Project / EDS

Environmental Decision Support System.

EDS is MDNT's sensing-to-alerting pipeline for environmental conditions: it ingests environmental telemetry, models near-term change, and turns the result into structured forecasts, alerts, and operational awareness for the teams that need to act on them.

Active prototype Sensing / Forecasting

System Overview Discuss EDS Explore EDS

System Architecture

Signal to decision, with action in the loop.

EDS combines changing environmental inputs with near-term modeling, threshold evaluation, and outputs that can notify an operator or inform external field controllers.

Conceptual architecture / Current prototype model

01 / Sense Sky, weather, field sensors Time-aligned environmental signal
02 / Normalize Common environmental schema Quality, gaps, and sensor state
03 / Forecast Near-term conditions Statistical and ML model output
04 / Evaluate Thresholds and baselines Operational significance
05 / Alert + Act Awareness and control inputs Notification or external device response

System Overview

Turning environmental signal into operational awareness.

EDS started as a practical problem, not an abstract one: deciding whether a night is actually worth setting up for astrophotography, and keeping optics clear of dew once a session is running. That recurring, concrete decision depends on cloud cover, sky transparency, seeing, and dew point, all changing hour to hour. Those conditions shaped the system's core sense-forecast-alert loop well before it was generalized into a broader environmental decision-support pipeline.

EDS is built as a layered pipeline rather than a single monolithic tool. Sensing sources feed a normalization and modeling layer, which produces forecasts and evaluates them against configurable thresholds, which in turn drives alerting and awareness outputs. Each layer evolves independently. New sensing sources, refined models, and additional output channels can be added without reworking the system end to end.

The goal is continuous, structured visibility into changing environmental conditions, not a one-off report or a static map. EDS is designed for teams that need to track conditions over time and respond when they change. That includes environmental monitoring groups, infrastructure operators, and field research teams. It is not limited to a single fixed use case. EDS is currently an active prototype: core ingestion and modeling are running against real test data, and the surrounding system is still being shaped around that foundation.

Domain: Sensing / Forecasting
Status: Active prototype
Core loop: Sense → Forecast → Alert → Act
Primary users: Astrophotographers, environmental monitoring & infrastructure teams

Sensing Inputs

A layered set of environmental signal sources.

EDS treats sensing as a plural, extensible layer rather than a dependency on any single feed. Sources are normalized into a common schema so the modeling layer can reason over them consistently, regardless of where a given signal originates.

Astronomical Observing Conditions

Cloud cover, sky transparency, seeing, dew point, and moon phase make up the specific combination of conditions an imaging session actually depends on.

Meteorological Telemetry

Wind, precipitation, temperature, humidity, and pressure readings from ground stations and public weather feeds.

Hydrological & Water-Level Data

River, reservoir, and coastal water-level sensors used to track hydrological change over time.

Remote Sensing & Imagery

Satellite and aerial imagery layers used to contextualize and cross-check ground-level sensor readings.

Field Sensor Networks

Distributed, IoT-class sensor nodes reporting localized environmental conditions from fixed or mobile deployments.

Historical & Climatological Data

Long-run historical datasets used to establish baselines, seasonal patterns, and anomaly thresholds.

Open Environmental Data Feeds

Public and institutional environmental datasets integrated as supplementary context around primary sensing sources.

Forecasting & Decision Engine

Four domains. Two windows. One reviewable state.

EDS combines local telemetry, retained history, and station health to estimate 1-hour and 3-hour advisory risk across astronomy, rain, dew / fog, and storm conditions. Data quality, thresholds, confidence, and physics-aware safeguards remain visible before the system exposes a reviewable state.

01 / Local signal
Observed here
- Temperature
- Humidity
- Pressure
- Light
- Station health
02 / Retained context
Meaning over time
- TimeAligned
- HistoryRetained
- Gaps / stalenessVisible
- Station qualityChecked

03 / Forecast matrix

Four domains

Representative advisory outlooks across four domains and two forecast windows
Domain	1 h	3 h
Astronomy	Outlook	Outlook
Rain	Outlook	Outlook
Dew / fog	Outlook	Outlook
Storm	Outlook	Outlook

Advisory windows, not live values

04 / Decision gate
Evidence before state
- Data quality
- Thresholds
- Historical baseline
- Confidence
- Physical safeguards
- Conservative hold
05 / Reviewable advisory
Advisory state Review Reasons visible Advisor / dry-run
Proposed action only

Representative system model. No live data. Explore the complete EDS system

Operational Awareness Outputs

Awareness built for operators, not just data for analysts.

