A geographical approach for on-the-fly prioritizing social-media messages for flood risk management based on sensor data

Abstract

Flood risk management requires updated and accurate information about the overall situation in vulnerable areas. Social media messages are considered to be as a valuable additional source of information to complement authoritative data (e.g. in situ sensor data). In some cases, these messages might also help to complement unsuitable or incomplete sensor data, and thus a more complete description of a phenomenon can be provided. Nevertheless, it remains a difficult matter to identify information that is significant and trustworthy. This is due to the huge volume of messages that are produced and which raises issues regarding their authenticity, confidentiality, trustworthiness, ownership and quality. In light of this, this paper adopts an approach for on-the-fly prioritization of social media messages that relies on sensor data (esp. water gauges). A proof-of-concept application of our approach is outlined by means of a hypothetical scenario, which uses social media messages from Twitter as well as sensor data collected through hydrological stations networks maintained by Pegelonline in Germany. The results have shown that our approach is able to prioritize social media messages and thus provide updated and accurate information for supporting tasks carried out by decision-makers in flood risk management.

Geo-spatial text-mining from Twitter – a feature space analysis with a view toward building classification in urban regions

“…One of the applications of social media and remote sensing is weather-caused disasters like floods. For example, the combination of remote sensing data and Twitter messages results in improvement of flood detection and predicting (Wang, Skau, Krim, & Cervone, 2018), and flood risk management (de Assis, Herfort, Steiger, Horita, & Porto de Albuquerque, 2015). The general applicability of social media text messages to the building instance classification task has been shown previously using techniques such as LDA (Blei, Ng, & Jordan, 2003) and LSTM recurrent neural networks on georeferenced tweets from Munich (Huang, Taubenböck, Mou, & Zhu, 2018).…”

Service-oriented middleware for dynamic, real-time management of heterogeneous geosensors in flood management

“…In this context, this paper aims to present an approach for supporting flood risk management by means of the near real-time combination of social network messages and sensor data streams. It extends our previous works (ASSIS et al, 2015a;ALBUQUERQUE et al, 2015) by adopting a workflow analysis which structures and defines an automated near real-time prioritization of social network messages based on the sensor data stream. Furthermore, it describes the formal representation of the problem statement, and makes an evaluation of the approach through case studies.…”

“…Due to the high cost and complexity of deploying real sensors, simulation was the chosen method to evaluate the proposal. The performed simulations are based on 2 scenarios that consider a real-world application of a wireless stationary sensor network to monitor a river water level (HORITA et al, 2015) and mobile sensors that are able to perform cooperative coordinated missions (DOERING et al, 2014). The stationary sensors have a small physical size and a lowcost long-range communication device, while the mobile sensors are small autonomous aircrafts (UAVs) containing a Mission and Vision Control Module composed of additional hardware as shown in Figure 11.…”

“…The combination of information provided by a collaborative platform (esp. Ushahidi) and sensor data collected via a wireless sensor network have been built for decision-making in flood risk management (HORITA et al, 2015).…”

“…temperature of the water or the volume of rainfall). Although these data are useful for supporting decision-making, further information is required for estimating the overall situation at an affected area (HORITA et al, 2015). Social networks like Twitter, Facebook, and Instagram, can overcome this issue either by providing information from areas which are not covered by sensors or complementing semantically the data provided by them STAMBERGER, 2010;VIEWEG et al, 2010;ZIELINSKI et al, 2013;HORITA et al, 2015).…”