(IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 9, No. 1, 2018 577 | Page www.ijacsa.thesai.org OpenMP Implementation in the Characterization of an Urban Growth Model Cellular Automaton Alvaro Peraza Garzón Instituto Tecnológico de Mazatlán Universidad Autónoma de Sinaloa Sinaloa, México René Rodríguez Zamora Universidad Autónoma de Sinaloa Instituto Tecnológico de Mazatlán Sinaloa, México Wenseslao Plata Rocha Facultad de Ciencias de la Tierra y el Espacio Universidad Autónoma de Sinaloa Sinaloa, México Abstract—This paper presents the implementation of a parallelization strategy using the OpenMP library, while developing a simulation tool based on a cellular automaton (CA) to run urban growth simulations. The characterization of an urban growth model CA is shown and it consists of a digitization process of the land use in order to get all the necessary elements for the CA to work. During the first simulation tests we noticed high processing times due to large quantity of calculations needed to perform one single simulation, in order to minimize this we implemented a parallelization strategy using the fork-join model in order to optimize the use of available hardware. The results obtained show a significant improvement in execution times in function of the number of available cores and map sizes, as a future work, it is planned to implement artificial neural networks in order to generate more complex urban growth scenarios. Keywords—Cellular automata; parallel programming; simulation models; OpenMP; urban growth I. INTRODUCTION The evolution in the land use of the territory is a fundamental element in our society, since it manifests different variables that affect our daily life, for example, accessibility to different points of interest within the city, slopes of the land, etc. This evolution has gained interest mainly fueled by the different environmental problems especially those in urban areas [1]. Thanks to the advances in the computing field and the development of important analytical tools such as Geographic Information Systems (GIS) or simulation models, the study of the changes taking place in metropolitan areas has been promoted [2]. The analysis of the environmental alterations that result from these changes and the development of new planning instruments, has caused that different disciplines, specifically the Artificial Intelligence (AI), approaches from a computer and mathematical point of view to give alternative solutions to this problem [3]. Numerous modeling tools have emerged in recent years. In the case of urban growth, the models based on cellular automata (CA) are the most widely used [4]. Regression models, artificial neural networks (ANNs), multi-criteria evaluation techniques (MCE), and still incipient, agent-based models (ABM) can also be found. The CA based models are oriented fundamentally towards the representation of the attributes of a given geographic region in a two-dimensional lattice, in which a neighborhood radius is defined and a certain rule of evolution is applied in order to define the behavior of the CA. With the use of these models it has been possible to generate territorial scenarios prospectively [5]. To generate these scenarios, a characterization of a CA is needed, this has different components, such as the size of the study area, maps of urban uses, map scales, neighborhood radius, evolution rules, slopes, and others geographical factors [6]. The developing of a CA based simulation tool to generate territorial scenarios prospectively in order to implement future simulation techniques, bring us to address some challenges. One of them was, the huge amount of mathematical operations needed in one single simulation, because the complexity of the algorithm to do such operations results to be exponential. One key calculus in the whole simulation process is, the transition potentials (TPs) of each cell in the map, these TPs show the probability of a cell to change from one state to another. The amount of these TPs have a direct impact on the computation cost needed to perform the mathematical calculations. To optimize these calculations, we enhanced sequential algorithms with parallelization strategies in order to maximize computational hardware. The library OpenMP (Open Multi- Processing), widely used in parallel programming, helps to implement a parallel strategy called fork-join. This allows to take advantage of hardware resources for the execution of processes in shared memory [7]. The present work aims to implement the fork-join strategy to speed up the necessary TPs calculations and to compare the results against the first sequential algorithm used in the simulation. The base maps for the experiments where generated from the study area of Culiacan, México. Being the faster growing city in the State of Sinaloa, we plan to use the simulation tool to understand the dynamics of the urban changes and to forecast for planning urban development as a future work. The remainder of this paper is structured as follows. All material and methods such as, the study area, digitation process, CA model and OpenMP are defined in Section II. Calculus of transition potentials for each pixel using the fork- join model are explained in Section III. Also proposed