Assessing large-scale almond field experiments using a mobile orchard phenotyping platform

The South Australian Research and Development Institute (SARDI) maintains a suite of almond experiments at the Almond Board of Australia's experimental orchard. These plantings exceed 17 ha (9000 trees) and explore almond response to density, training, scion cultivar, rootstock genotype and soil amendments. Repeated measures of development metrics and light interception were needed, but the scale of the field experiments made traditional manual methods impractical. A ground based mobile orchard phenotyping platform (MOPP) was developed to facilitate the rapid measurement of canopy metrics. The MOPP integrates multiple sensors upon an all-terrain vehicle including a computer with bespoke data capture and processing software, a global positioning system with inertial measurement unit, light detection and ranging, normalised difference vegetation index sensors, and a towed array of light sensors (ceptometer) adjustable to row widths from 4.5 to 7 m. Data are collated using geofencing and is reported on a per-tree basis. Since its development in 2020, the MOPP has been deployed on a four-to-six-week cycle across multiple experiments measuring canopy metrics such as tree height and width, trunk diameter, light interception and canopy NDVI. MOPP data collection, at around 1.5 ha h^-1, is faster than traditional methods allowing efficient repeated orchard assessment. It has demonstrated strong correlations between LiDAR and manual measurements of trunk diameter (R²=0.71, RMSE=72 mm) and canopy height (R²=0.82, RMSE=0.11 m). The ceptometer has quantified differences in canopy light interception in response to genotype, planting density and tree age. Repeated surveys show interactions between genotype and planting density over time. The precision of SARDI's orchard phenotyping platform continues to be refined as experimental trees, and their associated data sets, grow. These data sets are compared against ground and aerial remote sensing systems with a view to securing a rapid, repeatable and reliable method for characterising canopy development metrics.