6.8.2 — Multi-Hazard Warning Systems — maturity: live

Cascading Risk Modelling

Using multi-source satellite observation to model how one hazard triggers the next — flood undermines slope, slope fails, debris blocks river, river backs up into city.

When flood, landslide, and grid failure strike simultaneously, nations that own their sensor infrastructure see the full picture first — and act while renters are still waiting for data.

Disasters rarely arrive alone. An earthquake loosens hillside material; three days of rain saturate it; the resulting landslide dams a river whose backwater inundates a town that has already lost its power grid. National emergency managers who model only single-hazard events are planning for the wrong disaster. Satellite data — radar-derived soil-moisture, InSAR surface-deformation, optical flood extent, thermal anomaly mapping — gives the continuous, wide-area observational thread that lets a cascading risk model update itself as the chain unfolds rather than being run once as a static pre-event estimate.

The satellite stack required is deliberately heterogeneous. C-band SAR provides all-weather flood and landslide mapping at 5–10 m resolution. InSAR time-series (Sentinel-1 cadence or better) tracks precursor ground deformation in the hours to days before a mass-movement event. Multispectral imagery captures post-event land-cover change and infrastructure damage. Medium-resolution thermal data flags wildfire ignition that a flood-weakened, debris-laden watershed can spread further than pre-disaster fuel models predict. No single commercial vendor provides this combination under a single sovereign access agreement, which is precisely the problem.

The operational output is a continuously updating probabilistic risk graph — nodes are hazard states, edges are conditional trigger probabilities — delivered to the national emergency operations centre before and during a multi-hazard event. Response commanders can see not just where the current hazard is, but which downstream nodes are about to light up and how long the window is before they do. That foresight shifts resource pre-positioning from reactive to anticipatory, directly reducing casualties and infrastructure loss in a disaster sequence that would otherwise overwhelm reactive coordination.

What matters

Cascading sequences cause disproportionate casualties: the 2018 Palu, Indonesia earthquake-tsunami-liquefaction chain killed over 4,000 people, a toll dominated by the third-order hazard that static models missed.
InSAR deformation time-series can detect precursor slope instability at centimetre scale days before failure — but only if the national operator has persistent, uninterrupted tasking authority over the relevant orbit track.
Commercial data-access agreements routinely exclude priority tasking during declared foreign disasters, precisely when cascading risk model inputs are most urgently needed.
Multi-hazard risk chains cross ministry boundaries; only a sovereign platform with a single data authority can enforce real-time data fusion across meteorology, geology, hydrology and civil protection feeds without inter-agency licensing friction.

Quick facts

Median latency of SAR-based flood extent maps (commercial tasking): 18–36 hours (2024) — Copernicus Emergency Management Service Activation Statistics
Additional mortality risk when secondary hazard follows primary within 72 h: up to 40% higher (2022) — UNDRR Global Assessment Report on Disaster Risk Reduction 2022
Lives saved per year globally if early warning coverage were universal (UN estimate): ~23,000 (2022) — UN Early Warnings for All Initiative — Executive Action Plan

Sovereignty score: 9/10 — Cascading risk modelling depends on uninterrupted, priority access to heterogeneous satellite feeds at the exact moment a foreign vendor is most likely to ration or deny them — sovereign ownership is the only operationally reliable solution.

Commercial priority-tasking agreements contain force-majeure and national-security carve-outs that allow vendors to redirect capacity during large-scale disasters, removing access precisely when cascading risk inputs are most critical for national emergency decision-making.
Export-control regimes (US EAR, EU dual-use regulation) can restrict delivery of high-resolution SAR and InSAR products to certain nations or during certain geopolitical periods, breaking the continuous data thread that cascading risk models require to remain calibrated.
Cascading risk model outputs inform decisions to evacuate cities, shut down nuclear or hydro infrastructure, and pre-position military assets — their classification, integrity and dissemination must be controlled by national civil protection and defence authorities, not third-party platforms with foreign legal exposure.
Multi-hazard chains evolve faster than procurement cycles: a sovereign operator can retask satellites, adjust revisit geometry and fuse new data streams within hours; a nation dependent on commercial subscriptions must re-negotiate terms while the cascade is already propagating.

