Prediction of protein content in malting barley using proximal and remote sensing

• Published in: Precision agriculture 2010-12, Vol.11 (6), p.587-599
• Main Authors: Söderström, Mats, Börjesson, Thomas, Pettersson, Carl-Göran, Nissen, Knud, Hagner, Olle
• Format: Article
• Language: English
• Subjects: Agricultural Science; Agriculture; Agronomy; Atmospheric Sciences; Barley; Biomedical and Life Sciences; Chemistry and Earth Sciences; Computer Science; Crops; Fjärranalysteknik; Food Science; Harvest; Harvesting; Jordbruksvetenskap; Life Sciences; Livsmedelsvetenskap; malting barley; Nitrogen; Physics; Prediction models; Proteins; Proximal sensing; Ratios; Reflectance; Regional differences; Remote Sensing; Remote Sensing/Photogrammetry; Satellite image; Satellite systems; Satellites; Sensors; Soil Science & Conservation; Statistics for Engineering; Studies; Vegetation index; Vegetation mapping; Weather; Wheat; Yara N-Sensor
• ISSN: 1385-2256; 1573-1618; 1573-1618
• DOI: 10.1007/s11119-010-9181-6

This paper examines the prediction of within-field differences in protein in malting barley at a late growth stage using the Yara N-Sensor and prediction of its regional variation with medium resolution satellite images. Field predictions of protein in the crop at a late growth stage could be useful for harvest planning, whereas regional prediction of barley quality before harvest would be useful for the grain industry. The project was carried out in central Sweden where the variation in protein content of malting barley has been documented both within fields and regionally. Scanning with an N-sensor and crop sampling were carried out in 2007 and 2008 at several fields. The regional data used consisted of weather data, quality analyses of the malting barley delivered to the major farmers' co-operative, crops grown and field boundaries. Satellite scenes (SPOT 5 and IRS-P6 LISS-III) were acquired from a date as close as possible to the N-sensor scans. Reasonable partial least squares (PLS) models could be constructed based on weather and reflectance data from either the N-sensor or satellite. The models used mainly reflectance data, but the weather data improved them. Better field models could be created with data from the N-sensor than from the satellite image, but a local satellite-based model based on a simple ratio (middle infrared/green) in combination with weather was useful in regional prediction of malting barley protein. A regional prediction model based only on the weather variables explained about half the variation in recorded protein.