Relating satellite imagery with grain protein content

Abstract

Satellite images, captured during the growing seasons of barley, sorghum and wheat were analysed to establish a relationship between the spectral response and the harvested grain protein content. This study was conducted near Jimbour (approx. 151°10’E and 27°05’S) in southern Queensland. Grain protein contents of the geo-referenced samples, collected manually during the harvest, were determined using a laboratory-based near-infrared spectrophotometer. Grain protein contents in grain varied between 7.4–15.2% in barley, 6.2– 10.6% in sorghum and 13.1–15.6% in wheat. The Landsat images of 18 September 1999 (a week after barley flowering), 5 March 2000 (three weeks before sorghum harvest), and 15 August 2001 (two weeks before wheat flowering) were analysed. Additionally, an ASTER image of 24 September 2001 (three weeks after wheat flowering) was also examined. Digital numbers, extracted from raw image bands and derived indices, were correlated with grain protein contents. The grain protein content in barley was correlated strongly (r>0.80) with bands 2, 4 and 5 of the Landsat scene, first principal component, and the tasselled cap brightness and greenness indices. Similarly, wheat protein content was well correlated (r>0.75) with the near infrared band (band 4) of the Landsat scene, first principal component, and the tasselled cap brightness, greenness and wetness indices. The band 3 (near infrared band) of the ASTER image, captured well after flowering, was moderately correlated (r<0.5) with the protein content of the wheat. However, the grain protein content in sorghum was found poorly correlated (r<0.20) with Landsat image bands and indices. Results indicate that it may be possible to use certain bands and indices of the satellite images, captured around the time of flowering, to predict grain protein content of barley and wheat crops.