Surviving the data deluge: geostatistical and signal processing methodologies for smart farm sensor networks

Abstract

There is a trend towards the deployment of more and more soil, plant, animal, asset, and machinery performance sensors on the farm. More sensors mean more data, especially if it is generated in live streams and it remains a significant challenge for it to be distilled to a manageable size and rendered in a useable form. Seemingly simple and intuitive 'front-ends' require specialised and complex algorithms and software working behind the scenes. The SMART FARM sensor network is an example of a future farm technology which can generate very large data sets. This network monitors meteorological and soil conditions over an area of approximately 500 acres. There are 100 nodes equipped with multiple sensors all transmitting data back to a central server at 5 minute intervals, 24 hours a day, 7 days a week. In this paper we will survey the range of statistical and computing tools being developed by the SMART FARM team to render this information rich data field into management-relevant information including visualisation, detection and understanding of trends, and generating critical state alarms.