Supporting Data and Code for "Managing to Climatology: Improving semi-arid agricultural risk management using crop models and a dense meteorological network"

Without reliable seasonal climate forecasts, farmers and managers in other weather-sensitive sectors might adopt practices that are optimal for recent climate conditions. To demonstrate this principle, crop simulation models driven by a dense meteorological network were used to identify climate-optimal planting dates for U.S. Southern High Plains (SHP) un-irrigated agriculture. This method converted large samples of SHP growing season weather outcomes into climate-representative cotton and sorghum yield distributions over a range of planting dates. Best planting dates were defined as those that maximized median cotton lint (April 24) and sorghum grain (July 1) yields. Those optimal yield distributions were then converted into corresponding profit distributions reflecting 2005-2019 commodity prices and fixed production costs. Both crop’s profitability under variable price conditions and current SHP climate conditions were then compared based on median profits and loss probability, and through stochastic dominance analyses that assumed a slightly risk-averse producer.

Methods
The main data sets used here are simulated rainfed cotton and sorghum yields generated via the DSSAT CROPGRO-Cotton and CERES-Sorghum crop models. The model's weather inputs were provided by 21 Texas Tech University mesonet stations over an 11-year period.  This weather data can't be re-distributed per an agreement with Texas Tech University.

Given the 231 station-years of weather input data, the models are used to generate similarly dense yield distributions. Model simulations for both crops are repeated over 32 planting dates to find those that maximize median lint and grain yields. After yield scaling to adjust the aggregate median of simulated yields over all planting dates to agree with median reported Southern High Plains (SHP)  rainfed cotton and sorghum yields, the resulting yield distributions are converted to profit distributions. These distributions are formed based on fixed production costs, but variable lint and grain commodity values representative of market conditions since 2005. The resulting simulation chain thus transforms dense samples of growing season weather variability into similarly dense distributions of yield and profit outcomes that are consistent with current SHP summer growing conditions and recent market conditions. The yield and profit distributions produced by this chain can be used to determine optimal planting dates of both crops, estimate the profit and risk effects of management, andcompare the profitability of rainfed cotton and sorghum over a range of commodity prices.