GIS, Remote Sensing and AI for Biodiversity Monitoring Training Course

Participants can structure ecological spatial data for reliable analysis and reuse.

• Vector, raster and tabular data roles within biodiversity monitoring workflows
• Coordinate reference systems selected using extent, measurement and data-source requirements
• Geodatabases organised with naming conventions, metadata and version control
• Attribute domains standardise habitat classes, survey methods and observation status
• Spatial joins connect sightings, plots, management units and administrative boundaries
• Topology checks identify gaps, overlaps, duplicates and invalid monitoring polygons

Participants can design spatial sampling schemes and prepare field data for geospatial analysis.

• Probability sampling designs compared for species, habitat and threat surveys
• Stratified random sampling allocates effort across ecological and management strata
• Transect, plot and point-count locations generated from access constraints
• Detection probability considered when interpreting absence and abundance records
• Mobile data schemas capture coordinates, effort, observer and taxonomic evidence
• Field validation plans connect image classifications with reference observations

Participants can select appropriate remotely sensed data for specified biodiversity monitoring objectives.

• Passive and active sensors compared for habitat structure and condition
• Sentinel-2 spectral bands matched to vegetation, water and bare-ground questions
• Landsat archives support long-term land-cover and disturbance time series
• Spatial, spectral, temporal and radiometric resolution evaluated against monitoring needs
• Cloud cover, seasonality and phenology assessed before image selection
• Scene metadata records acquisition dates, processing levels and known limitations

Participants can prepare satellite imagery and derive documented ecological indicator layers.

• Surface reflectance products selected instead of uncorrected top-of-atmosphere imagery
• Cloud, shadow and cirrus masks remove contaminated Sentinel-2 pixels
• Band resampling aligns pixel sizes before multi-band analysis
• Normalized Difference Vegetation Index measures relative green vegetation response
• Normalized Difference Water Index identifies open water and moisture patterns
• Index limitations documented for dense canopy, soil background and seasonal variation

Participants can create habitat classification schemes that serve monitoring and management decisions.

• Habitat classes defined through observable spectral, structural and ecological characteristics
• Classification legends align field categories with reporting and management needs
• Object-based and pixel-based approaches compared for habitat mapping
• Training samples separated from independent accuracy-assessment reference samples
• Class imbalance addressed through targeted sampling and aggregation decisions
• Minimum mapping units prevent misleading detail in final habitat products

Participants can build and assess supervised classifications for habitat and land-cover monitoring.

• Random Forest classifiers trained using labelled spectral and environmental features
• Support Vector Machine parameters tuned for separable habitat classes
• Feature selection removes redundant bands, indices and terrain variables
• Confusion matrices calculate producer, user and overall classification accuracy
• F1 scores assess performance where habitat classes remain imbalanced
• Post-classification smoothing applied without erasing ecologically significant patches

Participants can apply AI-assisted image interpretation with appropriate validation and governance controls.

• Machine-learning outputs treated as evidence requiring ecological and spatial validation
• Training-data provenance records source, date, interpreter and confidence level
• Human-in-the-loop review targets low-confidence and high-consequence classifications
• Bias checks examine geographic, seasonal and observer-related sampling gaps
• Sensitive-location controls restrict outputs that could expose threatened species
• Model cards document intended use, limitations, performance and review ownership

Participants can derive terrain, hydrological and landscape metrics relevant to biodiversity condition.

• Digital elevation models generate slope, aspect and topographic wetness indices
• Hydrological flow direction identifies catchments, drainage paths and wetland connections
• Distance-to-edge surfaces quantify fragmentation around habitat patches
• Patch size, core area and connectivity calculated using landscape metrics
• Least-cost paths model movement routes across resistance surfaces
• Ecological interpretations distinguish structural connectivity from functional connectivity

Participants can detect, quantify and communicate significant habitat changes over time.

• Image dates matched by season to reduce phenological change errors
• Post-classification comparison quantifies transitions between mapped habitat classes
• Spectral change metrics identify disturbance before full land-cover conversion
• Time-series charts distinguish persistent trends from isolated anomalous observations
• Change polygons attributed with area, date range and confidence evidence
• Minimum detectable change documented against pixel size and classification accuracy

Participants can prepare occurrence records for species modelling while managing observation bias and sensitivity.

