High-resolution 3D crop evapotranspiration modeling pipeline
A data pipeline integrating multispectral reflectance, leaf temperature data, and 3D reconstruction to model crop evapotranspiration (ET) and leaf biochemical concentration at sub-leaf spatial resolution. Uses NeRF-based rendering and machine learning to create digital twins of crop canopies for precise water productivity analysis.
Overview
This solution offers a high-resolution data pipeline for 3D crop evapotranspiration (ET) modeling, enabling precise estimation of crop water use and leaf biochemical concentrations at sub-leaf spatial resolution. By integrating multispectral reflectance and leaf temperature data with advanced 3D reconstruction and machine learning techniques, the pipeline creates a digital twin of crop canopies or individual plants. This digital twin visualizes ET distribution within the target canopy, providing farmers, agronomists, and researchers with actionable insights for irrigation management, yield optimization, and crop water productivity assessment. The approach addresses the limitation of treating crop canopies as uniform surfaces by capturing the complex structure and variable radiation patterns that influence water loss and plant health.
Technical specifications
Key features:
- Multi-angle data collection capturing thermal and multispectral images at over 100 sensor viewing angles, enabling characterization of non-Lambertian canopy reflectance behavior
- 3D reconstruction using Neuron Radiance Fields (NeRF) to generate wavelength-specific scene renderings from multi-angle image sets
- Positional encoding and Instant NGP techniques for faster NeRF training and improved reconstruction accuracy
- Deformable NeRF exploration for dynamic crop scenes across growth stages
- Combined energy balance and empirical ET model adapted to 3D space, integrating canopy structure, wavelength-specific reflectance, and leaf temperature distribution
- Versatile visualization of ET distribution at arbitrary viewing angles within the canopy or individual plant
- Optional continuous raw data collection including LiDAR point clouds and sensor data at bi-weekly intervals throughout the growing season
Technology readiness level
The pipeline builds on one year of validated multi-angle field data collected during the 2020 growing season for maize and soybean canopies. The underlying ET model integrating physical and empirical methods has demonstrated promising results for plot-scale water productivity comparison. Future validation efforts will focus on implementing 3D reconstruction with NeRF, adapting the ET model to 3D space, and expanding the multi-angle dataset library across major crops throughout their growth cycles. The technology is currently at an advanced research and development stage with field-validated components ready for integration into a complete 3D ET modeling system.
About University of Nebraska, Lincoln
The University of Nebraska–Lincoln is a comprehensive public research university and the flagship campus of the University of Nebraska system, combining land-grant reach with a collaborative, industry-engaged culture. A research and technology park adjacent to campus provides modern wet and dry labs, greenhouses, offices, and conferencing, enabling companies to co-locate with faculty and access shared equipment and pilot environments. A statewide extension network links university expertise with producers and communities, creating rapid pathways for field trials, demonstrations, and workforce pipelines. Research is supported by competitive federal funding from NSF, USDA, DOE, and NIH. A dedicated technology transfer office manages IP, licensing, agreements, and startup formation with industry-friendly terms.
delivered to your inbox.