Resumen de Integrating stkeholder knowledge in cellular automata models of land use change
- español
Introducción: En los últimos años, se han desarrollado muchas investigaciones en las que se aplica la modelización geográfica para estudiar las recientes transformaciones en los usos y ocupación del suelo. Sin embargo, son pocos los modelos de este tipo que incorporan de manera explícita la información y el conocimiento de los agentes conocedores del territorio, en muchas ocasiones sin vinculación a la investigación científica. La falta de inclusión de estos agentes en la modelización puede limitar la fiabilidad del modelo a la hora de representar el territorio estudiado, pero sobretodo, puede reducir su utilidad y aplicabilidad en la planificación territorial, ya que estos modelos pretenden servir de apoyo en la toma de decisiones políticas.
Objetivo: Esta tesis tiene como objetivo demostrar que el conocimiento de los agentes conocedores del territorio, recopilado a través de un proceso participativo, puede integrarse con la modelización cuantitativa de los cambios del uso de suelo (el caso concreto de los modelos de autómatas celulares), y de este modo generar modelos geográficos mejorados para su aplicabilidad en los estudios de caso, y la divulgación de sus resultados. En el Capítulo 1, se presenta la hipótesis principal de la tesis y se definen una serie de objetivos específicos para comprobar dicha hipótesis. El Capítulo 2 se dedica al estado del arte. En el Capítulo 3, se presenta el marco conceptual de la modelización geográfica en general, y en el tipo de modelo que se aplica en la tesis en particular, que trabaja en base a una técnica conocida como autómatas celulares (AC). En el Capítulo 4, se presenta una revisión bibliográfica de algunos de los ejemplos de modelos del tipo considerado en esta tesis en los que se han incorporado algunos elementos que van mas allá de la investigación pura con el fin de apoyar en la toma de decisiones (DSS-Decission Support Systems) y en la elaboración de políticas (POSS-Policy Support Systems). En los Capítulos 5, 6 y 7, se presenta la metodología aplicada para desarrollar el modelo integrado resultado final de esta tesis, presentado en el Capítulo 8. En el Capítulo 9 se presenta una breve discusión de los objetivos principales de la tesis y se evalúa si se han cumplido de manera satisfactoria partiendo de la hipótesis principal de la tesis establecida en el Capítulo 1. Las conclusiones del trabajo se presentan en el Capítulo 10. En el Capítulo 11, la tesis termina con una discusión sobre futuras lineas de investigación que pueden surgir a partir del trabajo aquí presentado.
Conclusiones: Las aportaciones principales de este trabajo, a través de la metodología propuesta en esta tesis, se orientan en turno a dos líneas. Por una parte se ha propuesto un proceso de modelización integrado para generar un modelo más cercano a la realidad con capacidad de responder a las preocupaciones reales de la sociedad sobre el espacio investigado. Por otra parte, se ha convertido a los agentes en usuarios verdaderos de dicho modelo, plenamente involucrados y capaces de criticar y opinar sobre el tema representado a través del modelo. En un mundo en el que la necesidad de dialogar y consensuar es cada vez más acusada, si la modelización pretende ser relevante en la planificación y gestión del territorio, la integración de los procesos participativos en la modelización técnico-analítica ofrece un camino muy prometedor para el estudio y gestión de los impactos del ser humano sobre el medio ambiente hacia el futuro
- English
Since approximately half a century ago, the pressure exercised by the human race on the biosphere has driven scientific activity in a variety of disciplines. A great number of diverse activities relating to monitoring, documentation and research have emerged under the banner of global change research. Changes in earth surface cover, among other aspects, may affect the control of factors such as erosion, water supply and climate. For this reason, the monitoring and analysis of land use and land cover changes are of crucial importance in understanding the processes of change to the earth's surface and its ecosystems.
Geographical Information Technologies (GIT) offer a broad range of methods for the study of land use and land cover change, and research applying techniques of geographic analysis to investigate recent land use and land cover change is extensive. Examples include the extremely strong urban growth experienced in the Madrid Autonomous Community in recent decades (the first case study in this thesis), or the transformation of the land surrounding the Doñana Natural Protected Area since the 1950's (the second case study presented in this thesis), mainly as a result of agricultural intensification of the area. In both situations, in spite of the great differences between them, the question that is most frequently asked is: Where are we going? In order to answer this question, and to encourage a land planning regime focused towards a more sustainable use of the limited natural resources available to us, it is necessary to progress beyond monitoring or static geographical analysis. This is the aim of this thesis. To achieve it, dynamic modelling techniques were applied to the investigation of land use change for the two case studies previously mentioned. In the work, detailed attention was given to the integration of quantitative modelling with participatory processes that introduced a qualitative perspective. By doing so, the aim was to improve the model, its applicability to the case studies and the dissemination of the results. Also, more specifically, in the second case study (the Doñana Natural Area), a process of social learning was initiated, in which stakeholders from different sectors and disciplines (including the research team), exchanged ideas and knowledge acquired about the future of the protected area and the threats that it faces in the short and medium term (Aim 1).
