Weather · grids · space

Decision intelligence for weather, grids, and space operations.

Echofold builds evidence-backed forecasting and engineering decision-support products with transparent provenance and human review at the point of action.

Mission operations and dynamics assurance

Managed delivery. Benchmarks, prospective validation, and pilot engagements are scoped with explicit evidence and decision boundaries.

Product families

One evidence system.
Three operating domains.

Each product is offered as a managed engagement today, with maturity and authorization boundaries kept visible as the stack expands.

01Managed benchmark

Weather + grid intelligence

EarthWeather

A managed weather-intelligence family spanning short-window hazard nowcasting, revision-aware long forecasts, and Texas grid wind forecasts.

  • Short-Window Nowcast
  • Long Forecast
  • Wind Intelligence
  • Issue-time provenance
View EarthWeather
02Managed pilot

Space-weather decisions

StormShield

A human-reviewed space-weather stack connecting solar-eruption probabilities, geomagnetic onset and lifecycle estimates, ionosphere disturbance ranking, and operations review.

  • StormTrigger
  • Direct Kp + native Hp
  • Lifecycle + recovery
  • GNSS/Ionosphere
View StormShield
03Managed engagement

Mission environment + dynamics

MissionControl

Traceable mission-environment and dynamics-assurance workflows for planning, screening, numerical cross-checking, and operator review.

  • Environment bridge
  • Dynamics assurance
  • Reference propagation
  • NEO / three-body research
View MissionControl
01 / EarthWeather

Weather + grid intelligence

One weather family. Three distinct decision surfaces.

EarthWeather combines independently evidenced Short-Window Nowcast, Long Forecast, and Wind Intelligence products under shared issue-time provenance and immutable issuance controls. Their targets, comparators, and evidence denominators remain separate.

  • Short-Window Nowcast1–6-hour hazard probabilities
  • Long Forecast1–48-hour multi-hazard planning
  • Wind IntelligenceERCOT system total + five regions
  • Evidence controlsTarget-specific, locked comparisons

Applicable toWeather-sensitive planning, renewable-generation analysis, operational awareness, infrastructure review, and research.

Primary launch deliverableStandardized hourly, versioned ERCOT system and five-region wind forecasts, with separately scoped short-window and long-range weather intelligence.

Each surface is offered through a managed benchmark or prospective validation; live availability and latency are established before operational use.

02 / StormShield

Space-weather decision support

From solar onset to recovery and operational review.

StormShield connects probabilistic solar-eruption monitoring, geomagnetic onset and intensity, episode lifecycle and recovery, site-level ionosphere disturbance ranking, and satellite-operations review in one traceable workflow.

  • OnsetStormTrigger + Direct Kp pathways
  • IntensityNative Hp30/Hp60 horizons
  • LifecyclePeak, duration, recovery, all-clear
  • Impact reviewGNSS/Ionosphere + Satellite Ops

Designed forSatellite operators, GNSS-dependent operations, and grid teams reviewing space-weather risk.

Managed deliverableTraceable, human-reviewed space-weather issue packages with source and decision provenance.

StormShield complements official space-weather services and supports human-reviewed operational decisions.

03 / MissionControl

Mission environment + dynamics assurance

Reproducible review for difficult mission decisions.

MissionControl turns verified environmental advisories, approved asset profiles, planned operations, reference propagation, and difficult dynamics cases into traceable operator-review workflows.

  • Environment bridgeAdvisory-to-asset review context
  • Dynamics assuranceRegime-aware routing + invariant audits
  • Cross-checksReference-integrator comparison
  • Research extensionNEO and three-body screening

Designed forMission-planning and engineering teams reviewing difficult environment and dynamics cases.

Managed deliverableReproducible environmental and dynamics-assurance review with operator-facing evidence.

MissionControl supports engineering review while retaining operator authorization for flight decisions.

Product structure + evidence

Platforms first. Component claims kept separate.

The parent panels show how each product family operates. The cards below them report selected, non-pooled results with a named target, comparator, sample, evidence status, and scope.

EarthWeather platform

Shared forecast controls. Three independently evidenced decision surfaces.

Managed benchmark + validation

Short-Window Nowcast, Long Forecast, and Wind Intelligence belong to one managed product family without borrowing performance from one another.

