Iowa State University Capstones, Theses and Graduate Theses and Dissertations Dissertations 2011 WRF simulations of mesoscale convective systems at convection-allowing resolutions Jeffrey Dean Duda Iowa State University Follow this and additional works at:https://lib.dr.iastate.edu/etd Part of theEarth Sciences Commons Recommended Citation Duda, Jeffrey Dean, "WRF simulations of mesoscale convective systems at convection-allowing resolutions" (2011).Graduate Theses and Dissertations. 10272. https://lib.dr.iastate.edu/etd/10272 This Thesis is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please [email protected]. WRF simulations of mesoscale convective systems at convection-allowing resolutions by Jeffrey Dean Duda A thesis submitted to the graduate faculty in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Major: Meteorology Program of Study Committee: William A. Gallus, Jr., Major Professor Raymond W. Arritt Xiaoqing Wu Iowa State University Ames, Iowa 2011 Copyright © Jeffrey Dean Duda, 2011. All rights reserved. ii TABLE OF CONTENTS ABSTRACT iv CHAPTER 1. GENERAL INTRODUCTION 1 1. Background and motivation 1 2. Thesis organization 3 CHAPTER 2. SENSITIVITIES OF WRF PRECIPITATION TO MODIFICATIONS TO THE KAIN-FRITSCH SCHEME 4 1. Abstract 4 2. Introduction 4 3. Methodology 7 4. Results 9 1. Sensitivity to the consistency between horizontal and vertical grid resolution 9 2. Sensitivity of the partition of precipitation between KF and microphysics schemes – the ramp zone 11 3. Sensitivity of simulated precipitation to modifications to the KF scheme 13 5. Summary and conclusions 17 6. Acknowledgements 19 7. References 20 8. Tables 24 9. Figures 27 CHAPTER 3. THE IMPACT OF LARGE-SCALE FORCING ON SKILL OF CONVECTIVE INITIATION AND UPSCALE EVOLUTION WITH CONVECTION-ALLOWING GRID SPACINGS IN THE WRF 40 1. Abstract 40 2. Introduction 41 3. Methodology 42 1. Case selection 42 2. Model configuration 43 3. Analysis of model output 44 4. Results and analysis 47 5. Case studies 52 1. Case 11 – 1200 UTC 22 May 2008 to 1200 UTC 23 May 2008 52 2. Case 7 – 1200 UTC 01 June 2010 to 1200 UTC 02 June 2010 56 3. Case 24, I – 1200 UTC 01 April 2006 to 1200 UTC 02 April 2006 59 4. Case 34 – 1200 UTC 17 June 2006 to 1200 UTC 18 June 2006 61 6. Discussion and conclusions 64 7. Acknowledgements 66 8. References 67 9. Tables 71 iii 10. Figures 75 CHAPTER 4. GENERAL CONCLUSIONS 134 1. Summary 134 2. Conclusions 134 3. Acknowledgements 136 4. References 136 iv ABSTRACT Mesoscale convective systems (MCS) were studied using both idealized and real data WRF simulations using grid spacings in the range from 0.5 km to 12 km with an emphasis on 3 km to determine the necessity of a convective parameterization scheme. The Kain-Fritsch (KF) convective parameterization scheme was used as it is considered to give the best forecasts of precipitation in mesoscale models. The idealized simulations were used to conduct three sets of sensitivity tests. One set tested the ability of the model to adequately resolve typical two-dimensional squall line structure by varying the vertical grid resolution. It was determined that using 81 vertical grid levels was sufficient to model squall lines. A second set tested the sensitivity of the partition of precipitation into microphysics and convective scheme components to horizontal grid spacing. A zone of grid spacing values from about 1 km to about 6 km was identified over which the partition shifts from approximately 10% of precipitation from the KF scheme to anywhere from 60% - 100%. This zone was found to be insensitive to microphysics scheme and somewhat sensitive to initial conditions. The amount of precipitation produced per activation of the deep convective part of the KF scheme was also found to shift significantly across this range of horizontal grid spacing values. The third set tested the sensitivity of precipitation forecasts to five treatments of the scheme. While two of the treatments included the lack of a convective scheme and the unmodified KF scheme, the other three treatments involved modifications to the scheme. These modifications included removing the linear dependence of grid-resolvable vertical velocity on grid spacing, coarsening the vertical motion, temperature, and water vapor mixing ratio fields before the KF scheme ran, and coarsening the heat and moisture tendencies as well as convective scheme precipitation after the KF scheme ran. When applied to a set of three-dimensional real data cases, it was found that the use of no convective scheme and the unmodified KF scheme generally performed the best. However, due to a small sample size, the spread of