Enhancement of maize yield mapping through remote sensing
This dissertation develops a new approach to crop yield mapping that integrates the standard yield monitor data with hyperspectral remote sensing data. This approach demonstrates potential to improve the accuracy and spatial resolution of yield mapping which has important implications for studying the effects of densely measured site specific management parameters. A Literature review establishes the intra-field yield mapping context in which this new technique appears, finding that techniques dependent on yield monitor data alone have not sufficiently satisfied the requirements necessary to demonstrate the efficacy of site specific management. Because of its importance in preprocessing one of the primary hyperspectral remote sensing data sets for this research, significant attention is given to a vicarious calibration technique for mitigating the effects of relative radiometric error. The technique is based on median filtering. This median filtering approach could easily be extended to other hyperspectral along-track scanners for relative radiometric calibration and validation. Finally, the new yield mapping procedure called Discretization, Extraction, Classification and Regression (DECR) is described and evaluated using a robust range of maize crop conditions and cultivation practices, three different hyperspectral sensors, and two different combines. Validating against hand harvested crop yield, the DECR procedure is consistently accurate and greatly reduces sensitivity to yield monitor error when compared with other methods. Inspection of the resulting yield maps clearly reveals micro-effects on yield consistent with an improved spatial resolution. Using yield maps of consistent accuracy and improved spatial resolution should enable better measurement of response to precisely applied site specific inputs.
Nielsen, Purdue University.
Agronomy|Remote sensing|Agricultural engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our