Shared control layer · forecast-available inputs · issue-time provenance · immutable versioning

  1. Weather / 01Short-Window Nowcast

    f01, f03, and f06 hazard probability.

  2. Weather / 02Long Forecast

    f01–f48 multi-hazard planning priority.

  3. Grid / 03Wind Intelligence

    ERCOT system and five-region wind generation.

Target-specific delivery · separate comparators · independently scored evidence

Platform scopeThe platform provides common provenance and engagement controls. Each surface retains its own target, model authority, comparator, denominator, and promotion status.

EarthWeather / component evidence

Selected non-pooled evaluation records

Texas grid wind intelligence

EarthWeather · Wind Intelligence

Two frozen independent holdouts

29.25% / 27.37% lower f01 / f03 system MAE

f01: 865.07 versus 1,222.66 MW. f03: 1,118.01 versus 1,539.37 MW.

Target
ERCOT system-total realized wind generation at the one- and three-hour forecast gates.
Comparator
Same-gate ERCOT Short-Term Wind Power Forecast (STWPF).
Sample
Separate locked 60-date evaluations at f01 and f03; their evidence rows and absolute error levels are not pooled.
Evidence
Two frozen independent holdouts

Scopef03 is the primary commercial forecast and f01 the secondary near-term view; system and five-region outputs are available.

Supported-source route

EarthWeather · Short-Window Nowcast

Promoted frozen holdout

14.39% lower Brier error

0.029753 versus 0.034752, with lower date-level error on 57 of 60 dates.

Target
Forecast-time hazard probability at f01, f03, and f06 on a fixed CONUS grid.
Comparator
Validation-selected EarthWeather-built probability control using forecast-available HRRR fields; not raw HRRR or a NOAA-issued probability product.
Sample
Locked 60-date, 524,070-row CONUS-grid holdout with GOES ABI or HREF-family support on every row.
Evidence
Promoted frozen holdout

ScopeEvery scored row had GOES ABI or HREF-family support. A separate source-timed replication spans 125 dates; denominators are not pooled.

Established-source route

EarthWeather · Short-Window Nowcast

Frozen full-surface validation

44.51% lower Brier error

0.130864 versus 0.235834 on the same frozen surface, with 0.8160 AUC.

Target
Forecast-time hazard probability at f01, f03, and f06 on the original full validation surface.
Comparator
Fixed EarthWeather-built probability translation of forecast-available HRRR fields; not a NOAA-issued probability product or the newer validation-selected control.
Sample
10,000 event-focused case/tile rows, including 2,059 positives.
Evidence
Frozen full-surface validation

ScopeThe current source-aware router preserves this result exactly when both dedicated source families are unavailable.

Multi-hazard planning forecast

EarthWeather · Long Forecast

Frozen historical product

48.61% Brier skill for planning-window priority

0.105981 Brier error versus 0.206238 on the same frozen case-lead surface.

Target
Probability that a candidate multi-hazard window warrants refinement across f01–f48.
Comparator
Frozen HRRR-derived probability lane scored on the same case-leads.
Sample
10,259 deduplicated case-leads; separate from the Short-Window Nowcast and wind denominators.
Evidence
Frozen historical product

ScopePrioritizes act, review, or watch queues as qualifying forecast issues update; prospective absolute-location delivery is validated separately.

HRRR-derived controls are EarthWeather-built probability translations of NOAA HRRR forecast fields. They are not probabilities issued by NOAA. Results use separate, non-pooled evaluation populations and are not directly comparable.

StormShield platform

One traceable decision stack—from event onset through operational impact review.

Monitored shadow + customer review

StormTrigger is one upstream component. StormShield adds geomagnetic routing, native Hp intensity, explicit episode state, lifecycle and recovery, and downstream impact review.

Source-qualified solar, CME/ENLIL, solar-wind, Kp, and Hp state

  1. 01 / OnsetStormTrigger + Direct Kp

    Solar-event and geomagnetic-entry probability.

  2. 02 / IntensityNative Hp30 / Hp60

    Short-cadence escalation from 30 to 360 minutes.

  3. 03 / EpisodeLifecycle + recovery

    Survival, duration, recovery, all-clear, and reset.