the data was large and more tests are needed. v The conclusions from the sensitivity tests were that the KF scheme becomes less active as grid spacing decreases below horizontal grid spacings of 6 – 8 km. Below 1 km grid spacing, the KF scheme certainly should not be used. However, from 1 km to 3 km, it likely is not necessary to use the KF scheme. Using it at 3 km does not hurt the forecast, however. As other research has shown, there is some use for the KF scheme above 3 km. Three-dimensional real data WRF simulations were conducted at 3 km horizontal resolution for a set of 39 cases involving MCSs across the United States. The KF scheme was not used. Convective initiation was found to err by approximately 150 km in the west- southwesterly direction with a nearly zero mean timing error. Large scatter was found between the strength of large-scale forcing and the model skill at forecasting initiation, but traditional skill measures (ETS and bias) showed that stronger-forced cases were better forecast in the upscale evolution of the MCSs. Case studies were performed for a few cases to illustrate the ways in which the WRF can succeed or fail to accurately predict convective initiation. 1 CHAPTER 1. GENERAL INTRODUCTION 1.1 Background and motivation Deep, moist convection is one way in which the atmosphere is brought from a gravitationally unstable thermodynamic stratification to a neutral stratification. Deep, moist convection occurs when an air parcel rises due to an upward buoyancy force. This buoyancy force exists because the air parcel is less dense than the air around it. As the parcel accelerates upward, it expands, cools, condenses, and releases liquid water. This causes clouds and precipitation to develop. The parcel continues to rise and condense liquid water until it reaches the tropopause where it becomes as dense as the surrounding air, thus remaining there in the anvil of the resulting thunderstorm. When many thunderstorms erupt close together and persist, latent heating and merging of cold pools from the individual thunderstorm downdrafts can result in an evolution of the convection from storm scale to mesoscale. Given enough time, a mesoscale convective system (MCS) is born. MCSs are primarily a warm-season entity in the United States since the amount of instability and moisture needed to form them is generally not present during the cold season. Warm-season precipitation is a challenge to predict (Olson et al. 1995; Fritsch and Carbone 2004). A large portion of the annual rainfall for areas of the central U. S. comes from MCSs (Fritsch et al. 1986). MCSs and the precursor convection, which frequently starts as individual storm cells (in some cases, supercells), also produce a large portion of severe weather (Doswell et al. 1996; Wheatley et al. 2006; Wakimoto et al. 2006; Gallus et al. 2008; Duda and Gallus 2010). Therefore, predicting MCSs is both a crucial and difficult activity. There are two key areas in which much of the difficulty in predicting MCSs lies: convective initiation and upscale evolution. The difficulty with these two areas has two sources. One source is related to the need to parameterize convection since models using a high enough resolution to resolve convection have only recently come about. Weisman et al. (1997) concluded that 4 km grid spacing is sufficient to not require a convective parameterization scheme to resolve deep convection and squall lines. Bryan et al. (2003) concluded that grid spacing of O(100 m) is needed to fully resolve convection. Research and operational NWP models now have the capability to run over a large region at grid spacings 2 of 1 to 4 km, but not at O(100 m). Thus, traditional convective parameterization is not requried, yet convection is not fully resolved. The technology required to run models that can fully resolve all aspects of convection over large regions will probably not be around for many years, so issues with not using a convective parameterization in high resolution models will continue. One issue is the need for grid-scale saturation to occur for a microphysics parameterization scheme to resolve convection. This is physically unrealistic at grid spacings of 3 to 4 km. Also it is difficult to bring a grid box with area of 9 or 16 km2 to saturation. However, since convective parameterization schemes were not designed to be used at 1 – 4 km resolution, either new or modified schemes will have to be implemented or the inadequacy in using unmodified existing schemes will need to be studied. The other source comes from the issues with observational data used to feed models. Observational data in the U. S. are