Atmopheric Electrical Activity and the Prospects for Improving Short-term Weather Forecasting

How might lightning measurements be used to improve short-term (0-24 hr) weather forecasting? We examine this question under two different prediction strategies. These include integration of lightning data into short-term forecasts (nowcasts) of convective (including severe) weather hazards and the assimilation of lightning data into cloud-resolving numerical weather prediction models. In each strategy we define specific metrics of forecast improvement and a progress assessment. We also address the conventional observing system deficiencies and potential gap-filling information that can be addressed through the use of the lightning measurement. PRESENT STATE OF KNOWLEDGE In simplest terms, lightning is an electrical manifestation of thermodynamic and mechanical work performed by storm updrafts. Updrafts determine the supply, growth and transport of water condensate to the upper regions of storms, and directly control the dynamics of charge separation that lead to lightning. While there is considerable complexity in the microphysical charge separation process itself, the larger-scale physics involved are reasonably well understood and straightforward. This was exemplified by early theoretical work by Vonnegut [1963], who expressed the electrical power available for lightning generation in scaling-law form, dependent upon storm updraft velocity, charge density, area and electric dipole separation (height). This simple scaling approach, reexamined by Williams [1985] and Boccippio [2002], confirms that basic scaling limitations can be found even in instantaneous measurements of storm properties. Empirical data from ground-based field campaigns corroborate the links between lightning flash rates and storm updrafts. Figure 1 shows the tight relationship between total lightning rates, the precipitation and ice phase development, and updraft velocity during the evolution of an airmass thunderstorm in Alabama. The physics of charge separation and lightning channel breakdown are sufficiently well understood that 3-D cloud models, being developed at a growing number of laboratories [e.g., Mansell, 2000], have matured to include explicit microphysical charging and breakdown. Explicit microphysics in these models yields large scale relationships consistent with both observations and theoretical prediction, e.g., the connection between total flash rate and total ice mass (itself a direct product of storm updrafts). Similar relationships between lightning and precipitation ice are found when spaceborne Lightning Imaging Sensor (LIS) data are compared with 85 GHz microwave and 13 GHz Precipitation Radar measurements [Petersen and Rutledge, 2001; Goodman and Cecil, 2002]. The close coupling between lightning activity and storm updrafts and ice content implies that increases in lightning activity should be observed prior to severe weather, as many events such as damaging winds, tornadoes and hail are direct by-products of extreme updrafts and ice production aloft. This has been confirmed in case studies for decades, as reviewed by MacGorman et al. [1989]. Lightning jumps associated with severe weather events (Fig. 2) such as mesocyclones, tornadoes, damaging winds, hail and waterspouts were more recently observed in Florida by Williams et al. [1999] and in Alabama by McCaul et al., [2002]. In addition to increases in total lightning rate, MacGorman et al. [1991] have hypothesized that stronger updrafts will loft the main storm electric dipole to higher levels in a storm, thus favoring IC over CG discharges. This hypothesis is supported by evidence from electric field balloon soundings. Consistent with this hypothesis, the dominant component of the severe weather lightning “jump” described above is often found to be from IC lightning [Goodman et al., 1988; MacGorman et al., 1991]. In the most severe storms, the ratio of IC to CG lightning can be much greater than its mean values of ~3:1. While prediction, modeling and observation find close correspondence of lightning flash rates with convective properties, a significant degree of scatter, and dependence upon local convective regimes, is common [Petersen and Rutledge, 1998; 2001]. It is thus important to establish the forecast model physics deficiencies, resolution limitations, or initialization data inadequacies that can be addressed by the additional information content provided by lightning. Alexander et al. [1999] (Fig 3) demonstrated improved forecasts of surface pressure and precipitation through continuous assimilation of lightning data (from the National Lightning Detection Network) into models of the March 1993 southern U.S. Superstorm. The high temporal resolution of the lightning data (which were correlated with instantaneous estimates of rainrate to adjust model latent heating) was critically important for the model to correctly forecast the large scale development of the extratropical cyclone, including key parameters such as precipitation and minimum pressure. Notably, comparable improvements over control runs were not achieved upon less frequent assimilation of satellite infrared or passive microwave estimated rainfall rates. Another success was achieved by Chang et al. [2001] through the assimilation of continuous low frequency VLF measurements of lightning, again calibrated by intermittent satellite estimates of rainrate (through a probability matching technique). Rogers et al. [2000] produced an improved 24-h forecast of the rainfall pattern for a summertime mesoscale heavy rain event. Only the presence of deep convection (as might be indicated by lightning) at a model grid point triggered the convective parameterization scheme, on or off. Such an approach has the advantage that the convective precipitation rate and heating profiles generated by the parameterization are compatible with the local (model) environment. The effectiveness of the technique is enhanced in weakly forced environments, common in the summertime, where convective initiation and organization are governed by previous convective activity and the resulting temperature and moisture discontinuities (i.e., boundaries). Such methods, however, must continuously assimilate the convective parameter (lightning is this case); otherwise the model eliminates the imposed disturbance through convective adjustment. These lightning data assimilation strategies all rely on the relationship (correlation) between convective rainfall and lightning flash rate [Cheze and Sauvageot, 1997], and constant lightning detection efficiency within the forecast domain. Errors will be amplified if the relationship is non-constant (i.e., rainfall-lightning relationship varies with storm type or lifecycle, or the ratio of cloud flashes to ground flashes varies).