YEA Lab and YEA Journal: A Data Science Portal and Hypermedia Publication for Place-Based Ecological Monitoring

Document ID: CNL-FN-2026-029
Version: 2.0
Date: March 7, 2026
Author: Michael P. Hamilton, Ph.D.

AI Assistance Disclosure: This field note was developed through extended working dialogue with Claude (Anthropic, Opus 4 and Opus 4.6). The concepts, vision, and architectural decisions are the author's; Claude contributed to articulating the four-layer framework, the Journal's publication loop model, the three-tier data architecture, the instrument workflow model, and synthesizing the discussion into structured form. The author takes full responsibility for the content.

Abstract

This note captures the design rationale for two interconnected extensions to the Your Ecological Address (YEA) platform: YEA Lab, a data science portal for interactive ecological monitoring and analysis, and YEA Journal, a hypermedia publication layer that narrates the findings emerging from the platform's accumulating data. Where the field guide presents an interpretive reading experience organized around curated natural areas, the Lab provides the analytical workspace where monitoring data can be explored across time, space, and ecological gradients. The Journal sits between them as the publication layer — a place-based naturalist's magazine that synthesizes trends, highlights discoveries, and links readers into the Lab to explore the evidence behind the stories. Together with the existing field guide, these form a four-part architecture: Guide (reference), Journal (publication), Lab (analysis), and Archive (documentation) — each serving a distinct cognitive mode while sharing a common data substrate.

The Lab is organized as eight instrument stations along a research workflow, not as independent tools selected from a menu. Each station follows a consistent assess-provision-analyze pattern, supported by a domain-specific Strata AI collaborator. The output of each station feeds the next: a curated place list from the Site Finder becomes the input context for the Climate Analyst, which feeds the Panorama Analyst, and so on through habitat characterization, biodiversity analysis, and pattern detection. A Research Notebook runs alongside the entire traversal, documenting the analytical path in sufficient detail for reproducibility. The Journal & Publication Designer at the end of the workflow transforms research findings into curated public narratives.

The instruments operate across a three-tier data architecture: Level 1 (ecological address data from 21+ geospatial APIs), Level 2 (curated place metadata and enrichment), and Level 3 (continuous monitoring streams from physical, virtual, and human instruments). Each Lab instrument draws from some combination of these levels depending on its analytical purpose.

1. Motivation

The YEA field guide was designed as an interpretive layer: you stand at a place, you want to understand it. Cards tell you what's here — the geology, the ecoregion, the climate envelope, who curates it. Monitoring widgets show you what instruments are currently observing. Gallery media show you what the place looks like across seasons and years. This works well for its purpose.

But the monitoring system has quietly become something larger. Physical sensors like the Ecowitt soil and microclimate array and the Tempest reference weather station collect data continuously, while Open-Meteo virtual weather stations deliver modeled hourly data for any coordinate. BirdWeather acoustic stations detect and classify species around the clock. The iNaturalist integration tracks new biodiversity observations as they accumulate. Panoramic photo stations capture seasonal change at permanent viewpoints. Each of these instruments generates time series data that compounds daily.

The field guide's card-and-drawer interface cannot do justice to this accumulating corpus. A monitoring widget squeezed into a 280-pixel-tall drawer can show the current state, but it cannot support the kind of exploratory analysis that the data invites: What does the temperature record look like across an entire year? How does the bird community shift week by week through spring migration? When did the first Oregon white oak leaves emerge this year compared to last? These are workbench questions, not reading-experience questions. They require a different interaction model — adjustable time windows, zoomable charts, comparative overlays, data export.

2. Four-Layer Architecture

The YEA platform is reconceived as four distinct layers, each serving a different cognitive mode:

2.1 Field Guide — The Reference Layer

The existing public-facing application. Card-based, scroll-driven, designed for comprehension. You select a place or enter coordinates and receive an ecological profile: terrain, climate, ecoregion, land cover, biodiversity, conservation status. Monitoring widgets provide status indicators. The gallery shows the visual record. AI narratives offer interpretive lenses. This layer answers: What is this place?

The Guide retains a compact Field Log on each curated place page — a brief running summary of recent activity and notable observations. But the full journal experience lives elsewhere.

2.2 YEA Journal — The Publication Layer

A standalone hypermedia publication, accessible from a main tab on the YEA home page. Not a blog. Not a linear diary. A place-based naturalist's magazine that the system itself helps write.

