Set up?
Does the local evidence justify the effort and energy?
Field-built environmental intelligence
Know what is happening here.
Decide what to do next.
EDS observes changing local conditions, estimates near-term environmental risk, and supports reviewable decisions and proposed connected actions. Physical execution remains future work.
Open the EDS Control CenterReview target configured for the current Console origin. This site does not assert that the future Console hostname is deployed.
Persistent local telemetry
Reviewable history + health
1 h and 3 h advisory windows
Observable decision state
The decision gap
A regional forecast is context.
The instrument site is the decision.
For astronomical observing, the useful window can change faster than a broad forecast suggests. Humidity can tighten toward the dew point. Local light and pressure can shift. A false positive wastes setup time and energy; a false negative can waste a rare clear interval or leave equipment exposed.
EDS began with a practical need: judge whether to begin, continue or stop an observing session—and gain earlier warning of conditions that could threaten optics and remote equipment.
Continue?
Is the trajectory stable enough to keep observing?
Protect?
Is dew or rapidly changing weather becoming a risk?
Stop?
Should the operator fail closed rather than assume safety?
Measure locally. Preserve the history. Make the uncertainty visible.
One continuous system
From observation to operating guidance.
A weather station reports conditions; a dashboard displays them. EDS preserves the local history, interprets near-term risk and keeps the resulting decision state reviewable by an operator.
-
01
Observe
Local temperature, humidity, pressure, ambient light, battery state and signal health.
Maintained field station -
02
Preserve
Timestamped telemetry, latest state, history, aggregates, gaps and staleness.
Reviewable local record -
03
Interpret
Current-condition rules and 1-hour / 3-hour advisory risk windows.
Confidence stays visible -
04
Decide
Operator views, policy state, reasons, events and proposed actions.
Advisor and dry-run today
The operating loop
Evidence stays attached to the decision.
Each stage preserves enough context to show what changed, what the system inferred, and why a conservative hold may be safer than a confident-looking answer.
Verified scope
What EDS can do—and where the boundary is.
Operational
Validated
Experimental
Expanding
Validated
Persistent local observation
Minute-scale MQTT telemetry, retained station status, bounded offline buffering, historical records and station-health checks.
Validated
History and replay
Latest observations, raw history, aggregates and client-side replay views keep trajectories available for review.
Validated
Observable decision support
Policy, state, reasons, overrides, events and proposed actions are available in advisor or dry-run workflows.
Validated · partial
ASCOM Alpaca conditions
Supported local measurements are available through a partial ObservingConditions surface; wind, rain-rate and cloud-cover properties are not implemented.
Advisory forecast rail
Four domains. Two decision windows.
The 3-hour value represents the highest or worst assessed class within the next three hours—not an exact event time. Outputs are advisory and trained against rule-derived labels.
Astronomy1 h3 hObserving quality
Rain1 h3 hLikelihood
Dew / fog1 h3 hRisk
Storm1 h3 hEvidence-limited risk
Candidate-model evaluation
Offline evaluation, release gates and a non-serving Astro 1-hour shadow candidate support controlled improvement. Candidate output is not the current serving truth.
Connected-device execution
The safety state machine can propose actions. Direct hardware execution and a verified N.I.N.A. workflow remain future work.
More stations, more context
The software can represent multiple stations. A validated distributed physical network and external weather-feed ingestion have not yet been demonstrated.
Where local truth matters
Designed for decisions that happen at a place.
The current evidence base is strongest for one coastal station and weather-sensitive astronomical operations. Other uses require their own sensors, thresholds and validation.
- 01
Astronomical observatories
Session go/no-go guidance, dew awareness, equipment protection and remote-site context.
- 02
Field research
Persistent local observations and a reviewable record around an event or experiment.
- 03
Remote installations
Environmental and station-health context for systems that cannot be watched continuously.
- 04
Independent R&D
An end-to-end testbed for sensing, data engineering, model governance and inspectable decisions.
Field study · 12 August 2026
Two witnesses recorded one changing environment.
The EDS station at Espinho preserved one-minute environmental observations while a telescope record was captured in the Vila Praia de Âncora area. The archive keeps those clocks, locations and evidence types distinct, then aligns them on one working timeline. It is the strongest current field validation of the EDS observation and publication chain—not the system's only intended use.
11 luxRecorded minimum
+8 sAfter predicted maximum
111 / 111Complete event-window samples
12Nearby comparison evenings
The 11 lux reading landed in the one-minute sample eight seconds after Espinho predicted maximum; the chart also records a 99.0% five-minute light reduction. The study does not validate every EDS forecast domain or prove secondary thermal, humidity, pressure or dew-point effects.
Field hardware + operations
Built to keep observing when no one is watching the station.
The system joins physical design, embedded firmware, telemetry transport, storage, advisory models, operator tooling and operational policy. Each layer matters only when the chain preserves the meaning and quality of the observation.
Documentary
Station and enclosure
Post-event hardware views identify the maintained environmental observer and its integration. They provide context, not eclipse-day evidence.
Documentary
Independent field power
A compact solar supply supports the documented field installation. The image records power context, not system-wide deployment scale.
Representative QA capture · fixture values · not live
Representative
Control Center + advisory output
A sanitized QA capture shows the current operator surface joining 1-hour and 3-hour guidance, current conditions and trends. Its values are fixtures—not live station state.
Independent R&D by MDNT
Environmental intelligence should end in a decision someone can inspect.
EDS is engineered by Teófilo Couto as an independent MDNT research and development system spanning station hardware, firmware, data services, model evaluation, operator tooling and field publication.