Camera Trap Processing

name camera-trap-processing description Processes camera trap image records into structured detection data, activity patterns, and trap effort summaries. Use this skill when the user mentions camera traps, wildlife cameras, trap nights, detection events, diel activity patterns, camtrapR, temporal overlap indices (Dhat), RAI (relative abundance index), camera station data, detection history generation, or independence thresholds for photo events. skill_version 1.0.0 Skill: camera-trap-processing Domain: Camera traps · Detection events · Activity patterns · Occupancy · Abundance indices Purpose Guides the agent through processing raw camera trap image data into validated detection records, calculating activity pattern metrics, and preparing outputs for occupancy and abundance estimation. Covers detection event definition, independence filtering, trap effort computation, diel activity analysis, and integration with the occupancy-and-detection skill via detection history matrices. When to Invoke Invoke this skill when: A user provides a directory of camera trap images or a raw detection CSV for processing The goal is to calculate wildlife activity patterns, diel overlap, or relative activity indices Detection history matrices are needed as input for occupancy models The user asks about independent detection events, trap effort, or camera operation Photographic mark-recapture or N-mixture models are planned trigger_keywords: camera trap , wildlife camera , detection event , trap night , activity pattern , diel activity , photographic index , camtrapR , detection history , occupancy camera , camera operation , independent detection Inputs Input Format Required Image directory (organised by camera station) Directory tree Required Camera metadata CSV (station, lat, lon, setup date, retrieval date) CSV Required Species list for filtering TXT or CSV Recommended Independence threshold (minutes) Integer (default: 30) Optional Detection record table (if images already processed) CSV Conditional Outputs Output Description record_table.csv One row per independent detection event per species per camera detection_history.csv Binary site × occasion matrix ready for unmarked camera_operation.csv Daily operation status per camera (1=active, 0=inactive) trap_effort_summary.csv Trap-nights per station and per species records_per_species.csv Count of independent events per species activity_plot.png Kernel density of diel activity with 95% CI activity_overlap.csv Diel overlap index Δ between two groups circular_stats.csv Mean activity time, concentration (κ), Rayleigh test Steps Validate camera metadata Confirm all stations have setup and retrieval dates, coordinates, and unique IDs. Run process_camtrap_data.R with the image directory and metadata CSV. The script creates the camera operation matrix from station dates. Extract EXIF and build record table camtrapR::recordTable() reads image timestamps and species labels from directory structure. Check that directory hierarchy matches: <station>/<species>/<images> . Output: record_table.csv with columns station, species, datetime, filename. Apply independence filter Events from the same species at the same station within indep_threshold_min (default: 30) are collapsed into a single event. Record the threshold in decision_log.md . If threshold is changed from 30 min, justify the choice using home range data. Calculate trap effort Compute trap-nights per station from the camera operation matrix. Flag stations with < 100 trap-nights in trap_effort_summary.csv . Do not use stations with < 20 trap-nights in occupancy estimation. Generate detection history matrix Use camtrapR::detectionHistory() with the chosen occasion length (default: 1 week). Output: binary matrix with rows = stations, columns = occasions. Pass to occupancy-and-detection skill for model fitting. Estimate diel activity patterns (optional — invoke estimate_activity.R ) Fit von Mises kernel to circular time-of-day data. Calculate Dhat4 overlap index between two groups (e.g., dry vs. wet season). If n_independent_events < 10 per species, report relative activity index only. Validate outputs Confirm all output files are non-empty. Check that detection history dimensions match (n_stations × n_occasions). Verify no station has all-NA rows. Record decisions in decision_log.md . Decision Points Condition Diagnosis Recommended Action Time between photos < independence threshold Events are not independent — same animal Collapse to one event; use 30 min default unless home range data justifies otherwise n_independent_events < 10 per species Insufficient data for occupancy estimation Report relative activity index (RAI) only; do not fit occupancy model Camera operational time < 100 trap-nights Insufficient sampling effort Flag station; exclude from occupancy analysis; include in effort summary Stations < 15 Below minimum for occupancy model Use naive occupancy with explicit caveat; see occupancy-and-detection skill Missing EXIF timestamps Images cannot be time-ordered Request re-processing; use filename timestamps as fallback if consistent Key Decisions to Document Record the following in decision_log.md after running this skill: Which independence threshold (minutes) was used and why How many events were collapsed due to non-independence Which stations were excluded due to low effort and why What occasion length was used for detection history and why Whether any species had insufficient events for occupancy (RAI reported instead) Tools and Libraries R suppressPackageStartupMessages ( library ( camtrapR ) )