The Journal is organized around curated places, but it reads like a publication you follow rather than a reference you consult. Nobody opens a field guide to browse. Nobody opens a data workbench without a hypothesis. But a journal is something that comes to you with what's new, what's changing, what's surprising.

Each curated place generates a running column populated by four kinds of content:

Place Cards — compact ecological address summaries. The TL;DR version of the full field guide profile. What biome, what elevation, what climate envelope, what's notable. A reader encountering a place for the first time gets oriented in a single card.

Trend Digests — instrument-generated summaries of what's happening. "Spring is running 8 days ahead of the 10-year phenological mean at Canemah Bluff." "Acoustic diversity peaked in the second week of May — 47 species detected, the highest weekly count since monitoring began." "The February temperature anomaly (+4.2 degrees F) was the strongest in the Open-Meteo reanalysis record for this grid cell." These are written by the monitoring system's AI summarizer, reviewed and verified by human curators.

Q&A Highlights — drawn from the Science persona's interpretive output, reframed as questions a curious naturalist would ask. "Why are the Oregon white oaks leafing out two weeks early?" "What explains the three-week gap in Barred Owl detections?" "How does this basalt bluff create a microclimate that supports species 200 miles north of their expected range?" The question format turns AI narrative output into something that reads like inquiry rather than report.

Lab Callouts — links to specific experiments, visualizations, and datasets in the Lab. "Explore the full acoustic migration timeline →" "Compare spring green-up across all Pacific Northwest sites →" "Download the hourly temperature record for your own analysis →" These create the publication loop: instrument collects data, Lab visualizes trend, Journal narrates the finding, reader follows the link to explore the evidence.

The Journal can be read at two scales: scan across all places for a landscape-level sense of what's happening in the network, or dive into a single place for deep seasonal narrative.

2.3 YEA Lab — The Workbench Layer

A standalone application launched from a main tab on the YEA home page. Place-centric but not place-confined. Eight instrument stations organized as a research workflow (see Section 5), each with full interactive workspaces: time series with adjustable windows, cross-place comparisons, the ecoSPLAT terrarium models running in a proper viewport, data export, experiment configuration. This layer answers: What is this place doing, and how does it compare?

The Lab does not share the field guide's panel-based GUI. It is a purpose-built data science interface designed for exploration, analysis, and discovery.

2.4 CNL Archive — The Documentation Layer

The existing technical document series (CNL-TN, CNL-SP, CNL-FN, etc.). Research findings, system specifications, protocols, and working papers that document how the platform itself works and what it has discovered. The Archive is the institutional memory — the place where methodology is recorded, design decisions are justified, and results are reported in formal scientific style.

The Archive feeds the Journal (a technical note's findings can be summarized as a Journal entry) and the Lab (a protocol document specifies how an instrument should be configured). But it operates at a different register — written for peers and posterity, not for the casual naturalist reader.

2.5 The Knowledge Cycle

The four layers form a generative loop:

1. Lab instruments collect, provision, and analyze data continuously.
2. Journal narrates the findings — what changed, what's surprising, what it means.
3. Guide provides the reference context — the ecological address that makes the findings interpretable.
4. Archive records the methodology and the formal analysis.

A reader might enter at any point. A birder checking the Guide's species list for Canemah Bluff sees a monitoring widget showing recent acoustic detections. She taps through to the Journal entry about early Swainson's Thrush arrival. The entry links to a Lab visualization of migration timing across five years of acoustic data. The methodology behind the acoustic classifier is documented in a CNL Technical Note in the Archive. Each layer is self-contained but linked, and the reader follows her curiosity through whatever depth she wants.

3. Three-Tier Data Architecture

The Lab instruments operate across three levels of information within the YEA platform. Each level represents a different depth of ecological knowledge, and each Lab instrument draws from some combination of these levels depending on its analytical purpose.

3.1 Level 1 — Ecological Address

For any location on Earth, the YEA platform compiles a coherent, spatially aligned ecological profile by querying 21+ external geospatial APIs in parallel. The resulting profile spans anthropogenic (land cover, land protection), geographic (terrain, geology, coordinates), environmental (climate classification, climate history, water balance, current weather), and biotic (ecoregion, vegetation type, biodiversity, conservation-listed species) categories. The profile is generated on request and cached in the lookup_cache table with source-appropriate TTL values.