Reference architecture

Payload: Primary: C-band SAR, 5m stripmap and 20m ScanSAR modes, 250km swath in wide-area mode for flood and landslide mapping. Secondary: multispectral imager, 10m resolution, 8 bands including NIR and SWIR for land-cover change and burn-scar mapping. Optional augmentation: GNSS-RO receiver for atmospheric profiling to feed precipitation cascade triggers.
Bus class: ESPA-class microsat, 150–200kg per SAR unit; 6U–16U cubesat for the multispectral and GNSS-RO auxiliary nodes. Mixed-class constellation keeps per-unit cost low while maintaining the payload diversity cascading risk models require.
Orbit: Sun-synchronous LEO at 520–560km, dual-plane 8-satellite SAR constellation providing 4–6 hour revisit over national territory; 12-satellite multispectral constellation in complementary planes. InSAR coherence requires consistent repeat-pass geometry, so orbital maintenance budget must hold ground-track repeat to within ±1km.
Ground segment: 3-station national network (X-band SAR downlink, S-band TT&C) co-located with national meteorological and seismological network nodes to minimise new infrastructure cost; SatNOGS-compatible UHF/VHF backup for housekeeping telemetry. Direct injection into the national emergency operations centre data bus via fibre.
Data pipeline: On-board L0 compression and checksumming → ground L1 radiometric calibration → automated L2 products: flood extent (random forest classifier), InSAR displacement time-series (SBAS algorithm), land-cover change (change vector analysis) → L3 cascading risk graph updated on a sovereign GPU cluster using a Bayesian network engine → REST API and event-driven webhook triggers.
End-user delivery: Interactive probabilistic risk graph console for the national emergency operations centre, showing hazard-state nodes, conditional trigger probabilities and time-to-next-node estimates. Push alerts via national CAP broker (links to §6.8.4) when a cascade threshold is crossed. Classified downlink of precursor deformation maps to defence and critical-infrastructure protection authorities on a separate network.
Time to launch: First 2-satellite SAR demonstrator in 24 months from contract signature, sufficient for initial InSAR baseline collection; full 8-satellite SAR plus 12-satellite multispectral constellation operational in 42 months. Risk graph model training begins on historical Sentinel-1/Copernicus archive from month 6.
Caveats: InSAR time-series requires a stable orbital repeat track — any constellation operator who cannot guarantee orbital maintenance to sub-kilometre ground-track tolerance will degrade coherence and produce unreliable deformation products. SAR bus and payload primes with no US-origin ITAR components include European (Airbus, OHB, SSTL) and Indian (ISRO commercial arm, Pixxel) suppliers; procurement strategy must specify this from RFP stage to avoid late-programme export-control delays.

Frequently asked

What exactly is a 'cascading risk' and why does it need its own satellite application?

A cascading risk occurs when one hazard triggers or amplifies a second — for example, heavy rainfall saturating slopes already destabilised by an earthquake, producing landslides that block evacuation routes during a cyclone. Standard single-hazard early warning systems are blind to these interactions. Satellite-derived data layers (SAR deformation, soil moisture, thermal anomalies, precipitation) need to be fused into a single probabilistic model that explicitly represents trigger-pathway chains — that requires dedicated integration architecture, not off-the-shelf flood alerts.

Why should a nation own this capability rather than subscribe to a commercial cascade-modelling service?

Commercial providers can withdraw, reprice, or impose export controls on data and algorithms at the moment of a national emergency — exactly when continuity matters most. A sovereign system means the government controls sensor tasking priority, data retention policy, algorithmic parameters, and dissemination channels. It also means classified infrastructure vulnerability data (power grid topology, dam safety records) never leaves national jurisdiction to feed the model.

How many satellites does a minimum viable sovereign cascade-monitoring constellation require?

A practical minimum is a mixed constellation: 6–8 microsatellite SAR nodes for surface deformation and flood extent, 4–6 optical/multispectral microsatellites for vegetation stress and wildfire precursors, and access to at least one precipitation-radar feed (either owned or through a bilateral data-sharing agreement such as with the GPM mission). Total: 10–14 satellites gives sub-6-hour revisit over a medium-sized nation's territory, sufficient to update cascade models before a second event window.

Can this system work if a country already has Copernicus or SERVIR access?

Copernicus Emergency Management Service activations and SERVIR regional products are valuable baselines, but they are demand-driven and prioritised by the activating organisations — they cannot be tasked unilaterally during a national crisis competing with events in other regions. A sovereign system runs continuously and autonomously, pushing model updates on a scheduled cadence rather than waiting for an international activation decision.

What is the typical end-to-end lead time a cascade model should provide to be operationally useful?

UNDRR guidance recommends a minimum of 72 hours for population evacuation and 24 hours for critical infrastructure pre-positioning. Cascade models that fuse satellite soil moisture anomalies and SAR-detected slope creep with numerical weather prediction can realistically achieve 48–96 hour probabilistic warnings for landslide-flood sequences in well-instrumented catchments. Lead times for earthquake-triggered secondaries remain shorter (6–12 hours) because the primary event itself is not yet forecastable.

How do cascade models handle uncertainty, and how is that communicated to decision-makers?

Best-practice implementations (following WMO probabilistic forecast guidelines) output ensemble probability distributions — e.g., '70% probability of secondary landslide within 48 hours given observed precipitation anomaly' — rather than binary alerts. These are mapped to impact thresholds calibrated against national emergency-response protocols. The challenge is translating ensemble spread into actionable colour-coded products without overwhelming civil protection authorities with caveats.

Does this application overlap with compound event forecasting?

The two are closely related but distinct in focus. Compound Event Forecasting (§6.8.1) concentrates on the meteorological co-occurrence statistics that prime the environment for multi-hazard events. Cascading Risk Modelling (§6.8.2) begins at the moment a primary hazard is detected and models the physical trigger pathways to secondary and tertiary events in near-real-time. The outputs of compound forecasting feed the prior probabilities used in cascade models.

