The land transformation model-cluster framework: Applying k-means and the Spark computing environment for large scale land change analytics. (January 2019)

• Record Type: Journal Article
• Authors: Omrani, Hichem; Parmentier, Benoit; Helbich, Marco; Pijanowski, Bryan
• Is Part Of: Environmental modelling & software. Volume 111(2019)
• Page Start: 182
• Page End: 191
• Publication Date: 2019-01
• Subjects: Clustering -- Parallel processing -- Spark environment -- Land use change; LTM land transformation model -- LTM-cluster land transformation model-cluster framework -- HPC high-performance computing -- ML machine learning -- ANN artificial neural network -- LCS land change science -- PCM percent correct match -- TOC total operating characteristic -- AUC area under the curve -- ER error rate -- MSE mean square error; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2018.10.004
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: This study introduces a novel framework for land change simulation that combines the traditional Land Transformation Model (LTM) with data clustering tools for the purposes of conducting land change simulations of large areas (e.g., continental scale) and over multiple time steps. This framework, called "LTM-cluster", subsets massive land use datasets which are presented to the artificial neural network-based LTM. LTM-cluster uses the k -means clustering algorithm implemented within the Spark high-performance compute environment. To illustrate the framework, we use three case studies in the United States which vary in simulation extents, cell size, time intervals, number of inputs, and quantity of urban change. Findings indicate consistent and substantial improvements in accuracy performance for all three case studies compared to the traditional LTM model implemented without input clustering. Specifically, the percent correct match, the area under the operating characteristics curve, and the error rate improved on average of 9%, 11%, and 4%. These results confirm that LTM-cluster has high reliability when handling large datasets. Future studies should expand on the framework by exploring other clustering methods and algorithms. Highlights: We addressed the challenges of simulating land change across large regions and spans of time. Our model, called LTM-Cluster, is a scalable modeling framework aimed to benefit research involving large datasets. The Spark environment is used to reduce the burden of high computational time when handling a huge amount of data. The implementation of LTM-Cluster is available free-of-charge. Our results consistently showed significant performance improvements of the LTM-Cluster compared to the traditionally parameterized LTM.