Xz. Liang et al., Development of a regional climate model for US midwest applications. Part I: Sensitivity to buffer zone treatment, J CLIMATE, 14(23), 2001, pp. 4363-4378
A regional climate model (RCM) is being developed for U.S. Midwest applicat
ions on the basis of the newly released Pennsylvania State University-NCAR
Fifth-Generation Mesoscale Model (MM5), version 3.3. This study determines
the optimal RCM domain and effective data assimilation technique to accurat
ely integrate lateral boundary conditions (LBCs) across the buffer zones. T
he LBCs are constructed from both the NCEP-NCAR and ECMWF reanalyses to dep
ict forcing uncertainties. The RCM domain was chosen to correctly represent
the governing physical processes while minimizing LBC errors. Sensitivity
experiments are conducted for the Midwest 1993 summer flood to investigate
buffer zone treatment impacts on RCM performance.
The results demonstrate the superiority of the buffer zone treatment that c
onsists of the physically based domain choice and revised assimilation tech
nique. Given this treatment, the RCM realistically simulates both temporal
variations and spatial distributions in the major flood area (MFA). This su
ccess is identified with the accurate representation of both the midlatitud
e upper-level jet stream and Great Plains low-level jet (LLJ). The RCM repr
oduces different climate regimes, where observed rainfall was identified wi
th the periodic (5 day) passage of midlatitude cyclones in June and persist
ent synoptic circulations in July. The model also correctly simulates the M
FA rainfall diurnal cycle (with the peak amount at 0900 UTC), which follows
the LLJ cycle by approximately 3 h. On the other hand, RCM performance is
substantially reduced when the southern buffer zone extends to the Tropics,
where large forcing errors exist. In particular, the RCM generates a weake
r LLJ and, as a consequence, a decreased amount and delayed diurnal cycle o
f the MFA rainfall. In addition, the MM5 default LBC data assimilation tech
nique produces considerable model biases, whereas the revised technique imp
roves overall RCM performance and reduces sensitivity to domain size.