Using machine learning models with historical weather, Genetics, and management to identify rice respond to future climate warming
In this study, we model and predict rice yields by integrating molecular marker variation, varietal productivity, and climate, focusing on the Southern U.S. rice growing region. This region spans the states of Arkansas, Louisiana, Texas, Mississippi, and Missouri and accounts for 85% of total U.S. rice production. By digitizing and combining four decades of county-level weather, genetic, and management data (1970 - 2015), we develop ten machine learning models for yield prediction. A two-layer meta-learner ensemble model that combines all ten methods is externally evaluated against observations conducted in the same states. Finally, the ensemble model is used with forecasted weather from the Coupled Model Intercomparison Project across the 110 rice-growing counties to predict production in the coming decades.
This study is currently being peer-reviewed in the PNAS journal.
In another project related to Rice production in the southern U.S., we tested the performance of novel genetic combinations under untested environmental scenarios and management practices can be virtually examined using process-based crop models. This study addresses the long-standing interest in the crop modeling community to expand the utility of process-based models to broader germplasm panels (e.g., breeding lines or diversity panels). We leverage a dataset from a long-term trial on advanced experimental lines and released varieties from the Louisiana rice breeding program to calibrate and evaluate the decision support for agrotechnology transfer (DSSAT) CSM-CERES-Rice model. Next, we use data collected by the same program on a large collection of breeding lines to generate numerous in silico genotypes and evaluate their performance across different management practices (different planting dates) and three climatic conditions (current climate and two future scenarios based on CMIP6-SSP5-8.5 climate projections). Our simulations indicate that shifting the current planting date (i.e., March) back by 1–2 months (to January) under moderate warming conditions (+1.3°C warmer and 41% higher CO2 level), and 2–3 months (to December) under extreme warming conditions (+4.1°C warmer and 133% higher CO2 level) could potentially offset the negative impacts of the increased future temperature.
This study has been published in the Crop Science journal (https://doi.org/10.1002/csc2.21036).
