In the present study, the WRF-ARW model V3.5.1 has been used to simulate the monthly average temperature of 33 stations over Bangladesh in the pre-monsoon season. Final Reanalysis (FNL) (1ox1o ) data collected from National Centers for Environment Prediction (NCEP) is used as initial and Lateral Boundary Conditions (LBCs), which is updated at six hours interval i.e. the model is initialized with 0000, 0600, 1200, and 1800 UTC initial field of the corresponding date. The model has been configured in a single domain, 6 km horizontal grid spacing with 161×183 grids in the east-west and north-south directions and 30 vertical levels. In this research, the WSM6-class microphysics scheme coupling with the Kain-Fritsch cumulus parameterization scheme have been used. The model has been integrated by using initial and LBCs from NCEP-FNL analysis at six hourly intervals.
The model has been run 107 days for long-range prediction starting with the initial condition of 0000 UTC of 17 February up to 0000 UTC of 1 June for the period 2010-2014. The model is also run for 72-h with every day 0000 UTC initial conditions for 94 days starting with the initial condition of 0000 UTC of 27 February 2014 for the prediction of 24, 48, and 72-h lead time temperature in the month of March, April and May 2014. However, the results are presented for March-May to avoid the spin-up effects of the first 15 days. A spin-up period of 1 month may be sufficient for the atmospheric and upper soil layer, but it is not enough for deep soil layer variables (Christensen 1999). A great disparity in spin-up time between models, variables, and regions is often reported. Important reductions in spin-up time are expected in humid regions (Rodell et al. 2005) and, in particular, subtropical South America is a region where a quick equilibrium is expected (Silva and Berbery, 2006).
There is a limited number of meteorological observatories in the NE and SW regions of Bangladesh. Due to less number of stations in the NE and SW regions 8 more points have been added in the Bangladesh Map to study temperature data. 2m level temperature has been extracted at 3 hourly intervals then made daily and monthly average temperature data for 24, 48 and 72-h, and 92 days during the studied period. From the WRF Model run, 3 hourly outputs are made during the study period. These 3 hourly temperature data are converted into monthly temperature data of March, April, and May during 2010-2014. This simulated data has been compared with the observed temperature at 33 meteorological stations of BMD.
In this study, 24, 48, and 72-h have been considered as day 1, day 2, and day 3 of the model run. The monthly data has been plotted for March, April, and May using 24, 48, and 72-h lead-time prediction for the year 2014 and also using 92 days prediction during 2010-2014. The RMSE and MAE have been calculated for 33 meteorological stations because there is not any observational data for others. The CC, RMSE, and MAE for 24, 48, and 72-h and 92 days predicted temperature have been plotted using SURFER software for the year 2014.