Abstract
Currently, studies on climate dynamics and land use and land cover changes are essential to identify the magnitude and speed of what has been called global change and the main driving forces of these changes. These forces are not homogeneous in time or space, so it is important to know their spatiotemporal patterns and the effects they have on, among others, natural and agricultural ecosystems. Evenly, aspects of physical geography and socioeconomic factors must be considered. To carry out these studies in large territories, but with a level of detail that does not entail the risk of superficial visions, it is necessary towork with extensive datasets of information (Big Data) that must be appropriately processed (not every approach is equally valid). For this reason, it is important to establish methodological solutions that correct problems and maximize the quality of the information before it is used in the analysis of global change.
The main objective of this thesis is to contribute to the analysis of the dynamics of global change in the Iberian Peninsula (IP) through the study of the climatic time series of meteorological stations and the forest expansion dynamics based on information obtained through remote sensing. To address this objective, it will previously be necessary to implement strategies to improve the quality of the information produced from sensors since this information will allow us to analyze the trends and generate the metrics to quantify the changes and their impacts.
Initially, the performance and influence of different gap-filling methods are evaluated from the climatic trends of temperature and precipitation, identifying spatiotemporal patterns at the IP scale, and the applicability of methods to reduce spatiotemporal inconsistencies in derived climatic surfaces. The different gap-filling methods show similar results when we compare the temporal trends of change, but, as we will indicate below, with differences at the spatial level, which is not a minor contribution. It is also observed that the gap-filling
entails a reduction in the variability of the climatic series, although, in return, it is possible to operate with more homogeneous and comparable data.
Secondly, filtering rules are developed to be applied to ground truth areas as a previous step to perform land cover classification processes from remote sensing data. This task is necessary when there is a possibility that this "truth" contains inconsistencies that are not minor, nor that they are semantic concerning our objectives. Moreover, this is especially frequent when this ground truth has been derived from auxiliary databases. The ultimate
goal, in this case, is to obtain a multitemporal land cover mapping of the highest quality and spatiotemporal coherence. Applying the proposed filters to ground truth areas resolves the most common errors associated with natural and agricultural categories, showing a substantial increase in the overall accuracy of the cartography obtained.
These methodological improvements make it possible to address with greater guarantee analysis of change for important areas of IP. Thus, it is observed that climate trends indicate an overall increase of 1.45 °C in the mean temperature for all IP, but a more complex spatial pattern for precipitation. Regarding forest expansion, the dynamics along a latitudinal gradient of large amplitude (~185 km) and representative of the IP (from the Pyrenees to the Sierra Nevada mountain range) are analyzed, also applying methodological improvements for the extraction and filtering of the occurrence locations of change, the
sampling and statistical analysis (Boosted Regression Trees), which also allows reducing possible biases in the interpretation of the results.
From the diversity of results obtained, the main factor of the forest expansion has been related to the distance to previous forests, indicating a process of initial densification and subsequent forest expansion. The expansion of broadleaf deciduous forest formations has also been conditioned by water availability and specifically by the distance to the hydrographic network. On the other hand, broadleaf evergreen forests have been related to topoclimatic factors (precipitation, temperature, and solar radiation) in contexts of lower
water availability, as well as coniferous forests, also favored by a greater distance from the provincial capitals. The factors showed variability across bioclimatic regions and at the temporal level, especially in the case of grasslands. The abandonment of agricultural activities was the leading cause of new forests’ appearance in the Mediterranean regions, while the new forests related to natural succession dynamics occurred mainly in Mediterranean and Eurosiberian mountain areas. The driving factors of forest expansion have shown an irregular spatial distribution, according to the water requirements of the
forest physiognomic types, in addition to a lack of stationarity in the inducing factors, which indicates that the driving forces operate differently at the spatial and temporal level.
In conclusion, this thesis provides novel information on global change and its driving forces derived from robust methodological strategies, also proposed as novelties, shows that working on a detailed scale and spatial and temporal attention is possible, allowing richer and more rigorous territorial analyses and leaving open new questions and possibilities of study for science, planning and management of the IP.
| Date of Award | 8 Jul 2022 |
|---|---|
| Original language | Spanish |
| Awarding Institution |
|
| Supervisor | Xavier Pons Fernandez (Director), Miquel Ninyerola Casals (Director) & Pere Serra Ruiz (Director) |