Date of Award

2024

Document Type

Master Thesis

Degree Name

Master of Engineering (Research)

Department

Process, Energy, and Transport Engineering

First Advisor

Dr. Michael D. Murphy

Second Advisor

Dr. Elodie Ruelle

Abstract

The increasing need for highly productive and eco-friendly agriculture is driving demand for decision-support tools to assist agricultural operators in managing systems more precisely and sustainably. Although grasslands provide a cost-effective option for ruminant production and can offer many environmental benefits (biodiversity, carbon storage etc.), grasslands management is challenging due to the inherent climate and soil sensitivities and their potential negative impact on water quality. In Ireland, the Moorepark Saint Gilles Grass Growth (MoSt GG) model was developed for the dual purposes of research-based simulations and on-farm grass growth prediction for farmers. The objective of this thesis was to improve the model’s grass growth prediction accuracy and to better simulate the interaction between soil, water and nitrogen (N) fluxes and their associated effects on grass growth and N leaching. The first step was the conceptual improvement of the model through the further development of the soil sub-model, the addition of a new soil water pool and the new consideration of day length in the radiation component. The improved model underwent a two-step evaluation process, involving the simulation of virtual paddocks with diverse soil types to assess the impact of model improvements on grass growth and N leaching over two years, followed by the comparison of model biomass estimation with measured data from four experiments conducted over two to fourteen years in Ireland and France. Model adaptations enhanced simulation accuracy across all evaluated experiments, reducing root mean square error (RMSE) from 322-1,011 to 312-671 kg of DM/ha. This novel consideration of soil characteristics resulted in a higher variability in grass production and N leaching depending on soil type and weather conditions, leading to improved correlations between the simulated and actual grass growth. The addition of the new soil water pool representing the topsoil layer improved the accuracy in drier years (for French data) due to the more realistic simulation of grass growth recovery after a drought. The second step of this work was to improve the model accuracy for predicting on-farm grass growth (kg of DM/ha/day) through the calibration of the MoSt GG model’s most influential parameters. Through a sensitivity analysis carried out using the Morris method, ten parameters were selected for model calibration using a dataset of grass growth measurements from 14  commercial farms across four years. The calibration process improved the model accuracy to predict grass growth for this existing dataset: Mean absolute percentage error (MAPE) reduced from 30.0 to 19.8%, RMSE reduced from 17.23 to 13.37 kg of DM/ha and R2 increased from 0.58 to 0.71. The improvement of the model accuracy through calibration was further evaluated using a new dataset of ten commercial farms over four years. The calibration process improved the model accuracy for this new dataset: MAPE reduced from 30.1 to 19.1%, RMSE reduced from 18.28 to 13.40 kg of DM/ha and R2 increased from 0.53 to 0.67. The resulting calibrated model better predicts varying soil types, climates, and fertilisation management practices and can be used for a wider range of farms. Further improvements are still possible for the MoSt GG model. For example, the incorporation of a clover sub-model or the further development and evaluation of existing components of the model (crude protein content of the grass, N leaching, gaseous emissions).

Access Level

info:eu-repo/semantics/openAccess

Download

Share

COinS