Stream and rain gauge telemetry for flood warning programs

Reference article · 7 min read

A flood warning network earns its keep during a handful of hours every few years. The rest of the time it is a maintenance obligation: batteries to replace, vegetation to cut back, sensors to verify, and data to publish that almost nobody reads. Agencies rebuilding an aging ALERT network usually find that the original design assumptions no longer hold. A repeater path is blocked by new construction, the base station software runs on an operating system nobody supports, radio licenses need attention, and the staff who knew the network by heart have retired.

What follows is a specification guide written for the agency side of that rebuild: a county flood control district, a regional flood authority, or a city public works department that owns gauges, shares data with the National Weather Service, and has emergency managers who act on alarms. The goal is a network that stays up in the storm that matters and that a small team can maintain in the quiet years between storms.

Choosing sensors and siting them honestly

Rainfall. Tipping bucket gauges remain the workhorse and are commonly specified at 0.01 inch per tip. They are known to undercatch at high intensity and in wind, so siting matters more than brand: open exposure, away from trees and structures, with a wind shield where conditions warrant. Heated or weighing gauges are worth considering where frozen precipitation is part of the hazard.
Stage. Options include vented pressure transducers, non-contact radar, bubblers, and float and shaft encoders in a stilling well. Non-contact radar avoids debris and sediment contact and is often easier to maintain on bridge structures. Whatever you choose, tie readings to a surveyed datum and install a staff plate for independent verification.
Discharge. Only publish discharge if you maintain a rating. Where the channel shifts often, index velocity instruments provide a more defensible basis than a stage rating alone.
Survivability. The site must survive the flood it is meant to measure. Mount enclosures, antennas, and solar panels above the design flood elevation, anchor against debris impact, and consider bank erosion and channel migration over the life of the installation.

Backhaul: radio, cellular, and satellite

Licensed radio. A self-owned VHF or UHF path has no recurring carrier fee and no dependence on a commercial network. ALERT2 was developed to address the collision and error-detection weaknesses of the legacy ALERT protocol. Radio requires licensing, path studies, repeater sites, and someone who owns that infrastructure.
Cellular. Inexpensive, two-way, high bandwidth, and fast to deploy. It also depends on carrier towers with their own battery and generator limits, and on network capacity during a regional emergency.
Satellite. GOES Data Collection System transmissions require authorization from NOAA and assigned platform identifiers and transmit windows, which constrains how often a site reports. Commercial satellite services offer two-way messaging at a per-message cost.

Specify dual path on your most consequential sites, and be honest about what redundancy means. Two paths that both terminate at the same repeater, or both ride the same carrier, are one path.

Power budgets and reporting intervals

Size the power system for the worst case, not the average. The trap is that transmit duty cycle rises during an event at the same time solar input falls, so calculate several consecutive overcast winter days at full event reporting rate. Specify a charge controller with low-voltage disconnect, and telemeter battery voltage and panel current as data points so you can trend a failing battery before it strands you.

Separate reporting modes in the specification. Rain gauges typically report per tip or per accumulation increment. Stage sensors report on change of state plus a scheduled heartbeat. That heartbeat is your failure detector: a site that misses a defined number of consecutive check-ins should raise a maintenance alarm that is visibly different from a hydrologic alarm.

Thresholds, alarms, and escalation

Set thresholds against consequences, not round numbers. Useful anchors include low water crossing and road overtopping elevations, first floor elevations of known structures, rate of rise over a short window, and rainfall intensity across multiple duration windows.

Escalation needs to be written down and tested. Define primary on-call, secondary, and supervisor tiers; use more than one notification channel; require positive acknowledgement; and log every notification with timestamps for the after-action review. Exercise the roster on a schedule rather than discovering during an event that a phone number changed. Nuisance alarms destroy trust, so allow tuning, but record who suppressed what and why.

