Skip to Main content Skip to Navigation
Theses

Data-driven approach for addressing global agricultural issues : application to assess productivity of conservation agriculture under current and future climate

Abstract : In this thesis, we present a machine learning (ML) pipeline that produces data-driven global maps to address the global agricultural issues, such as assessing the spatial distribution of the productivity conservation agriculture (CA) versus conventional tillage (CT) under current and future climate. Our approach covers the selection and comparison of ML algorithms, model training, tuning with cross-validation, testing, and results global projection. We demonstrate its relevance using a global dataset we conducted which comparing the crop yields of conservation agriculture (CA) and no tillage (NT) vs. conventional tillage (CT) systems with a wide range of crop species, farming practices, soil characteristics and climate conditions over crop growing season. Through this ML pipeline, various models for classification, regression and quantile regression are trained based on 12 mainstream ML algorithms. The models are used to map the crop productivity of CA and its variants vs. CT at the global scale under different farming practices and climate conditions in the past (1981-2010), current (2011-2020) and future (2051-2060) scenarios. We reveal large differences in the probability of yield gains with CA across crop types, agricultural management practices, climate zones, and geographical regions. We show that CA stands a more than 50% chance to outperform CT in dryer regions of the world, especially with proper agricultural management practices. In conclusion, CA appears as a sustainable agricultural practice if targeted at specific climatic regions and crop species.
Document type :
Theses
Complete list of metadata

https://pastel.archives-ouvertes.fr/tel-03530959
Contributor : ABES STAR :  Contact
Submitted on : Tuesday, January 18, 2022 - 1:02:11 AM
Last modification on : Friday, August 5, 2022 - 2:41:12 PM
Long-term archiving on: : Tuesday, April 19, 2022 - 6:35:08 PM

File

ContratDif__ED581_SU_Yang_1028...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-03530959, version 1

Citation

Yang Su. Data-driven approach for addressing global agricultural issues : application to assess productivity of conservation agriculture under current and future climate. Agronomy. Université Paris-Saclay, 2021. English. ⟨NNT : 2021UPASB048⟩. ⟨tel-03530959⟩

Share

Metrics

Record views

107

Files downloads

11