Soybean yield gaps and water productivity in the western U.S. Corn Belt

Patricio Grassini, Jessica A. Torrion, Haishun S. Yang, Jennifer Rees, Daryl Andersen, Kenneth G. Cassman, James E. Specht

Research output: Contribution to journalArticlepeer-review

Abstract

Although the U.S. accounts for 38% of global soybean production, and produces 85Mt annually, quantification of U.S. soybean yield gaps and assessment of underpinning causes have not been performed. Likewise, no attempt has been made to estimate the attainable water productivity (kg seed per mm of water supply). Here we make an initial attempt at such analyses by evaluating yield gaps and water productivity of soybean in the state of Nebraska (western U.S. Corn Belt), which ranks fifth among U.S. soybean producing states with a 6.5Mt annual production. Rainfed and irrigated production are both important in Nebraska, accounting for 46% and 54% of total state soybean output. We evaluated a database containing on-farm field yields, applied inputs, and management practices collected from 516 soybean fields in the eastern and central regions of Nebraska during three successive crop seasons (2010-2012). Yield gaps were estimated as the difference between yield potential and on-farm field yields. Yield potential was estimated by two approaches: (i) a soybean simulation model coupled with field-specific weather and management data, and (ii) derivation of a boundary function for the relationship between producer-reported soybean yields and seasonal water supply using quantile regression. Actual yields were evaluated in relation to applied inputs and management factors to identify the most likely causes of yield gaps. Across regions, average yield gap of irrigated soybean ranged from 10 to 30% of the simulated yield potential. Water supply set an upper limit to productivity, not only in rainfed fields, but also in irrigated fields in a drought year in which irrigation did not fully satisfy crop water requirements. The boundary function for the relationship between yield and seasonal water supply had a slope (≈attainable water productivity) of 9.9kgha-1mm-1 and x-intercept (≈soil evaporation) of 73mm. A seasonal water supply of ca. 650mm appeared sufficient to maximize seed yield. Average rainfed and irrigated yields were 31 and 20% below their respective yield potential estimated from the boundary function. Sowing date also set an upper limit to producer yield, and this management practice alone explained the largest portion of the observed yield variation among fields. Soybean yield linearly decreased with sowing date delay, with the magnitude of the yield penalty varying among regions. Irrigated yields were not higher in no-till fields, and, in fact, a yield penalty was observed in no-till fields compared to reduced- or disked fields, especially in region-years with cooler early-season temperatures. Higher irrigated and rainfed yields were observed in fields that received starter fertilizer or P fertilizer application. Finally, application of in-season fungicide enhanced irrigated crop yields though this effect was inconsistent over years. Analysis of this database indicated that (i) soybean producers in this region obtained yields close (70-90%) to the estimated yield potential ceiling, and (ii) future agronomic on-farm yield improvement might be achieved by fine-tuning of current management practices, including earlier sowing date coupled with judiciously chosen tillage to achieve warmer soils in the springs plus suitably applied nutrient fertilizer application and in-season fungicide.

Original languageEnglish (US)
Pages (from-to)150-163
Number of pages14
JournalField Crops Research
Volume179
DOIs
StatePublished - Aug 1 2015

Keywords

  • Glycine max L.
  • On-farm data
  • Soybean
  • Water productivity
  • Yield gap
  • Yield potential

ASJC Scopus subject areas

  • Agronomy and Crop Science
  • Soil Science

Fingerprint

Dive into the research topics of 'Soybean yield gaps and water productivity in the western U.S. Corn Belt'. Together they form a unique fingerprint.

Cite this