Earth Observation Early Warning & Response

An open, contributor-friendly project to turn Earth observation into earlier warnings and better response decisions for catastrophic risk.

Status: Active

Started: December 2025

Contributors: 3

Project Manager:

Introduction & Rationale

Earth observation (EO) systems already monitor floods, fires, drought, storms, conflict-related destruction, and more. But in many high-stakes scenarios, the limiting factor is not "more imagery". It is the end-to-end chain from sensing to decision-ready signals: clear definitions of what to detect, how quickly, at what spatial scale, with what confidence, and how those signals translate into actions.

This project aims to produce an open, practical early-warning and response framework for catastrophic risk. The unique contribution is a decision-first EO architecture: start from the response decisions that matter, derive the minimum viable signals and latency requirements, then map to feasible sensing and processing pipelines.

What makes this project new

Decision-first requirements. We start from concrete response decisions and work backwards to sensing requirements (latency, coverage, resolution, confidence, false alarm tolerance).
Multi-hazard, multi-signal. We treat EO as a portfolio of signals (optical, SAR, thermal, RF, atmospheric) rather than a single sensor type.
Failure-mode aware. We explicitly design for degraded infrastructure: cloud cover, comms disruption, model drift, adversarial conditions, and data access constraints.
Contributor-friendly task packaging. Work is broken into discrete, mergeable tasks with dependencies and clear outputs.

Research Objectives

Define a short list of catastrophic-risk scenarios where EO can realistically change outcomes in days-to-weeks timescales.
Specify decision-ready signals and requirements (what to detect, by when, and with what confidence).
Map those requirements onto feasible architectures (sensors, orbits, ground segment, processing, dissemination).
Produce an open reference playbook: signals, pipelines, evaluation protocol, and a pathway to pilots.

Methodology (high level)

Scenario selection: pick 3–5 scenarios with clear decision points and credible data sources.
Signal specification: define the minimal signals and their requirements (coverage, revisit, latency, spatial scale, uncertainty).
Pipeline design: propose an end-to-end pipeline from raw data to alert product.
Evaluation: define metrics and backtesting approach against historical events and open datasets.
Operationalisation: produce a pilot plan and partner map.

EO Early-Warning Scenarios: Use-Case Shortlist

Completed

Inputs:

MPIL mission and catastrophic risk framing
Existing EO capabilities and public datasets
Response decisions that plausibly change outcomes (days-to-weeks)

Process:

Select 3–5 scenarios and define the key decision points and stakeholders for each.
Write a 1-page brief per scenario: what to detect, why it matters, and the rough decision timeline.
Explicitly flag limits where EO can’t support a scenario without overclaiming.

Outputs:

eoew_use_cases.md (scenario briefs + rationale)

Completed by:

Rory Dick

Decision-Ready Signal Requirements (Latency, Coverage, Confidence)

Completed

Inputs:

Selected scenarios from the use-case shortlist
Examples of EO alert products (public)

Process:

For each scenario, define minimal signals and requirements: revisit, latency, spatial scale, false alarm tolerance, uncertainty reporting.
Make requirements decision-first: tie each requirement to a real action that can be taken.

Outputs:

eoew_requirements.md (requirements table + narrative)

Completed by:

Ellroi

Dataset & Sensor Inventory (Open Sources First)

Outstanding

Inputs:

Selected scenarios from the use-case shortlist
Public EO data catalogues (Sentinel, Landsat, MODIS/VIIRS, etc.)

Process:

Build an inventory of candidate datasets/sensors per scenario, prioritising open access.
For each source, record modality, latency, coverage, resolution, and key limitations.

Outputs:

eoew_dataset_inventory.md (inventory table + notes)

Coverage/Revisit Simulation + Chart

Outstanding

Inputs:

Decision-ready latency and revisit targets from requirements
First-order constellation assumptions

Process:

Run a lightweight revisit model to estimate feasibility bands for revisit/latency.
Produce a chart for fast communication and a CSV for reproducibility.

Outputs:

Latency Budget (End-to-End)

Outstanding

Inputs:

Scenario deadlines and operational needs
Typical EO product delays and processing stages

Process:

Decompose end-to-end latency into stages and identify dominant terms.
Define target latency bands and minimum metadata requirements (timestamps, caveats).

Outputs:

Reference Architecture Sketches (End-to-End Pipeline)

Outstanding

Inputs:

Decision-ready requirements
eoew_dataset_inventory.md (completed inventory table)

Process:

Sketch an end-to-end pipeline per scenario: ingest – preprocess – model/heuristics – alert product – dissemination.
Include failure modes and degraded-ops assumptions.

Outputs:

eoew_architecture_sketches.md (diagrams + narrative)

Evaluation Protocol (Backtesting + Metrics)

Outstanding

Inputs:

eoew_architecture_sketches.md (agreed reference architecture)
Historical event data for selected scenarios

Process:

Define metrics: time-to-detect, false alarms, spatial accuracy, uncertainty calibration, operational latency.
Specify a reproducible backtesting approach using open data where possible.

Outputs:

eoew_evaluation_protocol.md (protocol + metrics)

Pilot Plan + Partner Map

Outstanding

Inputs:

eoew_architecture_sketches.md (agreed reference architecture)
Candidate stakeholders and implementers

Process:

Propose a realistic pilot: scope, timeline, data access, success criteria, and governance.
Map credible partners: existing EO programmes, NGOs, research groups, and operational agencies.

Outputs:

eoew_pilot_plan.md (pilot outline + partner map)

Project Repository

This repository contains materials generated from the project.

eoew_overview.md - Project overview and framing
eoew_use_cases.md - Scenario shortlist (decision-first briefs)
eoew_requirements.md - Signal requirements table
eoew_dataset_inventory.md - Dataset and sensor inventory
eoew_latency_budget.md - End-to-end latency model
eoew_modelling_notes.md - Modelling scope and artifacts
eoew_revisit_sim.py - Revisit/coverage simulation
eoew_revisit_results.csv - Simulation results
eoew_revisit_chart.svg - Chart (SVG)
eoew_architecture_sketches.md - Reference architectures
eoew_evaluation_protocol.md - Backtesting and metrics
eoew_pilot_plan.md - Pilot outline and partner map