Estimating Development Stages of Corn from Spectral Data — An Initial Model

When applying remote sensing technology to crop area estimation or crop condition assessment, a knowledge of the crop development stage at the time the remotely sensed data is acquired is important. In addition to historical statistics and weather driven development stage models, remotely sensed spectral data provides a potential tool for estimation of crop development stage. A model based upon spectral data should also provide advantages in terms of simplicity, timeliness, and applicability to areas as small as individual fields. In this investigation, a model for determining development stage of corn (Zea mays L.) from spectral measurements in the four Landsat bands of 0.5 to 0.6, 0.6 to 0.7, 0.7, to 0.8, and 0.8 to 0.11 <m is proposed. Estimated corn development stages using the model were determined from Landsat data acquired over sample areas 01 the U.S. Corn Belt as well as agricultural research plots at the Purdue University Agronomy Farm, West Lafayette, Indiana. The estimated corn development stage using the model was compared to ground observations made with the Hanway scale (1966). Corn development stage estimates made from Landsat data had a correlation of R² = 00964 with the ground observed stages, being most accurate at stages earlier than Hanway Stage 6.0 (blister). Analysis of spectral data obtained from corn research plots at Purdue Agronomy Farm revealed that the model worked well with variation in plant population and soil background differences as well as arbitrary division into grain yield levels due to various experimental conditions. For all treatment variations of soil type and plant population and yield groupings, the R² of calculated vs. observed development stage exceeded 0.95. This model's advantage for remote sensing is that it uses only spectral data and does not require meteorological data such as temperature, precipitation, and daylength. The model can be applied to individual corn fields rather than a general application to a region as large as crop reporting districts (or a state) and was found to work well for fields sampled over a large geographic area.