High-Resolution Yield Models for Spatial Variability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for yield prediction and modeling in agriculture are not sufficiently accurate or scalable, particularly due to the failure of National Commodity Crop Productivity Index (NCCPI) values to account for spatial variability in crop yield patterns driven by complex interactions of soil, topography, fertility, water, and pest pressures, leading to inaccurate assessments and inefficiencies in High-Resolution Land Management (HRLM) techniques.
Innovation Solution
The development of high-resolution yield models using remote sensing data and machine learning techniques to predict crop yields at subfield locations, leveraging electro-magnetic reflectance data and constructing training vectors from yield measurements and remote sensing data to estimate yield quantities, enabling more widespread application of HRLM techniques like Variable-Rate Technologies and Integrated Landscape Management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If NCCPI values are used for yield prediction, then the method is simple and widely available, but the accuracy is insufficient due to failure to account for spatial variability
Solution Approach 1:
The patent segments the field into multiple subfield locations and creates high-resolution yield models for each segment rather than using a single uniform NCCPI value for the entire field. This segmentation allows capture of spatial variability in soil, topography, fertility, water, and pest pressures across different locations within the field, thereby improving yield prediction accuracy while maintaining computational feasibility through localized modeling.
Solution Approach 2:
The patent applies local quality by developing location-specific yield models that incorporate local environmental conditions (soil properties, topography, fertility, water availability, pest pressures) for each subfield location. Instead of using a uniform approach, the model adapts to local conditions at each spatial location, improving prediction accuracy by accounting for the heterogeneous nature of agricultural fields.
2Measurement precision
If machine yield data is collected for training, then model accuracy improves, but scalability is limited due to requirement of direct producer interface and complex data transfers
Solution Approach 1:
The patent introduces remote sensing data as an intermediary that bridges the gap between limited machine yield data and broader spatial coverage. Remote sensing provides auxiliary information about environmental conditions (soil moisture, vegetation health, topography) that can be integrated with machine yield data to train models for locations where direct yield measurements are unavailable, thereby improving scalability without sacrificing accuracy.
Solution Approach 2:
The patent creates a multi-functional modeling system that can operate in multiple modes: (1) using machine yield data where available, (2) using remote sensing data where machine data is unavailable, and (3) combining both data sources. This universal approach allows the system to scale across different fields and producers without requiring direct machine yield data from every location, while maintaining model accuracy through flexible data integration.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach allows for the creation of accurate, high-resolution yield models that can predict yield variability even without machine yield data, enhancing the scalability and accuracy of HRLM techniques, thereby improving crop management and reducing costs and environmental impacts.
Implementation Method 1
utilizing remote sensing data... electro-magnetic reflectance data
Data Source
AI summary
Disclosed here are systems, methods, apparatus, and/or non-transitory computer-readable storage comprising machine-readable code for the development and application of high-resolution crop yield models. The disclosed yield models may be captured yield data and corresponding remote sensing data covering relatively limited areas. Embodiments of the disclosed yield models may be capable of estimating spatial yield characteristics in areas for which accurate yield data are not available (and/or not practical to acquire), thereby enabling more widespread application of integrated land management techniques.