Uncertainty assessment of soya bean yield gaps using DSSAT‐CSM‐CROPGRO‐Soybean calibrated by cultivar maturity groups

Abstract

Soya bean yield gap can be caused by different factors resulting in uncertainties when the objective is to use such information for farm decision‐making and reference yield determination. Thus, this study aimed to quantify the soya bean yield gap for four sites, located in Southern and Midwestern Brazil, as well as the uncertainties of that related to cultivars, sowing dates, soil types and reference yields. The crop simulation model DSSAT‐CSM‐CROPGRO‐Soybean was calibrated for cultivars with similar maturity groups, based on the data obtained from the best farmers at the county level. The yield gap by water deficit (YGWD) was obtained through the difference between potential and attainable yields, and that one caused by sub‐optimum crop management (YGCM) by subtracting actual yield of each county, obtained from official statistics between 1989/90 and 2014/15 growing seasons, from the estimated attainable yield. The yield was simulated using four sowing dates, three soil types and two soya bean maturity groups by county. The reference yield uncertainty was quantified using yield reference from crop model and regional winners of the soya bean yield context, conducted by CESB (Brazilian Soybean Strategic Committee), for the growing seasons from 2013/14 to 2015/16. The crop model showed a good agreement between measured and simulated crop development and growth using calibration by maturity group, with low root mean square error (347 kg/ha). Southern sites had a mean YGWD of 1,047 kg/ha, while in the Midwest, it was lower than 100 kg/ha. The YGCM was 1,067, 528, 984 and 848 kg/ha, respectively, for Castro, PR, Mamborê, PR, Montividiu, GO and Primavera do Leste, MT, representing the opportunity for yield gain when having the best farmers as reference. The maturity groups, sowing dates and soil types showed to be an important source of uncertainty for yield gap determination, being recommended to investigate the farms in detail for an appropriate quantification. The reference yield showed expressive uncertainties, with some farmers presenting conditions to increase their soya bean yields by more than 3,000 kg/ha, when considering as reference the yields obtained by the winners’ farmers. These results show that uncertainties must be reduced when assessing farm yield gaps, in order to ensure that expected rate of soya bean yield growth could be reached by adopting the same technologies from CESB winners and best farmers in the county as a reference.

Estimation and forecasting of soybean yield using artificial neural networks

“…The RMSE values for both estimation and forecasting were below those found by Teixeira et al (2019) who used the CROPGRO -soybean (DSSAT) model to estimate soybean yield for localities in Brazil they found an average RMSE of 347 kg ha −1 . Khaki et al (2020) using Convolutional Neural Networks (CNNs) and RNNs, together with Random Forest (RF), Deep Feedforward Neural Networks (DFNN), and the Least Absolute Shrinkage and Selection Operator (LASSO), obtained corn (Zea mays L.) and soybean yield forecastings throughout the Corn Belt (including 13 states) in the United States obtaining RMSE of 427.04 kg ha −1 .…”

Assessment of agricultural efficiency and yield gap for soybean in the Brazilian Central Cerrado biome

“…In these cases, the approach used was the same; however, Sentelhas et al (2015) used yield from cultivars trial as reference from growing seasons between 2008 and 2011 (maximum yield below 6,000 kg•ha -1 ); while Battisti et al (2018) used yield from soybean contest during growing seasons between 2014 and 2017, with yield level between 4,800 and 8,800 kg•ha -1 . This yield gap range helped to create different scenarios of yield project by management improvement (Teixeira et al 2019).…”

“…The ratio between AE and CE was analyzed by fitting an adequate regression to verify the effects of water deficit on management efficiency. In this analysis, AE was obtained by adding technological tendency yield at AcY in the previous growing seasons from 2016/17, following the approach of Teixeira et al (2019). The AcY rate was obtained by linear adjust, which was multiple by the number of years between the growing seasons and 2016/17, adding this value to AcY.…”

Assessment of sowing dates and plant densities using CSM-CROPGRO-Soybean for soybean maturity groups in low latitude

“…Soybean crop management for new regions has been based on the management for current production areas, which can lead to low efficiency to explore new environmental conditions, for example, using sowing dates, plant density and cultivars from other production regions that limit potential yield (Teixeira et al, 2019). The management needs to be adapted for different environmental conditions to reduce risks associated with climate and production costs (Battisti et al, 2020a) and to increase crop resilience (Halsnaes and Taerup, 2009).…”

“…The management needs to be adapted for different environmental conditions to reduce risks associated with climate and production costs (Battisti et al, 2020a) and to increase crop resilience (Halsnaes and Taerup, 2009). Crop management practices that can be used in new environments to improve yield and crop resilience, include sowing dates (Hu and Wiatrak, 2012;Spehar et al, 2015), maturity group (Battisti et al, 2018;Teixeira et al, 2019) and irrigation (Justino et al, 2019;Battisti et al, 2020b).…”