Hay yield estimation using ultrasonic and LiDAR sensors

Authors: LEE, KYUHO - University Of Missouri; Sudduth, Kenneth; ZHOU, JIANFENG - University Of Missouri; KITCHEN, NEWELL - Retired ARS Employee

Submitted to: Journal of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/6/2025
Publication Date: 3/1/2026
Citation: Lee, K., Sudduth, K.A., Zhou, J., Kitchen, N.R. 2026. Hay yield estimation using ultrasonic and LiDAR sensors. Journal of the ASABE. 69(1):37-52. https://doi.org/10.13031/ja.16023.
DOI: https://doi.org/10.13031/ja.16023

Interpretive Summary: Yield monitoring systems are widely used commercially in grain crops to map yields at a scale of a few meters. However, such high-resolution yield monitoring and mapping for hay and forage crops is not yet available to farmers, as commercial hay yield monitoring systems only obtain the weight of individual bales. This makes it difficult to map and understand the spatial variability in hay yield, as needed by farmers interested in applying precision agriculture principles to hay production. This study investigated the feasibility of using two ground-based sensor types, ultrasonic and Light Detection And Ranging (LiDAR) for estimation and mapping of hay yield by regression and machine learning models. Data were obtained from sensors mounted to the mowing tractor and related to 110 manual measurements of hay yield from throughout the test field. Decision tree machine learning models provided the most accurate yield estimates, but represented only slightly over 50% of the hay yield variability across the field. However, the results of this research will provide information to improve data collection and processing methods for application in future studies.

Technical Abstract: Yield monitoring systems are widely used in grain crops, but high-resolution systems for hay crops are not yet commonly adopted. In this study, we investigated methods to estimate hay yield using proximal ultrasonic and light detection and ranging (LiDAR) sensors to measure plant height above ground level. Data were collected during hay harvest in June 2021 from mature grass and red clover hay in a 35-ha field using sensors mounted in a nadir view on a boom extending in front of the cutting width of a hay mower. One hundred and ten biomass sampling areas were located across the field using 1-m^2 quadrats and plant height of hay was measured manually inside each quadrat at multiple points. Hay in each sampling area was harvested manually to determine wet and dry biomass. Simple regression and machine learning (ML) models were built to estimate biomass using various descriptive statistics of sensor-based plant height data from each sampling area, e.g., median and quantile 90%. Before running regression models, sub-datasets were established based on four conditions, including 1) sensor type selection (ultrasonic or LiDAR), 2) data selection (all, north, or west area), 3) sensor data range selection (1 m, 2 m, or 3 m), and 4) processed type selection (raw, statistical and combination data). Statistical analysis was conducted to compare the performance of different modeling methods and identify optimal conditions. Raw ultrasonic data obtained over a 3 m range centered on the quadrat and ML analysis produced the most accurate model. For dry biomass estimation, the R2 ranged from 0.31 to 0.52. The highest R2 values for each sensor were 0.56 for LiDAR plant height estimation of wet mass and 0.52 for ultrasonic plant height estimation of dry mass. Prediction results might be improved with a lower driving speed and/or higher data collection frequency.