Twitter and Weather. Dynamical precursors of flooding in Indonesia assessed using satellite observations and Twitter activity. (#233)
The Maritime Continent is an archipelago within Indo-Pacific warm pool characterized by the largest precipitation amount, globally. Such environmental conditions combined with complex topography make it favorable for extreme precipitation events.
Such events not only disrupt communities but also enhance weather related communications. These include television coverage and local journal articles but also social media. Due to popularity of Twitter in Indonesia, extreme weather events cause spikes in its communication. Such spices in number of Twitter posts related to flooding can be analyzed and attributed to individual events, for example over Sumatra, an equatorial island in the western part of the archipelago. These data are combined with satellite remote sensing data about precipitation, outgoing longwave radiation and soil moisture. Therefore, crowd sourced, local measures of a disruptiveness of an event can be analyzed in a context of local atmospheric forcing.
We present an analysis of Twitter-based flooding events in relation to subseasonal-to-seasonal variability in the atmospheric circulation. Seasonal circulation is due to monsoon, intraseasonal (30-90 days periodicity) is related to the Madden-Julian Oscillations (MJO) and synoptic scale variability is represented by activity of equatorial waves, primarily convectively coupled Kelvin waves (CCKW). Because both MJO and CCKWs propagate eastward from Indian Ocean and strongly affect Sumatra.
The results show that major flooding events are preceded by extreme 5-day precipitation accumulation. Such accumulation often occurs during prolonged periods of above normal precipitation, typical for monsoon or MJO conditions. However, many of flooding events are actually triggered by synoptic scale CCKWs embedded in those large scale circulations. In some cases CCKW events alone can trigger a flood. Since CCKWs affecting Sumatra are often initiated over central Indian Ocean, there is extended range predictability of such events, which may enable early warning for to-be-affected communities by various channels, including social media.