Four classes of models of the lightning return stroke are reviewed. These four classes are: (1) the gas dynamic models; (2) the electromagnetic models; (3) the distributed-circuit models; and (4) the "engineering" models. Validation of the reviewed models is discussed. For the gas dynamic models, validation is based on observations of the optical power and spectral output from natural lightning. The electromagnetic, distributed-circuit, and "engineering" models are most conveniently validated using measured electric and magnetic fields from natural and triggered lightning. Based on the entirety of the validation results and on mathematical simplicity, we rank the "engineering" models in the following descending order: MTLL, DU, MTLE, BG, and TL. When only the initial peak values of the channel-base current and remote electric or magnetic field are concerned, the TL model is preferred. Additionally discussed are several issues in lightning return-stroke modeling that either have been ignored to keep the modeling straightforward or have not been recognized, such as the treatment of the upper, in-cloud portion of the lightning channel, the boundary conditions at the ground, including the presence of a vertically extended strike object, the return-stroke speed at early times, the initial bi-directional extension of the return stroke channel, and the relation between leader and return stroke models. Various aspects of the calculation of lightning electric and magnetic fields in which return stroke models are used to specify the source are considered, including equations for fields and channel-base current, as well as a discussion of channel tortuosity and branches.

This review paper, prepared for this second special issue on lightning of the IEEE Transactions on Electromagnetic Compatibility, summarizes major publications on lightning and lightning protection since the first special issue published in November 1998, i.e., during the last decade. The review is organized in the following five sections: lightning discharge-observations, lightning discharge-modeling, lightning occurrence characteristics/lightning locating systems, lightning electromagnetic pulse and lightning-induced effects, and protection against lightning-induced effects.

Analyses of electric and magnetic fields measured at distances from tens to hundreds of meters from the ground strike point of triggered lightning at Camp Blanding, Florida, and at 10 and 20 m at Fort McClellan, Alabama, in conjunction with currents measured at the lightning channel base and with optical observations, allow us to make new inferences on several aspects of the lightning discharge and additionally to verify the recently published “two‐wave” mechanism of the lightningMcomponent. At very close ranges (a few tens of meters or less) the time rate of change of the final portion of the dart leader electric field can be comparable to that of the return stroke. The variation of the close dart leader electric field change with distance is somewhat slower than the inverse proportionality predicted by the uniformly charged leader model, perhaps because of a decrease of leader charge density with decreasing height associated with an incomplete development of the corona sheath at the bottom of the channel. There is a positive linear correlation between the leader electric field change at close range and the succeeding return stroke current peak at the channel base. The formation of each step of a dart‐stepped leader is associated with a charge of a few millicoulombs and a current of a few kiloamperes. In an altitude‐triggered lightning the downward negative leader of the bidirectional leader system and the resulting return stroke serve to provide a relatively low‐impedance connection between the upward moving positive leader tip and the ground, the processes that follow likely being similar to those in classical triggered lightning. Lightning appears to be able to reduce, via breakdown processes in the soil and on the ground surface, the grounding impedance which it initially encounters at the strike point, so at the time of channel‐base current peak the reduced grounding impedance is always much lower than the equivalent impedance of the channel. At close ranges the measuredM‐component magnetic fields have waveshapes that are similar to those of the channel‐base currents, whereas the measuredM‐component electric fields have waveforms that appear to be the time derivatives of the channel‐base current waveforms, in further confirmation of the “two‐wave”M‐component mechanism.