Theses and Dissertations

Issuing Body

Mississippi State University


Varco, Jac J.

Committee Member

Wallace, Teddy P.

Committee Member

Ambinakudige, Shrinidhi S.

Committee Member

Henry, W. Brien

Committee Member

Cox, Michael S.

Date of Degree


Document Type

Dissertation - Open Access



Degree Name

Doctor of Philosophy


College of Agriculture and Life Sciences


Department of Plant and Soil Sciences


Fertilizer N is one of the most costly inputs in corn (Zea mays L.) and cotton (Gossypium hirsutum L.) production and is a strong yield determining factor. Variable rate N fertilization has the potential to improve resource use efficiency, profitability, and help to minimize adverse environmental impacts. Vegetation indices (VIs) may be useful for in-season crop health monitoring to assist in fertilizer N management and yield prediction. This research determined the utility of aerial imagery in detecting corn and cotton response to varying N supply using five selected VIs. The VIs derived from aerial images, chlorophyll readings and tissue N for corn from V5 to V9 growth stages and cotton beginning the 1st week of flowering through to latelowering were used to relate to fertilizer N rates and plant N status and yield. The results showed that VIs derived from aerial imagery could be used to differentiate N supply and in-season grain yield of corn beginning at V5 to V6; however, models from later growth stages had greater r2 values than earlier growth stages. Single variable models that used VI, chlorophyll content, or plant N concentration as an independent variable were overall stronger than 2 variable Multiple Linear Regression models (MLRs). Three independent variables used in MLRs contained multicollinearity. For cotton, the use of VIs derived from aerial imagery to differentiate N supply may depend on environmental factors such as soil and weather. However, VIs may be useful for in-season lint yield prediction beginning the 1st week of flowering although later stages improved accuracy. The MLRs that were developed with 2 independent variables may be more suitable for in-season lint yield prediction than single independent variable models.