Validation of CALMET/CALPUFF models simulations around a large power plant stack

Abstract

CALMET/CALPUFF modeling system is frequently used in the study of atmospheric processes and pollution, and several validation tests were performed until now; nevertheless, most of them were based on experiments with a large compilation of surface and aloft meteorological measurements, rarely available. At the same time, the use of a large operational smokestack as tracer/pollutant source is not usual. In this work, first CALMET meteorological diagnostic model is nested to WRF meteorological prognostic model simulations (3x3 km2 horizontal resolution) over a complex terrain and coastal domain at NW Spain, covering 100x100 km2, with a coal-fired power plant emitting SO2. Simulations were performed during three different periods when SO2 hourly glc peaks were observed. NCEP reanalysis were applied as initial and boundary conditions. Yong Sei University-Pleim-Chang (YSU) PBL scheme was selected in the WRF model to provide the best input to three different CALMET horizontal resolutions, 1x1 km2, 0.5x0.5 km2, and 0.2x0.2 km2. The best results, very similar between them, were achieved using the last two resolutions; therefore, the 0.5x0.5 km2 resolution was selected to test different CALMET meteorological inputs, using several combinations of WRF outputs and/or surface and upper-air measurements available in the simulation domain. With respect to meteorological aloft models output, CALMET PBL depth estimations are very similar to PBL depth estimations using upper-air measurements (rawinsondes), and significantly better than WRF PBL depth results. Regarding surface models surface output, the available meteorological sites were divided in two groups, one to provide meteorological input to CALMET (when applied), and another to models validation. Comparing WRF and CALMET outputs against surface measurements (from sites for models validation) the lowest RMSE was achieved using as CALMET input dataset WRF output combined with surface measurements (from sites for CALMET model input). Following, CALPUFF model was applied to simulate the local atmospheric diffusion of SO2 (as an inert tracer) from a large power plant smokestack (with four parallel independent liners), considering two different stack configurations (one single point source as a summa of four liners vs. one point source per liner) and two different CALMET meteorological simulations (using as input dataset only the WRF model output vs. only surface and upper-air meteorological measurements). Comparison of those CALPUFF simulations results against the hourly average ground level concentration (glc) measurements shows that the best model performance was obtained by using only WRF model output as CALMET input; also, better glc results are obtained considering one point source per liner in CALPUFF simulations.