Some of these data are dynamically changing (weather, recent species observations) and some are derived from historical databases (geology, ecoregion boundaries, climate normals). But all are available for any coordinate, whether or not that location is a curated place. This is the universal baseline — the ecological address that works everywhere.

3.2 Level 2 — Curated Places

The second level is the growing network of named, enriched sites stored in yea_places and its related tables. These began as the original 33 biological field stations from the OBFS Virtual Field project and have expanded to 1,000+ sites across multiple categories: field stations, ecological reserves, preserves, LTER and NEON sites in the United States, International LTER sites worldwide, bird conservation centers, observatories, and sanctuaries. Additional programs are under review for inclusion.

Each curated place carries information that no API can provide: site abstracts, stewardship descriptions, history, facilities, access information, organizational affiliations, associated people, media galleries, photopoints, and AI-researched enrichment narratives. This institutional knowledge deepens interpretation at known sites and provides the place-level context that makes cross-site comparison meaningful.

3.3 Level 3 — Monitoring Widgets

The third level is the continuous data streams attached to curated places, classified by observer_class in yea_monitoring_sources:

Physical instruments — hardware deployed in the field. Ecowitt soil and microclimate sensor arrays, Tempest reference weather stations, BirdWeather PUC acoustic monitors, STRATA stratification arrays, webcams, and other specialized sensing platforms. These produce the highest-quality data but scale slowly and require maintenance.

Human instruments — observations generated by people. Photo-monitoring imagery at permanent viewpoints, species observations submitted through iNaturalist and eBird, station records maintained by site staff. The quality depends on observer effort and expertise.

Virtual instruments — the conceptual equivalent of a data logger for globally generated and modeled data. A virtual instrument defines a source (e.g., Open-Meteo), a location (coordinates), and a sampling interval. A cron job continuously captures the data and saves it to the database. Open-Meteo climate and weather data, WeatherFlow NearCast normalized measurements, and NOAA point location models can all be configured as virtual instruments. WeatherFlow, the parent company of Tempest, offers NearCast as a network-derived product that normalizes current measurements — currently rainfall, with plans to extend to all weather variables — for any location encompassed by their station network. Unlike Level 1's on-demand API calls, a virtual instrument produces an accumulated time series — the same API, but a fundamentally different data product.

The distinction between Level 1 and Level 3 matters. When a researcher queries an ecological address, she gets the current Open-Meteo snapshot. When she opens the Climate Analyst in the Lab, she gets the accumulated hourly record from a virtual instrument that has been running continuously for months or years. Both use the same API, but the Level 3 data supports time series analysis, anomaly detection, and trend identification that the Level 1 snapshot cannot.

3.4 Level Interactions

Each Lab instrument draws from these levels differently. The Site Finder (Lab 01) works primarily at Levels 1 and 2 — filtering curated places by their ecological address attributes. The Climate Analyst (Lab 02) needs all three — Level 1 for baseline climate classification, Level 2 for place context, and Level 3 for the continuous monitoring streams that power time series analysis. The Biodiversity Analyst (Lab 05) uses Level 2 for place context and Level 3 for accumulated observation records, while relying on Level 1's species count data as a coverage indicator.

4. Spatial Scales

The Lab operates at three spatial scales, each unlocking different kinds of ecological inquiry:

4.1 Point Scale

A single curated place. The monitoring instruments as they exist today, but given the interface they deserve. Full-width time series charts. Annotation tools. Date range selectors. Side-by-side comparison of co-located instruments (temperature overlaid with acoustic detections, for example). Data export in CSV and JSON. This is the baseline — every curated place gets this automatically once it has active monitoring sources.

4.2 Landscape Scale

The place plus its ecological neighborhood. Define the spatial extent by radius, watershed boundary, or ecoregion polygon. Pull virtual weather grids across the area using Open-Meteo's gridded data. Map iNaturalist observations as a density surface. Plot elevation-temperature gradients from valley floor to ridgeline. Overlay BirdWeather station detections within the landscape to map acoustic coverage. This is where the work at James Reserve with the Center for Embedded Networked Sensing (CENS) was heading in the early 2000s — except now the sensor network is virtual and the spatial coverage is continental.