The output layer is designed around how conditions actually get used day to day: quick situational read, spatial context, timely notification, and a record to look back on.

Situational Summaries

Condensed, human-readable summaries of current and forecasted environmental state.

Geographic Overlays

Map-based views that place sensor data, forecasts, and alerts in spatial context.

Alert Delivery

Configurable notification channels for qualifying conditions and threshold breaches.

Historical Trend Views

Time-series views for tracking how conditions evolve against baseline patterns.

Exportable Reporting

Structured exports for downstream analysis, recordkeeping, or integration with other systems.

Adaptive Device Control

External control builds can consume EDS observing telemetry and turn it into device-local regulation and physical action.

Reference build / Automatic Optical Dew Control

Prototype Status

An active prototype, built in layers.

EDS is under active development. The system is being built layer by layer, with each stage validated against real test data before the next is added. It is not being shipped as a single fixed release.

Data Ingestion Pipeline

Implemented

Core ingestion and normalization for initial sensing sources is in place and running against live test data.

Forecasting Models

In Progress

Baseline statistical forecasting is operational; machine-learning model refinement is ongoing.

Alerting & Delivery

Prototype

Threshold-based alert generation exists; delivery channels are being expanded and hardened.

Operator Interface

Prototype

A dedicated EDS Console now runs current conditions, history, and advisory reasoning as structured, reviewable views, not a fixed dashboard product.

Field Deployment

Pilot

A pilot station is field-deployed and has been tested across real seasonal events; expanding to additional sites is the next milestone.

Note EDS now has its own dedicated site with interface captures, system detail, and field studies — explore.eds.mdnt.pt .

Field Study / 01

One event, two witnesses: the August 2026 solar eclipse.

The Espinho EDS station preserved 111 / 111 complete one-minute event-window observations on 12 August 2026. A separate CosmoNutz optical record was captured in the Vila Praia de Âncora area. The public archive keeps the two locations, clocks, and evidence types distinct, then aligns them on one working Replay timeline.

Ambient light reached 11 lux at 19:32:22 WEST, eight seconds after Espinho predicted maximum. The record supports high-confidence eclipse detection through ambient light. Secondary temperature, humidity, pressure, and dew-point effects were not uniquely attributed; this was one event, at one station, during one sunset.

Event: 97.9% partial solar eclipse
Location: Espinho, Portugal · Aug 12, 2026
Minimum illuminance: 11 lux · +8 s after Espinho predicted maximum
Dataset: 111 / 111 observations · 12 comparison evenings

Open the Eclipse Archive Replay the Event

Field Note / 02

Repeated local transitions across a winter sequence.

The published Winter Storm Field Note examines the Espinho station record from 20 January through 10 March 2026. It preserves five approved local pressure windows, applies one frozen comparison method, and keeps supported, below-threshold, and negative evidence visible.

Three official sequences met the study's 10 hPa / six-hour screening contract. Ingrid remained below that threshold, while the March period showed no major local pressure signature. This is an observational record only: it does not establish historical forecast skill, warning lead time, national storm severity, or a forecasting threshold.

Period: 20 Jan–10 Mar 2026 · UTC
Coverage: 71,464 / 72,000 minute bins · 99.256%
Supported sequences: 3 official sequences across 5 retained windows
Largest local 12-hour fall: −20.223 hPa
Acquisition: 60-second median interval

Read Field Note Data & Method

Collaboration

Technical collaboration on environmental sensing and decision support.

MDNT is open to focused collaboration on EDS: sensing integrations, forecasting model design, environmental data partnerships, and structured field pilots. If your work touches environmental monitoring, infrastructure operations, or applied environmental research, get in touch.

Sensing & data integrations
Forecasting & alerting model design
Environmental data partnerships
Field pilot opportunities

labs@mdnt.pt

PGP encrypted contact (optional)

Fingerprint: 20F4 1435 3E7E A1A6 7EC0 EA11 8DFF F119 FF5B 8464
Algorithm: ECC (Curve25519)
Function: Encryption, decryption, signing, verification
Created: 2026-07-10

Public Key