Colloquium: Classification of Early Warning Systems

Posted on July 29, 2008  /  0 Comments

Nuwan noted that this leads on from the earlier coversation that were had in regard to Early Warning Systems (EWS) and explained why classification is importat for people in this field. This is important mainly for comparisons between countries, institutions and technologies. And so a ranking should be established.

Four examples were identified,

  • Community based last mile hazard warning system
  • Traceability of agriculture markets- trying to maximize the profit of the produce.
  • Dam failure EWS
  • Financial EWS- Looks at currency and banking crisis within a country.

Looking at early warning systems in a abstract way.

There are some ambiguous schemes that have been used..

  • According to a space of time
  • Multiplicity of events
  • Decision model
  • Domain: financial, flood, tsunami, cyclone etc

Therefore it can be noted that ICTs are used to try and minimize or distance the threat of a impending hazard.

He went on to note explain his classification tree. Mainly divided into operational orientation, complexity of the system and entropy of expected state ( the amount of information that flows. Each one is a process, one wants to knw how long it takes to enter, process and exit the system).

The Govt is planning to put sirens in the villages then this is the no observer model (this is a complete control system) but if we feed information to the people this would constitute a with observer model. The no observer controller model can lead to panic and disarray. With the observer system the people will be able to prepare and hence not panic. They will also knw most details about the impending tariff such as the time for evacuation.

Also should be able to classify whether its a forward or feedback path.

He went on to talk about the EWS at the Yala sancutary. This was a natural system that exists in the sanctuary.

EWS process is as below:






Between each of these stages there is a process using some sort of technology.

Nuwan goes on to describe the different types of sensors and the way in which these sensors are trusted. It deals with factors such as precision, sensor error, accuracy, resolution, linearity, hysteria etc. Under the LM-EWS the sensor would be the either the internet or SMS. Information on dependability and availability of the sensor as well. This can be quantified.

He goes on to explain how evidence theory is used to measure the plausibility of multi-systems.

He uses a Gherkin detection example, to show how the technology/ person detects the good or bad gherkins.

There are four detection possiblities,

1. True noise

2. True alarm

3. False alarm

4. Missed surprise

This would allow to seperate sensoring from detection (extracting information from a pool).

Decision systems – Have to make a decision in a limited time. The decision is crucial and has to be made before the warning horizon.

Broker systems – is the intermediary btw decision stage and . Within this system there are other functions such as translation engine, transportation with routing, control flow, rules engine, warehouse for store0n-forward, plugin adapters. He also went on to explain the diff types of topologies.

Within the brokers systems there are providers and consumers. The broker performance is calculated using throughput. Mechanisms for an abstract measure for broker systems is an open process.

There are 5 types of complexities.

1. Time independant – zero, real, imaginary

2. Time dependant – combinatorial, periodic

Axiomatic design framework – And design has the customer domain, functional domain, design domain and process variables.

Nuwan then goes on to explain queing theory.

There has been some argument that detection and sensoring can be coupled but this is not the situation in Nuwan’s opinion and he argues that he needs to be able to prove this.

Proffessor Samarjiva commented saying that Nuwan would need to position himself and the work itself within a space. Firstly, set it within an abstract space ( within systems, queing etc theory) and then move out into academics and practitioners.

Helani Galpaya noted that diff areas can be classified depending on the details that were collected. This would be the ideal outcome for the classification process.

Nuwan noted that terminology also seemed to be a problem as there is not much literature or consensus on theories/ terms.

Prof. Samarajiva noted that for it to be a fair, it should entail that like with like comparision be done. He argued that the most important factor is the outcome of the system itself, regardless of location, technology or process.

Nuwan noted that there are many field that

Comparisons were also drawn between building evacuation systems as well. And therefore there should be a lot of literature in this regard.

Nirmali questioned what the goal of classification is? Nuwan responded sayin that it carries importance for planner and policy makers, be able to recommend a particular system for a scenario, create interest in the field, eliminate ambiguity in the terminology and achieve the optimal design that can be used.

Prof. Samarajiva noted that the design has a weakness in it that it does not take into account any financial impediments. Capital/cost is a large factor in hazard warning.

Sujatha Gamage questioned of an example can be given that is a feed forward, time dependant and where waiting time is greater than service time. Nuwan responded that this can be done… and went on to explain the scenario of where a dam is about to breach.

Comments are closed.