Why it matters
Without AI, unidentified death patterns and undocumented causes lead to misinformed deaccession, delayed replacements, and gaps in wild-origin material preservation.
From design to safety
Design
Design mortalities into policy and data schema
Define mortality categories, data standards, provenance types, and validation rules to shape collection planning and risk-aware target species for ex situ conservation, while anchoring policies to regulatory guidelines and insurance requirements.
- Accession index cabinet
- Tree health risk log
- Irrigation control hub
Build
Ingest data and establish provenance
Capture accession histories, mortality events, site-factor records, and replacement plans; ensure data lineage and verifications align with Index Seminum and curatorial policies, plus genetic traceability where available.
- Accession index cabinet
- Tree risk assessment forms
- Root zone sensor array
Operation
Monitor mortality and alert workflows
Continuously read mortality signals, correlate with site factors, trigger alerts for imminent risk and recommended actions to plant records and curators; integrate with notifications and escalation protocols.
- Plant health surveillance notebook
- Glasshouse climate log
- Root zone sensor array
Maintenance
Plan deaccession and replacement cycles
Assess deaths by cause, schedule replacements, update records, and adjust site-factor metrics to improve future survival probabilities, while aligning with breeding programs and seed bank policies.
- Accession index cabinet
- Irrigation control hub
- Tree risk assessment forms
Safety
Risk governance and biosecurity checks
Verify decisions with the plant health officer and garden director; document safety reviews for deaccession, new plant introductions, and visitor risk considerations, including emergency response planning.
- Plant health surveillance notebook
- Tree risk assessment forms
- Glasshouse climate log
Part by part
- 1
Accession index cabinet
- Watch for
- Untracked deaccessions or undocumented mortality causes obscure true trend signals.
- AI action
- Reads mortality events and correlates with site factors; flags anomalies for review.
- 2
Tree health risk log
- Watch for
- Late-stage decline without explained cause reduces detection of systemic issues.
- AI action
- Aggregates mortality by species and site; triggers investigations.
- 3
Irrigation control hub
- Watch for
- Soil moisture anomalies co-occuring with sudden losses indicate irrigation failure.
- AI action
- Monitors irrigation data; sends maintenance alerts when thresholds breached.
- 4
Root zone sensor array
- Watch for
- Unexpected soil moisture gradients near plant loss sites.
- AI action
- Detects anomalies; forecasts drought stress or overwatering risk.
- 5
Glasshouse climate log
- Watch for
- Microclimate shifts preceding tissue necrosis or disease outbreaks.
- AI action
- Compares daytime/nighttime climate data; prompts climate adjustments.
- 6
Tree risk assessment log
- Watch for
- Emerging structural faults near public access without quick action.
- AI action
- Evaluates risk scores; recommends removal or protection plans.
Worked example
Catching a silent mortality pattern before deaccession may escalate
Illustrative scenario with sample data.
A mid-sized arboretum detects rising losses among a subset of Quercus within poorly drained beds, suggesting site-factor stress and potential root issues.
Signal in
gardenArea: 40 acres; accessions: 3200; glasshouses: 5, sensors: 28; mortalityEvents: [{species: 'Quercus robur', site: 'Bed 7', date: '2025-09-14', count: 2, cause: 'unknown'}]; siteFactors: {soil: 'clay loam', drainage: 'poor', rainfall: 'seasonal'}; records: {provenance: 'ex situ', indexSeminum: true}
System out
Mortality trend flagged for Quercus robur in Bed 7 with probable root zone hypoxia; deaccession risk assessment initiated; replacement plan proposed; site drainage improvement recommended; plant health officer notified.
Human sign-off: Head Curator of Living Collections reviews the mortality cluster, approves deaccession of affected trees, and signs off on drainage work and replacement plan.
How to roll it out
Phase 1
Weeks 1 to 4
Establish data schemas, calibrate sensors, and align governance for mortality tracking.
Phase 2
Weeks 5 to 12
Ingest historical records, set alert rules, train staff on interpretation and follow-up actions.
Phase 3
Weeks 13 to 20
Run pilot in one living-collection bed and validate deaccession workflows.
Where people stay in charge
- Decisions on deaccession or replacement must be signed by the Head of Living Collections or Garden Director.
- Why: ensures alignment with conservation priorities, governance, and risk tolerance.
- Any: All mortality-caused decisions rely on verified data and plant health officer concurrence.
Questions owners and crews ask
How does AI identify causes of mortality when records are incomplete?
AI cross-references mortality events with site-factor data, provenance details, and historical trends to infer likely causes, while flagging uncertainties for curator review.
Who validates deaccession recommendations produced by the system?
Deaccession recommendations require approval from the Head of Living Collections and the Garden Director, with plant health officers consulted for safety and biosecurity.
What data sources are essential to feed the mortality model?
Key sources include accession records, mortality event logs, site-factor databases, soil and irrigation sensors, climate logs, and Index Seminum provenance records.
How does the system handle false positives in mortality alerts?
Alerts are routed to curators for verification; confidence scores and supporting observations are documented before any action is taken.