Based on the available data, forecaster comments, and impacts (damage, fatalities) for each storm, the TC QPF group has chosen Tropical Storm Ernesto from 2006 as an initial case for further study. We expect Ernesto to have a stronger boundary that can be contrasted with the weaker, in-situ boundary present in Tropical Storm Hanna. The following is a proposed outline of questions to address as the TC QPF group begins studying the first case and subsequent cases.
I. Determine what type of boundaries are present? Where are they located?
– look at surface analyses (HPC, NWS local office analyses)
– perform manual analyses using a dense observation network
– categorize boundaries as shallow diabatically generated boundaries, synoptic-scale baroclinic zones, or other
II. How much rain fell? Where did rain fall in relation to the boundaries?
– look at Stage IV QPE products (HPC) with sufficient detail for region of interest
– if possible, overlay QPE and analyzed boundaries
– analyze radar and satellite data (archived at local NWS offices, NCDC, NC State)
– look for cloud, precip structure indicative of boundary interaction
III. Did operational models capture the boundaries? Did they capture the areal extent, amount of precipitation adequately? If so, did they capture the boundary interaction? At what time preceding the event?
– look at archived model data (NWS local offices, HPC, NC State)
– quantify the deviation from the observed data at different times preceding the event
IV. What did the actual forecasts predict preceding the event? Did they provide adequate lead time for emergency managers or other flood control agencies?
– look at archived forecast discussions (available at NWS local offices, HPC?)
– look at archived QPFs (NWS local offices, HPC)
V. What processes or factors contributed most to forecast uncertainty and ensemble spread?
– run WRF-ARW with different parameterizations, MP schemes, initial and lateral conditions
– quantify the results for different runs in relation to operational model runs and observed data
VI. Which model runs produced the most accurate forecast? Is there an ensemble of model runs that best resolves the QPF/boundary interaction? Do these model runs provide realistic improvements that are useful for improving forecasts and lead times for emergency managers or other flood control agencies?