• Occurrence records screened for coordinate errors, duplicates and taxonomic synonyms
• Coordinate uncertainty retained rather than falsely precise spatial locations
• Spatial thinning reduces clustered records and sampling-bias effects
• Presence-only, presence-absence and abundance records require different modelling approaches
• Background points selected to reflect accessible areas and survey effort
• Data sensitivity classifications govern access to threatened-species occurrence records

Participants can develop and interpret species distribution models using occurrence data and environmental predictors.

• MaxEnt models estimate relative habitat suitability from presence-only records
• Environmental predictor layers screened for ecological relevance and collinearity
• Variance inflation factors identify strongly correlated predictor variables
• Cross-validation partitions occurrence records for independent model testing
• AUC, sensitivity and omission rates assessed with ecological judgement
• Suitability thresholds selected according to survey purpose and decision risk

Participants can select monitoring approaches that account for imperfect detection and repeated observations.

• Occupancy models separate occurrence probability from observation detection probability
• Repeated surveys provide evidence for estimating detection across sampling occasions
• Abundance indices distinguished from population estimates and density measures
• Camera-trap effort standardised using trap nights, placement and operating periods
• Acoustic monitoring data linked to recording schedules and species-call validation
• Inference limits stated when sampling effort cannot support population conclusions

Participants can develop spatial indicators for habitat condition, degradation and restoration progress.

• Reference-condition baselines define realistic habitat condition comparison points
• Vegetation cover, bare ground and invasive extent form condition indicators
• Restoration trajectories compare intervention areas against reference and control sites
• Before-after-control-impact designs strengthen attribution of restoration outcomes
• Remote-sensing indicators paired with field observations for ecological interpretation
• Condition scorecards distinguish implementation activity from biodiversity response

Participants can prioritise locations for action using spatial biodiversity risk and value criteria.

• Risk layers combine habitat value, threat exposure and receptor sensitivity
• Multi-criteria decision analysis weights criteria through documented stakeholder assumptions
• Constraint maps identify legally protected, culturally sensitive and inaccessible areas
• Cumulative impact mapping considers existing, approved and proposed disturbances
• Avoidance, minimisation, restoration and offset options mapped along impact pathways
• Priority rankings tested through sensitivity analysis of weights and thresholds

Participants can establish quality assurance and reproducibility controls for biodiversity geospatial products.

• Quality assurance plans assign checks, acceptance criteria and approval responsibilities
• Metadata capture source lineage, processing steps, assumptions and data limitations
• Reproducible workflows separate raw data, scripts, intermediates and final outputs
• Spatial accuracy assessed against reference data and intended map scale
• Uncertainty maps communicate confidence alongside habitat and suitability results
• Peer review questions test ecological plausibility and decision fitness

Participants can communicate biodiversity monitoring findings accurately to technical and non-technical audiences.

• Map layouts apply scale bars, legends, north arrows and source statements
• Classification labels use plain ecological terms without concealing uncertainty
• Indicator dashboards link spatial trends to targets, thresholds and actions
• Technical methods sections describe data sources, algorithms and validation results
• Executive summaries state findings, limitations, decisions and recommended actions
• Accessible visual design avoids misleading colour ramps and exaggerated symbols

Participants can plan the people, data, governance and review cycles required for sustained monitoring.

• Monitoring calendars coordinate satellite acquisition, field seasons and reporting deadlines
• Roles define data stewardship, technical review and management decision authority
• Standard operating procedures control collection, processing, approval and publication
• Data retention schedules preserve raw evidence and documented analytical lineage
• Budget assumptions cover imagery, field effort, software, storage and specialist support
• Adaptive management cycles connect monitoring results with intervention adjustments

Participants can build, review and present a complete biodiversity monitoring project package for a defined management decision.

• Project scope defines decision owner, ecological receptors and monitoring boundaries
• Monitoring design links indicators, sampling methods, imagery and review intervals
• Geospatial workflow produces habitat maps, change analysis and priority areas
• Validation evidence documents accuracy, uncertainty, data gaps and review findings
• Management recommendations connect mapped evidence with feasible response actions
• Final presentation defends methods, limitations and implementation priorities to reviewers