12A.5 Validating a WRF Ensemble Using IMPOWR Field Data for Two Fire Weather Days in Central New Jersey

Thursday, 7 June 2018: 9:00 AM
Colorado A (Grand Hyatt Denver)
Keenan R. Fryer, Stony Brook University, Stony Brook, NY; and B. A. Colle and J. J. Charney

Wildland fires represent a potentially dangerous and costly forecast problem for the Northeast U.S. (NEUS). While their occurrence less than the western U.S., the high density of population in the NEUS can cause significant problems. Past studies have shown that mesoscale models for NEUS fire weather days (FWDs) have large biases in warm season PBL temperatures. During the IMPOWR (Improving the Mapping and Prediction of Offshore Wind Resources) field study, which took place in the spring and summer months of 2013–2014, a number of flights were done using a Long-EZ aircraft to collect observations over the Pine Barrens region of New Jersey.

The focus of this work is two of these Long-EZ flights completed on the 10th and 14th of April, 2013 over the central New Jersey pine barrens. The flights consisted of a number of spiraling vertical profiles as well as a series of vertically stacked level flight legs which are used to give a cross-sectional view of the PBL. The Weather Research and Forecasting (WRF) model are validated against these aircraft observations, along with observations from buoy, land surface stations, and NWS balloon soundings. Two different ensembles of WRF results are compared. The first group is a small 3-member ensemble down to 1.33-km grid spacing using the RAP for initial and boundary conditions, with each member employing one of the YSU, ACM2, or MYNN2 PBL schemes. The second is a large (40 member) ensemble which uses the MYNN2 PBL scheme and was initialized using analyses obtained from a continuously cycled ensemble Kalman filter (EnKF) Data Assimilation (DA) system. For the DA system, 6 hourly cycling began 10-14 days prior to an event, where a 40 member, 12km resolution WRF ensemble was advanced 6 hours, then observations from NCEP’s Meteorological Assimilation Data Ingest System (MADIS) assimilated using an implementation of an ensemble adjustment Kalman filter in the Data Assimilation Research Testbed (DART). The physics ensemble provides an insight into the PBL sensitivity for these events while the DA ensemble provides insight into the sensitivity to initial conditions.

This presentation will summarize the Long-EZ observations, along with any biases in temperature, wind speed, and moisture profiles. Results from the physics ensemble show a small sensitivity to the various PBL schemes, with a slight warm bias, which is different from the large cool biases shown in previous work of fire weather days in this region. Preliminary results from the EnKF ensemble have more of a cool bias. The PBL ensemble is clustered together with weaker than observed wind speeds, while the DA ensemble has a more accurate prediction of the winds. This presentation will suggest some of the possible reasons for the different PBL and ensemble performance.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner