Crop Yield Estimation Using Remote Sensing and LAI Model Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing crop yield prediction methods rely on local calibration using ground data, limiting their applicability to large areas and requiring extensive in-situ measurements, which are costly and impractical.
Innovation Solution
A method that combines remotely sensed data with crop growth models like APSIM, using LAI as a linking parameter, to estimate crop yield without ground data calibration, by fusing imagery data to detect sowing dates and generate LAI simulations that match remotely sensed LAI, thereby predicting yield.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional yield prediction methods use local calibration with ground data, then measurement precision is improved, but device complexity and ease of operation deteriorate due to requiring extensive in-situ measurements
Solution Approach 1:
The patent uses crop growth models (APSIM, DSSAT, WOFOST, AquaCrop) as intermediaries between remote sensing data and yield prediction. These models simulate crop development processes and are calibrated once at local level, then applied regionally without requiring extensive ground data collection for each area, thus reducing operational complexity while maintaining precision
Solution Approach 2:
The patent creates universally applicable crop growth models that can be used across multiple regions and crops. The models are designed to work with standard remote sensing data (LAI, vegetation indices) and climate data, making them transferable from local calibration to regional prediction without requiring region-specific ground data collection
2Measurement precision
If locally calibrated prediction methods are used, then measurement precision is improved, but adaptability deteriorates as methods are limited to calibration areas
Solution Approach 1:
The patent segments the prediction process into two distinct phases: (1) local calibration phase where models are trained with ground data from representative locations, and (2) regional prediction phase where the calibrated models are applied across large areas using only remote sensing and climate data. This segmentation allows models to learn local characteristics once and then generalize to broader regions
Solution Approach 2:
The patent uses crop growth models that can dynamically adjust parameters based on remote sensing observations (LAI, vegetation indices) and climate data. The models simulate crop development by changing physiological parameters over time, allowing them to adapt to different environmental conditions across regions while maintaining the structural framework calibrated locally
3Measurement precision
If extensive in-situ measurements are collected for calibration, then measurement precision is improved, but loss of time and loss of substance increase due to costly data collection
Solution Approach 1:
The patent performs preliminary calibration of crop growth models using ground data from representative locations before deploying them for regional prediction. This preliminary action captures the essential local characteristics once, and the calibrated models can then be applied repeatedly to multiple regions and time periods without requiring repeated ground data collection
Solution Approach 2:
The patent creates simplified representations (crop growth models) that copy the essential behavior of crop systems based on limited ground data. Once calibrated, these model copies can simulate crop development across regions without requiring physical measurement campaigns, replacing expensive and time-consuming in-situ data collection with computational simulations
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
Enables accurate crop yield estimation at pixel and regional scales, providing yield maps months before harvest, with high spatial resolution and reduced reliance on ground data, improving food security forecasting.
Implementation Method 1
crop growth models can be used to simulate key physiological processes including phenology, organ development (of leaf and grain), water and nutrient uptake, biomass
Implementation Method 2
Optical remote sensing cannot see through the crop canopy or the soil surface but can, for example, provide valid information about canopy chlorophyll content
Data Source
AI summary
A method for estimating crop yield of an analyzed area, which is a region of interest, according to which imagery data is acquired from one or more remotely sensed sources, using a remote sensing platform or one or more satellites. A remotely sensed LAI of the analyzed area is generated by fusing the acquired imagery data and the sowing date of the analyzed field is detected by processing the imagery data. The detected sowing date and a sowing date window are fed into a crop simulator (e.g., APSIM) and a set of predetermined parameters representing the state of the analyzed area is then generated, and fed into the crop simulator. The crop simulator generates a plurality of LAI simulations, each corresponding to a different combination of parameters. Only LAI simulations in which the simulated LAI best matches the remotely sensed LAI are identified and selected, while omitting all other LAI simulations. Finally, the yield prediction that corresponds to the simulation(s) with the best match, are selected.