Multi-label streaming: exploitation of aggregated environmental, sensor and management data sources ========================================================= This project focuses on the problem of managing and aggregating multiple data sources into a modelling framework that can harness the power of machine learning to fully characterise the development of complex environmental systems such as crop orchards. The challenge consists in integrating multiple heterogeneous data sources in an optimal framework. Crop yield forecasting is a crucial global challenge due to the increased demand in volume and quality from a growing world population. In addition, climate change, and the limited land available for horticulture highlight the importance of devising optimal and sustainable usage strategies of the resources available now to bridge the yield gap, i.e. the difference between the actual and the optimal potential yield in crop orchards. To address this issue, PlantTech, an MBIE-funded Regional Research institute is carrying out a large data-driven strategy leading to the characterisation of these complex agricultural environments. This data collection campaign features a heterogeneous coverage both spatially and temporally. Data sources include remote sensing (e.g. satellite or airborne imaging), weather stations, LIDAR scanners, proximal sensing, imaging of canopy for fruit counting and sizing, etc. A crop orchard represents a complex dynamic system that is affected by a variety of factors across different scales: the environment, plant phenotype, soil chemistry, microbiome, and orchard management decisions. Hence, attempting to capture the development and key features of an orchard system means to be able to collect data from various disparate sources spanning different temporal and spatial scales, and, most importantly, have the ability to aggregate this data in a unified framework to allow its optimal exploitation. This project, co-supervised by researchers from PlantTech Research Institute and the University of Waikato, aims at tackling this problem by developing machine learning methods in a multi-label streaming methodology with direct applications to data-driven horticulture and environmental science. How to apply To apply for this PhD position, please send the following by email to the contacts listed below: CV Two letters of recommendation Cover letter describing your research interests and plan Deadline for applications: 30th of November 2020. Contact Dr Alvaro Orsi PlantTech Research Institute alvaro@pri.co.nz Prof. Albert Bifet University of Waikato abifet@cs.waikato.ac.nz