In the past, increases in agricultural production were directly linked to increased energy, land, and capital utilization. Mechanization was the primary factor driving increases in agricultural production in the early 20th century.  By the mid to late 20th century fertilizers, pesticides, and genetics were driving advancements in agricultural production.  Today, another wave of technological innovation known as Digital Agriculture is poised to revolutionize farming.  

In general, production agriculture has been slower and more cautious to adopt big data technology than other industries.  This presentation will discuss some of the promises and pitfalls related to Digital Agriculture, and then examine some real-world examples of current research and development work at OSU.  One example to be reviewed will be on-going efforts to create robust and intelligent control systems and sensors to help monitor, optimize, and manage autonomous agricultural machinery systems.  Integrated sensing for agricultural machinery control and automation is a complex and challenging topic.  Augmenting a priori spatial information with real time sensor data is needed to achieve the next level of automated agricultural machines.  A excellent source of a priori data is the Oklahoma Mesonet.  A recent project analyzed 21 years of historical weather data (~106 data files) from the Oklahoma Mesonet system.  The data examined the practicality of flying unmanned aircraft for various agricultural purposes in Oklahoma.  Current research is looking at how to best link real-time sensor data with the vast amount of a priori data.

Mobile, real-time agricultural sensors need "look ahead capability," centimeter-level resolution, and entire root zone coverage.  Most data points need to be geo-located and time stamped.  The difficulty of obtaining soil property data quickly and cheaply remains one of the biggest sensing challenges for digital agriculture.  Autonomous agricultural machines need to assess the terrain they are driving through and typically scan for obstacles such as wet or soft soil.   Soil trafficability is the capacity of soils to support vehicle travel.  Basic trafficability factors include, soil strength, stickiness, slipperiness and variation with weather.  In order to develop accurate variable-rate prescriptions for crop inputs like seed and fertilizer, it is necessary to estimate soil parameters such as; soil moisture versus depth, organic matter, pH, CEC, soil texture, nitrogen levels.  Issues involved with this data include collection speed, data accuracy (sensor reading vs. actual lab value), sensor resolution (temporal & spatial), local vs. wide-area (regional or global) calibration, etc.  These data sets quickly become very large.  Work is on-going with regard to algorithms and methods of data processing and analysis.

Presenter(s)

Paul Weckler