Yield estimation for corn with multitemporal and multisensoral remote sensing data

Abstract

To establish relationships between remote sensing data and agricultural yield a sufficient ground-truth is needed. Plant parameters (green LAI (leaf area index) and biomass) of corn (Zea mays) fields were measured weekly in two different test sites during three vegetation periods (1990, 91, 92). Also an extensive landuse mapping of the test sites took place every year. One requirement for remote sensing data is to cover all developmental stages with increasing and decreasing LAI of corn crops. Eight LANDSAT-TM images, one airborne multispectral GER-image, and four field spectrometer measurements from different years and seasons, which fulfill the given requirement, were chosen. The data were calibrated to absolute reflectance values in the bandwidth of TM and geocoded. Using the ground data, a functional relationship between dry biomass and the green LAI was established. The green LAI was first transformed into a new term called DLP (day of the year, LAI and phenology). The DLP contains the green LAI, the season by multiplying the green LAI with the day of the year and the phenology on the day of measurement. This new term DLP gives an accurate tool for estimating the dry biomass of a corn field on any date of interest. To enable the use of remote sensing data to determine DLP, the relationship of LAI to spectral measurements was used. The green LAI correlates strongly with the arNDVI (absolute reflectance normalized difference vegetation index) calculated from the absolute reflectance values of the infrared and red wave bands of the different sensors. Deriving the green LAI of corn from LANDSAT-TM coupled with the landuse map, the dry biomass and the yield of each corn field in the testsite were calculated. >