Machine Learning Models for Hyper-Local Weather Prediction

Location: Surry Hills, NSW

Duration: 4-5 months

Project Background

The primary factor driving variability in Agricultural production is the weather. Weather affects the growth of crops but also, importantly, the activities that are required to grow, harvest and process produce. Gridded weather products are most often quite inaccurate at predicting the climate on farms, especially in cases of semi-covered agriculture. The Yield has built systems to instrument properties and collect and normalise weather and related phenomena from farms to provide hyper-local data about growing conditions. This data is then combined with gridded weather forecasts using various ML techniques to provide hyper-local forecasts. The models developed so far have worked well for The Yield’s MVP. The Yield has gathered actuals from the field across a wide range of climates and is now using new sources of forecast data.

Research to be Conducted

Investigate, develop and report on ML models to create hyper-local predictions for a range of weather phenomena across Australia using The Yield’s sensor data collection and high-resolution gridded forecast data source. This may include contributing to research and development of an automated feature-engineering and model selection framework.

Investigate, develop and report on ML models to create hyper-local mid-range (30 – 60 days) predictions of rainfall likelihood using The Yield’s sensor data collection and high resolution gridded forecast data sources as well as sources (to be determined by this project) of long range wide field climate forecast indices.

Skills Required

We are looking for a PhD student with the following:

ESSENTIAL

  • Demonstrable experience coding in Python with working knowledge of keras, pandas, numpy, and scikit-learn.
  • Strong problem-solving skills with an understanding of the constraints on developing machine learning solutions.
  • Working knowledge of querying and manipulating spatial and temporal datasets.
  • Strong organisational, communication and collaboration skills to build and strengthen working relationships in both academic and commercial environments.
  • Ability to thrive in a dynamic and fast-paced environment of product discovery and development.

DESIRABLE

  • Experience developing models using machine learning to achieve outcomes in areas such as environmental, climatic, geospatial, marine or agriculture modelling.
  • Experience working with weather and climate datasets.
  • Experience visualising data with Power BI or Tableau.

Expected Outcomes

  1. Investigate, obtain and normalise data from a range of long-range weather indices.
  2. Develop one or more predictive model(s) for mid-range rainfall likelihood using tools and data obtained in (1) and other data provided by The Yield.
  3. Present a research report on the model containing a comparative analysis showing why this model is being put forward over others. This must include an error analysis across all micro-climates in our network.
  4. The model may then be “productionised” and integrated into our production system and must use agreed tools and interfaces to enable this.

Additional Details

The intern will receive $3,000 per month of the internship, usually in the form of stipend payments.

It is expected that the intern will primarily undertake this research project during regular business hours, spending at least 80% of their time on-site with the industry partner.  The intern will be expected to maintain contact with their academic mentor throughout the internship either through face-to-face or phone meetings as appropriate.

The intern and their academic mentor will have the opportunity to negotiate the project’s scope, milestones and timeline during the project planning stage.

Applications Close

25 September 2019

Reference

APR – 1187