Combining ultrasonic sward height and spectral signatures to assess the biomass of legume-grass swards

• Authors: Thomas Fricke;Michael Wachendorf
• Venue: Computers and Electronics in Agriculture
• Year: 2013

In binary mixtures of either white clover (Trifolium repens L.), red clover (Trifolium pratense L.) or lucerne (Medicago sativa L.) with perennial ryegrass (Lolium perenne L.) as well as in pure swards of each single species, biomass has been assessed measuring sward height with an ultrasonic distance sensor and spectral-radiometric reflections. Measurements and sampling of reference data were conducted along a wide range of biomass levels on 0.25m^2 subplots at 17 dates in 4 growth periods defined by 4 main cuts between September 2005 and September 2006. To improve biomass (BM) predictions on exclusive ultrasonic sward height (USH) by complementation of vegetation indices (VIs), a collection of existing and hyperspectral VIs have been evaluated in combination with USH. While red/NIR-based VIs performed suboptimal, indices representing bands related to water absorption or the NIR-shoulder showed better predictions. Best prediction accuracies were achieved by a combination of USH with sward-specifically selected 1nm bands using the normalized spectral vegetation index (NDSI) reaching R^2-values of 0.83 in common swards and 0.88-0.90 for species-specific calibrations, respectively. Broadening of bands up to 100nm did only marginally reduce prediction accuracies. Using fixed NDSI bands selected from common swards instead of sward specific selected ones, did not significantly reduce prediction accuracy. It is identified that VIs act complementarily to USH and can avoid overestimations of BM frequently observed in grassland by the exclusive use of USH. Both, bandwidth flexibility and fixed NDSI band configurations can facilitate a configuration of sensors for legume-grass swards in a wide range of yield levels.