Remote Sensing of Almond and Walnut Tree Canopy Temperatures Using an Inexpensive Infrared Sensor on a Small Unmanned Aerial Vehicle

Improving water use efficiency in agriculture will become increasingly important in the face of decreasing water resources and a growing population. Increasing water use efficiency, or water productivity, has been shown to greatly reduce irrigation water usage in many orchard crops with little to no impact on yield. In some specialty crops, improving water productivity can even lead to a higher value crop. Current irrigation practices depend largely on uniform applications of water over large fields with varying degrees of heterogeneity. As a result, much of the field receives more water than it needs. A system to monitor the needs of each plant or smaller groups of plants within the field would be helpful in distributing irrigation water according to each plant or group of plants' needs. Such a system would help conserve water resources. Stomatal conductance is a good indicator of plant water-based stress, as it is the main response a plant has to limit transpiration-related water losses. The difference between leaf temperature and air temperature, when adjusted for environmental conditions, can give a good indication of stomatal conductance. Recent efforts at UC Davis have employed a handheld sensor suite to measure leaf temperature and other environmental variables like wind speed, air temperature, and humidity in almond and walnut trees. Though effective, this method requires walking or driving through the orchard and measuring several leaves on a given tree, so it is impractical for large-scale monitoring. Satellite and aircraft can measure canopy temperatures remotely, but these applications typically do not have the spatial resolution for precise monitoring or the temporal resolution necessary for irrigation decisions, and they are too expensive and impractical for smaller-scale farms. A smaller unmanned aerial vehicle (UAV) could employ the same methods as satellite and larger aircraft-based systems, but relatively inexpensively and at a scale catered to the needs of a given field for more precise monitoring.;The goal of this study was to explore the feasibility of using an inexpensive temperature sensor (Melexis MLX90614; NV Melexis SA, Rozendaalstraat 12, 8900 Ieper, Belgium) on a small UAV (Mikrokopter OktoXL; Hisystems GmbH Flachsmeerstrasse 2, 26802 Moormerland, Germany) to sense the canopy temperatures of almond and walnut trees. To accomplish this goal, we installed an infrared temperature sensor and a digital camera on a small UAV. The camera provided a spatial awareness of the IR temperature measurements which would otherwise require a very expensive thermal imager to obtain. The UAV was flown above almond and walnut trees recording images and temperatures, which were aligned temporally in post-processing. The pixels of each image were classified in to four classes: sunlit leaves, shaded leaves, sunlit soil, and shaded soil. Assuming that the measured temperature could be described as a weighted sum of each class in the field of view of the IR sensor, a linear system of equations was established to estimate the temperature of each class using at least several measurements of the same tree. Results indicated a good correlation between the temperatures estimated from the linear system of equations and the temperatures of those classes sampled on the ground immediately following each flight. With leaf temperatures ranging from about 12 to 40 degrees Celsius between 23 flights over two years, the linear solver was able to estimate the temperature of the sunlit and shaded leaves to within several degrees Celsius of the sampled temperature in most cases, with a coefficient of determination (r2 value) of 0.96 during the first year, and 0.73 during the second year.;An additional study was undertaken to detect spatial temperature distribution within the orchard. Ground measurements were taken of every other tree in two walnut rows and one almond row using the handheld sensor, and the UAV was flown over those rows immediately following each ground sampling. An interpolated temperature map of the UAV's temperature measurements indicated a very similar temperature distribution as that measured with the handheld sensor, but the UAV was much faster and, in parts of the rows, it provided a higher spatial resolution than the handheld sensor.