A national ranking of beach water quality and sanitation, built on OODS: what should it measure, and how do we keep it fair?

[h0ffmann]

🇧🇷 Resumo: quando o OODS (nosso armazenamento aberto de dados de balneabilidade) estiver consolidado, queremos publicar um ranking nacional de qualidade da água e saneamento das praias brasileiras. Antes de construir, queremos ouvir vocês: o que medir, em que nível (praia, município, estado), como comparar órgãos com métodos diferentes, e como evitar um ranking injusto ou enganoso. Comente em português ou inglês!

Why

Bathing-water results are public in Brazil, but scattered: each state agency publishes its own way (PDF bulletins, maps, CSVs), on its own schedule, with its own indicator. Nobody can answer simple questions like "which coastal cities have improved over the last ten years?" or "where does poor sanitation show up on the beach?"

The Open Ocean Data Store (MIP-0075) is consolidating these results into one open, versioned dataset, starting with IMA/SC, INEA/RJ and INEMA/BA. Once enough states are in, a national ranking becomes possible, and could be one of the most useful things marola publishes.

What we have in mind (a starting point, not a decision)

• Water quality: per point, the share of samples classified própria over the last N bulletins (our point_fitness, under CONAMA 274's classification), rolled up to beach → municipality → state.
• Sanitation context: municipal sewage collection and treatment rates from the national sanitation information system (SNIS/SINISA, to be confirmed), joined by IBGE municipality code. We already store the code for each sampling point.
• Transparency: how often, how completely and how openly each agency publishes. A state that doesn't publish shouldn't rank well by default.
• Trend: improving, stable or worsening, which may matter more than a single position.

Open questions, where we need you

1. Unit: should we rank beaches, municipalities or states? Or publish all three?
2. Comparability: agencies use different indicators (E. coli, enterococci, thermotolerant coliforms), sampling frequencies and seasons. How do we compare them honestly? Separate rankings per indicator, normalisation, or only the agency's own própria/imprópria verdict?
3. Small samples: a beach with 4 samples a year vs one with 52. Minimum counts, confidence intervals, Bayesian shrinkage?
4. Sanitation link: how do we show beach quality next to sewage coverage without implying causation the data can't support?
5. Fairness and impact: rankings can shame small municipalities with little budget, or push agencies to publish less. How do we design it to encourage transparency and investment instead? Should we rank "most improved"?
6. Coverage: Brazil has 17 coastal states. What's the minimum coverage before a "national" label is honest?
7. Presentation: a page on marola.dev, a yearly report, an open dataset with a reproducible notebook, or all of them?
8. Partners: universities, NGOs, journalists or public bodies who'd want to co-design or validate the method?

Principles we'd like to keep

• Open and reproducible: every number traceable to an agency bulletin in OODS, with the code to recompute it.
• Honest about limits: missing data shown as missing, never as clean water.
• Agency verdict first: we show the official classification and add our own summary next to it, never in its place.

How to help

Reply with your answers to any of the questions, examples of rankings you think are done well or badly, data sources we're missing (state agencies, sanitation datasets), or if you'd like to help build it. No code needed: method, domain knowledge and local knowledge are just as valuable.

This comes after OODS is consolidated (MIP-0075); nothing will be built until the method is discussed here and written up as a MIP.