Notice how the thunderclap does not happen in a point in time, but lasts for a length of time. This afternoon, most of the thunderclaps were long-drawn rumbles instead of one instantaneous clap. This is explained by modelling the lightning bolt as a number of explosions strewn across its path. As such, sound originates from all points on the lightning bolt instead of just at its point of contact with the ground (Figure 1). We can assume also, for the moment, that the lightning goes straight up (Figure 2). This is justified by the fact that many of the thunder rumbles generally appear to the listener to travel upwards from the horizon.
Figure 1. The sound comes from the entire length of the lightning bolt.
Figure 2. Straight-bolt lightning model
By this set up and by nailing down the start and end time of a thunder rumble, it would be possible to determine the height of the thunderbolt as well as its point of contact with the ground. Given knowledge of the bearings and a map, it would be possible to pinpoint where exactly the thunderbolt has touched ground. And given a good grasp of trigonometric tables, it would be possible to determine, with a quick mental calculation, the maximum ascent of the thunderbolt. (Figure 3)
Figure 3. Measuring the ascent of the straight bolt using rumble time; sound amplitude-time graph included for reference
Now, as we all know, the same clouds that produce thunderstorms, the cumulonimbi, also produces tornadoes. Tornadoes on land are unheard-of in Singapore, but we have had one or two waterspouts (tornadoes on water) off the south coast in my lifetime. Now suppose there is a correlation between cumulonimbus size (measured by height) and the likelihood of tornadoes produced from its bottom face, and then suppose also that the maximum ascent of a thunderbolt is correlated to the height of the cumulonimbus. (Figure 4) Then, by listening to the rumbles originating from the azimuthal angles covering Singapore's southern waters and mapping the data according to their geographical positions (see point 3), a listener would be able to produce a three-dimensional image of a cumulonimbus complex hovering over the Straits of Singapore during a thunderstorm (Figure 5).
Figure 4. Correlation of cloud height to maximum ascent of thunderbolt
Figure 5. Thunderbolt mapping procedures
Then, by knowing the correlation between tornado likelihood and cloud height and also whatever qualitative details we might have gathered, we can set up an automated waterspout forecast service by using light sensors for lightning, two sound sensors (a sound interferometer) for recording and pinpointing the location of the thunderclap, and a running software analysing and mapping every thunderclap, feeding every one of the recordings into a metric so that, when it exceeds a value, the system can alert all the waterspout junkies in the country and lure them to the seasides for a probable feast for the eyes.