Data Reduction

Overview

During the period between launch and mid-January 2000 the PDD had recorded some 1.8 million events using all trigger modes. Of these 1.8 million recorded events, 995,000 were captured in the internal trigger mode (see section 2.3), with the compensation set to the night/slow compensation mode/rate (see section 2.2). Of these 995,000 internal triggers, night/slow mode events, 58% of these events were detected between 1800 and 0600 local time (LT) (i.e., "mostly night"), while the remaining events were detected between 0600 and 1800 LT (i.e., "mostly day"). This bias in the number of event detections between local night and local day (at the subsatellite point) results from a convolution of the sampling preferences (not accounted for in this work) and the day/night detection efficiency of the PDD.

Types of Optical Signals Observed

We have examined the waveforms of several thousand randomly selected PDD events. We found a few percent of events to be due to energetic particles, which are recorded when the PDD is set into the internal trigger mode and the duration test is disabled or set to a small (less than five samples) value. We also found a class of events whose waveforms exhibit a character more like noise (e.g., due to glint or instrument electronic noise) than the characteristic curve obtained from observations of lightning. A typical "noise" event is characterized by a highly oscillatory signal, having a RMS amplitude that varies with the magnitude of the DC offset upon which the signal is superposed. We endeavored to flag the particle and noise-induced triggers in the PDD data set.

Figure 4 illustrates the three basic types of PDD waveforms. Figure 4a shows a typical signal obtained from an optical lightning emission, Figure 4b shows a typical slowly varying (across the 1.9-ms record) signal classified as "noise induced," and Figure 4c shows a typical signal classified as "due to energetic particle."

Signal Discrimination

We endeavor to identify and remove from consideration the triggers attributed to noise (Figure 4b) and energetic particles (Figure 4c). We describe rejection criteria for these two types of signals in the following.

In the case of waveforms that exhibit the noise behavior shown in Figure 4b, we impose a requirement that the maximum signal amplitude exceed the minimum signal amplitude by more than an order of magnitude. This requirement eliminated most of the noise-induced false triggers and removed 29% of the internal trigger mode, night/slow compensation mode data set. This criterion does have the effect of rejecting low-power lightning signals but does not significantly affect the peak amplitude distribution of PDD events, as will be shown in section 5.2.

In the case of false triggers generated by energetic particles, we are able to isolate these signals by first imposing the noise rejection test and removing signals like that shown in Figure 4b. We consider that terrestrial lightning signals are broadened by cloud droplet scattering, while energetic particle signals are not. To exploit this fact, we estimate the slope of the decay portion of the light curve and exclude energetic particle signals by excluding rapidly decaying signals. Thus we impose the requirement that the sampled signal may not decrease by more than a factor of 8 within any four consecutive samples (60 s) in the 100 s following the trigger point. Further, the signal amplitude may not fall below the trigger level within the first 100 s following the trigger point. These criteria remove much less than 1% of the internal trigger, night/slow compensation mode data set. We found that most of these particle triggers occurred during the early portion of the mission when the duration test was set to low values (less than five samples; see section 2.3).

After imposing these particle and noise rejection criteria, we are left with 70% (687,000) of the total number of internal trigger, night/slow compensation mode PDD events obtained since launch. What now remains is a reduced population of events generally attributed to optical lightning emissions that were autonomously detected by the PDD and recorded with negligible distortion due to the background compensation. This population of triggers will form the data set from which we derive our statistics and that we compare with previous observations of optical emissions from lightning in section 5. However, while we attribute the down-selected data to lightning, we require a comparison with NLDN data to lend credibility to the claim that the PDD detects lightning.

Next: Comparison of PDD an Up: Optical observations Previous: FORTE Photodiode Det