Resilience Early Warning

RESILIENCE ANALYTICS • OSONWANNE GROUP LLC

Resilience Early Warning: explainable warning for water, food and energy stability from public data

A working research prototype prepared in response to an Army Geospatial Center broad agency announcement topic on AI techniques for publicly available data. It ranks areas by risk, explains every score with its evidence, and states forecast uncertainty honestly.

Measured, not just claimed

Every number below comes from the running prototype and its own hindcast tests, using only free, keyless public data.

13

Areas scored on water, food, energy and acute-hazard stress: 9 international and 4 US communities

7

Public sources: NASA POWER, World Bank, GDACS, USGS, NOAA, U.S. Drought Monitor and FEMA

35%

Lower 30-day rainfall-index error than a no-change forecast, across 1,794 hindcasts since 2021

67→77%

Held-out coverage of the 80% forecast interval after calibration, with the remaining gap stated openly

Resilience Early Warning console: areas ranked by composite risk, a world map with live hazard alerts and stream gauges, and the explanation panel with pillar scores and a 30-day rainfall outlook
The prototype console: areas ranked by risk, live hazard alerts and gauges on the map, and an explanation panel for the selected area.

What it does

Explainable risk scores

Each area gets water, food, energy and acute-hazard scores from stated rules. The five biggest drivers are listed as reasons, each citing its source, so a single crisis is never averaged away.

Anomaly detection

Standardized rainfall indices against the 1991–2020 normal, heat anomalies, high or low streamflow, and nearby hazard alerts flag areas that are drifting out of their normal range.

Honest 30-day outlook

An ensemble forecast gives low, median and high bands for the next 30 days, calibrated and tested on held-out hindcasts so the stated uncertainty can be trusted.

Cited analyst briefs

One click writes an analyst brief for any area in which every statement cites the data it came from. It runs in demo mode with templates, so no AI model invents anything.

Standards-based outputs

Results are served as GeoJSON and CSV and through an OGC API – Features service, so they can feed existing geospatial systems instead of standing alone.

Data audit trail

Every data pull is logged with its address, time, data currency and a fingerprint in a hash-chained log you can verify in the browser.

Who it helps

PLANNERS

Prioritize early

An auditable ranking of where stress is building, with lead time and stated confidence.

WATER AND EMERGENCY MANAGERS

Drought and flood watch

US communities are scored with USGS streamflow, Drought Monitor and FEMA risk data alongside live weather alerts.

ANALYSTS

Defensible sourcing

Every score and brief traces back to a logged, fingerprinted data pull.

PARTNERS

Shareable by design

All inputs are public, so results can be shared with humanitarian and international partners.

See it running

Explore the working prototype with no login: pick an area, read why it scored the way it did, check the 30-day outlook, generate a cited brief, or browse the data audit log.

Questions? bon@osonwanne.com • 602-632-9589

Resilience Early Warning is an independent research prototype by Osonwanne Group LLC that uses only publicly available data. It is a research demonstration, not an official assessment, and it is not affiliated with or endorsed by the U.S. Army, the U.S. Army Corps of Engineers or any agency.