Observing a river basin: a reference design

01What a river basin needs to see

A river basin observing site supports one decision above all: when to warn people downstream, and how strongly. The international manual on flood forecasting names water levels in rivers and lakes, river discharge and sometimes groundwater as the key hydrological observations¹.

Rain is the other half of the picture. Rainfall intensity, duration and forecasts are prerequisites for a forecasting system, and they are needed in real time to make the most of the warning lead time¹.

Water level ranges at a gauge can be linked to the extent of flooding, so a series of trigger levels can raise warnings through telemetry¹. The relationship between levels upstream and downstream, and the time a peak takes to travel between them, has long been a key means of prediction¹.

Observation is one link in a longer chain. The international early warning initiative rests on four pillars, with detection, observation, monitoring and forecasting as one of them alongside risk knowledge, dissemination and preparedness².

The reference site places water-level gauges along a reach, a weather station on the valley floor and an edge node with a radio tower, all joined to a forecasting centre.

02Siting a gauge where level stands for flow

A water-level gauge is useful for warnings when its level can be turned into flow. The stream gauging manual lists what makes a good site: a straight reach, flow confined to one channel, a stable bed, and a control that keeps the stage sensitive to flow³.

It also asks that the reach be straight for about ten times the stream width above and below the gauge, where the control is the channel itself³. Instruments and shelters must stand above all flood levels, and the site must be within reach of a suitable telemetry system³.

Rarely will an ideal site be found, and judgement is needed to choose among imperfect ones³. A gauge just upstream of a confluence or within tidal reach, for example, records the other stream or the tide as well as its own flow.

Level becomes flow through a rating, the stage-discharge relation built from field measurements of discharge. The manual covers how ratings are drawn, extended beyond measured flows, shifted as the channel changes and corrected for ice⁴. Field discharge measurements themselves follow published techniques, now including acoustic Doppler instruments⁵.

03Lead time is the hard constraint

The basic principle for lead time is the minimum advance warning needed for people to act effectively¹. Households may need one to two hours to move belongings and place barriers, while evacuating people along a large river may need days¹.

Lead time is won upstream. Telemetered rain gauges and radar add warning time beyond extrapolating river levels, and a rainfall forecast fed into a catchment model can add several hours more¹. In small, urban catchments the response can be so fast that effective warning is extremely difficult¹.

Gauges far upstream are often far from power and mobile coverage, so they report over low-power wide-area radio. Such networks are standardised internationally⁶, and in shared bands the harmonised European standard caps how much of the time a device may transmit⁷.

The simulator shows the trade. A slower, longer-reaching radio setting spends more time on air for each message, so fewer messages fit within the duty-cycle limit, which matters most when a flood makes every reading urgent.

04From sensor to warning

Flow gauges at forecast points should have telemetric links to the operational control centre, and forecast points usually sit at gauges near towns or other places where flood impact would be high¹.

In the reference design, the edge node gathers gauge and rain data, checks them against trigger levels and forwards them to the forecasting centre. Local triggers can sound before the centre has run its models, which buys minutes where minutes count.

Warnings leave in an open, non-proprietary message format for all kinds of alerts, with geographic targeting, several languages and support for updates and cancellations⁸. Sirens, phones, broadcasters and road signs can then act on the same message.

05What travels with each value

A discharge figure depends on the rating that produced it, so the rating travels with the value. An open information model for water observations covers time series and, in a separate part, ratings, gaugings and channel sections⁹.

Each value leaving the site therefore carries its stage, the rating version used to convert it, the time, the gauge and its datum, and its quality flags. When a rating shifts after a flood, past values can be recomputed and republished as corrections, not silently overwritten.

Storage at the gauge serves the same end. A telemetry system should buffer measured data until they can be sent on for processing¹, so a reading taken during an outage keeps its own time.

06When links and power fail

Floods bring down the links and power that flood warning needs. The manual asks that telemetry for flood warning be designed for continuous operation in heavy rain and floods, with duplicate power or communication lines and backup generators for key functions¹.

Where there is no mains power, batteries, solar or wind supply the stations, with backup batteries sized for the local weather; the manual puts typical backup periods between one week and one month¹. It also recommends a backup operations centre on a completely different site¹.

The reference design follows the same logic at the edge. Local triggers keep running when the link to the centre is lost, and stored readings are sent on in order once it returns. A silent gauge is shown as silent, with its reason, rather than as a calm river.