Submitted to: Precision Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/15/2002
Publication Date: 3/15/2004
Citation: SHUFENG, H., SCHNEIDER, S.M., EVANS, R.G., DAVENPORT, J.R. BLOCK ESTIMATING OF SPATIAL YIELD DATA AND ITS UNCERTAINTY. PRECISION AGRICULTURE, 5, 73-84, 2004. Interpretive Summary: Farmers have begun using crop harvesters equipped with yield monitors, a device that records yield and location at thousands of points in the field. These measurements are used to develop a map of the field which shows areas of high and low yield which can be used for future management decisions. Under perfect conditions, every part of the field would be harvested with a yield monitor-equipped harvester and every data point would be valid. Unfortunately this is not always the case. If a harvester breaks down, another harvester that is not equipped with a yield monitor might be used for part of the field, resulting in no yield information in that part of field. There might be problems with the yield monitor itself that result in bad information, such as a negative weight for yield. This information must be thrown out. How can a map be developed when there are gaps in the yield information? If estimations are used, how can a farmer know how much faith to put in the map? Our work showed that two different estimation methods could be used. One method was designed for use when only a few points are missing. The other method is used for large gaps in information. An additional map, called an uncertainty map, was developed that showed which parts of the field had enough good information to be certain of the accuracy of the yield, and where estimates had been used to fill in large information gaps which might lead to more uncertainty. These maps can be used when making management decisions. Expensive, potentially risky management options might be considered in areas of the field with high information accuracy and low uncertainty; whereas inexpensive, low risk options might be considered in areas of less accuracy and greater uncertainty.
Technical Abstract: On-the-go yield monitors have been available for both grain and bulk crops. Most of the yield monitors today provide yield measurement at a fixed time interval. Conversion of these "point" yield data into raster yield maps for further analysis is necessary. In this study, a data blocking procedure is proposed to create raster yield maps from "point" yield data. The blocking procedure includes: (1) converting the fixed-time-interval data into a fixed-distance-interval data; (2) using a moving average algorithm to estimate a cell value when there are sufficient data points within the cell; (3) using a geostatistical algorithm to estimate a cell value when there are not enough data points within the cell but values of its neighboring cells are known; and (4) calculating an uncertainty index for each cell value estimation. An example application of the yield blocking procedure with potato harvest data in 1996 was given.