What international frameworks oblige nations to develop this kind of capability?

The Sendai Framework for Disaster Risk Reduction 2015–2030 (Target G) requires nations to substantially increase the availability of and access to multi-hazard early warning systems by 2030. The UN Secretary-General's Early Warnings for All initiative (2022) operationalises this with country-level action plans. The WMO 2023 State of Climate Services report identifies cascading risk modelling as one of the highest-priority capability gaps globally.

Glossary

Cascading hazard: A sequence in which a primary hazard event triggers or amplifies one or more secondary hazards through physical, infrastructure, or social pathways.
SAR (Synthetic Aperture Radar): A microwave imaging radar carried on satellites that produces high-resolution surface imagery regardless of cloud cover or daylight conditions, enabling flood mapping and land-deformation detection.
Probabilistic ensemble forecast: A set of multiple model runs with slightly varied initial conditions or parameters that represents the range of plausible outcomes and their relative likelihoods, rather than a single deterministic prediction.
InSAR (Interferometric SAR): A SAR processing technique that compares two radar images taken at different times to detect millimetre-scale ground deformation, used to identify slope instability before a landslide occurs.
CAP (Common Alerting Protocol): An OASIS open standard message format that allows a single warning to be distributed simultaneously across multiple communications networks in a machine-readable, interoperable form.
Trigger pathway: The documented physical or systems mechanism by which one hazard event creates the conditions necessary for a secondary hazard — for example, rainfall-induced soil saturation enabling slope failure.
Soil moisture anomaly: A satellite-derived measure of how much current surface soil water content deviates from the long-term seasonal average, used as a precursor indicator for both flood and landslide risk.
Ground segment: All the Earth-based infrastructure — antennas, command systems, data processing centres — required to operate a satellite, downlink its data, and deliver processed products to end users.
GPM (Global Precipitation Measurement): A joint NASA/JAXA satellite mission that provides near-global precipitation estimates every 30 minutes, widely used as a satellite rainfall input layer for cascade models.
MHEWS (Multi-Hazard Early Warning System): An integrated system that monitors, forecasts, and communicates risks from multiple hazard types within a unified governance and dissemination framework, as defined by WMO and UNDRR.

References

Sendai Framework for Disaster Risk Reduction 2015–2030 — Target G of the Sendai Framework calls for substantially increasing the availability of multi-hazard early warning systems and disaster risk information by 2030; cascading risk modelling is explicitly identified as a gap in country-level implementation reviews.
Global Assessment Report on Disaster Risk Reduction 2022 — The GAR 2022 documents that mortality risk increases by up to 40% when a secondary hazard follows a primary event within 72 hours, underscoring the operational requirement for real-time cascade modelling rather than single-hazard alert systems.
Early Warnings for All — Executive Action Plan — The UN Secretary-General's initiative targets universal early warning coverage by 2027 with an estimated investment of $3.1B; the plan explicitly calls for satellite-derived multi-hazard data fusion as a core technical pillar.
Copernicus Emergency Management Service — Activation Statistics — CEMS activation records show median product delivery times of 18–36 hours after tasking request for SAR-based rapid mapping; sovereign constellations with dedicated ground stations can reduce this latency to under 4 hours for national-priority events.
NASA GPM Mission — Precipitation Measurement Overview — The Global Precipitation Measurement Core Observatory provides 30-minute global precipitation composites at 0.1-degree resolution; these are the primary freely available satellite rainfall input layer for operational cascade models globally.
ESA SNAP (Sentinel Application Platform) — InSAR Processing for Landslide Monitoring — ESA's open-source SNAP toolbox enables InSAR processing of Sentinel-1 data to detect millimetre-scale slope deformation; this pipeline is increasingly embedded in national cascade-monitoring workflows as a precursor layer for secondary landslide risk.
OASIS Common Alerting Protocol v1.2 — CAP v1.2 defines the machine-readable alert message standard adopted by WMO and ITU as the interoperability layer for distributing multi-hazard warnings across heterogeneous national and international communication networks.
World Bank — Disaster Risk Finance Analytics: Cascading Risk and Parametric Tools — World Bank analysis finds that nations with integrated multi-hazard satellite monitoring reduce post-disaster reconstruction costs by 15–30% relative to GDP impact, owing to earlier pre-positioning of resources and reduced secondary-event losses.

Related applications

6.8.4 — Common Alert Protocol Distribution (Multi-Hazard Warning Systems)
6.8.5 — Multi-Hazard Risk Atlases (Multi-Hazard Warning Systems)
6.1.1 — Flood Extent Mapping (Flood Intelligence)
6.1.2 — Flash Flood Forecasting (Flood Intelligence)
6.1.3 — Urban Flood Modelling (Flood Intelligence)
6.1.4 — River Stage Monitoring (Flood Intelligence)
6.1.5 — Coastal Storm Surge Prediction (Flood Intelligence)
6.1.6 — Flood Insurance Claims Verification (Flood Intelligence)