Convective Modeling Group Links

The Chicago Flood of 17-18 July 1996

Table of Contents

Overview | Previous Research | Observational Data | Developments| Publications| References|

Principle Investigators

Dr. Steven Peckham
Dr. Bob Wilhelmson

Overview

The forecasting of heavy rainfall events is a challenging problem as the convective precipitation guidance that is supplied by numerical models is often inadequate (Funk, 1991; Junker and Hoke, 1990). Key elements in the Chicago event include the frontal boundary, convective outflow boundaries, cloud boundaries, and the low-level jet. Their role in the initiation and subsequent evolution (including their relationship to the training of convection along the frontal and outflow boundaries) are not fully understood.

Objective

to further address the issues of convective initiation and training along frontal and outflow boundaries together with the interaction of these boundaries and subsequent convective formation.

Initial focus - the initialization strategy.

  • One initialization methodology is the use of four-dimensional data assimilation (e.g., Stauffer and Seaman, 1990) which has been shown to produce realistic mesoscale features in the model's initial conditions (Colle and Mass, 1998; Chen et al., 1998; Parsons and Dudia, 1997).
  • Another methodology is the incorporation of artificially constructed data to augment the observations (Stensrud and Fritsch, 1992; Zhang and Fritsch, 1986).

    Questions

  • the role of the low-level jet Observational studies (Angel et al. 1997, Changnon and Kunkel 1999, Angel and Huff 1999) suggest that the orientation of the low-level jet was partially responsible for the focusing of the second period of intense rainfall.
  • The strong southwesterly flow supported the strong storms by transporting high equivalent potential temperature air over the warm front boundary.
  • The impact of the outflow boundaries
  • possible mid-level vortex from the previous event.
  • The impact of the moisture from Lake Michigan

    Previous Research

    Jewett (1999) and Laird (1999) have conducted preliminary simulations of the Chicago flood of 17-18 July 1996 using the Penn. State/ National Center for Atmospheric Research (PSU/NCAR) mesoscale model (MM5) (Grell et al., 1995). These simulations examined the ability of the model to reproduce the heavy rainfall event under a variety of initial conditions, convective parameterization schemes and grid resolutions. Initial analysis of the simulations reveals that the numerical model does well at capturing several details of the flooding event. The location and motion of frontal boundaries are close to those observed during 17-18 July 1996. In addition, several model simulations produce convective precipitation totals in excess of 16 inches in northern Illinois, comparing favorably with the observations. Inaccuracies in the location of convective initiation and the onset of precipitation in the model are being investigated.

    COMET: case study 006

    Brian Jewett's MM5 simulations of the Chicago flood of July 1996

    Neil Laird's MM5 simulations of the Chicago flood of July 1996

    Observations

    Observational data

    Developments

    Current Simulation

    This is a 24 hour, 3 grid simulation initiated at 12Z on 17 July 1996 using the NCAR
    re-analysis fields. The     three grids have 59 vertical levels and 27-km, 9-km, and 3-km horizontal resolution.

    No nudging to analysis fields is conducted during the simulation.
    Each simulation uses the Blackadar PBL scheme and the outter grid uses the Grell convective parameterization.

    Initialization methodology

  • First guess fields obtained from RUC analysis fields for 16-19 July 1996.
  • Several bogus upper-air soundings are introduced across Iowa and northern Illinois at 1200 UTC 17 July. The purpose is to provide additional information regarding a decaying MCS, or cluster of thunderstorms situated along the Iowa/Illinois border (e.g., cloud locations). The reminants of this decaying MCS move across the Chicago metroplex during the morning hours bringing some light rainfall (highest observed totals around 1 inch during the morning hours). The outflow from these convective showers appears to be important in deterimining the location of storm re-development later that afternoon.

    Simulation results

    The following links display select fields for 1600 UTC on 17 July 1996 from:

    MM5 Simulation Composite Radar image
    Total rainfall
    Rainfall tendency
    Vertically integrated cloud water
    Vertically integrated rain water
    		 LOT Composite radar image

    Light rainfall has developed across northeastern Illinois, southern Wisconsin and eastern Iowa. The coverage is sparse in the simulation. In addition, the convective storms have not developed along the full length of the Iowa-Illinois border. One possible explanation is the relatively coarse resolution (3-km) for convective simulations and the shallow depth (weak forcing) of the outflow boundary in the model. Perhaps an additional grid using 1-km resolution will need to be introduced to reproduce the full coverage of the morning convective event. However, the evolution of the afternoon appears to not be impacted by the inability of the model to accurately reproduce the precipitation coverage during the morning hours.

    The following links display select fields for 2300 UTC on 17 July 1996 from:

    MM5 Simulation Composite Radar image
    Total rainfall
    Rainfall tendency
    Vertically integrated cloud water
    Vertically integrated rain water
    		 LOT Composite radar image

    An east-west band of thunderstorms extends from eastern Iowa, across northern Illinois into Indiana at this time. This structure is similar to the observed band of thunderstorms. However, the western edge of the convective storms in the MM5 simulation extend back into eastern Iowa instead of terminating in northwestern Illinois. IR satellite imagry at 2200 UTC (see observational data above) reveals that some convective clouds (enhancement suggests that little, or no precipitation is falling) exist along the Missour-Iowa border. Hence, the simulation appears to have shifted the western edge of the precipitation about 100 km to the west. Storm motion in the MM5 simulation across northern Illinois is to the east-southeast, or nearly parallel to the resolved convection.

    Rainfall totals during the afternoon (1800 UTC to 2300 UTC) reveals that heavy rainfall is generated in the model across Eastern Iowa, southern Wisconsin, northern Illinois and northwestern Indiana. The largest change in total rainfall being across northwestern Illinois where approximately 10 cm of rain (4 inches) has fallen since 1600 UTC. Rainfall totals are highest in extreme northeastern Illinois and northwestern Indiana (roughly 14-cm to 22-cm, or 5 to 9-in) when convective storms are repeatedly crossing the region.

    Comparing to observations, raingauges located across Chicago recorded between 2-cm and 13-cm of rainfall between 1800 and 2300 UTC. Convective storms during this time period develop in northwestern Illinois and train to the east-southeast along an outflow boundary. This behavior, in addition to the relatively heavy afternoon precipitation, is qualitatively reproduced by the simulation.

    Original Simulation

    This is a 36 hour, 3 grid simulation initiated at 12Z on 17 July 1996 using the NCAR
    re-analysis fields. The three grids have 59 vertical levels and 81 km, 27-km, and 9-km
    horizontal resolution. The grids are nudged toward the gridded analysis data over the
    first 12 hours of the simulation. For the remaining 24 hours no data assimilation
    is done. Each simulation uses the Blackadar PBL scheme and the Kain - Fritsch
    convective parameterization.

    The following links display select fields at 3 hour increments from:
  • grid number 1
  • grid number 2
  • grid number 3
    .

    Publications