Date of Award:

8-2011

Document Type:

Dissertation

Degree Name:

Doctor of Philosophy (PhD)

Department:

Civil and Environmental Engineering

Committee Chair(s)

Christopher M.U. Neale

Committee

Christopher M.U. Neale

Committee

Gary P. Merkley

Committee

Wynn R. Walker

Committee

Bruce Bugbee

Committee

V. Philip Rasmussen

Abstract

Multispectral aerial and satellite remote sensing plays a major role in crop yield prediction due to its ability to detect crop growth conditions on spatial and temporal scales in a cost effective manner. Many empirical relationships have been established in the past between spectral vegetation indices and leaf area index, fractional ground cover, and crop growth rates for different crops through ground sampling. Remote sensing-based vegetation index (VI) yield models using airborne and satellite data have been developed only for grain crops like barley, corn, wheat, and sorghum. So it becomes important to validate and extend the VI-based model for tuber crops like potato, taking into account the most significant parameters that affect the final crop yield of these crops.

This research involved developing and validating yield models for potato crop in southern Idaho fields using high-resolution airborne and satellite remote sensing. High resolution multispectral airborne imagery acquired on three dates throughout the growing season in 2004 was used to develop a VI-based statistical yield model by integrating the area under the Soil Adjusted Vegetation Index (SAVI) curve. The model was developed using hand-dug samples collected in two center pivots based on soil variability and crop growth patterns to account for variability in the leaf area duration and yields. The three date Integrated SAVI (ISAVI) model developed was then validated using 2005 spot yield samples collected from two center pivot fields and also tested for 2004 and 2005 whole field data over dozens of center pivot fields. The three- date model was applied using 2004 and 2005 satellite images and tested. The eight-date ISAVI yield model was also extended to satellite images to estimate the potato yield. The overall yield estimation using the eight-date ISAVI model was better than the three-date model as the image inputs covered the complete growth cycle of the crop from emergence to harvest.

Actual Evapotranspiration was also used as another independent variable in the model to improve the yield predictions. The actual ET was calculated using canopy reflectance based crop coefficient method for all the spot yield locations in 2004 and regressed with actual yield. Both actual yield and ET correlated very well. Multiple linear regression analysis was performed using two independent variables, namely, ISAVI and actual ET to predict the actual potato yield. The results showed a significant improvement in the correlation and the new model developed was validated using 2004 and 2005 whole field data. The results showed a reasonable RMSE and low MBE as well as a good linear correlation for both the years and a great improvement over yield estimated using only the three-date ISAVI in the simple linear regression model. A spatial variability analysis was also performed at different scales using airborne and satellite images to understand the typical spatial correlation within potato fields.

Checksum

4efa815ba79335045f8097fc21f445ba

Comments

This work was made publicly available electronically on September 29, 2011.

Download

Share

COinS

Copyright for this work is retained by the student. If you have any questions regarding the inclusion of this work in the Digital Commons, please email us at .