Weather forecasts dictate our daily activities and allow us to respond properly during extreme
weather events. However, weather forecasts are never perfect, but differences with model
output and with observations can be minimized. Discrepancies between meteorological
observations and weather model outputs are often caused by resolution differences (point vs.
grid comparisons) and by the parameterizations used in the model. Atmospheric model
parameterization refers to substituting small-scale and complicated atmospheric processes by
simplified ones. In order to make weather forecasts more accurate, one can either increase the
model resolution or improve the parameterizations used. Increasing model resolution can
simulate small-scale atmospheric processes better, but takes a longer simulation time. On the
other hand, improving model parameterization schemes involve in-depth measurements,
analysis and research on numerous atmospheric processes. However, one can find a
combination of existing parameterization schemes that would minimize observation-model
differences. It is therefore essential to ask the question, “What model resolution and
parameterization scheme combinations at a particular location and at particular seasons produce
model output that has the smallest difference with observations simulated at a reasonable
amount of time?”

Northern Thailand is a meteorologically active and unstable region especially during the
summer and monsoon months (e.g. intense thunderstorms, hail storms, etc). It is also where
high concentrations of air pollutants occur during the dry months (e.g. haze). It is therefore
essential to have model forecasts close to observations for this region to reduce risk from
weather and from air quality degradation. This study aims to find the optimum model resolution
and parameterization scheme combinations at particular provinces in northern Thailand with
available data during the wet and dry seasons that produces minimum differences with
observations.

Nested model simulations were performed using the Weather Research and Forecasting (WRF)
model (v. 3.6) ran in the High-Performance Computer (HPC) cluster of the National
Astronomical Research Institute of Thailand (NARIT) for northern Thailand (2 km spatial
resolution and hourly output), for the whole of Thailand (10 km spatial resolution and hourly
output), and for the entire Southeast Asia (50 km spatial resolution and 3-hourly output).
Combinations of the WRF Single-Moment 3-class, the WRF Single-Moment 5-class, the Lin et
al. (Purdue), the WRF Single-Moment 6-class and the WRF Double-Moment 6-class
microphysics parameterization schemes, as well as the Betts-Miller-Janjic, the Kain-Fritsch
scheme, the Grell-Freitas (GF) ensemble and the Grell 3D cumulus parameterization schemes
were utilized to determine the optimum resolution and parameterization of the model when
compared to observations. Measured data came from the Thai Meteorological Department
(TMD) weather stations in Chiang Mai, Chiang Rai and Lampang in northern Thailand from
December 1-15, 2014 (cool dry season), from May 1-12 (hot dry season) and from August 1-7,
2015 (wet season). Results showed a seasonal dependence on the optimum microphysics and
convective parameterization combination scheme. It was also found out that cloud resolving
model grid sizes still failed to capture convective process as indicated by the derived optimum
resolution for the hot-dry and wet seasons.