A comparison of growing season agrometeorological stress and single-date Landsat NDVI for wheat yield estimation in west central Saskatchewan

Abstract

Modeled moisture stress from three weather stations and normalized difference vegetation index (NDVI) values from historical Landsat images for an area of west central Saskatchewan were utilized to assess the yield potential of several wheat fields, some of which were suspected to have suffered residual herbicide damage. The study was initiated to assess whether the crop yields in recent years had been adversely affected by residual herbicide damage. There was a wide variation in wheat yields between fields within a growing season and between growing seasons. NDVI values from individual fields could not provide a reliable estimate of wheat yield. However, a linear regression of mean annual producer-reported wheat yields on mean growing season moisture stress from three weather stations that surrounded the study area was significant with a coefficient of determination (r2) of 0.85. Although mean annual producer-reported wheat yields showed a direct relationship with mean NDVI, the regression was not significant. Mean NDVI was also poorly correlated with mean growing season moisture stress. This may be partly related to the inability to source cloud-free Landsat images from late July when the wheat would normally be near peak growth. A within-year comparison of individual yields and NDVI values for fields that had been treated and those that had not been treated with herbicide showed that there was no statistical difference between them.