low convective adjustment. At the same time, the BMJ scheme is particularly The character of the scheme’s influence changed lit- amenable to critical examination because the shape and tle over the next 2 hours (1500 – 1700 UTC). In particu- character of its “footprint” are much easier to identify lar, thermodynamic profiles within the shallow-cloud than characteristic profiles produced by other schemes – layer retained the same shape, although solar heating and there is value in knowing when and where a convec- caused cloud base to rise with time and this allowed tive parameterization has been active in a model! Most cloud top to rise as well (Fig. 1c). By 1800 UTC, low- importantly, a detailed examination of BMJ behaviors is level convergence associated with the advancing cold warranted because model-forecast soundings from the front was quite strong (not shown) and boundary layer Eta Model have come to play an important role in prepar- depth was increasing rapidly. Boundary layer moisture ing forecasts for many types of weather. was not decreasing significantly as the layer deepened, Forecasters concerned about devel- suggesting the presence of strong moisture conver- opment can benefit from knowing how to identify when gence. At the same time, there appeared to be little if parameterized shallow convection has been active. The any upward motion aloft. BMJ shallow component warms and dries model layers Since the computed shallow cloud base was near near the LCL while cooling and moistening the model the top of the boundary layer, the scheme effectively sounding near the computed cloud top. It nudges the transported moisture from the top of the well-mixed layer environment towards profiles characterized by linear θ (or just above the top) into the lower to middle part of the decreases in and qv as a function of decreasing pres- shallow cloud layer (Fig. 1d). Note the substantial moist- sure. Oftentimes this process distorts the shape and ening of the shallow cloud layer between 1700 and 1800 structure of the CIN layer, sometimes completely elimi- UTC (cf. Figs. 1c and d). At the same time, shallow con- nating a stable layer that can be critical for inhibiting con- vection was effectively communicating with the convec- vective development. This tendency can be very evident tive boundary layer (through turbulent diffusion), over the Great Plains of the U.S., where elevated mixed removing moisture from low-levels. This communication layers create strong capping inversions with some regu- between parameterized boundary layer turbulence and larity during the warm season (Carlson et al. 1983; Lan- shallow convection is frequently reflected in Eta Model icci and Warner 1991). When the LCL is close to the top simulations. of a convective boundary layer, within which turbulent Between 1800 and 2000 UTC, boundary layer depth mixing is parameterized separately in the model, the began to peak. Parameterized shallow convection con- combination of convection and turbulence parameteriza- tinued to moisten the cloud layer and cool its upper half. tions can effectively mix moisture out of the boundary Yet, above about 650 mb, the remained dry, layer up towards the shallow cloud top while mixing high precluding the development of parameterized deep con- θ air downward into the boundary layer. When some or vection (Fig. 1e). After 2000 UTC, upward motion com- all of these processes are active in the model, convec- menced aloft within the model and BMJ deep convection tive parameters such as CAPE and CIN are significantly activated between 2100 and 2200 UTC. By the latter impacted. If forecasters can learn to identify these char- time, the remnants of the shallow convective structures acteristic tendencies associated with BMJ shallow con- were still evident, but it can be seen that the sounding vection, they can make more informed assessments of was evolving toward the familiar BMJ deep convective the likelihood of convective initiation and intensity. reference profiles (Fig. 1f), where the temperature profile It is also important for forecasters to be able to rec- is slightly less stable than moist adiabatic, and moisture ognize when BMJ deep convection has been active. Unlike parameterized shallow convection, deep convec- REFERENCES tive activity is easy to confirm by examining the convec- tive rainfall field. Its characteristic thermodynamic Baldwin, M. E., J. S. Kain, and M. P. Kay, 2002: Proper- ties of the convection scheme in NCEP’s Eta model profiles are easy to recognize as well. The temperature that affect forecast sounding interpretation. profile is slightly unstable from cloud base to the freezing Accepted for publication in Wea. Forecasting. [Also level, then marginally stable ( slightly less than available at http://www.nssl.noaa.gov/mag/ moist adiabatic) up to the computed cloud top. The dew- bmj_paper_all.pdf] point depression is specified to vary linearly from about Betts, A. K., 1986: A new convective adjustment 3-5o C at cloud base, to 