The main core of this thesis consists of three scientific articles. The first article deals with obtaining temporal series of land-use and land-cover maps for large areas of the Mediterranean fringe of the Iberian Peninsula. To do this, a classification scheme of Landsat imagery has been implemented. Training and test areas were extracted from the Land Occupation Information System of Spain 2005 (SIOSE, Sistema de Información sobre Ocupación del Suelo en España). The article includes a procedure for obtaining these areas on dates lacking such kind of reference information. An overall accuracy of over 90% has been obtained for all maps, and accuracy by categories is above 85% in most cases. The second article analyzes the main successional stages and transitions between the following vegetation-cover categories: grasslands, shrublands, coniferous forests, broadleaf evergreen forests and broadleaf deciduous forests. Three areas of Spain were included in this study as different scenarios for model development. Ambits correspond to land-cover classifications of three entire Landsat scene and the study period is 1987-2012. Conditions were described by topography derived variables and drought-occurrence variables, which were obtained through a Geographical Information System (GIS). The machine-learning technique Boosted Regression Trees (BRT) were used to identify the most important variables and describe the relationships between the forest vegetation transitions and key factors. Main findings: Transitions to Mediterranean and sub-Mediterranean broadleaf forests are positively associated with drought occurrence while transitions to conifers are negatively affected by drought; Important interactions between topography derived variables and drought have been found. The study provides robust evidence that drought occurrence plays an important role in the decline of conifers and the expansion of Mediterranean and sub-Mediterranean broadleaves, which could become the dominant species in many areas of the Mediterranean if climate model forecasts are met. The third article focuses on the abandonment of rainfed and irrigated herbaceous crops in order to derive specific explanations according to the crop for two different study areas: the Central Spanish Pyrenees and the Central Ebro Basin. The methodology is also based on remote sensing, GIS and BRT. Variables quantifying the cost distance to urban centers and roads and other variables related to socioeconomic data at a municipal level have been included in this analysis. Topography derived variables are the main determinants, except for irrigated crops in the Ebro Basin, where locational factors play a more important role. BRT models have allowed identifying other significant patterns such as: the vulnerability of irrigated crops to drought; the higher dependence of agricultural activity in the Pyrenees on internal networks; pattern shifts of land abandonment in the analyzed sub-periods, and; evidence of the importance of economic diversification for maintaining cropland. The abandonment probability does not always has a linear relationship with the explanatory factors and, in some cases, it shows polarized response patterns that reflect, for a single factor, different contexts prone to be abandoned.
Análisis de factores determinantes de las transiciones forestales y el abandono de cultivos en el ámbito mediterráneo : un modelo de dinámica de cubiertas del suelo basado en teledetección, SIG y boosted regression trees
Vidal Macua, J. J. (Author). 14 Feb 2018
Student thesis: Doctoral thesis
Student thesis: Doctoral thesis