Work

Cloud-Hosting of Agricultural Crop Simulator and Optimizer for Model Calibration and Farm Input Recommendations

Year: 2024

Type: article

Abstract: Agricultural models have traditionally been designed for Windows desktop computers, limiting their accessibility and usability across different operating systems and devices. In contrast, cloud-based tools have the potential to improve accessibility, lower individual computational requirements, and enable transparency of results and ease of collaboration. In this paper, we report the first cloud-hosting of a widely used agricultural crop model, Root Zone Water Quality Model (RZWQM2), along with our own-developed optimizer for model calibration and farm decision recommendation, on the MyGeoHub cyber-infrastructure for the benefit of the agricultural research and farming community, and other stakeholders. The optimization is implemented based on a randomized block Coordinated Descent (rbCD) search that is shown to outperform the previous RBFOpt-based optimizer utilized within the MISSION framework for farm input management. The paper also presents a new equivalent formulation of the optimization function introduced in MISSION that renders its MINLP to a simpler NLP, removing the integer variables, reducing the number of decision variables to half and improving the optimization result as well as speed. To demonstrate the effectiveness of the cloud-hosted RZWQM2 model and the new optimizer, we also conducted a case study using a test dataset of maize growth in Greeley, CO. MISSION′ predicts higher crop yield and profit than both the expert recommendation (25.8% yield, 39.8% profit) as well as the published RBFOpt-based optimizer (2% yield, 3% profit) in 2008. We also report up to 26% higher yield and 40% higher profit of MISSION′ over expert recommendations for 4 different seasons over 2008-2011. The NLP reformulation of MISSION′ combined with the rbCD algorithm provided speedup that allows searching for input applications on all days of the season, compared to the earlier MISSION implementation that optimized by limiting to having 7 gap days between any two applications, while still took 7.6x less time on each optimization.

Source: Computers and Electronics in Agriculture

DOI: 10.1016/j.compag.2025.110364

Authors: Balaji Sesha Srikanth Pokuri, Anupam Bhar, Ratnesh Kumar

Institution: Iowa State University

Language: English

Topic: Smart Agriculture and AI

Subfield: Plant Science

Field: Agricultural and Biological Sciences

Domain: Life Sciences

SDG: Zero hunger

Open Access status: hybrid

Estimated APC paid: $3,680