There is a great diversity of land use models, some already incorporated into commercial software or available over the internet. In particular, this thesis has focussed on a type of model based on a technique known as cellular automata (CA). In this model, a set of cells are defined, having states (for example, land uses) at a given moment (T0). Cells are able to transform from one state to another over the course of a time sequence (T0, T1, T2... Tn) on the basis of the relationship between their own state and the state of the cells that are immediately adjacent or in proximity, known as the cell neighbourhood (N). However, the potential of each cell to transition is not determined exclusively by its neighbourhood, but Accessibility (A), the influence of lines of communication (e.g. transport, irrigation) is also taken into account. In the same way a cell's transition potential takes into account Suitability (S), the biophysical eligibility (e.g. rainfall, slope) of each land area to adopt a particular use, and Zoning (Z), the incorporation of legislative and planning restrictions in place (for example, protected areas, urban spatial plans). Finally, since human activity in the landscape is not purely deterministic, a stochastic parameter is added (v). This type of model has been referred to throughout this thesis as N,A,S,Z. This model is very effective for generating realistic land use growth patterns, something that has been greatly facilitated in the research presented as part of this thesis thanks to the participatory process, that led, at the same time, to increased dissemination of the results. In the same way, the N,A,S,Z model has been presented in this thesis as a conceptual framework for understanding land change dynamics prior to embarking on model development. It has been proposed that this model can be applied more broadly in the context of land use change research (Aim 2).
In order to generate simulations of possible future land use it is necessary to identify the changes that have taken place by comparing historic maps of the region under study, and, subsequently, trying to replicate then in a simulation by establishing a set of rules known as transition rules. The process of generating the transition rules for the model is known as calibration. The calibration of N,A,S,Z models is normally approached using trial and error procedures, supported by analysis of historic change undertaken previously. Undertaking the calibration process in this way can be arduous, but it is considered to be necessary, since the process obliges the user to look for, and try to understand, the real change process that have taken place in the territory in order to be able to replicate them (Aim 3).
One of the most innovative aspects of this thesis lies in the attempt to demonstrate that the most relevant end user in the process is she/he who actually develops the model, not some other end user present in the researcher's imagination who may not exist in reality. For a model of this kind to be used by anyone outside of the scientific community, it is advisable to involve local stakeholders to contribute the knowledge of the region in which the model is applied. In the model presented in this thesis, stakeholders were involved in model development from the beginning; they helped to define the study area, the data base and land use categories to employ in the model, and they carried out reclassification of the land use data set to determine the most important land use dynamics for modelling their region. Stakeholders also gave their opinions and knowledge about the effect of the different biophysical factors on each land use category and contributed criticism of the way the model worked. The process was concluded with participatory activities dedicated to evaluating the participatory modelling process itself. (Aim 4).
Finally, the generation of suitability maps in N,A,S,Z type models is difficult to attain exclusively through trial and error approaches. In this thesis a new methodology has been developed for calculating suitability using raster map algebra, combining Artificial Neural Network-based transition analysis with participatory evaluation of the influence of each suitability factor on each of the modelled land uses (Aim 5).
The main contributions of this work, by means of the methodology proposed in this thesis are orientated around two distinct axes. On the one hand, an integrated modelling procedure has been proposed, in order to generate a model that is closer to reality and with the capability to respond to the real concerns of society about the territory under study. On the other hand, the stakeholders have been converted into the model's true users, fully involved and able to criticise and give their views about the themes represented within it.
Although the format has been adapted to present a consistent and coherent narrative, the work, the doctoral thesis presented here comprises a series of eight scientific articles, most of which have been published in impact factor journals. Moreover, the research carried out gave rise to the development of a series of innovative methodologies for the integration of discursive information into quantitative models, all of which have been disseminated and accompanied by a large quantity of data which can be downloaded from the project website (http://www.geogra.uah.es/duspanac), created and maintained by the PhD candidate. The geographic analysis also generated a geodatabase, likewise publicly available at the same website. The intention has been to provide the model and the data generated the maximum utility, visibility and accessibility possible.
In a world in which the need for dialogue and consensus is increasingly urgent, the integration of participatory processes in technical-analytical modelling offers a promising direction for the study and management of human impacts on the environment into the future.
