AI-IoT integrated framework for tree tilt monitoring: A case study on tree failure in Hong Kong

Urban forestry is a challenging task in densely populated cities. Tree failure accidents often induce casualties and property damage that should be prevented in advance. In light of such a need, an AI-IoT integrated framework for tree tilt monitoring is proposed in this study, and consists of three stages: (I) data collection through a LoRaWAN-based IoT tree monitoring system, called Internet of Tree Things (IoTT) and data processing; (II) AI-enabled data analytics (anomaly detection); and (III) diagnosis of tree stability. With the tree tilt data wirelessly captured by the IoTT, associated anomalies in the measurement can then be automatically detected by the vanilla LSTM model, if the prediction error exceeds the threshold for the normal responses. A four-level likelihood of tree failure ratings based on the number of anomalies detected within a time-window is introduced accordingly. The proposed framework is further validated using data from five failure trees in Hong Kong, which had been constantly monitored lasting from ∼1 month to ∼22 months before their final failure during typhoons. The resultant lead time for categorizing these trees into the lowest level of the failure likelihood ranged from ∼8 hrs to ∼35 hrs and that into the highest level of the failure likelihood was from ∼1 hr to ∼30 hrs. By comparing the tree tilt angles in response to similar weather conditions, together with the mechanical responses, such as elastic or elastoplastic responses revealed in the temporal wind tipping curves in the subsequent inclement weather events, long-term tree stability trends can be examined with biomechanical evidence and can complement visual tree inspection in which the early signs of tree failure might be missed between inspections. By including the knowledge and judgement of arborists in the decision loop, the framework provides an evidence-based, objective, and effective approach to enhance urban forestry safety.