Skip to content

Arrival traces

scripts/arrival_traces/generate_arrival_traces_v2.py derives the Campaign-2 (August 4–28, 2025) arrival streams used in the systems analysis from the Darwin Core tables. It has no arguments: run it from the scripts/ directory of a table set, and it reads ../events_all_modalities.csv, ../occurrences_all_modalities.csv, and the drone Occurrence table, and writes ../analysis/arrival_traces_v2/.

cd "$SMARTWILDS_DWC_OUT"/scripts
python generate_arrival_traces_v2.py

Two layers, one stream per modality:

Layer Unit of arrival Streams
Acquisition One recording or image capture acoustic_wav_recording, camera_image_capture
Inference One retained Occurrence camera_operational_occurrence, bioacoustic_machine_occurrence, drone_machine_detection_crop, citizen_science_curated_occurrence

Every output row links back to its source Event or Occurrence (sourceID), carries a second-level timestamp and a timeEligible flag (date-only records are kept but excluded from inter-arrival statistics), and names the release it came from. arrival_trace_contract_v2.json records, per stream, the unit, inclusion rule, date window, counts, rate, coefficient of variation, and burstiness; input_manifest_v2.json carries SHA-256 checksums of inputs and outputs.

Rates are time-eligible arrivals divided by the fixed 600-hour window; CV is the population standard deviation over the mean of successive inter-arrival intervals; burstiness is B = (CV - 1) / (CV + 1). The arrival statistics are in the supplemental material.