Simulating Soybean Yield Potential under Optimum Management

Abstract

Core Ideas Soybean yield and crop growth was measured in high‐yield contest fields. Radiation use efficiency and nitrogen uptake were greater than recorded values. Yield was grossly under‐predicted using default growth parameters. Yield predictions were greatly improved using measured growth parameters. Crop models have played important roles in identifying potential constraints for crop growth and yield. A relatively simple soybean [Glycine max (L.) Merr.] model consisting of a daily C, N, and water budget was used to simulate yields for optimum production environments at Fayetteville, AR, and at a farmer's contest field in Missouri for 2012 to 2013. Data were collected on radiation use efficiency (RUE), N accumulation rate, specific leaf nitrogen (SLN), and the dry matter accumulation coefficient (DMAC) as a measure of whole‐crop seed growth rate. In Fayetteville, measured yields ranged from 4977 to 7144 kg ha−1, and simulated yields averaged 34.0% below the measured yields using the default model parameters. Using measured parameters in a modified model, predicted yields were 2.8% above observed. Default parameter simulations for the contest fields were 39.0% below measured yield and were 18.7% below measured yield when using measured parameters. Sensitivity analyses indicated that lower DMAC values increased yields due to slower seedfill rates, allowing additional N accumulation and a slower translocation of N to the growing seeds. Simulating an increased SLN and RUE increased predicted yields in 2012 and 2013 when N accumulation rates were great enough to supply the required N for new biomass. Alternatively, increasing N accumulation rates increased yield up to a plateau when all N requirements were met. These results illustrate the importance and interconnectivity of the crop growth processes relating to C and N metabolism and that current bounds for crop growth characteristics should be reconsidered.

A surrogate model based on feature selection techniques and regression learners to improve soybean yield prediction in southern France

“…Statistical approaches search and explore the relations between data to explain the variables of interest whereas mechanistic models are based on the description of biophysical processes. Dynamic crop models simulate daily growth and development in relation with environmental resources and agricultural inputs; they allow the testing of functional hypotheses and the identification of potential constraints to crop growth and yield (Purcell & Roekel, 2019). However, mechanistic and statistical approaches can be combined in order to improve the crop modeling predictions (Casadebaig et al, 2011(Casadebaig et al, , 2020.…”