Resumen de Tesis Doctorales

Abstract

© 2020 Asociacion de Geografos Espanoles. All rights reserved. The management of waste deposited in illegal landfills (ILs) is a significant problem in contemporary societies due to the large volume of refuse deposited and the respective impacts on the environment and human health. This thesis addresses the construction of IL occurrence models based on predictive modelling and simulation. The modelling applied enabled a precise mapping of potential IL occurrence zones to be obtained, and to determine the socioeconomic and environmental features controlling them, the aim being to support officials responsible for applying environmental protection and territorial planning policies (Biotto et al., 2009). A total of 286 and 153 IL respectively located during the year 2016 on the islands of Gran Canaria (GC) and La Palma (LP) through fieldwork and orthophoto interpretation were characterised and modelled (Quesada-Ruiz et al., 2018). The characterization and modelling of the IL also used a geospatial database generated from a sampling of IL and information derived from various feature subsets (117): waste type, control and vigilance, socioeconomic, accessibility, distance to elements of interest, visibility and physiography. Multivariate analysis techniques were applied, such as principal component analysis (PCA), logistic regression (LR) and discriminant analysis (DA). PCA determined the interrelationships between features and simplified the feature space. LR and DA enabled causal relationships to be established between the independent features and IL occurrence. Both models highlighted the NDVI, slope, land-use changes and proximity to urban centres as decisive features for IL occurrence in LP, while in GC the influence of the coastline, land-use transitions to artificial coverage or greenhouse density stood out. DA was used to obtain maps of the probability of IL occurrence and of areas potentially affected by IL (Quesada-Ruiz et al., 2019a). The development of a new methodology is probably the most significant contribution of the thesis. Area Feature Guided Regularised Random Forest (AFGRRF) is a new feature selection method applied to binary phenomena and supported by the Gaussian Regularised Random Forest (GRRF) algorithm (Rodriguez-Galiano et al., 2018; Deng et Runger, 2013). A novelty of AFGRRF is the joint optimisation of overall mapping accuracy and minimisation of the affected area by success ratios. Finally, IL occurrence was investigated in two pilot zones on the island of GC, norwest (Zone A) and southeast (Zone B), considering IL as a dynamic and complex system (Quesada-Ruiz et al., 2019b). This enabled evaluation of IL progression between periods, analysing the time dimension of IL growth and its relationship to land-use change processes such as the' housing bubble'. Unlike studies that focus on predictive IL modelling, this approach for the first time went beyond mere analysis of their location and enabled joint estimation of the spatial and time dimension of IL occurrence, simulating IL occurrence between 2000 and 2018 based on cellular automata models (Eastman, 2015). The analysis of surfaces affected by IL for the period between 2000 and 2006 indicated an inter-annual growth rate of 4.5% and 9.5% in Zone A and Zone B, respectively for the period between 2006 and 2012 the inter-annual growth rate was 6.6% and 6.7%. Those changes represented an increase in absolute terms in the year 2006 of 20.5 ha and 97.8 ha, and in 2012 of 97.8 ha and 11.6 ha, for Zone A and Zone B, respectively. The best model simulation forecast an increase of 52.3 ha and 81.5 ha affected by IL for the year 2018 in Zone A and Zone B, respectively. The analysis of the relationship between the IL and land uses showed that the IL of both zones were mainly located in agro-livestock zones and areas of vegetation not far from urban areas, accessible and with neither control nor vigilance for all the periods. Growth of IL was greater in urban areas, construction spaces and industrial zones during the period between 2000 and 2006, and may be highly related to the urban expansion process linked to the "housing bubble". However, the inter-annual IL growth rates in urban environments fell during the period after the 2008 economic crisis and the appearance of new urbanisation regulations.