Deep Learning Cropland Layer Prediction for In-Season Crop Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for generating crop-specific land cover maps, such as the USDA Cropland Data Layer (CDL), are not timely, with data for the current year not being released until the following year, limiting applications that require real-time or near-real-time data, especially for monitoring and verifying agricultural management practices.
Innovation Solution
A deep learning convolutional neural network is trained using historical CDL data and satellite imagery to predict cropland data layers and agricultural management zones, enabling early prediction of crop types, planting and harvesting dates, and regenerative practices based on satellite imagery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional methods are used to generate crop statistics, then data accuracy is maintained through ground truth verification, but data timeliness deteriorates as results are only available in the following quarter
Solution Approach 1:
The system performs preliminary analysis of satellite imagery throughout the growing season to generate early predictions of crop statistics before the traditional ground truth verification process is complete. This allows data to be available in advance while maintaining accuracy through the subsequent verification process.
Solution Approach 2:
A deep learning neural network serves as an intermediary between satellite imagery and final crop statistics, generating preliminary predictions that bridge the gap between early data availability and verified accuracy. The network processes imagery to produce intermediate results that can be refined later.
2Productivity
If deep learning neural networks are employed to predict crop data, then data timeliness is improved with predictions available during the growing season, but system complexity increases
Solution Approach 1:
The deep learning neural network is designed to perform multiple functions: classifying crop types, estimating yield, and detecting management practices within a single unified system. This multi-functionality reduces overall system complexity compared to using separate specialized systems for each task.
Solution Approach 2:
The system uses satellite imagery that is already available and processed by existing infrastructure, eliminating the need for additional data collection equipment. The neural network processes this existing data to generate predictions, reducing hardware complexity while maintaining high productivity.
3Ease of operation
If satellite imagery alone is used for prediction, then operational ease is improved by eliminating ground truth collection, but measurement precision deteriorates compared to methods using extensive ground verification
Solution Approach 1:
The system performs preliminary predictions using satellite imagery during the growing season, then applies ground truth verification afterward to refine and validate results. This two-stage process maintains operational simplicity while improving measurement precision through sequential processing.
Solution Approach 2:
The system incorporates feedback loops where ground truth data, when available, is used to verify and refine the satellite imagery-based predictions. This feedback mechanism improves measurement precision while maintaining the operational ease of using remote sensing as the primary data source.
Data Source
AI summary
A computer-implemented method for predicting a cropland data layer (CDL) for a current year includes: retrieving a first set of records from a historical CDL database, where the first set corresponds to sampled areas of a region taken over a period for a number of years; retrieving a second set of records from a historical imagery database, where the second set corresponds to the sampled areas of the region, the period, and the number of years; employing the second set as inputs to train a deep learning network to generate the first set; retrieving a third set of records from a current imagery database, where the third set corresponds to a prescribed region, and where the third set corresponds to the time period and the current year; and using the third set as inputs and executing the trained deep learning network to generate a predicted CDL for the current year.