Large testing of AI-based camera traps on a forest defoliating pest in Europe

This repository contains the raw and compiled datasets used in the study comparing conventional pheromone traps and automated TrapView traps for monitoring the pine processionary moth (Thaumetopoea pityocampa) in France, Italy, Portugal and Slovenia.

Folder descriptions

raw_data_flight_France, raw_data_flight_Italy, raw_data_flight_Portugal, raw_data_flight_Slovenia

These folders contain the raw flight-monitoring data for each country. Each Excel file corresponds to an individual trap and includes daily moth catches throughout the monitoring period.

• Files ending with _DAILY contain daily catches obtained by summing the moth counts recorded from the three automated images captured each day.
• The remaining files contain the original trap records used to derive the daily datasets.

raw_hourly_climate_data_France, raw_hourly_climate_data_Italy, raw_hourly_climate_data_Portugal, raw_hourly_climate_data_Slovenia

These folders contain the hourly climatic data recorded by each automated TrapView trap throughout the monitoring period, including:

• Air temperature (°C)
• Relative humidity (%)

These data were used to derive the climatic variables included in the analyses.

compiled_data_all_countries.xlsx

This file contains the compiled dataset used for statistical analyses. Each row corresponds to the daily observations for one trap and includes:

• Country
• Site
• Habitat
• Trap ID
• Daily catches recorded by automated TrapView traps
• Weekly catches recorded by conventional traps
• Nightly mean temperature (°C)
• Nightly mean relative humidity (%)

Nightly climatic variables were calculated as the average of hourly measurements collected between 21:00 and 06:00, corresponding to the nocturnal flight activity period of the pine processionary moth.

Nest_density_per_ha_France_Italy.xlsx

This file contains winter nest density per hectare used to evaluate the relationship between adult trap catches and local infestation levels in Italy and France. For each sampling site , the dataset provides the number of pine processionary moth winter nests per hectare associated with the corresponding trap type.

 Supplementary Material

It presents confusion matrices for low and high moth activity, based on manual validation of  TrapView images,. Representative TrapView images illustrate true positives, true negatives, false positives, and false negatives.