University of Wisconsin-Extension logo

The United States is a major soybean-producing country that supplies 34 percent of global annual soybean production. Most U.S. soybean-producing regions are rain-fed and thus are highly vulnerable to extreme weather events.

Drought and elevated air temperatures, now more increasingly frequent due to climate change, are important constraints in crop production across major agricultural areas globally. Thus the challenge to increase crop yields to meet future demand can be achieved by increasing the rate at which climate-change-adaptation practices are identified and adopted.

Vapor-pressure deficit is a measure of atmospheric water demand with a strong influence on plant transpiratory water loss. Increasing vapor-pressure deficit values are generally associated with drought and heat. Improved genetic traits and crop-management strategies could help mitigate the projected negative impacts of climate change on crop yields.

Drought-tolerant traits introduced through conventional breeding resulted in soybean-transpiration rates that plateaued at vapor-pressure-deficit levels greater than 1.40 to 2.10. Crop-management strategies such as earlier-than-typical planting have also been proposed in regional studies as a strategy to increase yields. But soybean exhibits different sensitivities to weather during varied developmental stages. Therefore the sensitivity of a crop to climate-adaptation strategies and their effectiveness in mitigating drought-induced yield reduction remains unclear.

An important step toward adapting to climate change and mitigating its impact on yield is accurate identification of weather conditions that most affect crop yield. One option is planting-date adjustment. Regional trials have shown the benefits of earlier planting – but there’s a limit to how much the regional field trials can extrapolate results. Our objective was to examine crop sensitivity to varying in-season weather conditions. We wanted to model optimal planting dates and associated yield and monetary benefits due to planting-date adjustment across the United States. To-date there’s no similar previous work.

Materials and Methods

We used data from soybean seed-yield cultivar trials performed by agricultural-university personnel during 2007–2016 in 27 states. Those multiple site trials were conducted each year in representative soybean-production areas. Within each state the trial planting-date data bracketed the 50 percent planting-date progress reported by U.S. Department of Agriculture’s National Agricultural Statistics Service for each state and year. Those 27 states accounted for about 99 percent of total U.S. soybean cultivated area – 2007–2016 average. State-wide average yield and weather conditions were calculated, resulting in 186 state year soybean yield and weather-condition data.

To identify weather variables during the growing season that had the strongest impact on soybean seed yield in the 10-year 27-state data set, we used conditional inference regression tree analyses.

Our chosen candidate predictors were

  • the state in which a trial was performed
  • cumulative precipitation and solar radiation
  • average vapor-pressure deficit
  • maximum temperatures, and
  • relative humidity.

We divided the growing season into six successive 30-day time windows, from three days before planting to 150 days after planting. We calculated predictor values for each window, thereby resulting in 31 total candidate weather predictors.

We used several methods in predicting the yet-to-be observed soybean yields.

  • Machine-learning regression analysis utilized coordinates and the aforementioned 30-day-specific and season-wide variables to capture differential weather sensitivities at different development stages and season-wide weather variables.
  • For all locations and years in the study, seven different weather data sets were created by changing planting date from minus-30 to plus-30 days from typical, in 10-day increments.
  • The machine learning model was applied in all weather-data sets to simulate soybean yield in each location from 2007 through 2016, for a total of 1,890 simulations.
  • The simulated yields were fitted in a multilevel model to quantify the effect of variable planting date on soybean yield.
  • The monetary effect of optimum planting was calculated by considering the percentage of yield change due to optimum planting within each state, and the total state-wide non-irrigated soybean-production change in each year.
  • The estimated state-year-specific total income was adjusted for inflation to 2016 US dollar values.
  • Frost probabilities were calculated using binomial distribution of event occurrence – spring frost vs. no frost – for different daily minimum-temperature thresholds in all locations of the study. The previous 46 years of weather data for each location were used to calculate probabilities.

Results and Discussion

The conditional inference tree analyses revealed that vapor-pressure deficit during 61 to 90 days after planting was the most important predictor of soybean yield, which was consistent with a finding in a previous study that focused on just three Midwestern states.

The worst trial yields were observed in state-years in which vapor-pressure deficit was greater than 2.44 kilopascals from 61 to 90 days after planting, and vapor-pressure deficit from minus-30 to zero days before planting was greater than 1.79 kilopascals.

The best-yielding trials were those in which vapor-pressure deficit was less than 2.44 kilopascals from 61 to 90 days after planting, in 13 states, and with precipitation greater than 3 inches from 61 to 90 days after planting. Those results show that the state and amount of precipitation from 61 to 90 days after planting are important yield-limiting factors – mainly in non-drought conditions.

