Data Management for Site-Specific Farming

Data warehousing and data mining is a new research area for agricultural research. Our goal in this research program is to develop effective technologies to process the massive data set generated by precision farming production and research. In this study, we are optimizing both the sensing process and data to knowledge (D2K) conversion process. Automatic and supervised learning processes have been applied on large d atabase of agricultural crop systems. The objectives are to eventually understand the complicated system by means of processing a massive database with the state-of-the-art high performance computing systems.

Dr. Georing (professor emeriti) from our Department has initiated the precision farming data mining work. Preliminary results on the automatic learning process for crop systems are very encouraging. Data from the University of Illinois Morrow Plots was used, for a 30-year period for which consistent data involving modern production practices was available. A feed-forward, completely connected, back-propagation artificial neural network (ANN) was designed to approximate the nonlinear yield function relating corn yield to factors influencing yield. The relatively simple back-propagation, feed-forward neural network predicted corn yields with 80% accuracy; when an example with abnormally low yield was discarded, accuracy rose to 83.5%. We will set up a high-tech data management environment for the application of ground-based and remote sensing data to precision farming.

One of our proposals on geo-spatial data processing has been selected by the NSF-NCSA Faculty Fellow program to fund for 2000-2001. Dr. Tian will work in the National Center of Supercomputer Applications as a Faculty Fellow for one year to develop agriculture production related data mining tools. We propose to collect and share multi-location, multi-layer data and setup a web-based knowledge discovery in database (KDD) environment on a high-performance computing system for remote sensing based precision farming studies.