Predicting developed land expansion using deep convolutional neural networks. (December 2020)

• Record Type: Journal Article
• Authors: Pourmohammadi, P.; Adjeroh, D.A.; Strager, M.P.; Farid, Y.Z.
• Is Part Of: Environmental modelling & software. Volume 134(2020)
• Publication Date: 2020-12
• Subjects: Developed land expansion -- Pixel-wise segmentation -- Convolutional neural network; 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.2020.104751
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Many aspects of land-use management and policy making require information regarding how and where land cover and land-uses will change in the future. In this research, we propose a method for modeling and predicting developed land expansion using the idea of pixel-wise semantic segmentation through deep convolutional neural networks. This analysis is done on a watershed scale with a focus on where developed lands are predicted to expand. We introduce a method to construct data cubes of the land patches which represent important information related to diverse characteristics of the area under consideration. We model the developed land expansion using an encoder-decoder network, and then perform prediction using a simple sigmoid layer. Our results indicate a performance accuracy of 98% on the test data. The proposed technique could thus play an important role in improving our understanding, mapping, and modeling of spatially explicit landscape changes, and in facilitating land-use decision making. Highlights: Spatial data representation, based on multispectral data cubes. Application of deep convolutional pixel-wise segmentation for land change modeling and prediction. Application of Graphics Processing Units (GPUs) from the Pittsburgh Supercomputing Center (PCS) to model large multispectral data cubes.