  4. 04 / ImpactGNSS + operations

    Site ranking, satellite review, and BES context.

Human-reviewed release · immutable issue package · Satellite Ops and MissionControl handoff

Platform scopeThe episode controller assigns authority by state and horizon; it does not average incompatible probabilities. Outputs fail closed when required provenance or freshness is unavailable.

StormShield / component evidence

Four selected, non-pooled evaluations

Geomagnetic onset + intensity

StormShield · Native Hp

Frozen final holdout

11.7% / 10.4% lower Hp30 / Hp60 onset Brier

Average precision improved in every independently fitted onset cell on the final holdout.

Target
Native Hp30 and Hp60 onset probabilities from 30 through 360 minutes.
Comparator
Validation-locked delayed-state references at the same cadence, threshold, and horizon.
Sample
2024–2025 final holdout; 35,088 Hp30 and 17,544 Hp60 timestamps, totaling 1.93 million scored cell rows.
Evidence
Frozen final holdout

ScopeThe frozen router promotes replicated cells and retains the reference where a challenger did not qualify.

Persistence + recovery

StormShield · Episode Lifecycle

Promoted human-review product

0 / 85 premature formal all-clears

Survival and recovery Brier error improved at every authorized horizon in both evaluation periods.

Target
Survival, remaining duration, safe recovery, and all-clear for an already-active Kp ≥ 5 episode.
Comparator
Accepted lifecycle v2 on the same episode-blocked evaluation rows.
Sample
32 locked 2023 and 72 post-lock 2024–2025 episodes; 28 plus 57 formal all-clear decisions.
Evidence
Promoted human-review product

ScopeConditional active-episode decision support; it does not independently declare initial storm onset or automate control.

Measured ionosphere disturbance

StormShield · GNSS/Ionosphere

Frozen external comparison

0.8863 AP next-hour ROTI disturbance ranking

Versus 0.8017 for a WAM-IPE WRS +1h translation and 0.8157 for a lagged GloTEC translation.

Target
Station-level next-hour ROTI ≥ 0.5 disturbance ranking.
Comparator
Incumbent-field translations scored on the same frozen target and issue rows.
Sample
1,313 matched station-hours, eight stations, and ten Jan–Mar 2026 episodes.
Evidence
Frozen external comparison

ScopeUncalibrated next-hour ranking score; probability calibration and receiver-integrity authorization are separate outputs.

Solar-eruption probability

StormShield · StormTrigger

Production-shadow probability model

0.49–7.54% lower Brier error across four horizons

Average precision also improved at 1, 6, 12, and 24 hours; grouped lower 95% bounds cleared zero for both metrics.

Target
Future X1/R3-class solar-eruption onset probability.
Comparator
Frozen development climatology.
Sample
75 independent 2017–2025 X1/R3 episodes in the frozen rolling test.
Evidence
Production-shadow probability model

ScopeProbabilistic onset performance; customer-watch and alert operating points are evaluated and authorized separately.

Evidence reviewed August 2026. Results shown are from defined, frozen historical evaluations. Live availability, latency, and authorized use are established through prospective managed validation. Detailed regional, routing, model, and statistical evidence is reserved for qualified technical diligence under appropriate confidentiality.

Deliberate expansion

Future modules stay with the product they extend.

StormShield

Planning and grid impact

CycleHorizon and GIC workflows remain visibly labeled as prospective shadow or partner-validation modules.

MissionControl

Dynamics and NEO assurance

Regime-aware three-body and N-body work extends managed screening and numerical review—not autonomous navigation.

Echofold Research

Observational model audit

Licensed or customer-supplied photometry can support reproducible lensing-model comparison without implying discovery confirmation.

Echofold

Technical enough for diligence. Clear enough for decisions.

Echofold develops forecasting and engineering decision-support systems for environments where provenance, issue-time integrity, and explicit human authorization matter.

The launch site is intentionally focused. Detailed methods, customer data exchange, and delivery credentials move through a scoped engagement—not a public upload form or open API.

Start a focused conversation

Bring the decision target.
We’ll define the evidence path.

For a managed benchmark, prospective validation, or product-fit discussion:

hello@echofoldtech.comNo forms or customer-data uploads are used on this website.