4.3 Gradient Scale

Across places. Line up curated sites along a latitudinal transect, an elevation gradient, a precipitation gradient, a continentality gradient. Watch spring green-up march northward through phenological wave analysis. Compare bird arrival dates across sites. Overlay climate velocity data to identify places where environmental change outpaces species migration capacity — the climate refugia question. This requires multiple curated places with comparable monitoring configurations, which is exactly what the platform is designed to accumulate over time.

The first version of the Lab does not need all three scales. Point scale — the place-centric workbench — is sufficient to launch. Landscape and gradient scales grow naturally as more places are curated and more virtual instruments are deployed.

5. The Instrument Workflow

5.1 Workflow Model

The Lab's eight instruments are not independent tools selected from a menu. They are stations along a research workflow, where the output of each station feeds the next. A researcher enters the Lab with an ecological question, refines it through successive analytical stages, and emerges with documented findings ready for publication.

Each station follows a consistent three-phase pattern:

Assessment. Strata AI evaluates the working place list against the instrument's data requirements. Which sites have the necessary data? What monitoring sources are already configured? Where are the gaps?

Provisioning. Where data gaps exist, the instrument provides tools to fill them. Deploy virtual climate instruments at sites lacking weather monitoring. Import 360° panoramas from the researcher's own fieldwork. Configure SOMA meshes for sites with sufficient training data. This is the "Designer" half of the dual-named instruments.

Analysis. With comparable data available across the working place list, the instrument's full analytical capabilities engage. Time series comparison, PCA, gradient analysis, anomaly detection, cross-site correlation.

5.2 Strata AI

Each station has a domain-specific Strata AI collaborator that understands the instrument's analytical purpose, the data sources it draws from, and the methodological constraints that apply. At Lab 01, Strata helps decompose an ecological question into effective search parameters. At Lab 02, Strata performs data readiness assessment and can deploy virtual instruments. At Lab 05, Strata acts as a methodological gatekeeper, flagging sites where observation density falls below minimum thresholds for robust comparison and suggesting appropriate analytical approaches given the sampling limitations.

Strata is not a generic chatbot. It is an AI collaborator with knowledge of the specific instrument's domain, the three-tier data architecture, and the accumulated context from prior stations in the workflow.

5.3 The Eight Instruments

LAB 01 — Site Finder. Advanced search across the curated place network. Filter by Köppen climate, RESOLVE biome, ecoregion, elevation, network affiliation, monitoring sources, and conservation status. The AI-assisted Q&A helps the researcher refine ecological questions into effective search criteria: one site or many? Continental or regional? Specific species or functional group? The output is a working place list that becomes the input context for all subsequent instruments.

LAB 02 — Climate Analyst & Data-logger Designer. Full-resolution climate time series from physical sensors (Ecowitt soil and microclimate arrays, Tempest reference weather stations) and virtual instruments (Open-Meteo modeled data, WeatherFlow NearCast). Adjustable windows, overlays, anomaly detection, and data export. The Data-logger Designer half provisions the data infrastructure: assessing which sites in the working place list have active climate monitoring, which need virtual instruments deployed, and configuring the yea_monitoring_sources records and cron schedules for new virtual instruments. Historical reanalysis data (Open-Meteo archive, 1991–present) is available immediately; continuous forward logging begins once instruments are configured.

LAB 03 — Panorama Analyst & Terrarium Designer. 360° panoramic archive navigation across field stations with seasonal and annual switching. The ecoSPLAT Gaussian splat terrarium reconstruction viewer enables structural state vector extraction — a 10-parameter descriptor (linearity, planarity, sphericity, verticality, anisotropy, green fraction, hue entropy, red-to-green ratio, brightness, saturation variation) computed per terrarium cell and aggregated by spherical ring (see CNL-TN-2026-033). Network PCA on the 54-column structural matrix separates stations by biome and season along ecologically interpretable axes. The Designer half provides access to the panorama import pipeline for researchers contributing their own 360° data, and can trigger ecoSLAM reconstruction for panoramas without existing terrarium models.

LAB 04 — Habitat Analyst. Integrated habitat characterization from multiple viewing geometries and data sources. From above: drone orthomosaics and 3D models with AI-assisted individual canopy segmentation. From orbit: LANDFIRE EVT (1,068 NatureServe ecological systems), RESOLVE ecoregions, NLCD and Esri Sentinel-2 land cover classifications. From the ground: ecoSPLAT structural vectors carried forward from Lab 03. Whittaker-style gradient analysis plots vegetation community composition against environmental gradients from Lab 02 — temperature, moisture, elevation — to reveal how communities sort along ecological axes.

