Orchard Canopy Stress Monitoring with In-Field and Remote Sensing Technologies
Nutrient uptake is closely associated with the water status in the soil and plant. Water makes it easier for plants to take up nutrients, and a structured irrigation strategy will assist with sufficient nutrient uptake. Under-irrigating causes inadequate water uptake due to a lack of availability, which also affects the nutrient uptake of the plant. While more commonly, over-irrigation leaches nutrients from the soil, thus affecting plant nutrition.
The primary goal of the study is to monitor orchard canopy stress status with both in-field and UAV-based sensing technologies, thus providing an optimal decision-making system for nutrient and water management. In 2022, we focused on a field test on evaluating irrigation effect on nutrient uptake and production and a UAV (drone) based tree canopy measurement to indicate the tree health condition.
Assess canopy-level water stress and nutrient uptake
In the 2022 season, we set up an orchard block for irrigation and nutrient uptake experiment. Three irrigation methods were applied with 50%, 100%, and 150% of calculated evapotranspiration. Each method was used for two rows of trees in this 0.9-acre block. We can see the difference in water usage among these methods. While due to the freezing temperature in the early season, only very few fruits were left on the trees. While it is difficult to evaluate the irrigation and nutrient uptake without production comparison among these methods, we will continue this experiment in the 2023 season. Meanwhile, we have purchased a chlorophyll meter, which will be used to measure the canopy-level water/nutrient stress in the new test.
Using UAV-based remote sensing technologies for canopy health monitoring
Field experiments were conducted in a GoldRush apple cultivar orchard block at the Penn State Fruit Research and Extension Center (FREC) in Biglerville, PA, USA. The total area of the experimental field was approximately 0.9 acres, consisting of six rows with 36 trees in each row. The trees were planted in 2009 and trained in a tall spindle architecture with a tree-to-tree spacing of 1.2 m and a row-to-row spacing of 6.1 m. Images were acquired using a rotary-wing unmanned aerial vehicle (UAV) DJI Matrice 200 (DJI Technology Inc., Shenzhen, China) with a high-resolution RGB Zenmuse X5S camera (DJI Technology Inc., Shenzhen, China) on June 21, 2021 (Fig. 1).
To measure tree specifications, a total of 12 white paper boards were placed on reference trees. The board dimension was 0.51 m × 0.76 m (width × height). Trees at the left and right rows from the board were considered as references for canopy characteristics measurements. A total of 24 trees were randomly chosen for the manual and UAV-based tree height and canopy volume measurements. The canopy coverage was measured for all trees on the site to generate the canopy coverage map. A total of 12 ground control points (GCPs) were marked based on the locations of the 12 boards. An Inertial Navigation System-Global Navigation Satellite System (INS-GNSS) (INS-D, Inertial Labs, Paeonian Springs, VA, USA) with 1 cm position accuracy was used to collect the geographical location of the GCPs.
An orthomosaic map was generated with the acquired UAV images (Figure 2). A digital surface map (DSM) and digital terrain map (DTM) were then generated with the software by considering classified dense point clouds after correcting the location bias. For the DSM, all dense point clouds, including ground, trees, and other objects, were used, and only the ground point clouds were used for DTM generation. The accuracy of georeferencing was about 0.03 m (3 cm) in DSM and DTM generation.
Canopy height measurement
A subtraction between DSM and DTM generated a height map of apple trees. The DSM of the orchard included elevation information for the apple trees and ground relative to the mean sea level. The DTM included elevation information for only ground relative to the mean sea level. Therefore, the subtraction of these two models provided only heights for the apple trees.
The average tree canopy height was 3.07 m in manual measurement, whereas the average height was 3.05m in UAV-based measurement. The measurement errors ranged from 0 to 0.73 m with an average of 0.20m, equivalent to a 6.64% error relative to the manual measurement. These results indicated the potential of UAV-based apple tree canopy height measurement to quantify individual tree height with less than 10% error.
Canopy volume measurement
The total intensity height from all pixels within the ROI was calculated to measure the individual tree canopy volume. The canopy volume of the targeted trees was computed by multiplying the total height of the intensity of all pixels with the area of ground sample distance (GSD).
The UAV-based tree canopy volumes were compared with the ground-based LiDAR measurement of 24 targeted trees. A high correlation of R2 = 0.80 was obtained between the UAV-based measurements and LiDAR-based canopy measurements. While the results showed certain errors in the UAV-based measurement when compared with the LiDAR measurements, with the mean absolute error (MAE) of 0.3 m3.
Canopy coverage measurement
The percentage of canopy coverage for all trees located in the experimental orchard was calculated (Figure 4), which ranged from 0 to 100%. The highest percentage indicates the ROI included more canopies, whereas the lowest percentage indicates the ROI contained fewer or no canopies.
Summary
UAV-based imaging methods were established in this study to measure apple tree canopy characteristics by analyzing high-resolution aerial RGB imagery. This UAV-based method provided a relatively fast approach to calculating major tree characteristics in order to estimate the general health condition of the trees. Results showed that the study successfully measured tree canopy characteristics such as canopy height, canopy volume, and canopy coverage and were subsequently validated with ground measurements.
