Write the rule first.
Place, data source, model, season, baseline and threshold go into one canonical spec. Its hash is stored on-chain before any money is locked.
Built only on public, reproducible inputs.
ERA5 via Open-MeteoSRPN-v1 integer indexEIP-712 attestationBase Sepolia testnet
See it work.
No wallet needed.
Four clicks from rainfall data to an enforced, independently checkable payout. Campaigns below are read live from RilvaEscrow on Base Sepolia · testnet.
1
See the rainfall record
33 October–December seasons at Yabelo against the 1991–2020 normal, recomputable in your browser.
Open the Climate Explorer2
Verify a payout
A campaign the contract released after a signed, reproducible attestation. Every check runs in your browser.
3
Verify a refund
A wetter season: the attested index was above the rule, so the contract returned the funds to the funder.
4
Try to break it
Forge a signature, change the index, settle twice. Watch the contract refuse each one, with no transaction sent.
All seasons are retrospective replays. Testnet funds only.
Drought-relief triggers are often decided out of sight: which rainfall data, how the number was computed, and who decided it was met. Rilva makes all three public, checkable and enforced.
Public rules.
Checkable evidence.
Rilva makes a drought trigger a public, reproducible and enforceable artefact, in four steps that anyone can follow and rerun.
Compute the evidence.
Daily ERA5 rainfall becomes one integer: percent of normal, in basis points. Missing days make the evidence invalid. Nothing is gap-filled.
Sign what was checked.
A named attestor recomputes the index and signs an EIP-712 statement that binds the campaign, the spec hash and the evidence hash.
Let the contract decide.
The escrow verifies the signature, compares the index with the threshold itself, and pays the fixed amount once, to a fixed address.
Case study · Yabelo, Borena, Ethiopia
One grid cell.
Three short rains.
Borena's pastoral economy depends on the October–December Hagaya rains. At the ERA5 cell over Yabelo, here is how three recent seasons compare with the 1991–2020 normal of 234.0 mm.
OND 2021
Rule met33.63 %
of normal · 78.7 mm in 92 days
CHIRPS v3 cross-check: 52.5 % (informational)
OND 2022
Rule met44.61 %
of normal · 104.4 mm in 92 days
CHIRPS v3 cross-check: 54.8 % (informational)
OND 2023
Rule not met213.22 %
of normal · 498.9 mm in 92 days
CHIRPS v3 cross-check: 140.9 % (informational)
Open the Climate Explorer
Context from humanitarian monitoring: FEWS NET, Dec 2021 ↗. Rilva does not verify impact reports.
From evidence
to settlement.
The attestor signs typed data, not a blank cheque. The contract re-checks every field before a single wei moves.
- 1
Same campaign, same spec
The signature names the campaign id and the spec hash stored at creation.
- 2
Same period
Observation dates must match the campaign exactly.
- 3
Right signer
Only the campaign's designated attestor can produce a valid signature.
- 4
Rule applied on-chain
Index ≤ threshold releases to the beneficiary; otherwise the funder is refunded.
Attestation · EIP-712
- campaignId
- assigned on-chain at creation
- indexBps
- 3363
- observationStart
- 20211001
- observationEnd
- 20211231
- expiresAt
- signing time + 1 hour
0xf1f503df82d576ee2fef554aa4bbfad0253a9612304dc1e915643abb9075bf8b0xde7fdf48a6a608f3702848d95643440ceac87c4f411ccd64fc7776653be94f50Predicted outcome: Released
Preview fields come from the real OND 2021 evidence file. No campaign shown here.
What /verify checks, with no wallet
- 1The spec hash matches the published spec
- 2The spec agrees with the on-chain terms
- 3The index recomputes from ERA5
- 4The evidence hash recomputes
- 5The signer is the campaign's attestor
- 6The outcome follows the rule
- 7The funds went to the right address
Don't trust us.
Check the math.
Every campaign has a public verification report. Your browser re-downloads the SHA-256-pinned ERA5 data, recomputes the index and the hashes, recovers the signer, and compares everything with the chain.
- Cryptographically authentic: who signed which bytes.
- Data reproducible: same inputs, same hash.
- Contract conditions satisfied: the rule was applied on-chain.
- Not checked: whether the rainfall estimate is accurate, or whether anyone was helped.
From replay
to real use.
What would have to change before a real programme could rely on this. A roadmap, not a claim: today Rilva is a testnet demonstration.
- 1
More than one attestor
m-of-n signatures or a challenge window, so no single party can settle a campaign alone.
- 2
Calibrated triggers
SPI or locally agreed thresholds, designed with the people a programme serves, instead of an illustrative 60 %.
- 3
More places and seasons
The same method with new specs: one public spec per location and season, each reproducible on its own.
- 4
Audit, then a pilot
An independent contract audit, then a supervised pilot alongside an existing anticipatory-action programme.
What stays the same: a public spec, evidence anyone can recompute, and a rule the contract enforces itself.
Fair questions.
Plain answers.
What Rilva proves, and what it does not. More in the methodology.
Is this live drought monitoring?
No. Every season shown is a retrospective replay of a past season. Rilva is not issuing drought alerts.
Is 60 % of normal an official drought threshold?
No. It is an illustrative hackathon rule. In ERA5 it would have fired in 7 of the 30 baseline seasons; a real programme would calibrate it.
Who do I still have to trust?
The attestor is trusted to sign honestly; Rilva makes every attestation checkable, not automatically true. A wrong or conflicting signature can be detected by anyone, but the contract cannot prevent it.
Is this insurance, or are real funds involved?
No. This is a testnet demonstration on Base Sepolia. No real funds, no insurance, no relief delivered. The contract has tests but no professional audit.