Taranis: Neural networks and intelligent agents in the early warning against floods
Highlights
► This paper describes the implementation of an Early Warning System (Taranis). ► Taranis is based on neural networks of CounterPropagation type and agents. ► The system creates models to predict floods and provide information. ► Taranis also offers intelligent mechanisms for dissemination of mass alerts. ► The system also is prepared to handle earthquake, tsunami and wildfire alerts.
Introduction
The Early Warning System (EWS) is the first line of defence for the most vulnerable population against large-scale natural catastrophes. This is the main reason for the development of this application Taranis, which uses artificial intelligence in a hydroelectric structure to provide a better source of information, when the data of the threats are already known. With this type of system, if corrective measures were taken in time, these catastrophes could be avoided. If theses preventive measures are not adopted flooding will occur, as was the case in the Dominican Republic caused by the tropical storm Olga in 2007, which was based on an outdated operations manual and limited to a specific period of time for storms, without taking climate change into account and not taking the decision to empty the reservoir in time, which caused the floods in Santiago de los Caballeros in December 2007.
The greatest deficiency in the current EWS is the mechanisms to alert the public in a rapid, accurate and efficient manner. It is no use having the best prediction system if there are not any mechanisms to alert the population in case of emergency. At the present time warning mechanisms on a large scale are carried out through short text messages to mobile phones. The most representative of this practice is the United States and Japan. In addition, Chile has recently begun the implementation of this technology.
In this article we present a tool that has as its objective to use the capacity of the neural network (NN) pattern recognition, to predict the volume of precipitation expected at any given time, information that would be used to create models to predict floods and provide information to manage the hydroelectric reservoir in an effective manner to prevent material damage. Another advantage of NNs used by this system is the ability to adapt to the new meteorological values caused by climate change. On the other hand, the system provides an alternative means of communication and issuance of massive alerts, through the use of intelligent agents, which is prepared to handle earthquake, tsunami and wildfire alerts, although the current system focuses on alerting against flooding.
The EWS is a tool that consists of a set of mechanisms and procedures for the detection of hazards, monitoring of indicators, communication of alerts and alarms and evacuation of vulnerable populations to safe areas (Lyon & Fletcher, 2001).
The EWS is made up of institutions that are responsible for assessing natural phenomena through the instrumentation and broadcasting of alerts to ensure that the institutions of civil society are responsible for communicating to the civilian population of the possibility of such an extreme phenomenon occurring and proceed with the evacuation or procedures to mitigate the losses.
The EWS requires technical knowledge about the threat, what the causes are and the risks faced by society. This involves knowing the temporal and spatial behaviour of such phenomena through the modelling of physical quantities and the development of procedures to make the forecasts that are applied as part of the EWS. In this context, one looks for precursor signals that may relate to the likely magnitude of the event.
The issue of the EWS is one of the most commented on today. International agencies and other developed countries have been trying to prevent disasters and not only limit themselves to help once they have happened. We have knowledge of floods since the beginning of civilization, but as a result of climate change these are becoming increasingly frequent and aggressive, a situation which has led to the main organizations for international cooperation to search for alternatives to preserve life and property. This is the reason why the area of Central America and the Caribbean currently have 19 EWS against floods of a centralized type and 26 of a decentralized or community type, and the number continues to rise.
EWS requires 4 basic components (ISDR, 2006), these are:
- 1.
Knowledge of the risks,
- 2.
Technical monitoring and alert service,
- 3.
Communication and dissemination of the alerts,
- 4.
The ability of community responses.
The community EWS as well as the centralized one possess a great deficiency in the general structure. While the forecasts have improved the way information is structured to meet requirements, it turns out to be very expensive and the most of the cases do not have the 4 elements that an EWS should have.
These systems are implemented with fixed data structures and do not have the capacity to use new technological advances; on the other hand, the platforms are closed which increases the costs of implementation and maintenance; another factor that elevates the cost is that of communication platforms which are based on radio.
Another deficiency that the EWS has is that these systems are mostly used against the threat of flooding, while these countries have at least 3 other different risks, such as: tsunamis, earthquakes and volcanic eruptions. Likewise, they do not have mechanisms to alert the entire population of an area or region at the same time; the alerts and warnings are directly implemented by relief agencies.
On the other hand, the ability to integrate of different data sources is limited. In most cases they are based on forecasts of foreign agencies, mainly due to the scepticism meteorological organizations have of the countries concerned in the Central American area and the Caribbean.
