Unmanned aerial vehicles (UAVs) equipped with multispectral sensors present an opportunity to monitor vegetation with on-demand high spatial and temporal resolution. In this study we use multispectral imagery from quadcopter UAVs to monitor the progression of a water manipulation experiment on a common shrub, Baccharis pilularis (coyote brush) at the Blue Oak Ranch Reserve (BORR) ~20 km east of San Jose, California. We recorded multispectral imagery at several altitudes with nearly hourly intervals to explore the relationship between two common spectral indices, NDVI (normalized difference vegetation index) and NDRE (normalized difference red edge index), leaf water content and water potential as physiological metrics of plant water status, across a gradient of water deficit. An examination of the spatial and temporal thresholds at which water limitations were most detectable revealed that the best separation between levels of water deficit were at higher resolution (lower flying height), and in the morning (NDVI) and early morning (NDRE). We found that both measures were able to identify moisture deficit across treatments; however, NDVI was better able to distinguish between treatments than NDRE and was more positively correlated with field measurements of leaf water content. Finally, we explored how relationships between spectral indices and water status changed when the imagery was scaled to courser resolutions provided by satellite-based imagery (PlanetScope).We found that PlanetScope data was able to capture the overall trend in treatments but unable to capture subtle changes in water content. These kinds of experiments that evaluate the relationship between direct field measurements and UAV camera sensitivity are needed to enable translation of field-based physiology measurements to landscape or regional scales.

Extracting the shapes of individual tree crowns from high-resolution imagery can play a crucial role in many applications, including precision agriculture. We evaluated three CNN models—MASK R-CNN, YOLOv3, and SAM—and compared their tree crown results with OBIA-based reference datasets from UAV imagery for seven dates across one growing season. We found that YOLOv3 performed poorly across all dates; both MASK R-CNN and SAM performed well in May, June, September, and November (precision, recall, and F1 scores over 0.79). All models struggled in the early season imagery (e.g., March). MASK R-CNN outperformed other models in August (when there was smoke haze) and December (showing end-of-season variation in leaf color). SAM was the fastest model, and, as it required no training, it could cover more area in less time; MASK R-CNN was very accurate and customizable. In this paper, we aimed to contribute insight into which CNN model offers the best balance of accuracy and ease of implementation for orchard management tasks. We also evaluated its applicability within one software ecosystem, ESRI ArcGIS Pro, and showed how such an approach offers users a streamlined efficient way to detect objects in high-resolution UAV imagery.

Extracting the irregular and complex shapes of individual tree crowns from high-resolution imagery can play a crucial role in many applications, including precision agriculture. We evaluated three CNN models - MASK R-CNN, YOLOv3, and SAM - and compared their tree crown results with OBIA-based reference datasets from UAV imagery for seven dates across one growing season. We found that YOLOv3 performed poorly across all dates; both MASK R-CNN and SAM performed well in May, June, September, and November (Precision, Recall and F1 scores over 0.79). All models struggled in the early season imagery (e.g., March). MASK R-CNN outperformed other models in August (when there was smoke haze) and December (showing end of season red leaf senescence). SAM was the fastest model, and as it required no training, it could cover more area in less time; MASK R-CNN was very accurate and customizable. In this paper, we aimed to contribute insight into which CNN model offers the best balance of accuracy and ease of implementation for orchard management tasks. We also evaluated their applicability within one software ecosystem, ESRI ArcGIS Pro, and showed how such an approach offers users a streamlined, efficient way to detect objects in high resolution UAV imagery.

Monitoring individual tree crown characteristics is an important component of smart agriculture and is crucial for orchard management. We focused on understanding how UAV imagery taken across one growing season can help understand and predict the growth and development of pistachio trees grown from rootstock seedlings. Tree crown characteristics (i.e., height, size, shape, and mean normalized difference vegetation index (NDVI)) were derived using an object-based image analysis method with multispectral Uncrewed Aerial Vehicles (UAV) imagery flown seven times over 472 five-year-old pistachio trees in 2018. These imagery-derived metrics were compared with field-collected tree characteristics (tree height, trunk caliper, crown height, width and volume, and leaf development status) collected over two months in 2018. The UAV method captured seasonal development of tree crowns well. UAV-derived tree characteristics were better correlated with the field tree characteristics when recorded between May and November, with high overall correlations in November. The highest correlation (R² = 0.774) was found between trunk caliper and June UAV crown size. The weakest correlations between UAV and field traits were found in March and December. Spring leaf development stage was most variable, and mean NDVI values were lowest in March, when leaf development starts. Mean NDVI increased orchard-wide by May, and was consistently high through November. This study showcased the benefits of timely, detailed drone imagery for orchard managers.