Yellow River smart platform — thirteen business lines behind one flood-control portal

A smart application platform for a Yellow River authority: space-air-ground IoT sensing of hydrology, weather and river conditions, feeding pre-flood warning, in-flood dispatch and post-flood review for ecological protection and flood safety. One portal aggregates thirteen business modules, from flood monitoring, works management and the water-resources one-map to water administration, party building and device operations. De-identified presentation.

Industry
River authority / public-sector information systems
System type
Portal aggregation + business modules + IoT sensing + wall dashboard
Key technology
Space-air-ground sensing · video ingest · sluice water-level monitoring · AI alerts · GIS one-map · multi-unit data isolation
Jiding scope
Requirement shaping, portal and module architecture, front and back end, device integration, launch support
13business modules behind one portal: flood control, works, water, party building
7layers pinned on the water-resources one-map: works, video, quality, intake, patrol
3trace channels: human patrol, engineering alerts and AI alerts, one disposal loop
3flood phases: pre-flood warning, in-flood dispatch, post-flood review — one dataset
Data flows up into one map; permissions flow down to each unit Decision Business Alerts Sensing Wall dashboard live weather · alerts · flood info · real-time hydrology One-map summary & filing water-resources map · four-forecast entry Flood control 6 screens Works classified Water map one-map · 7 layers Water admin regulations Detention area 7 quick entries E-gov site public portal Party build 5 content types Culture culture · integrity Patrol app patrol · works · AI alerts Sluice monitoring water intake · level curves Device ops register · fault analysis Unified login accounts · unit switching Video river · works · sluice Sluice gauges level & layout sensing Weather & hydrology space-air-ground feeds Patrol terminals field station reports Modules can ship one at a time: the portal chassis comes first, and each new line reuses login, unit switching, the dashboard and device ops.
Four layers: sensing sees, the alert layer reads it correctly, business modules govern, the decision layer explains — thirteen modules share one sensing and alerting base, referencing each other instead of copying data

Three gaps: split lines, no time to spare, paper on the ground

The managed asset is a river hundreds of kilometres long, with works, sluices and floodplains along it. The business is not complex; what is complex is that several units, several lines and several generations of systems coexist. Those three breaks decide that integration must come before features.

  • Each line keeps its own silo. Flood control lives in one system, works registers in a spreadsheet, water quality and intake in another package, party building in a fourth. Leadership asks for reports one by one; reconciling the different definitions is itself a translation exercise.
  • Flood season runs on minutes. Rainfall is on one screen, response rosters on a sheet, sluice levels at the site, video in another client. During a flood, the decision-maker is never short of data — only of the half hour it takes to assemble it.
  • The field runs on paper. Patrols are noted in a notebook, photos sent over chat, works alerts arrive by duty phone, and what an AI could watch goes unwatched. How long a device has been faulty, and which devices fail repeatedly — without a register, operations runs on memory.
One decision shaped the platform: the hard part of river-authority digitalisation is integration itself. So the chassis comes first — unified login, unit switching, the wall dashboard and device operations — and thirteen modules hang off the same data chain. Modules can ship in phases; the chassis cannot be split.

Thirteen modules, one portal

The manual reads as thirteen parallel entries, visited in turn. Grouped by user, they are three kinds: flood and water business, sensing and alerts, governance and operations — one login, one chassis.

ModuleWhat is insideWho uses it
Flood controlUnits, flood mitigation, key flood-defence sites, rainfall statistics, flood information, response forcesFlood office: the screen opened daily before and during the season
WorksClassified works register per unit: embankments, sluices and danger points, each archivedWorks management: who owns this reach, of what class
Water resourcesOne-map with seven layers: works, video, running state, water quality, intake records, patrols, e-governmentWater resources and dispatch: seven dimensions at one pin
Water adminClassified water-administration regulations per unitEnforcement: legal basis by unit and category
Detention areaQuick entries to flood mitigation, works, water resources, enforcement, culture, the four-forecast platform and the one-mapDetention-area management: seven lines, one entrance
E-governmentIndependent public website for information releasePublic and communications: disclosure lives apart from the platform
Party buildingMember records, videos, the three-meeting one-lesson cycle, activities, analyticsParty affairs: organisation life and member data in one book
CultureOrganisational culture and integrity-building contentCommunications: culture results in one place
Patrol appPatrol records, engineering alert details, AI alert detailsField stations: what was patrolled, what is alerting, what the AI saw
Sluice safetyWater-intake sluice monitoring: statistics, curves, layout diagram, overview, water level, weatherSluice keepers: level curve and layout on one screen
Wall dashboardLive weather, alerts, flood information, real-time hydrologyCommand hall: the first screen of every flood shift
Unit switchQuick entry to switch between subordinate unitsEveryone: data scoped by unit, one tap to change view
Device opsDevice counts, alarm and fault analysis, daily statistics, videoOperations: the register decides which device gets replaced

Two easy-to-miss details: the privacy-policy checkbox at login and the unit-switch entry on the portal are not features — they are the chassis. One puts compliance into the front door; the other lets one system serve many units without data leaking across.

