Using Data Science for Understanding and Predicting Long-term Land Use and Cover Changes: A Study Case in Mexico

This thesis offers a comprehensive examination of the trends and changes in Land Use and Land Cover (LUCC) in Mexico. It employs a combination of bibliometric analysis, topic modeling, remote sensing, machine learning, and geographical machine learning methods. The study methodically reveals the patterns, factors, and outcomes of land use and land cover change (LUCC), providing crucial insights for the preservation of the environment and the development of policies.Chapter 2, titled "A Bibliometric and Topic Modelling Synthesis of Scientific Research on Land Use and Cover Change," offers a thorough analysis of the scholarly output pertaining to research on land use and cover change (LUCC). The primary objective of this study is to utilise bibliometric analysis and topic modelling approaches in order to ascertain key research themes, influential authors, journals, and institutions that have played a substantial role in creating the field. The analysis illustrates the gradual evolution of research foci over a specific timeframe. This chapter provides a thorough examination of the historical development of land use and land cover change (LUCC) studies, highlighting the increasing complexity and diverse nature of current research in this field. In this chapter, the findings of topic modelling are presented, which provide light on the underlying thematic patterns found in the literature. Additionally, the efficacy of including Large Language Models in the modelling process for deducing coherent narratives from unstructured data is emphasised.Chapter 3, titled "Investigating the Drivers of Deforestation in Mexico: An Analysis using Geographical Machine Learning," explores the socioecological factors that contribute to the occurrence of deforestation in Mexico. The study used a blend of traditional models and machine learning methods to infer the connections between changes in land cover and various socio-economic, biophysical, and geographical factors.In Chapter 4, titled "Detecting Changes in Mexican Vegetation over 40 Years: A Multi-temporal Segmentation Approach Using Landsat Imagery," a thorough examination of alterations in vegetation throughout Mexico is presented. This study utilises the Continuous Change Detection (CCD) method and analyses four decades of Landsat images to identify places with notable changes in vegetation. The findings provide valuable information about the timing and location of land use and land cover change (LUCC) in various regions of the country. This chapter showcases the efficacy of CCD in capturing intricate temporal variations, hence enhancing comprehension of the ecological transformations taking place over an extended period.Chapter 5, titled "Predicting Land Use Cover in Mexico: A Continuous Classification Approach Using Long-Term Landsat Data and Machine Learning," introduces a comprehensive predictive model for Land Use and Land Cover (LULC) over the entire country from 1985 to 2000. The provided continuous model accurately forecasts Land Use and Land Cover Change (LULCC) and establishes a strong foundation for future national-scale research on LUCC, including its causes and effects.Chapter 6 presents a concise overview of the findings derived from the research.Together, these studies provide a comprehensive examination of LUCC dynamics in Mexico, offering valuable insights into the patterns, drivers, and potential impacts of land use and cover changes. Overall, this thesis aims to contribute in the knowledge of LUCC processes and provide bases to inform evidence-based conservation and land management strategies.