Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Improve Yield Prediction with Uas-Based Leaf Area Index Estimation and a Hybrid Machine Learning- and Process-Based Model
Crop yield prediction using process-based crop models lacks accurate datasets with enough samples to calibrate. Previous studies used leaf area index (LAI) estimation to calibrate crop model. The LAI was estimated based on satellite remote sensing images with coarse spatial resolution and statistical or machine learning models. Accurate LAI measurements can be done using manually measured method, but it is time consuming and labor intensive to get enough samples for crop model calibration. The challenges are (1) LAI estimation models may not work for fields with different genotypes or environmental conditions, (2) crop model calibration cannot accurately predict crop yield with limited observed datasets. The overall goal of this study was to integrate data-driven machine learning modelling and a process-based crop model with LAI estimations derived from UAS images to improve crop yield prediction. The first objective was to improve LAI estimation performance in a quick and accurate way using multiple plant traits derived from ultra-high spatial resolution unmanned aircraft systems (UAS) images. Those traits were vegetation indices (VIs) and the plant structural traits (e.g., canopy cover, plant height). The second objective was to improve crop yield prediction using machine learning modelling and the calibrated crop model by assimilating the improved LAI estimation. Results showed that LAI estimation performance was significantly improved for fields with multiple genotypes by using several UAS-derived plant traits simultaneously in modelling compared by considering only individual plant trait. Such performance improvement was marginal for the fields involving one or two crop genotypes. Integrating a machine learning modelling and the calibrated crop model using LAI estimation based on UAS significantly improved the winter wheat crop yield prediction performance. The relative error was decreased from 40.4% using APSIM calibration method to -0.4% using the hybrid method for SY Wolf cultivar and decreased from 31.3% to 21.7% for Ruth cultivar. Using the integrated model potentially will save time and effort of improving breeding processes and management.
Agricultural engineering|Remote sensing|Computer science
Wang, Lin, "Improve Yield Prediction with Uas-Based Leaf Area Index Estimation and a Hybrid Machine Learning- and Process-Based Model" (2022). ETD collection for University of Nebraska - Lincoln. AAI29322341.