404-11 Improving Yield Data Quality with Data Integrity Zones.

See more from this Division: ASA Section: Agronomic Production Systems
See more from this Session: On-Farm Research: II. Advancing Precision Ag Tools

Wednesday, November 9, 2016: 2:00 PM
Phoenix Convention Center North, Room 223

Elizabeth M. Hawkins, Agricultural and Biological Engineering, Purdue University, Lafayette, IN and Dennis R. Buckmaster, Agricultural and Biological Engineering, Purdue University, West Lafayette, IN
Abstract:
The agriculture industry is transitioning from precision to decision agriculture and the importance of data on the farm is growing. For decades, the amount of data collected about field operations and crop performance has increased; however, for many farmers and researchers, the utility of this data has not yet been realized. Big data is currently being touted as the solution to this problem, but the application of these advanced analysis techniques present new challenges of their own. Data quality is a primary concern when using these techniques for decision-making and as the amount of data continues to grow, data quality will become even more crucial.

One approach is to use a smaller, more accurate subset of data collected in data integrity zones (DIZ) for analysis and interpretation of multi-temporal yield data. These zones are determined by utilizing descriptive metadata to identify areas within a field where errors and artifacts in the data are likely to be reduced. Data in DIZ were isolated and the reduced datasets (approximately 40% of the raw data) were used for comparative analysis of nine years of yield data. Data collected outside of the DIZ were compared to DIZ data to identify differences in not only yield but also combine operating conditions. The use of DIZ data increased the power to accurately detect and interpret yield differences within the field and reduced the error in the whole field yield estimate. The stability of the temporal yield response for each management zone was improved, making it possible to better understand in-field yield variation. This novel approach could enable the application of traditional statistical tools for on-farm research and larger scale yield trials, as well as give insight into improving data quality to feed future big data analytical models.

See more from this Division: ASA Section: Agronomic Production Systems
See more from this Session: On-Farm Research: II. Advancing Precision Ag Tools