7-9 o C at the freezing level, scheme. Part I: Observational and theoretical basis. back to 3-5 o C at cloud top. As with the shallow convec- Quart. J. Roy. Meteor. Soc., 112, 677-691. ______, and M. Miller, 1986: A new convective tive signature, the active convective layer is often first adjustment scheme. Part II: Single column tests recognizable by its characteristic lack of small-scale using GATE wave, BOMEX and arctic air-mass data structure. Small-scale vertical structures in both the tem- sets. Quart. J. Roy. Meteor. Soc., 112, 693-709. perature and moisture fields are transformed quickly into Black, T. L., 1994: The new NMC mesoscale Eta Model: curves with nearly monotonic decrease in temperature Description and forecast examples. Wea. and dewpoint with height when either deep or shallow Forecasting, 9, 265-278. Carlson, T. N., S. G. Benjamin, G. S. Forbes, Y.-F. Li, convection activates with the BMJ scheme. 1983: Elevated mixed layers in the regional severe It is hoped that this study will promote the direct storm environment: Conceptual model and case analysis of model-forecast soundings, rather than relying studies. Mon. Wea. Rev., 111, 1453–1474. on 2-D plan view plots of diagnosed quantities such as Doswell, C. A., III, and E. N. Rasmussen, 1994: The CAPE or CIN. Not only should this help in removing effect of neglecting the ambiguity on how such fields are computed (e.g., which correction on CAPE calculations. Wea. Forecasting, 9, parcel was lifted, whether or not the virtual temperature 625–629. correction (Doswell and Rasmussen 1994) was used, Giorgi, F., 1991: Sensitivity of simulated summertime etc.), but this will also lead to better understanding of the precipitation over the western United States to dif- characteristic behaviors of the BMJ scheme. Examina- ferent physics parameterization. Mon. Wea. Rev., tion of model-forecast soundings can provide valuable 119, 2780-2888. clues to help forecasters comprehend and interpret over- Hart, R. E., G. S. Forbes, and R. H. Grumm, 1998: The use of hourly model-generated soundings to fore- all model behavior. Yet, these soundings must always cast mesoscale phenomena. Part I: Initial assess- be used with caution. Developing a more complete ment in forecasting warm-season phenomena. Wea. understanding of the BMJ scheme can help forecasters Forecasting, 13, 1165–1185. distinguish between those characteristics of model Janjic, Z. I., 1994: The step-mountain eta coordinate soundings that have a meteorological origin and those model: Further developments of the convection, vis- that are more of a computational anomaly. cous sublayer, and turbulence closure schemes. Mon. Wea. Rev., 122, 927-945. Lanicci, J. M., T. T. Warner, 1991: A Synoptic Climatol- Acknowledgments ogy of the Elevated Mixed-Layer Inversion over the Southern Great Plains in Spring. Part I: Structure, This work was partially funded by a COMET Coop- Dynamics, and Seasonal Evolution. Wea. Forecast- erative Project, UCAR Award No. 099-15805, and ing, 6, 198–213. NOAA-University of Oklahoma Cooperative Agreement Lilly, D. K., 1960: On the theory of disturbances in a con- #NA17RJ1227. This manuscript was inspired by many ditionally unstable atmosphere. Mon. Wea. Rev., 88, 1-17. conversations on the topic of convective parameteriza- Manikin, G. and coauthors, 2000: Changes to the NCEP tion between the authors and numerous scientists and Meso Eta runs: Extended range, added input, added forecasters at SPC and NSSL, often held during daily output, convective changes. NWS Technical Proce- informal NSSL/SPC map discussions. In addition, the dures Bulletin No. 465, NOAA/NWS. Washington, authors acknowledge the COMET Program for their sup- DC. 85pp. [Available at http://www.nws.noaa.gov/ port, and especially for providing the first two authors om/tpb/465.htm and from National Weather Service, several opportunities to discuss this topic with NWS fore- Office of , 1325 East-West Highway, Sil- ver Spring, MD 20910] casters at their training symposia. We are also grateful Mesinger, F., 1996: Improvements in quantitative precipi- to Dr. Zavisa Janjic of NCEP/EMC for his comments and tation forecasts with the Eta regional model at the suggestions that helped greatly in clarifying and improv- National Centers for Environmental Prediction: The ing several aspects of the expanded version of this 48-km upgrade. Bull. Amer. Meteor. Soc., 77, 2637– paper that will appear in Weather and Forecasting. The 2650. authors would also like to thank Dr. Louis Wicker Molinari, J., and M. Dudek, 1986: Implicit versus explicit convective heating in numerical weather prediction (NSSL), Dr. Kimberly Elmore (CIMMS/NSSL), and three models. Mon. Wea. Rev., 114, 1822-1831. anonymous reviewers for their constructive comments Rosenthal, S. L., 1979: The sensitivity of simulated hur- and helpful suggestions on the expanded version of the ricane development to cumulus parameterization paper (see Baldwin et al. 2002 below). details. Mon. Wea. Rev., 107, 193-197.