rained.cloud

A Century of Weather Data Preserved

Status
Live
Stack
FastAPI, Next.js, Flutter, PostgreSQL, Azure, Supabase
Contents
  1. Notebooks
  2. What forty years looks like
  3. Eight platforms, one gauge
  4. Decile engine
  5. Privacy and ownership
  6. What’s preserved

Forty years of rain, written by hand, on paper that was failing.

The Bureau runs about 7,000 official stations, and the gaps between them span hundreds of kilometres. In those gaps the only record is private, handwritten, and held on paper that yellows.

Notebooks

A farmer in rural Australia has been writing down the rain every morning for forty years: the same walk to the gauge, the same pencil, the same ruled notebook. The numbers are small and precise: 3.2mm, 0.0, 14.5, Trace.

13.220.0314.54Trace567mm
Figure 1The same walk to the gauge, every morning.

That is not unusual. Thousands of landholders across the country keep private rainfall records that never reach the Bureau of Meteorology’s database. The Bureau tracks roughly 7,000 official stations, and the gaps between them can span hundreds of kilometres. In those gaps the only climate record is handwritten, private and increasingly fragile.

These notebooks had survived droughts, floods, shed relocations and mice. They would not last indefinitely: the paper was yellowing, the ink was fading and some entries were already illegible. Forty years of daily observations, the kind of continuous, location-specific record climate researchers would pay for, was slowly disappearing.

rained.cloud began as plain preservation: digitise the notebooks before they were gone. The scope grew from there.

What forty years looks like

Once the data was digital, it showed patterns the notebooks hid. A dry spell in 1994 that felt catastrophic at the time was unremarkable next to 1982. The “good years” of the early 2000s really were anomalous, 30% above the long-term mean. The last decade’s rainfall, which felt normal, was tracking below every prior decade on record.

Long-term data corrects memory.

Farmers make decisions based on what they remember, and what they remember is biased toward the recent and the dramatic. A forty-year dataset doesn’t have those biases.

Digitised rainfall record (sample station, annual totals)
Year   mm    Decile
─────────────────────────────────────────────────────────
1984   523   D3   ████████████░░░░░░░░░░░░░░░░░░░░░░░░░
1988   412   D1   █████████░░░░░░░░░░░░░░░░░░░░░░░░░░░░
1992   687   D5   █████████████████░░░░░░░░░░░░░░░░░░░░
1996   845   D8   █████████████████████░░░░░░░░░░░░░░░░
2000   921   D9   ██████████████████████░░░░░░░░░░░░░░░
2004   756   D7   ██████████████████░░░░░░░░░░░░░░░░░░░
2008   612   D4   ███████████████░░░░░░░░░░░░░░░░░░░░░░
2012   543   D3   █████████████░░░░░░░░░░░░░░░░░░░░░░░░
2016   478   D2   ████████████░░░░░░░░░░░░░░░░░░░░░░░░░
2020   502   D2   ████████████░░░░░░░░░░░░░░░░░░░░░░░░░
2024   561   D3   █████████████░░░░░░░░░░░░░░░░░░░░░░░░
─────────────────────────────────────────────────────────
Deciles calculated to BOM standards

The Bureau of Meteorology’s decile system ranks each period against the full historical record, so it only works with enough history: the standard minimum is thirty years. Most private stations don’t have that. This one does.

Eight platforms, one gauge

Entering a reading has to be easy enough that it happens every day without fail. Miss a day and the record has a hole. Miss a week and the monthly total is unreliable. Miss a month and the year is compromised.

rained.cloud covers as many platforms as it can. The same observation can be entered from:

  • A mobile app (iOS, Android) with offline support and GPS tagging
  • A web dashboard (Next.js) with interactive charts and station management
  • A voice command via Google Home — “Hey Google, record 4.2 millimetres”
  • A WearOS app for smartwatch entry at the gauge
  • Desktop apps for macOS, Linux, and Windows via Flutter

Every platform syncs to the same PostgreSQL database via Supabase. Conflict resolution handles the inevitable case where someone records on their phone and their partner records on the web. Background sync covers connectivity gaps, which are common in rural areas where mobile coverage is intermittent at best.

one record
Figure 2Many ways in, one record.
Platform architecture
Mobile (Flutter)    Web (Next.js)    Voice (Google Home)
        │                 │                  │
        └────────┬────────┴──────────────────┘
                 │
        Supabase (Auth + Realtime)
                 │
     PostgreSQL + PostGIS + TimescaleDB
                 │
      FastAPI (Analytics + BOM Deciles)
                 │
      Azure (CI/CD + Static + API)

The QR code setup at m.rained.cloud makes configuring a new device a matter of seconds: scan, authenticate, start recording. That matters when the target user is a farmer rather than a developer.

Decile engine

Raw rainfall numbers are useful. They are a lot more useful in context.

rained.cloud calculates rainfall deciles to the Bureau of Meteorology’s exact standard, the method used in official drought declarations. A decile ranking says where a period sits against the entire historical record: Decile 1 is the lowest 10% on record, Decile 10 the highest.

d1d2d3d4d5d6d7d8d9d10nowdriestwettest
Figure 3Where a quarter sits against the whole record.

A farmer whose current quarter is Decile 2 does not have to guess whether it has been dry: against a century of data, it has been drier than 80% of all recorded history. That changes decisions.

Where government records are available, the analytics layer merges them with private station data, which gives a view neither source has alone. A farm with forty years of private records can compare against the nearest BOM station’s hundred-year dataset, and the correlation analysis shows how representative, or how different, the farm’s microclimate is.

Privacy and ownership

Not every landowner wants their rainfall data public. Some consider it commercially sensitive, since a better-than-average rainfall profile adds value to a property. Others simply prefer privacy.

rained.cloud offers both public and private stations. Private data is encrypted and inaccessible to anyone without explicit access grants. Public stations contribute to a growing national dataset that fills the gaps between official BOM stations.

The workspace model supports multiple stations per user, with granular sharing controls. A farming family might keep their home station private but share a community gauge’s data publicly. A research institution might aggregate multiple private stations with landowner consent.

What’s preserved

The original notebooks aren’t going back into the shed. The data they held, forty years of a single location’s relationship with rain, is now permanent, queryable and contextualised against a century of national records.

The system also prevents future loss: every reading entered today goes into a record that will outlast the paper it would otherwise have been written on.

For the farmer who started writing numbers in a notebook four decades ago, the platform does what they were already doing, in a form that lasts.

rained.cloud is live at app.rained.cloud. Contact rained@drksci.com for partnership or data access.

  • Climate