LAB 05 — Biodiversity Analyst. Species accumulation curves, acoustic detection phenology, and seasonal community composition from iNaturalist, eBird, and BirdWeather. Cross-site comparison of biodiversity signatures along climate and elevation gradients. Strata AI plays a critical quality control role here: biodiversity data from citizen science platforms reflect sampling effort as much as actual diversity. A site with 200 iNaturalist observers will show a richer species list than a remote station with three visiting researchers. Strata assesses observation density at each site, flags locations where detection records fall below minimum thresholds for robust comparison, and recommends appropriate analytical methods — rarefaction curves, coverage-based estimators, or restriction to taxa with adequate detection histories.

LAB 06 — Pattern Analyst. Temporal pattern detection across sensor streams, phenological signals, and biodiversity time series. This instrument operates at two analytical levels. At the conventional level, MEO-style correlation analysis (see CNL-WP-2026-022) computes environmental relationships (temperature-humidity, temperature-bird activity), dawn chorus biological intelligence, and automated pattern discovery using statistical methods. At the embodied sensing level, SOMA-style Boltzmann machine meshes (see CNL-TN-2026-014, CNL-SP-2026-015) learn the joint probability distribution over environmental and biological variables, detecting cross-domain anomalies as energy tension rather than threshold violations. SOMA meshes can be deployed as virtual instruments — trained on accumulated data at each site, running continuous inference against live sensor feeds, with results stored in the thrml_inference and thrml_anomalies tables. The Pattern Analyst is the analytical engine that surfaces findings for the Journal.

LAB 07 — Research Notebook & Workflow Designer. Session-level research documentation that runs alongside the entire Lab traversal, not as a final step but as a companion process. The Notebook records the analytical path through each instrument station: decisions made, inputs provided, findings generated, and throughputs to the next station — with timestamps, data provenance, and the researcher's own annotations. The Workflow Designer allows researchers to save analytical paths as reusable protocols: "Douglas Fir climate-structure comparison" — same sequence of instruments, same filter logic, same analysis configuration — that can be re-run next season with updated data, or shared with colleagues. The output is a reproducible research record — someone else could read the Notebook and understand not just the findings but the analytical path that produced them.

LAB 08 — Journal & Publication Designer. The bridge between Lab analysis and public narrative. Trend digests, Q&A highlights, and Lab callouts are drafted from instrument findings and Research Notebook records. The curator review and verification workflow ensures that AI-generated entries are checked against the data before publication. Human field notes — geotagged, timestamped, linked to places — provide the ground-truth annotations that close the loop between what instruments detect and what humans observe in the field. This instrument manages the editorial workflow described in Section 8.

5.4 Illustrative Use Case: Douglas Fir Forest Ecology

A researcher enters the Lab wanting to study Douglas Fir ecosystems across their geographic range:

Lab 01 — Site Finder. Strata AI engages: "Are you interested in one site or a comparative study across many? Do you have a continental preference? Douglas Fir specifically, or any fir species?" The researcher specifies Douglas Fir across Pacific Northwest sites. The search filters curated places by LANDFIRE EVT vegetation classifications that include Douglas Fir, producing a working list of 14 sites spanning coastal to interior, low elevation to montane.

Lab 02 — Climate Analyst. Strata assesses data readiness: "Of your 14 sites, 3 have physical weather stations with continuous records, 5 have virtual Open-Meteo instruments already logging, and 6 have no monitoring sources configured. For those 6, I can deploy virtual climate instruments — Open-Meteo hourly at each coordinate. Historical reanalysis back to 1991 is available immediately." The researcher authorizes deployment. With comparable climate data now available across all sites, she characterizes temperature regimes, precipitation patterns, growing season length, and frost-free days along elevation and latitude gradients.

Lab 03 — Panorama Analyst. Strata reports: "Of your 14 sites, 4 have panoramic archives with terrarium reconstructions. 2 have panoramas but no splat reconstructions. The remaining 8 have no 360° data." The researcher imports her own drone-captured 360° imagery from three of her field sites. Structural state vector analysis reveals the closed-canopy conifer signature — near-zero zenith brightness, moderate upper-ring green fraction, low horizon green fraction — that characterizes mature Douglas Fir stands, and she compares structural variation across sites.