The public page takes its heaviest traffic in the same hour the network is most stressed. Keep it plain and fast, design for phones on a degraded connection, and host it separately from the acquisition system so a surge in readers cannot affect data collection. Every reading should display its observation time, and stale data should be marked as stale rather than shown as current. State plainly that data are provisional and unverified. Meet your accessibility obligations, which generally means WCAG-conformant markup and a text equivalent for every chart.

For sharing, coordinate early with your local Weather Forecast Office and River Forecast Center on ingest paths and station metadata, and with USGS on any co-located or cooperatively funded gauges. Publish a machine-readable feed so neighboring jurisdictions and regional networks can consume your data without scraping a web page.

Maintenance, rating curves, and degraded operations

Budget recurring site visits, not just capital. Each visit should verify rain gauge calibration, compare sensor stage against the staff plate, check the datum, clean intakes, and inspect antennas, grounding, and panel shading. Rating curves shift after channel change, and a curve extrapolated beyond its measured range is an estimate. Document shifts and the basis for them.

Write the degraded operations plan before you need it. Name the fallbacks: manual staff gauge reads by field crews, radar-based precipitation estimates, neighboring agency networks, and trained spotters. Define who declares the network degraded, how the public page reflects that, and how alarms are handled by hand in the meantime.

What to ask and what to require

Ask for a documented radio path study or cellular coverage verification at each proposed site, not a coverage map.
Require open, published protocols and a documented data model, and confirm the agency owns the raw observations and the historical record.
Require an export path and, where practical, source code escrow or an open interface for the base station software.
Ask how firmware is updated in the field and what the security posture is for any network-connected device.
Require itemized spare parts and service pricing for several years, plus lead times.
Make acceptance testing include a simulated event: forced alarms, escalation to a live on-call roster, and a failover to the secondary path.
Require training for field technicians and for the staff who will tune thresholds, and get the training materials as deliverables.

Confirm current requirements for radio licensing, satellite transmission authorization, and any grant program you plan to use directly with the responsible agency, since program rules and eligibility change.

HQT works with state and local agencies on systems engineering, integration, and software for field telemetry and public-facing data services.

Frequently asked questions

What is the difference between ALERT and ALERT2?

ALERT is the original one-way radio protocol widely used in United States flood warning networks, and ALERT2 is a later protocol developed to address its known weaknesses, including transmission collisions and the lack of error detection. ALERT2 uses time-division techniques and adds error checking so reports are less likely to be lost when many sites transmit during a storm. Many agencies operate mixed networks during a transition, so confirm that any base station and field equipment you buy can handle both.

Should a county flood warning network use cellular instead of licensed radio?

Cellular is cheaper to deploy, two-way, and easy to expand, but it depends on carrier towers that have their own power limits and can be congested during a regional emergency. Licensed radio costs more up front and requires repeaters, path studies, and license management, but the agency controls the path end to end. Many agencies use cellular for routine sites and keep radio or satellite as a second path on the gauges that drive life-safety decisions.

How often should stream and rain gauges report during a storm?

Most networks combine event-driven reporting with a scheduled heartbeat rather than fixed polling. Rain gauges report per tip or per accumulation increment, and stage sensors report when the water level changes by a defined amount, which keeps traffic low in dry weather and increases resolution during an event. The heartbeat interval is separate and serves as failure detection, so missed check-ins raise a maintenance alarm rather than being mistaken for stable conditions.

How do we share our gauge data with the National Weather Service?

Start by contacting your local Weather Forecast Office and the River Forecast Center that covers your basin, since they can tell you which ingest paths and metadata formats they accept for your area. You will generally need accurate station metadata, including surveyed datum, coordinates, and sensor descriptions, before data can be used operationally. Publishing a machine-readable feed also makes it easier for neighboring jurisdictions and regional networks to use your observations.

What should we do when the telemetry network fails in the middle of a flood event?

Have a written degraded operations plan that names who declares the network degraded, how the public gauge page communicates it, and what fallbacks are used. Common fallbacks include manual staff gauge reads by field crews, radar-based precipitation estimates, data from neighboring agency networks, and trained spotter reports. Decide in advance how alarms will be handled manually, and review the response in an after-action for every significant event.