Introduction

Extremely low frequency (ELF) radio waves (3 Hz-3 kHz) can propagate for long distances within the earth-ionosphere waveguide due to a weak ${\propto\,}R^{-1/2}$ dependency of amplitude on range, $R$ (Greifinger and Greifinger, 1976). This enables measurements to be made of CG current moments (current times height) at ranges well in excess of 1000 km. Numerous ELF measurements indicate that sprites are primarily correlated with positive cloud-to-ground (+CG) strokes that have slow tails produced by large continuing currents (Cummer and Inan, 1997; Reising et al., 1996; Bell et al., 1998; Reising et al., 1999; Sukhorukov and Stubbe, 1997; Marshall et al., 1998; Cummer and Stanley, 1999; Huang et al., 1999; Boccippio et al., 1995).

In his landmark paper which predicted the existence of sprites, Wilson (1925) also predicted that ``The discharges above the cloud would doubtless give rise to atmospherics''. There were several limited searches for these ``sprite sferics'' prior to 1996 (Winckler et al., 1996; Winckler, 1995; Franz et al., 1990). These studies failed to provide significant evidence that sprites produce detectable electric field waveforms. On the basis of photometer to VLF comparisons, Winckler (1995) concluded that there was no unambiguous evidence that sprites emit sferics, but admitted that the simultaneous presence of coincident sferics from intracloud or cloud-to-ground discharges made it difficult to address that question.

Farrell and Desch (1992) attempted to explain the apparent lack of VLF emissions by resorting to a longer time scale for sprite development. Their model placed the peak emission below $\sim$50 Hz; well into the ELF region of the spectrum. At the AGU Fall Meeting in 1996, New Mexico Tech (NMT) researchers reported slow ($\sim$500 Hz) electric field changes which were correlated with bright sprites on video but were temporally distinct from +CG slow tails. These slow field changes were unique in that they were not immediately preceded by the VLF radiation produced by a CG's return stroke. NMT researchers attributed these isolated slow field changes to vertical charge transfer within sprites and this was subsequently confirmed by photometric measurements which showed that the slow field changes were directly correlated with spatially-integrated sprite light output (Reising et al., 1999; Brook et al., 1997; Cummer et al., 1998). Examples of this correlation will be shown in this chapter and in Chapter 5.

When video is available at night, the isolated slow field changes have always been found to be correlated with bright sprite events and this correlation enables the detection of bright sprite events via ELF measurements when video is not available (Brook et al., 1997). Daytime sprite events were detected on August 14, 1998, via their ELF ``fingerprint''. These events were analyzed and published by the author (Stanley et al., 2000). The content of this published work is included in this chapter, along with some additional figures that were excluded from the original paper due to length constraints.

The lightning current moment (current times vertical channel length) for the daytime sprite events was extracted quantitatively by a technique described in detail by Cummer and Inan (2000). This technique had been used previously for sprite-producing discharges of both positive (Cummer and Stanley, 1999; Cummer et al., 1998) and negative polarities (Barrington-Leigh et al., 1999). In order to separate the probable lightning and sprite current (i.e., the second peak), it is assumed that the lightning current moment varies linearly after the first few milliseconds of the discharge. The charge moment changes were obtained by integrating the respective current moments according to: ${M_q}(t) = \int_{0}^{t}{M_i}(\tau)d{\tau}$.