ManuAI
VI

Simulation

Will the river reach Ben May?

Ben May is a made-up commune on the May River, at the foot of the hills. Rain that falls upstream takes a day or so to get here. You are the forecaster this week: set the rain, read the forecast, decide when to warn people.

Kit

I run the same river model 40 times with slightly different rain, because nobody knows next week's rain exactly. Where the runs agree, you can trust the forecast more.

This week upstream

Try a week

Rain upstream, mm per day

Drag a bar, or focus it and use the arrow keys.

20
D1
60
D2
240
D3
180
D4
40
D5
10
D6
0
D7

Wet ground can't soak up much more, so more rain runs straight into the river.

River level at Ben May

River level at Ben MayLine chart of the forecast river level over seven days. The shaded band holds the middle 8 of every 10 runs; the dark line is the middle run.0 m2 m4 m6 mD1D2D3D4D5D6D7
8 in 10 runs fall inside the bandMiddle runLowest homes (3.2 m)

Your call

Flood warning: move people

100% of runs flood the riverside homes, the wettest ones around Day 4, 06:00. Warn people by Day 3, 06:00 so they have a day to move children, older people and animals uphill.

Middle run peaks at 4.5 m around Day 5, 12:00. At the peak, most runs fall between 4.1 m and 4.9 m.

Ben May, seen from above

Day 5, 12:00

Ben May, seen from aboveMap of the village beside the river. Each zone shows the share of runs in which it is under water at the chosen time.from the hills ↓Rice fields100%Riverside homes100%Market95%School50%Clinic road5%Hill hamlet
  • Rice fieldswet in 100% of runs
  • Riverside homeswet in 100% of runs
  • Marketwet in 95% of runs
  • Schoolwet in 50% of runs
  • Clinic roadwet in 5% of runs
  • Hill hamletdry

How the real thing works

How AI helps warn people days earlier

Kit, the ManuAI fox, pointing

My river here is a toy with six numbers inside. Real forecasters work with far more data, and AI is one of the newer tools they use.

  1. 01

    It learns from the past

    A flood model can be trained on years of rain records, river gauge readings and past floods. It learns how a particular river answers a particular kind of storm, including the delay before the water arrives.

  2. 02

    Terrain matters

    Slope, soil, forests, dams and towns change how fast rain reaches a river. Maps from satellites and surveys give the model that shape of the land, so it can say which streets sit lowest.

  3. 03

    Many runs, one honest band

    Weather forecasts are unsure, and more unsure further ahead. Running the model many times with slightly different rain gives a spread. A wide band says: get ready, but we don't know yet.

  4. 04

    Days, not hours

    For big rivers, research and some public services now forecast floods several days ahead, even where there are few gauges. Flash floods from a short, very heavy storm are much harder to see coming.

What this toy leaves out

Real rivers have dams that release water, tides near the coast, landslides and blocked drains. A real forecast is checked by people who know the river, and the warning comes from the authorities. In Vietnam, follow the National Center for Hydro-Meteorological Forecasting and your local commune or ward.