Using LIDAR to measure forage yield of perennial ryegrass (Lolium perenne L.) field plots

Abstract

A LIDAR-based tool for non-invasive estimation of plant biomass in perennial ryegrass field plots was developed. This included designing and making a prototype of a machine for LIDAR data collection, and developing algorithms for data processing. The biomass estimates were validated with regression analysis against harvest data. The project was implemented in three phases. In phase 1, a prototype carrying frame and a light-excluding cover was constructed for the LIDAR scanner. An algorithm was developed for grass plot segmentation, ground surface detection and estimation of plant biomass. Phase 2 focused on developing the prototype tool further, including application-specific real-time capture end-user software for data capture and analysis. This included testing the algorithm and in-field testing of the software. An experiment was also conducted to study how the variation in ground level between different scans affected the measurement. It was found that the variation of ground level was significant (more than 20 mm) between adjacent scans and within each segment. An improved method with correction for soil surface variation was developed to estimate the ground level of each scan and increase the accuracy of biomass estimation. In phase 3, 86 segments in replicated field plots of a perennial ryegrass cultivar trial in Canterbury, New Zealand, were scanned with LIDAR at early, mid and late time points, with mechanical harvest and yield data collection at the late growth stage. Significant (P<0.0005) correlations were observed between processed LIDAR data and fresh and dry weights of plant foliage biomass with R2 values of 0.78 and 0.76, respectively. The late-growth calibrated data were used to explore ryegrass growth dynamics using LIDAR scans at early growth and mid-growth stages.