Agricultural land systems : modelling past, present and future regional dynamics

Abstract

This thesis arises from the understanding of how the integration of concepts, tools, techniques, and methods from geographic information science (GIS) can provide a formalised knowledge base for agricultural land systems in response to future agricultural and food system challenges. To that end, this thesis focuses on understanding the potential application of GIS-based approaches and available spatial data sources for modelling regional agricultural land-use and production dynamics in Portugal. The specific objectives of this thesis are addressed in seven chapters in Parts II through V, each corresponding to one scientific article that was either published or is being considered for publication in peer-reviewed international scientific journals. In Part II, Chapter 2 summarises the body of knowledge and provides the context for the contribution of this thesis within the scientific domain of agricultural land systems. In Part III, Chapters 3 and 4 explore remotely sensed and Volunteered Geographic Information (VGI) data, multitemporal and multisensory approaches, and a variety of statistical methods for mapping, quantifying, and assessing regional agricultural land dynamics in the Beja district. In Part IV, Chapters 5–7 explore the CA-Markov model, Markov chain model, machine learning, and model-agnostic approach, as well as a set of spatial metrics and statistical methods for modelling the factors and spatiotemporal changes of agricultural land use in the Beja district. In Part V, Chapter 8 explores an area-weighting GIS-based technique, a spatiotemporal data cube, and statistical methods to model the spatial distribution across time for regional agricultural production in Portugal. The case studies in the thesis contribute practical and theoretical knowledge by demonstrating the strengths and limitations of several GIS-based approaches. Together, the case studies demonstrate the underlying principles that underpin each approach in a way that allows us to infer their potentiality and appropriateness for modelling regional agricultural land-use and production dynamics, stimulating further research along this line. Generally, this thesis partly reflects the state-of-art of land-use modelling and contribute significantly to the introduction of advances in agricultural system modelling research and land-system science.