2019 ASHS Annual Conference

As unmanned aerial systems (UAS; “drones”) have become more accessible over the past decade, interest in their agricultural applications for both production and research have also increased. While considerable research has investigated the utility of UAS-based remote sensing in field crops, comparatively little research has focused on vegetable systems where variation in plant biophysical features and production system characteristics present unique challenges and opportunities. The primary objective of this research was to evaluate the potential for high-resolution multi-temporal aerial imagery to monitor growth characteristics and nitrogen status of fresh market tomatoes. A case study was carried out in 2018 in a tomato nitrogen fertility and cover crop trial at the Michigan State University Horticulture Teaching and Research Center. The background design was a factorial of four rates of N applied to the tomatoes (preplant urea; 0, 56, 112, 168 kg N ha^-1) and a legacy of four fall cover crops (none, oilseed radish, rapeseed, hairy vetch), arranged in a split plot RCBD with 4 replications. Tomatoes were grown staked on raised beds with black plastic mulch and drip irrigation. Color imagery was collected at approximately one-week intervals between transplanting and harvest using a DJI Phantom 4 Pro with a 20 MP camera at an altitude of 45 feet, and used to create color orthomosaics and digital surface models (DSMs) at 3-5 mm resolution. Initial processing involved the isolation of pixels which included vegetation using the binary ExGR index in order to exclude background soil/plastic pixels in the subsequent analysis. Three visible-spectrum vegetation indices (BGI, VDVI, NPCI) were calculated as estimates of relative N status, and the DSMs were used to estimate plant heights. Ground validation measurements included plant heights measured at five time points between late June and mid-August, and tomato N status measured by leaf tissue nitrogen concentration and SPAD meter in mid-July, approximately 8 and 9 weeks after planting, respectively. Height estimates were highly correlated overall with measured heights (R²=0.907, p<0.001), and both estimated and measured heights resolved significant N treatment differences over the study period (p=0.0024 and p=0.0067 respectively). N treatment differences were resolved using vegetation indices through July and August, with more significant p-values than SPAD measurements. While more research is required to fully realize UAS utility for tomatoes and other vegetable crops, robust monitoring of biophysical characteristics at high temporal resolution has the potential to support both research data-collection and farm decision-making.