Highlights Ultrasonic and LiDAR sensors were evaluated for estimating within-field hay yield variability. The 20 Hz ultrasonic sensor was more accurate than the 10 Hz LiDAR sensor. Machine learning yield estimation was more accurate than simple regression. Decision tree models generally provided the most accurate results. ABSTRACT. Yield monitoring systems are widely used in grain crops, but high-resolution systems for hay crops are not yet commonly adopted. This study evaluated the feasibility of using low-cost, tractor-mounted ultrasonic and fixed-beam LiDAR sensors for estimating hay yield and biomass in production field conditions. Ultrasonic and LiDAR sensors were used to collect plant height data for a mixed hay crop of red clover (Trifolium pratense L.) and timothy grass (Phleum pratense L.) within a 35-ha field as part of the harvesting operation in June 2021. Along the transects covered by the sensors, 110 sampling points were selected for data collection within a 1-m 2 quadrat. Data included manually measured plant height and dry biomass. Both sensors successfully estimated plant height using linear regression (R 2 = 0.75, normalized RMSE (RMSEn) 11%). Regression and machine learning (ML) models (i.e., decision tree, decision tree ensemble, Gaussian process regression, neural network, and support vector machine) were built using sensor-based plant height data and various descriptive statistics of that data to estimate dry hay mass. Data subsets were analyzed to compare sensor type, sensor data range, and data processing methods. The most accurate results were obtained with ultrasonic data collected over a 3-m range. The best analysis method was an automatically optimized decision tree ensemble model (R 2 = 0.52, RMSEn = 15% in the test dataset). These relatively low values reflect the variability of field conditions and limitations due to the low sampling rate provided by these low-cost sensors. Future research should consider optimized data collection protocols, including adjustments to driving speed, increased sampling rates, and refined sensor configurations, to improve accuracy. Keywords: Hay yield estimation, LiDAR, Precision agriculture, Proximal sensing, Ultrasonic sensor.
Lee et al. (Thu,) studied this question.