Lab 04 — Habitat Analyst. Remote sensing classifications from LANDFIRE EVT and Sentinel-2 are overlaid with the structural vectors from Lab 03 and the climate gradients from Lab 02. Whittaker-style gradient analysis shows how Douglas Fir community composition varies with temperature and moisture across the 14 sites.

Lab 05 — Biodiversity Analyst. Strata flags a methodological concern: "Seven sites have sufficient iNaturalist observation density for robust species comparison. Four have sparse records — fewer than 50 research-grade observations. Two have strong eBird coverage but minimal iNaturalist data, creating taxonomic bias toward birds." The researcher restricts cross-site biodiversity comparison to the seven well-sampled sites, using rarefaction curves to account for remaining differences in sampling effort.

Lab 06 — Pattern Analyst. MEO-style correlation analysis reveals temperature-bird activity coupling patterns at each site. For the three sites with physical weather stations and extended BirdWeather records, SOMA meshes are trained to learn the joint distribution of weather and acoustic biodiversity, enabling cross-domain anomaly detection.

Lab 07 — Research Notebook. Throughout this traversal, the Notebook has documented every step: the initial query, the Strata dialogues, the place list refinements, the virtual instruments deployed, the analytical parameters used, the findings at each station. The complete record is reproducible and shareable.

Lab 08 — Journal. The researcher's findings — that Douglas Fir structural signatures vary predictably along moisture gradients, with closed-canopy interior sites showing distinctive structural vectors compared to coastal sites — are drafted as a Trend Digest with Lab Callout links to the PCA visualization and gradient analysis. A curator reviews and publishes the entry.

6. Virtual Instruments

The concept of virtual instruments is central to the Lab's scalability. A physical instrument requires hardware in the field: a weather station bolted to a pole, a BirdWeather microphone mounted under an eave, a camera on a tripod at a permanent photopoint. These produce the highest-quality data but scale slowly and cost money.

A virtual instrument requires only coordinates and an API key. Open-Meteo delivers modeled hourly weather for any point on Earth, derived from ECMWF reanalysis and high-resolution forecast models. eBird and iNaturalist deliver crowd-sourced biodiversity observations for any bounding box. Sentinel-2 delivers 10-meter land cover classification on a five-day revisit cycle. USGS StreamStats delivers watershed boundaries and flow statistics for any pour point in the United States. None of these require physical infrastructure.

The virtual instrument concept extends beyond on-demand API queries. A configured virtual instrument is a yea_monitoring_sources record with observer_class = 'virtual' that defines a source, a location, a sampling interval, and a cron schedule. The instrument continuously captures data and stores it in the database — the equivalent of a data logger for globally modeled data. This distinction is critical: an on-demand API call (Level 1) gives you a snapshot; a virtual instrument (Level 3) gives you an accumulated time series.

The Lab 02 Climate Analyst & Data-logger Designer operationalizes this concept. When a researcher's working place list includes sites without climate monitoring, the Designer half provisions virtual instruments at those coordinates, creating the yea_monitoring_sources records and configuring the cron schedules. Historical reanalysis data fills the past; continuous forward logging builds the future record.

The NEON observatory (National Ecological Observatory Network) is the natural comparison point. NEON operates 81 fixed terrestrial and aquatic sites with standardized instrumentation, producing continental-scale ecological data for predetermined research questions. It is a magnificent achievement — and it is rigid by design. Its sites are fixed, its measurements are predetermined, its spatial resolution is coarse (81 points across a continent).

The YEA Lab inverts this model. Curated places are added by human decision, not committee process. Virtual instruments can be deployed to any place in minutes. The spatial resolution is determined by the density of curated places, which can be increased in any region of interest. The measurement portfolio at each site is configurable and extensible. The trade-off is obvious: NEON's physical instruments produce higher-quality, more standardized data. The YEA Lab's virtual instruments produce lower-quality but far more spatially extensive and rapidly deployable data. The two approaches are complementary, not competitive.

SOMA meshes (see CNL-TN-2026-014) represent a further extension of the virtual instrument concept. A trained Boltzmann machine mesh that runs continuous inference against sensor streams is itself a virtual instrument — one that transforms raw environmental and biological data into energy landscapes where anomalies manifest as mathematical tension rather than threshold violations. SOMA meshes are deployed and managed through the Pattern Analyst (Lab 06) with the same provisioning pattern as other virtual instruments.

