10108838Integrating machine learning into the management of Ips typographusClosedGrant for R&DInnovate UKThis project aims to develop and test the 'Remote Ips Trap' (RIT) for deployment in the field. _Ips typographus_ is the most significant pest of the highest throughput timber crop in the UK and has recently been identified in stressed and dying spruce in the southeast of England, windblown from outbreaks in continental Europe. Successful management of this pest requires early intervention and current policies aim for early detection and eradication. Monitoring work relies on a network of pheromone traps which capture _Ips_ flying into sites, providing vital data on phenology and dispersal distance, as well as indicating which sites are at greatest risk of _Ips_ establishment from propagule pressure. Sites which are positive for breeding _Ips_ are subject to intense eradication trapping over three years to remove over-wintering beetles, usually with 5-40 traps per site. Currently 10-20 new sites are found annually. However, timely intervention can be hampered by delays in sample acquisition and analysis which typically takes 3-6 weeks and, as the samples cover a two-week period, data interpretation can be difficult. We aim to develop a machine-learning model to identify _Ips_ in laboratory samples, significantly speeding up sample analysis. Once validated, this model will be used in a field-deployed RIT capable of photographing samples in the trap remotely, identifying _Ips_ and transmitting that data back to a GUI for interpretation. This will significantly reduce the time taken for sample analysis and will improve temporal resolution from two weeks to hourly. Speeding up sample analysis will allow identification work to scale efficiently with yearly increases in the number of sites requiring eradication trapping, ensuring surveillance remains cost effective. In addition, the RIT will have immediate impacts on the surveillance and monitoring of _Ips_ in the field, reducing the time delay between beetles flying into the trap and surveillance taking place to allow for real-time interpretation of results, and potentially allowing forest works to take place before any predicted summer flight which is not currently possible.