The knowledge of the risk and technical monitoring are the two basic components of an EWS, for which Taranis use NN to predict floods, and management of hydroelectric dams. It also offers intelligent mechanisms for communication and dissemination of mass alerts through agents, which are also supposed to control the reading of the hydrometeorological stations, with which the third structural of this type of system requirement is met and, likewise, through the knowledge of the level of risk of each community to evaluate the main requirements for each case. With the communication mechanism previously mentioned, you will obtain better control of the evacuation operations.
Section snippets
Neural networks
NNs applied to solve a variety of problems, according to Freeman and Skapura (1992), is a system of parallel processors connected to each other in the form of directed graph. Schematically each neuron of the network is represented as a node. These connections provide a hierarchical structure that attempts to emulate the physiology of the brain, searching for new models of processing to solve specific problems in the real world. What is important in the development of the technique of the NN is
Taranis application
The development of the application uses the concept of systems-based agents, because this requires the same to be implemented without the obligatory nature of agents, and in this way take advantage of the fundamentals of both paradigms (systems-based agents and multi-agent systems) (Corchado, Tapia, & Bajo, 2012).
Experiments and result
The first evaluations of the NNs were carried out using the data from the hydroclimatic stations in Santiago, Taveras, Jarabacoa and Manabao, which are located in the basin of the Yaque del Norte river. Data was extracted from the information system in real time the Instituto Nacional de Recursos Hidraulicos (INDRHI) in the Dominican Republic (Indhri station, 2010), but due to the fact that the daily sequence was relatively short and that after the pre-processing the information available was
Conclusions
In this article we presented a EWS that use NNs of CounterPropagation type and intelligent agents for analysis and assessment of the risk of flood caused by rain. The system creates models to predict floods and provide information to manage the hydroelectric reservoir in an effective manner to prevent material damage. The use of NNs provides an interesting alternative to adapt to the new meteorological values caused by climate change. On the other hand the addition of a second layer of NNs, to
Future applications
For a second implementation of the network layer scheme two new networks levels will be added, which will attempt to predict the evolution volume at the front and rear of the reservoirs. What is achieved is having more control and optimizing the process of drainage of hydroelectric dams, in addition to being able to improve the way in which resources are used in the event of drought. Likewise, enabling access to satellite information and the establishment of a hydrological database are intended.
References (19)
Competitive learning: From interactive activation to adaptive resonance
Cognitive Science
(1987)- et al.
Feature discovery by competitive learning
Cognitive Science
(1985) - Agricultural Research Services, (2011). USDA ARS,...
- Android developers. (2011). Supporting multiple screens,...
- Chama station, (2011)....
- et al.
A counterpropagation fuzzy-neural network modelin approach to real time streamflow prediction
Journal of hydrology
(2001) - Corchado, J. M., Tapia, D. I., & Bajo, J. (2012). A multi-agent architecture for distributed services and applications....
- et al.
Neural networks algorithms, applications, and programming techniques
(1992) - Heaton, J. (2010). Introduction to Encog 2.5 for Java, Heaton...
Cited by (15)
Artificial neural network approaches for disaster management: A literature review
2022, International Journal of Disaster Risk ReductionCitation Excerpt :Next, we review the ANN applications in these areas separately. Due to rapid climate change and global warming, floods are becoming a frequently occurring calamity [3,38]. This section reviews different ANN techniques for developing predictive models, early warning systems, and susceptible maps for predicting floods.
Development of a warning model for coastal freak wave occurrences using an artificial neural network
2018, Ocean EngineeringCitation Excerpt :Thirumalaiah and Deo (1998) and Sunkpho and Ootamakorn (2011) used an ANN to predict floods in urban areas. López et al. (2012) connected such a warning system to personal communication devices to communicate flood warnings as rapidly as possible. These previous studies show that ANNs are widely used to establish warning systems, especially for potential natural disasters.
Designing and Applying Flood Early Warning System Based on Waterfall and Water Level for Special of Palembang City
2020, Journal of Physics: Conference SeriesThe role of agent-based modeling and multi-agent systems in flood-based hydrological problems: A brief review
2020, Journal of Water and Climate ChangeAn analysis of flood forecasting criteria
2019, International Journal of Advanced Trends in Computer Science and EngineeringWSIA: Web ontological search engine based on smart agents applied to scientific articles
2019, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)