Development of a Regional Lightning Nox Parameterization for the Weather and Research Forecast Chemistry Model

Nitrogen oxides (NOx) produced by lightning (LNOx) play an important role in atmospheric chemistry, including the formation of tropospheric ozone (O3). Chemical transport models such as the Weather Research and Forecasting (WRF-Chem) model can simulate some aspects of ozone chemistry related to anthropogenic pollution, but to produce accurate O3 concentrations, it is important to accurately specify the LNOx. WRF-Chem currently only includes lightning and the production of the resulting NOx when run at cloud scale resolution. These cloud scale parameterizations are based on previous studies that have shown that lightning flash rate is strongly correlated with radar-derived storm height, updraft strength, the vertical flux of ice, and other storm parameters. We describe a new way to parameterize lightning occurrence and the formation of LNOx when WRF-Chem is run at the regional scale (e.g., 36 or 12 km grid spacing). We present a comparison of three regional scale lightning parameterizations during NASA's Intercontinental Chemical and Transport Experiment (INTEX-A, 2004). Two of the parameterizations use previously reported relations between lightning flash rate and radar-derived storm top height. We have developed the third parameterization which uses a relationship between convective precipitation, depth of the mixed phase layer, and flash rate. We investigated each lightning parameterization in the WRF-Chem model. After comparing them at 36 km and 12 km grid spacing for six different summer days during 2004, we found that our parameterization, called LCLIPER (Lightning Climatology and Persistence), produces total lightning (IC+CG) results that generally are comparable to "observations" of NLDN' total lightning. Two other parameterizations, Futyan and Del Genio (FDG) (2007) and Yoshida et al. (2009) were compared with LCLIPER. Since both parameterizations use a relationship with radar echo top and total lightning, they both produced similar results. Contingency table statistics, domain wide total flash counts and mean flash rate of coinciding observations and model lightning are presented to help determine which parameterization is superior. Results show that FDG and the Yoshida et al. (2009) schemes underestimated flash rate at 36 km grid spacing, and overestimated them at 12 km. LCLIPER also overall underestimates flash rate, but to a lesser extent. LCLIPER is the focus of the study since it provided the best results. The predicted flash rates and a NOx production term determine the two dimensional pattern of LNOx. The LNOx then is vertically distributed using our previously calculated climatological vertical distributions of lightning sources measured by the Lightning Detection and Ranging (LDAR) network at Kennedy Space Center that are functions of storm top. Warm season vertical distributions of lightning sources and flashes are presented using data from the Lightning Detection and Ranging (LDAR) network at Kennedy Space Center, FL. We emphasize the percentage of sources/flashes at each level compared to the vertical total and present the distributions as a function of storm top above ground level (AGL). The vertical profiles of sources and flashes are compared with each other and with those from previous studies. Results indicate that storms with tops higher than ~10 km AGL often have a bimodal or multiple peak distribution of percentage sources and flashes. However, distributions for storms with tops lower than ~10 km AGL exhibit only a single dominant peak. Temporal variations in the vertical distributions of flash percentages are examined for four clusters of storms occurring on different days. Results reveal considerable storm-to-storm and intra-storm variability. However, two similarities are observed between the four cases: 1) Maximum flash density (flashes km-3) occurs as the maximum storm top is reached, and 2) As the storms increase in intensity, both maximum flash density and flash percentage increase in altitude, and then both decrease in altitude as the storms decay. When compared to aircraft measurements during the INTEX-A field campaign, results of WRF-Chem LNOx indicated that the addition of lightning improved values of upper tropospheric NOx, but underestimated the values when considering the mean of the entire flight. At 12 km grid spacing, WRF-Chem without lightning predicted 61 pptv NOx in the layer between 7-9 km, while 248 pptv NOx was observed during the 12 July 2004 INTEX-A flight. WRF-Chem with the addition of lightning from LCLIPER increased NOx in this layer to 220 pptv. When the model produced lightning, LNOx was overestimated throughout the atmospheric column. Results indicate that radar derived echo tops may be underestimated causing the incorrect vertical distribution method to be selected.