Jiyul Chang

Document Type

Thesis - University Access Only

Award Date


Degree Name

Doctor of Philosophy (PhD)

Department / School

Plant Science


Management zones based on field history, yield maps, topography, remote sensing, and producer preferences have the potential to reduce sampling costs and improve fertilizer recommendations. The objectives of this study were: (i) to determine the influence of different approaches to define nutrient management zones based on soil nutrient and crop yield variability; (ii) to evaluate fertilizer recommendation errors; and (iii) to determine if remote sensing data combined with readily available soil attribute information can be used to predict crop yield. This research was conducted in three eastern South Dakota fields. Soil samples taken in grid sampling were analyzed for Olsen P and N03-N. An AgLeader 2000 yield monitor was used to measure com (Zea mays L.) and soybean (Glycine max L.) yields between 1995 and 2001. Remote sensing was collected in the spring, summer, and late summer in 1999, 2000, and 2001. Over 20 different approaches for identifying management zone boundaries were tested and principal component analysis was used to develop yield prediction models. Soil nitrate and Olsen P concentrations were spatially variable. Yields in summit/shoulder areas were limited by too little water, while in wet years yields in the footslope areas were limited by too much water. For all the methods tested to identify management zone boundaries, splitting the fields into 4-ha blocks had the lowest nutrient, yield, and fertilizer recommendations pooled variances. The impact of block sampling on fertilizer recommendations was attributed to field management and soil forming processes. These results suggest that if areas are not physically connected, then they should not be composited into a single sample, and that both intrinsic and prior management must be considered in developing nutrient management zones. Yield models based on remote sensing data explained the most yield variability when the models used several dates of information. Yield models based on only remote sensing data collected in the summer explained the least amount of yield variability. Adding soil attribute and plant information to the models had a small impact on the ability of the model to exp lam yield variability. Findings from this study can be used by land managers to improve fertilizer recommendations and to estimate crop yields prior to harvest.

Library of Congress Subject Headings

Precision farming Soils -- Sampling -- South Dakota Remote sensing -- South Dakota Crop yields -- Mathematical models



Number of Pages



South Dakota State University