One flood, six steps, one dataset

The value chain of flood work runs pre-flood warning, in-flood dispatch, post-flood review. Unpacked, it is six steps, each owned by a module — and crucially, none of the six changes systems: from rainfall to review, it is one flood dataset.

123 456 Pre-flood · warn In-flood · dispatch Post-flood Rainfall stats auto per station Key sites sites · forces Warning graded, to people Live monitoring hydrology · video Dispatch sluice · crews Review replay · debrief flood module refreshed hourly sites and forces counted by unit level from rain pinned on the wall dashboard hydrology sluice curves beside it intake records actions on the record every step replayed plans revised on facts Pre-flood rainfall, in-flood hydrology and post-flood review all read one flood dataset — nobody carries numbers between systems Step 4 is where the minutes are won: space-air-ground sensing puts hydrology, weather and river state on one screen, so dispatch shifts from assembling data to reading it. Review is not a written summary here; it is data replay — how heavy the rain was, where the level peaked, who ordered what, all on the record.
Flood loop: warning release and live monitoring are the two critical nodes; all six steps draw on the same four modules — flood control, dashboard, water resources, patrol app — with no system switching anywhere

Why rainfall statistics own the first screen. The first input to every flood decision is rain. Rainfall is never reported by hand; it aggregates from stations and refreshes hour by hour, setting the warning level and deciding whether response forces move forward. So rainfall statistics, key sites and response forces live in the same module: whoever reads the rain can see the sites and the crews without opening a second page. Every minute of a flood season is worth saving.

Why the four-forecast platform is an entry, not a system. Forecast, warning, rehearsal and plan are the standard agenda of water digitalisation, yet every one of them feeds on existing data: forecasts on rainfall and hydrology, rehearsals on works layout, plans on response forces. So the detention-area module presents the four-forecast platform as a quick entry alongside the rest. Data does not move and definitions are not re-invented — the capability grows on existing data instead of becoming another island.

The water-resources one-map: seven layers on one set of coordinates

The water-resources module is the platform's map base: works, video, running state, water quality, intake records, patrols and e-government — seven layers sharing one set of coordinates, so opening any pin surfaces all seven dimensions at once.

Embankment 01 Danger reach 02 Running state Water-intake sluice Video Video Water quality Works Video Running state Quality Intake Patrols E-government Seven layers Open any pin: the archive, running state, nearby cameras, latest quality and intake appear together The seven layers do not share tables — only coordinates. Data stays where it is owned; the map only references it Works / sensing point Key sluice Video
The one-map: a reference view of seven layers, not a seventh database — each line maintains its own module, and the map pins them onto one coordinate system

A trade-off: aggregate the map, or rebuild the data. The thorough path is a unified warehouse with works, quality and intake migrated in. But each line has its own systems under construction and its own maintenance rhythm; migration turns a data problem into an organisational one. We shipped map aggregation without moving data: the map references coordinates and key states, and each line keeps owning its data. The cost is that the map is not a live database and can go stale if a module stops updating — a visible, attributable cost, and cheaper than the organisational risk of migration.

Why video appears in three places. Video is a layer on the one-map, an operations object in device ops, and a review tool during flood duty — one capability, three uses: dispatch looks at the site, operations looks at uptime, alerts look at the replay. The implementation has one stream and one register; the three places are references. Add or remove a camera once and all three follow — the most direct dividend of building the chassis first.

The patrol app: three trace channels, one disposal loop

Field routine is patrol, report, repair. The patrol app splits it into three channels: patrol records produced by people, engineering alerts raised by thresholds, and AI alerts raised by video analytics — recorded separately, disposed together.

Human patrol field stations · scheduled and special Device thresholds sluice levels · running parameters Video AI camera feeds · model detection Patrol records time · person · place · issue Engineering alerts which device · by how much AI alerts which camera · what · when Disposal loop Handled bynamed Resultresolved / to repair Closed atsystem-written · fixed Three channels, recorded apart, disposed together: every alert shows who looked at it, what was done and when it closed — AI sees fast, people carry the responsibility
The patrol app: different sources, separate details, one loop — patrol is human judgement, alerts are machine judgement, and both meet in a disposal record that lands on a name

A trade-off: should AI replace patrols, or run alongside? AI watches tirelessly and could absorb much of the patrol load. But a river produces false positives — birds, glare, drifting debris. If AI dispatched crews directly, the field would drown in false alarms and then ignore the channel. So the two run in parallel: AI alerts form their own list for duty staff to screen and dispose; patrols continue by the book. AI widens the field of view; people confirm and take responsibility — the two lists sit side by side, and missed detections versus false alarms are visible for what they are.