generally too coarse to resolve some features that may be important to initiation of convection. While the density of observing stations at the surface is generally sufficient to resolve all but the smallest mesoscale features, observational data above the surface are comparatively sparse and are not sufficient to resolve any but the largest of mesoscale features. The International H O Project of 2002 (IHOP) focused on 2 mapping the distribution of water vapor and investigating its role in convective initiation. Many studies came out of IHOP that address the importance of accurate and high-resolution analysis of water vapor, temperature, and wind fields and the impacts on predicting convection initiation (Fabry 2006; Markowski et al. 2006; Martin and Xue 2006). These studies echo the findings of Crook (1996) who determined that very small magnitude differences in low-level temperature and moisture can affect whether thunderstorms form or not. Other studies have shown that the clarity with which mesoscale features are resolved in the observational data and the strength of forcing on the large scale have significant impacts on the temporal and spatial accuracy of numerical prediction of convection initiation (Stensrud and Fritsch 1994; Jankov and Gallus 2004; Szoke et al. 2004; Wilson and Roberts 2006). Using the results from the previous research, the present study looks for more insight into the skill of the WRF model at predicting convective initiation and upscale evolution of MCSs. The sensitivity of precipitation to vertical and horizontal grid spacing and to various 3 modifcations to the KF scheme will be tested. The results will be used to determine the need for the KF scheme at various resolutions. Case studies will also be performed to illustrate the complexity in 3 km model fields and the ways in which the WRF can succeed or fail to correctly simulate the initiation and upscale evolution of convection. 1.2 Thesis organization This thesis is organized in journal paper format. It contains two chapters, each of which is a stand-alone paper. Both will be submitted to the American Meteorological Society for publication in either Weather and Forecasting or Monthy Weather Review. The first paper, comprising chapter 2, investigates sensitivites of precipitation to vertical and horizontal grid spacing as well as modifications to the Kain-Fritsch (KF; Kain and Fritsch 1993; Kain 2004) convective parameterization scheme. The second paper, comprising chapter 3, investigates the impact of large-scale forcing on the skill of convective initiation using the WRF model. 4 CHAPTER 2. SENSITIVITIES OF WRF PRECIPITATION TO MODIFICATIONS TO THE KAIN-FRITSCH SCHEME A paper to be submitted to Monthly Weather Review Jeffrey D. Duda, William A. Gallus, Jr., Moti Segal 2.1 Abstract Two-dimensional idealized WRF model simulations were used to test the sensitivity of precipitation to various uses of the Kain-Fritsch scheme in this exploratory study. One set of sensitivity tests analyzed the range of grid spacings over which the partition of precipitation from microphysical and the KF convective schemes changed rapidly. Another set tested the sensitivity of precipitation to five different treatments of the KF scheme. Two of the treatments were the use of the unmodified KF scheme and the neglect of the KF scheme (explicit convection). A third treatment removed the linear dependence of grid- resolvable vertical motion on horizontal grid spacing from the trigger function. The other two treatments involved modifications that artificially coarsened the temperature, moisture, and vertical motion grids before the KF scheme ran and coarsened the heat and moisture tendency and convective precipitation grids after the KF scheme ran. The treatments were then applied to a set of three-dimensional real data cases. The goal of the study was to determine how active the KF scheme is in producing precipitation at grid spacings in the 0.5 km – 12 km range and whether modifications to the KF scheme, which would make its use at 3 km more appropriate, could be used to improve the forecast of precipitation. The results indicate that the KF scheme rapidly becomes less active as grid spacing decreases below 8 km and does not produce much precipitation below 1 km grid spacing. The 3D cases showed evidence that using the unmodified KF scheme may still be useful at 3 km grid spacing, especially in terms of the skill at forecasting convective initiation. However, the use of the unmodified KF scheme is problematic in the 1 – 8 km range since the closure and scale separation assumptions used to develop the KF scheme are violated.