Validating satellite monitoring of dairy pastures in Canterbury with Lincoln University Dairy Farm and commercial farm data

Abstract

Validation of satellite-based prediction of pasture cover for dairy farms in Canterbury (New Zealand) during the 2008 and 2009 milking seasons is reported. Satellitebased predictions made using the new Canterbury model were validated against estimates from a rising plate meter for the Lincoln University Dairy Farm (LUDF) weekly farm walks and from nine commercial farms, across 15 images. Validation against LUDF data showed high coefficients of determination (mean R2 = 0.85, range 0.69 to 0.97 kg DM/ha) and low residual standard errors (mean RSE =138 kg, range 53 to 244 kg DM/ha). Validation against commercial data showed a higher level of variability between farms and images than the LUDF data. The Canterbury model accounted for a large proportion of the observed variability in pasture cover of dairy pastures when matched to high quality data, and showed seasonal trends in the model coefficients. Higher variability associated with commercial data may be attributed to geographic distribution, timing and method of data collection as well as varying levels of competency in monitoring skills. Keywords: Pasture, monitoring, satellite data, paddock ranking, pastures from space