U.S. Rice Yield Predictions with Genetic and Climate Data for Southern Counties

Southern U.S. rice-growing counties, including Arkansas, Louisiana, Texas, Mississippi, and Missouri, provide the geographic scope for this dataset. It contains tabulated weather, genetic, and county-level rice yield information from 1970 to 2015, used to develop machine learning models for yield prediction. The study, supported by USDA NIFA grants and authored by Diane Wang of Purdue University, employs these models with forecasted climate data to predict future production across 110 counties.

Use Cases

• Predicting future rice yields based on forecasted climate data from the Coupled Model Intercomparison Project.
• Modeling the impact of molecular marker variation and allele frequencies on county-level rice productivity.
• Evaluating the performance of ensemble machine learning models using historical Uniform Regional Rice Nursery trial data (1980-2018).
• Analyzing the relationship between county-level weather patterns and rice yields in the Southern U.S. region.

Strengths

• Data covers a 45-year time span (1970-2015) for historical analysis.
• Focuses on a region accounting for 85% of total U.S. rice production, providing significant coverage.
• Includes three distinct predictive models (PCA and two meta-machine learning models) and associated Python scripts for reproducibility.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Purdue University West Lafayette (author Diane Wang), via paperswithcode.

Collection Method

Data digitized and merged from county-level variety acreage and genotyping-by-sequencing data, combined with weather and yield records.

Time Range

1970-2015 for core data, with external evaluation trials from 1980-2018.

Geography

Southern U.S. rice-growing region (Arkansas, Louisiana, Texas, Mississippi, Missouri).