Meteorological Cooperative › Warning systems › Why these systems exist

Why these systems exist

The Republic has three warning systems for hazards that are, by international standards, rare and mild. This page sets out the reasoning that produced them, the debt owed to the Japanese arrangements from which the earthquake system is derived, the difficulty of extending the same idea to landslide and flood, and the request the cooperative makes of the public about warnings that come to nothing.

The case that was made

The proposal reached the planning organs in 2019, some months after an event of magnitude 5.2 beneath the Belvacius hills reached REP 4 at Corcina and was felt in fourteen sub-regions. Nobody was injured. What the event exposed was the absence of any arrangement at all: no notice reached anybody, no train braked, no lift opened at a landing, and the first public information followed some forty minutes later.

The argument against building anything ran as follows. Damaging earthquakes occur here roughly once in three generations, damaging landslides affect a handful of dwellings a year, and serious flooding had struck once in living memory. Resources spent on warning are resources unavailable to the renovation programme, to restoration, or to the many obligations the ecological budgets impose.

The argument that prevailed rested on three points. Seconds of notice reduce injury by a large factor wherever such systems have been measured. The material and labour the systems consume are small in relation to what they protect, the seismic network having existed already for other purposes. And the rarity of the hazard is itself part of the problem, since a population that has never felt a strong earthquake has no practised response to one.

The debt to Japan

Japan operates the most developed earthquake warning system in the world, and REP-KJS is derived from it in substance, in method and in name. The cooperative states this openly, since the alternative would be to present borrowed engineering as invention.

What was taken and what was altered
Element	Japanese practice	As adopted here
Name	Kinkyu jishin sokuho, the emergency earthquake bulletin	Taken directly; abbreviated to KJS and used as REP-KJS
Physical method	Detect the P wave, estimate the event, broadcast ahead of the S wave	Taken without change
Intensity scale	Shindo, ten steps with suffixes at 5 and 6	Expanded into ten whole numbers; the correspondence is exact and reversible
Delivery	Flat recorded voice, fixed item order, two rising tones	Taken, with the tones altered slightly and the item order fixed by the same principle
Message length	Short; the schooling was done years earlier	Long; the broadcast performs the schooling as well
Targeting	Differentiated by region	Undifferentiated; the whole territory receives one broadcast with every sub-region stated in it

The two alterations follow from the same fact. In Japan the public understands what a warning means before it arrives, so four words suffice. In the Republic a resident may live a decade without feeling an earthquake, so the broadcast states the depth, the magnitude, the scale equivalents, the meaning of the step and the whole national distribution. The length is a consequence of rarity.

Extending the idea to landslide and flood

Japan warns for landslide and for flood as well, and the cooperative took the concept from there. Building the systems took considerably longer than the earthquake system did, because the physics differs.

An earthquake warning obtains its lead time for free. The P wave outruns the S wave, and a message outruns both, so the interval exists whether or not anybody models the ground. Detection is the whole of the problem, and detection is a solved problem.

Landslide and flood offer no such shortcut. Rain that has already fallen produces a slide or a flood some hours later, and the interval between the two exists only if somebody can compute what the rainfall will do. That requires a model of every catchment and every slope in the territory: gradients, soils, drainage, antecedent wetness, channel capacity, the state of a dune barrier. None of it could be borrowed, since the geology of the peninsula, its rainfall regime and its patterns of settlement resemble no other country closely enough.

Development, the three systems compared
System	Proposed	In service	Development	Principal difficulty
REP-KJS	2019	2021	2 years	Densifying the seismic network and certifying receivers
REP-FEW	2018	2025	7 years	Calibrating a rainfall-runoff model over every basin; gauging the reaches
REP-LEW	2020	2024	4 years	Dividing the territory into slope units and fixing a threshold for each

REP-FEW was proposed first, following the Sornorum floods of 2017, and entered service last. Six of those seven years went to model calibration and to installing gauges where none existed.

Warning in the same manner for all three

A deliberate decision made the three systems resemble one another as closely as the physics allows. They share the wire format, the delivery channels, the certification scheme, the client software, the fixed order of items in a message and the same flat voice. A person who has heard an earthquake broadcast recognises a flood broadcast immediately, and a device certified for one handles all three.

This uniformity cost something. The flood and landslide messages state a probability and a window, quantities an earthquake message has no need of, and fitting them into the same structure took several revisions of the format. The benefit is that the public learns one form of message and one response pattern.

On warnings that come to nothing

A request to the public

Over the period of record, REP-LEW issued 63 warnings followed by no movement and REP-FEW issued 88 followed by no flood. Those counts are the intended behaviour of the systems.

A threshold tight enough to eliminate them would miss real events. The cooperative has set the thresholds to favour notice, on the reasoning that acting on a warning consumes some minutes of inconvenience, and ignoring a real one causes injury.

A warning that came to nothing was correct at the moment it was issued. It stated that conditions had reached the level at which slides or floods occur. Whether one occurred at that particular place depends on details no method resolves.

The cooperative therefore asks that every warning be treated as real, including the tenth in a row that came to nothing. A population that learns to dismiss warnings loses the protection the system was built to provide, and that loss is invisible until the day it matters.

What the systems are worth

Honest accounting is difficult. The systems have existed for a short time, the hazards are rare, and the injuries avoided cannot be counted directly. What the cooperative publishes is the measurable part: the lead times achieved, the events warned, the events missed, and the warnings that came to nothing. Those figures appear on the page for each system and are revised as the record lengthens.

The greater share of the value falls on a day that has not occurred. The Republic has recorded no event above REP 8 since 1901. On the day one occurs, most of the eastern region receives between fifteen and forty seconds of notice, the western sub-regions rather more, and the schooling embedded in every ordinary broadcast will have been done many times over by then.

Published and open to forking

All three systems are published in full: source, wire format, conformance vectors, station firmware and the mechanical drawings of the sensor housings. Any country, cooperative or person may take them, adapt them and operate them, at no charge and under no condition.

The earthquake system exists because Japanese engineers published enough of their work for it to be reproduced. The flood and landslide models were built here at considerable cost, and countries with hazards resembling those of the peninsula should be spared that cost.

The porting guide, with the parts requiring local calibration