The sensitivity of soybean yield to variable in-season weather conditions was examined by creating weather-data sets that differed from the typical state-specific planting dates in 10-day increments, for all states and years in the study.

A machine-learning model calibrated to predict state-year-specific trial soybean yield across the United States – based on coordinates and weather variables – was applied to estimate yields for each hypothetical planting date in every state from 2007 through 2016. A clear trend of increased yields due to earlier planting was observed within most states across the 10 years of the study. Excluding Texas and Mississippi, planting 12 days earlier than what was practiced during the decade of 2007 to 2016 across the United States would have resulted in a 10 percent greater total yield.

Midwest soybean producers in states such as Iowa and Ohio could have theoretically experienced a small yield increase – 0.4 and 1.1 bushels per acre, respectively – during the past decade by eight to 10 days earlier planting, respectively. That result is in agreement with recent regional estimates of early planting-date effect on farmer fields. In other states with large cultivated areas such as Nebraska, Illinois and Wisconsin, producers appear to be already using near-optimum planting dates. It’s been reported that earlier planting dates resulted in a longer planting-to-first-trifoliolate growth-stage period – V1 –but also advances V1 occurrence on a calendar-date basis. That leads to earlier node accrual and floral induction, which can optimize the final number of main stem nodes and result in greater yield potential.

Using state-year-specific total income data and the previously calculated yield change due to planting-date adjustment, a 10-year cumulative monetary effect was estimated for each state. A substantial monetary gain from earlier planting was estimated in most soybean-producing states. Minnesota, North Carolina and Kansas would have experienced the greatest monetary gains that could have reached $0.9 billion to 1.5 billion. The gains would have been less in Southern and Southeastern states, despite the greatest yield change due to planting-date adjustment, mainly due to smaller cultivated area.

Planting-date adjustment overall across the United States from 2007 through 2016 would have resulted in a cumulative gain of $9 billion. We note that earlier planting may be associated with an additional cost for farmers to update or add additional planting equipment. But because such costs can be amortized out through time, we consider our estimates as an upper bound of hypothetical monetary benefits.

An important consideration in early planting is spring-frost occurrence, which can damage or destroy the crop – but only after emergence at 15 to 25 days after planting. The current common recommendation to soybean producers is to plant the first field when frost probabilities are less than 20 percent on or after emergence. Minor frost damage on emerged seedlings may occur when temperature is less than 32 degrees Fahrenheit, but becomes more damaging when temperature is less than 28 degrees for prolonged periods.

In Northern and Midwestern states, where the risk for early frost damage is more, optimum planting dates were observed for as many as 21 days earlier than what is typically used. Using the 21-days-earlier planting as a threshold, 2 percent of all 289 locations – all in North Dakota, South Dakota and Minnesota – exceeded the 20 percent probability threshold for daily minimum temperatures to be less than 32 degrees at emergence. Only 0.3 percent had exceeded the 20 percent probability for daily minimum temperatures to be less than 30 degrees.

Conclusions

Global temperatures are expected to continue increasing until the year 2100. And a 20 percent cumulative increase in vapor-pressure deficit in July in the Midwest is projected by 2040, driven by increased temperatures and reduced relative humidity. Climate simulations have estimated as much as a 30 percent reduction in precipitation during summer months in many U.S. regions, including the Midwest and much of the Corn Belt. It’s clear that soybeans exhibit variable sensitivities to weather during vegetative and reproductive development. To that end we show here that with state-specific planting-date adjustment, drought impact during sensitive developmental stages could be mitigated.

Overall our results agree with previously reported simulated future yield trends – a 7 percent to 15 percent increase – due to climate adaption in wheat, rice and maize. The results in our study complement the previously measured sensitivity of soybean-related long-term economic returns to regional climatic change. We have identified and quantified climate-change-related yield constraints. It’s evident that many progressive farmers in the north-central U.S. region continuously monitor and strive to plant crops earlier on an annual basis. Our results highlight the potential yield and monetary benefit that U.S. farmers can gain due to planting-date adjustment – by using the results we report as a point of reference of optimal planting date in each state. Our findings suggest that the USDA Risk Management Agency should consider updating its antiquated earliest planting dates for replant payments, to reflect current environmental and monetary factors.

Spyridon Mourtzinis engages in post-doctoral research in the department of agronomy at the University of Wisconsin-Madison. James E. Specht is an Emeritus professor in the department of agronomy and horticulture at the University of Nebraska-Lincoln. Shawn P. Conley is a professor in the College of Agricultural and Life Sciences at UW-Madison.