7. Relationship to Monitoring Widgets

The existing monitoring widget system (MW core + per-type modules: mw-pano, mw-ecowitt, mw-birdweather, mw-pano-video, mw-strata, mw-tempest) provides the data pipeline and the summary visualization for the field guide. These widgets will continue to serve their current role on the curated places page: compact status indicators showing current conditions, recent trends, and the most recent observation.

In the Lab, the same data sources power full-scale instruments. The MW module's endpoint and render functions can be extended with a labRender method that produces the full interactive workspace instead of the compact widget. Alternatively, the Lab instruments can be entirely separate modules that share only the API endpoints and data models. The choice depends on how much visual and interaction logic the two contexts share.

The monitoring card on the field guide gains an "Open in Lab" link for each instrument source. The compact widget is a window; the Lab instrument is the room.

8. The Journal as Publication Loop

The Journal serves a function that neither the Guide nor the Lab can: it creates a reason to come back. A reference work is consulted when needed. A workbench is used when you have a question. A journal is something you follow — it arrives with news.

The publication loop works like this:

Collection. Monitoring instruments run continuously. Weather stations record hourly. Acoustic classifiers process audio around the clock. iNaturalist observers upload sightings. Satellite revisits accumulate. The data compounds silently.

Detection. Periodic analysis routines (daily, weekly, seasonal) scan the accumulated data for noteworthy patterns: anomalies, milestones, firsts, records, trend inflections. "First neotropical migrant detection of the season." "Longest dry spell in the three-year monitoring record." "Tenth new fungal species observation this quarter." The Pattern Analyst (Lab 06) is the engine that surfaces these findings, using both conventional statistical methods and SOMA energy-based anomaly detection.

Narration. The AI summarizer generates a draft journal entry from the detected pattern. It draws on the Science persona's interpretive framework to frame the finding as a question worth asking, not just a statistic worth reporting. A human curator reviews, edits, verifies, or discards the draft. Verified entries are published to the Journal.

Invitation. Each Journal entry links to the specific Lab instrument, date range, and visualization that generated the finding. The reader who wants to see the evidence can follow the link and explore the data herself. The reader who just wants the story can stay in the Journal.

Response. A reader — or the curator — can write a human field note in response to what the instruments found. "I walked the bluff trail this morning and the thrushes are indeed singing from the big-leaf maple canopy along the north slope. Three singing males in the first quarter mile." This human ground-truth annotation feeds back into the system's understanding of what its instruments are detecting.

This is not a hypothetical workflow. The existing yea_field_log table already supports it: author_type distinguishes human from AI from verified entries, data_sources carries the JSON provenance, and verified_by records the human curator who approved an AI-generated summary. The Journal & Publication Designer (Lab 08) is the editorial interface to this data flow, and the Research Notebook (Lab 07) provides the analytical provenance that makes each published finding traceable back through the entire research workflow.

9. Implementation Sequence

The Lab instruments represent a substantial but modular development effort. Each instrument can be built independently while sharing common infrastructure (place selection, time controls, data export, Strata AI integration). A phased approach:

Phase 1: Lab Infrastructure & Climate Analyst. Build the Lab page infrastructure: URL routing, place selection, instrument navigation. Implement Lab 02 — the Climate Analyst & Data-logger Designer — as the first full instrument: weather time series for curated places using the existing Ecowitt/Open-Meteo data pipeline, adjustable date range, zoomable chart, temperature/humidity/precipitation overlays, data export, and virtual instrument provisioning. This establishes the layout paradigm, the three-phase assess-provision-analyze pattern, and the data pipeline patterns that every subsequent instrument reuses.

Phase 2: Site Finder. Implement Lab 01 — the Site Finder — with advanced search across the curated place network. Filter by ecological address attributes (Level 1), curated place metadata (Level 2), and monitoring source availability (Level 3). Strata AI-assisted Q&A for refining ecological questions into search parameters. The output — a persistent working place list — becomes the input context for all other instruments.

Phase 3: Panorama Analyst & Terrarium Designer. Implement Lab 03 with the ecoSPLAT terrarium viewer, structural state vector extraction (CNL-TN-2026-033), network PCA, and the panorama import pipeline. This builds on existing infrastructure: the 454 terrarium reconstructions, the describe_cell.py / harvest_matrix.py / analyse_matrix.py script suite, and the Pannellum viewer already integrated into the field guide.

