Improvement Of CA Model In The Parallel Simulation Of Land Use Change

Land use and cover change has become the global research thrend and hotspot, which is the most direct response to the interactions among the natural environment, humen society, and enconomic development. Cellulal automata (CA) is one of the favorite dynamic models in land use change simulation, due to the "bottom up" model mechanism, the capacity for simulating the global complex land use dynamic evolution based on the simple neighbor rules, and the good compatibility with GIS data and the flexible programming.After more than70years'development, though CA model has been improved in varying degrees in three aspects:evolution rules, cells and neighbor, and neighbor effect factors, it still has some shortages and problems urgently needed to be sovled. Such as, how to realize the asynchronous simulation of land use change in the large study area where there are great regional differences of the change characteristics? How to determine the parameters of the components of CA model according to data features of the study area? And the present CA model applied in LUCC is lack of the consideration of spatial distribution factors. Therefor, this dissertation attemped to improve the CA model considering cellular neighbor, evolution rules, and spatial-temporal scales of the model. Aiming at the problems, the theories and methods of solutions were put forwarded, whose feasibility and correctness have been verified in case study. The focus points and conclutions are chiefly as follows:1) Studing the theories and methods of CA model building rooted in local conditions. The key idea of the improvement of CA model building is "adjust measures to the local conditions":we built a geographical CA model on the basis of the features of land use data of the study area. The applications of LUCC and the existing problems of the CA model at present were considered comprehensively when the CA model was built in order to determine appropriate design parameters. So this dissertation proposed the corresponding theories and methods of the improvement in three aspects:cellular neighbor, spatial-temporal scales, and evolution rules.2) Studing the asynchronous and parallel simulation of CA model in land use change. The geographical CA model has a large amount of computation, which limites its usage within a small area. Furthermore, in a large area the great regional differences of the land use change characteristics could not be included by the same set of evolution rules. Thus, it is necessary to make geographical CA model asynchronous and parallel. This dissertation has discussed the methods of parallelization of geo-CA model from data-paraellelism and task-parallelism aspects respectively at both coarse-grained level and fine-grained level.3) Verifying the performance and feasibility of the improved CA model by case studies. Taking the four districts of Wuhan located in sourth of Yangtze River as study areas, namely Wuchang district, Hongshan district, Qingshan district, and Jiangxia district. Firstly, the driving factors of the land use change were selected based on the analyses of the regional characteristics in these areas; secondly, an improved CA model suiting for the case areas was built according to the improvement methods proposed by this dissertation; and lastly, we analyzed and discussed the correctness and feasibility of the improved CA model from six aspects, respectively, including five single improvement items and the synthetical improvement. The results showed that the improved CA model could efficiently improve the accuracy and the efficiency of the simulation, and it could better preserve the actual landscape shape of the parcels. Besides, owning to the asynchronous evolution, the geographical CA model could be used in a much larger region.