The WRF Lightning Forecast Algorithm: Sensitivities to Microphysics and Other Physics Schemes

The Lightning Forecast Algorithm (LFA), a simple, empirical proxy-based
method for converting selected fields from cloud-allowing models into
two-dimensional fields predictive of estimated lightning flash origin
density, has received considerable interest from the forecast community.
This has led to its incorporation into a number of high-resolution
regional forecast models, foremost among them the NOAA operational
High Resolution Rapid Refresh (HRRR) model. This has permitted more
testing and analysis of the LFA's performance across varying seasons
and locations. In recent years, the LFA has also been added to the
Weather Research and Forecasting (WRF) model runs performed at
the National Severe Storms Laboratory, the WRF ensembles run by
the Center for Analysis and Prediction of Storms at the University
of Oklahoma, and, more recently, in the experimental runs
disseminated by NOAA's Environmental Science Research Laboratory.
Results suggest that the LFA produces generally reasonable lightning
flash density fields, but that some recalibration might be desirable
in models using microphysics schemes more complex than the default
WRF single-moment 6-species scheme (WSM-6), which was in wide use
when the LFA was first designed.

To address the possible need for LFA recalibration within HRRR-like
WRF forecasts using the Thompson two-moment microphysics scheme,
we have examined output from LFA-equipped HRRR runs executed in
several research venues. In particular, we examine a selection
of 48-h long 10-member randomly initialized ensembles executed
each day at the National Center for Atmospheric Research (NCAR)
in Boulder, CO, and also output from daily 12-member matrices of
WRF simulations having four microphysics options mated with three
boundary layer options, as executed by NASA's Short-term Prediction
and Research Transition (SPoRT) team at the National Space Science
and Technology Center (NSSTC) in Huntsville, AL. Unmodified LFA
results from these diverse models are compared either to lightning
flash rate densities observed by the Earth Networks Total Lightning
Network (ENTLN), or to flash rate densities from model runs using
the LFA originally configured with WSM-6 microphysics.

Results from the SPoRT forecasts show that the Thompson 2-moment
microphysics scheme usually produces less graupel than WSM-6,
indicating a need to boost the LFA calibration factor in order
to produce lightning flash rates equal to those from well-calibrated
WSM-6 "reference" runs. Comparison of LFA peak flash rate densities
from Thompson-equipped runs with those using WSM-6 show that the
recalibration factor is approximately 2.01. For the other combinations
of microphysics and PBL scheme choices, the recalibration factors are
smaller, varying from 1.12 to 1.65. The recalibration factor inferred
from comparison of flash rate densities from the small available
sample of LFA-equipped HRRR runs and ENTLN observations is slightly
smaller than that mentioned above, which may be the result of
the smaller flash detection efficiency of ENTLN observations, as
compared with the more sensitive lightning mapping array data
used as ground truth in the original LFA study.