USING LIDAR TO MEASURE ALFALFA CANOPY HEIGHT

Alfalfa is a popular crop that is grown worldwide because it is a nutritious feed for livestock and fixes nitrogen in the soil. Profitable alfalfa production greatly relies on monitoring the status of the alfalfa crop. Traditionally, producers have used crop assessment techniques that rely on manual measurements of alfalfa plant height, which can be used to predict nutritive quality and yield. These manual measurements are often labor-intensive and provide low-resolution data that is not acceptable for field-scale monitoring. The goal of this study was to assess the capability of a simple LiDAR setup to accurately estimate the average canopy height of an alfalfa crop. To achieve this goal, we first developed predictive models of alfalfa canopy height using LiDAR-derived measurements as predictor variables. Second, we assessed the accuracies of the models and compared the properties of each model. Third, we determined the optimal LiDAR-derived measurements to use to accurately predict average alfalfa canopy height. The data used in our models were collected in two separate fields planted with two different cultivars of alfalfa. Data collection was performed on five dates spanning one entire growth cycle during the summer of 2019. A simple single-beam LiDAR sensor was used to scan the canopy of sample plots within the fields. Manual measurements of plant height and field observations of insect, disease, and weed pressure were also recorded. Of the data used in the predictive models, the 95th percentile of LiDAR-measured height was found to be the optimal predictor for estimating alfalfa canopy height. Using the 95th percentile as a single predictor in a linear regression model of measured average canopy height resulted in an R-2 of 0.90 and RMSE of 4.5 cm. Two other linear regression models using a combination of LiDAR measurements and LiDAR measurements with alfalfa health observations, respectfully, were developed for comparison. These models exhibited marginally better accuracies but required more inputs than the model only using the 95th percentile. This study shows how simple LiDAR configurations can be used for timely collection of accurate alfalfa canopy height data. From our findings, we suggest using the 95th percentile of LiDAR-derived canopy height to estimate alfalfa canopy height. This study lays the groundwork for research into more advanced LiDAR configurations for alfalfa applications, such as LiDAR-equipped UAVs.