High-throughput phenotyping and field-based biomass estimation for winter wheat, sugar beet and potatoes using UAV LiDAR

By Jelle ten Harkel

Abstract
Phenotyping of crops is important when estimating biomass and potential yield of new varieties of agricultural crops. Due to an increasing pressure on food production, accurate estimation of biomass during the growing season is important to optimise the yield. Using a UAV lidar system was first tested to estimate plant height for three different crops: potato (R²=0.50, RMSE=0.12), sugar beet (R²=0.70, RMSE=0.07) and winter wheat (R²=0.78, RMSE=0.03). Showing an accuracy for winter wheat comparable to other research. Using the 3DPI algorithm biomass was estimated for three different crops, namely potato (R²=0.24, RMSE=22.09%), sugar beet (R²=0.68, RMSE=17.47%) and winter wheat (R²=0.82, RMSE=13.94%). The algorithm proved to work well for sugar beet and winter wheat but underperformed for potato. Reasons for this underperformance were a dense canopy and the lack of a good Digital Terrain model. Next to high-throughput phenotyping four, different flight specifications were tested, such as flying height, flight speed and flight pattern. For these flights, the plant height and biomass estimations were compared to the flight used to construct the prediction model. For plant height no difference were found in spatial patterns within the field and the plant height is estimated well. Indicating that data could be acquired at a high altitude and flying speed. Accurately predicting biomass on the other hand requires a dense flying pattern at lower altitude. Here a homogeneous point density is needed, with numerous underneath canopy returns. This could be achieved by a cross-flight pattern. Although there are some limitations, using a UAV lidar system proved to be successful in predicting biomass for sugar beet and winter wheat.