Phase 4: Biodiversity Analyst. Implement Lab 05 with species accumulation curves, acoustic detection phenology (BirdWeather), observational biodiversity (iNaturalist/eBird), and seasonal community composition. Strata AI quality control for sampling effort assessment.

Phase 5: Habitat Analyst. Implement Lab 04 with integrated remote sensing classification, drone orthomosaic support, structural vector integration from Lab 03, and Whittaker-style gradient analysis against climate data from Lab 02.

Phase 6: Pattern Analyst. Implement Lab 06, integrating MEO-style correlation analysis with SOMA mesh deployment and inference visualization. This builds on existing infrastructure: the MEO v7 pattern analysis system and the SOMA proof-of-concept (CNL-SP-2026-015).

Phase 7: Research Notebook & Journal. Implement Labs 07 and 08. The Research Notebook requires session-level state management and integration points with all other instruments. The Journal & Publication Designer builds on the existing yea_field_log table and the curator review workflow.

10. Design Principles

Separation of concerns. The Guide is a reference. The Journal is a publication. The Lab is a workbench. The Archive is a library. Each layer has its own interaction model, its own visual language, its own cognitive demands. Trying to serve all four purposes in one interface produces mediocrity in all four.

Instruments as workflow stations. The eight Lab instruments are not independent tools picked from a menu. They are stations along a research workflow, where each station's output feeds the next. The working place list travels through the entire workflow, accumulating analytical context at each stop.

Assess-provision-analyze. Every instrument follows a consistent three-phase pattern. Before analysis, the instrument assesses data readiness across the working place list and provisions whatever is missing. The Climate Analyst deploys virtual weather stations. The Panorama Analyst imports 360° imagery. The Pattern Analyst trains SOMA meshes. The provisioning phase is not overhead — it is the "Designer" capability that makes the Lab a platform for building research infrastructure, not just consuming it.

Place-centric entry, gradient-capable growth. Every user enters the Lab through a specific place or set of places. Cross-place analysis emerges as a capability, not a requirement. The Journal follows the same principle: each place has its own column, and cross-place narratives emerge as the network grows.

Virtual instruments as first-class citizens. A modeled temperature time series from Open-Meteo is not a lesser substitute for a physical weather station — it is a different kind of instrument with different strengths (global coverage, historical reanalysis, no maintenance burden) and different limitations (spatial averaging, model error, no microclimate sensitivity). The Lab treats both with equal seriousness.

Strata AI as domain-specific collaborator. Each instrument's AI assistant understands that instrument's analytical purpose, data sources, and methodological constraints. At the Biodiversity Analyst, Strata is a methodological gatekeeper that flags inadequate sampling. At the Climate Analyst, Strata is a data readiness assessor that provisions virtual instruments. The AI is not generic — it is shaped by the instrument it serves.

The Journal is not a blog. Blogs are chronological and author-centric. The Journal is place-centric and event-driven. Content appears because something ecologically noteworthy happened, not because it's Tuesday. The publication rhythm is determined by the data, not a calendar.

Human editorial authority. AI generates. Humans verify. The Journal never publishes an AI-generated entry without curator review. The verified author type is the gold standard — it means a human ecologist read the AI's summary, checked it against the data, and affirmed it. This is not a bottleneck; it is the quality guarantee that makes the Journal credible.

Questions over statements. The Journal's Q&A format — drawn from the Science persona — frames findings as questions rather than declarations. "Why are the oaks leafing out early?" invites the reader into inquiry. "The oaks are leafing out early due to anomalous February temperatures" closes the conversation. The question format respects the reader's intelligence and acknowledges that ecological causation is rarely simple.

Data provenance everywhere. Every chart, every summary, every journal entry carries metadata about its sources: which API, which date range, which model version, which query parameters. The Research Notebook extends this to the analytical workflow itself — not just where the data came from, but what decisions were made at each instrument station. This is not bureaucratic overhead — it is scientific hygiene.

Bookmarkable, shareable state. Every view in the Lab has a URL that can be bookmarked and shared. yea.earth/lab/canemah-bluff/weather?from=2026-01-01&to=2026-03-07 drops you into exactly that view. Every Journal entry has a permalink. This is essential for cross-linking and for collaborative work.

Document History