Why disposal time is recorded. Anyone can build an alert list; the hard question is how long an alert lay unhandled. The closing time is system-written and cannot be edited by hand — the least glamorous and most valuable field in the system. With it, monthly reports can compute average closure time per alert class, which is the raw material of fault analysis. Without it, an alert system is just a louder phone.

Sluice safety and the wall dashboard: a curve for one point, a wall for the whole river

The sluice module watches one structure (the water-intake sluice); the dashboard watches one hall. One manages a point, the other a surface — and they share the same data.

ModuleWhat it showsWhat it solves
Sluice safetyPlatform statistics, curves, layout diagram, overview, water level, weatherA health sheet for one sluice: level curve and layout on one screen, adverse trends visible at a glance, weather explaining why the level looks the way it does
Wall dashboardLive weather, alerts, flood information, real-time hydrologyThe hall's first screen: no interaction required, the overview is always up; alerts pinned at the top where nobody can pretend not to see them

Why build a whole module around one sluice? Because a cross-province water-intake sluice carries a risk class far above ordinary works. Ordinary works need to be findable; a key sluice must be watchable — curves, layout, overview, level and weather organised around a single structure. The allocation principle behind it: sensing investment follows risk, not fairness. Six kinds of sensing on one critical sluice beat one kind spread across twenty.

The dashboard's value is not decoration but shared fact. Its four elements — live weather, alerts, flood information, real-time hydrology — come from the business systems in real time, not a slide show. During flood duty, command, duty staff and dispatch all read the same numbers; after the event, the screen doubles as the decision record.

Unit isolation and device ops: two ballast stones under the chassis

A river authority is organised as bureau, subordinate units and field stations. One platform serves them all, and two things decide whether it survives contact with reality: how well unit-level data is isolated, and how honestly devices are registered.

Unit switching: one system, many views

After account login and the privacy-policy confirmation, the portal offers a switch entry for subordinate units. Each unit sees only its own works, patrols and alerts; the bureau sees everything.

Because the switch lives in the portal rather than inside each module, the permission model is built once: thirteen modules share the same unit-role-data-scope model, new modules inherit it automatically, and "visible here, invisible there" cracks never appear.

Device ops: the equipment's medical record

Device ops maintains four things: device counts, alarm and fault analysis, daily statistics, video. Together they answer two questions: how many devices are working right now, and which ones fail repeatedly.

Fault analysis aggregates history per device: a level gauge that alarmed twenty times in a month is the one to replace, not whatever broke last week. Sensing hardware will fail; the module's job is to turn "it broke" into a planned, expected repair.

Three chassis pieces — unified login, unit isolation, device operations — are the least impressive in a demo and the most appreciated in year two. Business modules decide what the system looks like at launch; the chassis decides what it looks like three years later.

What the client got

  • A portal chassis — login with privacy confirmation, unit switching, unified navigation for thirteen modules.
  • The flood line: six flood-control screens, the detention-area quick entries including the four-forecast platform, and a wall dashboard of four live feeds.
  • The one-map: seven water-resources layers on one coordinate set, referenced rather than copied.
  • The alert line: patrol records, engineering alerts and AI alerts converging into one disposal loop with fixed closure times.
  • Key sensing: the water-intake sluice with level curves and layout, plus video along the line.
  • Governance and operations: the public e-government site, regulations, party building, culture, and device operations with fault analysis.
As agreed with the client, this page shows no unit names and no screenshots of production screens.

Building a multi-unit, multi-line platform?

Integration platforms for water authorities, government and campus estates rarely fail on a single feature. They fail on three questions: the portal chassis, agreed data definitions, and the permission model. We have ground these down across public-sector and IoT projects — bring your structure and we will tell you which of the three is your bottleneck. Related: smart factory & IoT and how we build.

MORE CASES

More Related Cases

Yellow River · smart application platform
Government SystemDigital Twin,Smart Yellow River

Yellow River · smart application platform

A smart application platform for a Yellow River authority: space-air-ground IoT sensing of hydrology...

View Details →
River sand-mining management platform
Government SystemSmart sand production site, smart manufacturing

River sand-mining management platform

An IoT management platform for river sand extraction and aggregate production: eight modules and 24 ...

View Details →

View All Cases →

Planning a similar production line or enterprise system?

Start with our pricing logic, then let's talk requirements — cost structure, technology comparison and acceptance criteria, all laid out in the open.