Cheia

A flood simulator for Curitiba and its metropolitan area that runs in a browser tab. Pick a rainstorm, in mm/h or from Curitiba's own intensity-duration-frequency curve, and watch the water run down the real terrain, fill the rivers and spill into the neighbourhoods. A live ranking lists the neighbourhoods and towns with the most land under water, and how many schools, hospitals and kilometres of street the water reached. Search for your street and it tells you at what time the water passes 30 cm there. Drop a storm cell on a neighbourhood, replay the storm of 18 February 2025 from the rain gauges, or run the next 48 hours of forecast. With a strong El Niño forming, a panel puts the next six months of forecast rain and heat next to what past El Niños did.

Open it · Nothing to install · Interface in Portuguese · Leia em português

2.4 million cells of water, one shader pass per step, 5.9 ms a step
81 fps on an integrated laptop GPU, 4 simulated minutes per second
storm of 18 Feb 2025: the model flooded 9 of the 18 reported spots
                      (by chance, with the same yardstick, about 4)
2-year storm (the every-summer one): 1.78 m at Wenceslau Braz x Linha Verde

On this page: what the simulation found · what you can do · checked against a real storm · the next six months · methods · what it can't see

A 100-year storm (98 mm in one hour), two simulated hours, recorded from the browser. Light blue is a few centimetres, purple is two metres and more.

I went looking for what already existed for Curitiba before writing any of this. Google's Flood Hub forecasts slow river floods and says plainly that it does not cover flash floods in large cities, which is what Curitiba gets. The state's drainage master plan has flood lines per river, as PDFs from 2002. Researchers have mapped flood risk for single basins. Nothing let you ask "what if it rains 80 mm in an hour, where does the water go?" and see the answer on the map.

What the simulation found

I ran Curitiba's design storms (2-, 10- and 100-year one-hour storms, and a 25-year three-hour one) over the whole region and measured three things: the water at the 18 spots that actually flooded on 18 Feb 2025, the share of each neighbourhood that passed 30 cm, and which schools, day-care centres, health units and streets got dangerous water. How it was done is under Methods; the full numbers are in docs/analise.json.

1. Parolin, Fanny and Novo Mundo flood even in the every-summer storm. A 2-year storm (52 mm in one hour) already puts 1.78 m of water at the Wenceslau Braz x Linha Verde junction, 1.39 m on the Henry Ford stream and 1.27 m on Rua Dr. Luiz Losso Filho, where water passes 30 cm within 5 minutes. At all three the current is already unsafe for anyone. They are valley bottoms that collect the water of several neighbourhoods.

2. The worst comes after the rain stops. In the bigger valleys the peak arrives 30 minutes to almost 2 hours after the rain starts, when the whole basin's water reaches the river. In Boqueirão (Carlos Essenfelder x Cleto da Silva) water passes 30 cm only 1h53 in, in a 100-year storm.

3. Intensity matters more than volume. The 25-year three-hour storm (more water in total) put less water on almost every spot than the 10-year one-hour storm. In Curitiba the danger is the short, heavy downpour.

4. Six of the 18 reported spots stay dry in any storm. Sítio Cercado (2), CIC (2), Fazendinha and Campo Comprido never pass 5 cm even in the 100-year storm, and the centre (Visconde de Nácar) never passes 12 cm. Those floods are drainage failures: blocked drains, undersized galleries, flat streets with nowhere to go. 30 m terrain can't see that, and the model must not be used to call those places safe.

5. The neighbourhoods with the largest share under water are outside Curitiba. In the 10-year storm, Alto Tarumã (15%), Parque das Águas (14%) and Jardim Claudia (11%) in Pinhais, Agaraú (14%) in São José dos Pinhais and Guaraituba (13%) in Colombo sit on the floodplains of the Iraí, Palmital, Atuba and upper Iguaçu. Inside Curitiba the top ones are Alto da Rua XV, Augusta, Hugo Lange and Santo Inácio (8 to 10%), partly the Belém and Barigui floodplains and parks, which are meant to fill.

The 18 spots of 18 Feb 2025, in four storms

Peak depth in a 3 × 3 cell box (about 100 m) around the spot; in brackets, how long after the rain started it passed 30 cm; H is the hazard class at the peak.

This is the beginning of the README. Read the rest at the original. Read the full README ↗