Simulating urban land use change by integrating a convolutional neural network with vector-based cellular automata

Zhai, Yuling; Yao, Yao; Guan, Qingfeng; Liang, Xun; Xia, Li; Pan, Yongting; Yue, Hanqiu; Yuan, Zehao; Zhou, Jianfeng

doi:10.1080/13658816.2020.1711915

Cited by 97 publications

(46 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Although the BPNN can obtain the land suitability of the environment quickly and accurately, it does not quantitatively reveal the individual contributions of the driving factors to LUCC during modeling, such as logistical regression and other methods [23]. How to interpret the neural network model is still a difficult issue in the field of computer science and needs to be explored in future work [63].…”

Section: Discussionmentioning

confidence: 99%

Spatiotemporal LUCC Simulation under Different RCP Scenarios Based on the BPNN_CA_Markov Model: A Case Study of Bamboo Forest in Anji County

Huang

et al. 2020

IJGI

View full text Add to dashboard Cite

Simulating spatiotemporal land use and land cover change (LUCC) data precisely under future climate scenarios is an important basis for revealing the carbon cycle response of forest ecosystems to LUCC. In this paper, a coupling model consisting of a back propagation neural network (BPNN), Markov chain, and cellular automata (CA) was designed to simulate the LUCC in Anji County, Zhejiang Province, under four climate scenarios (Representative Concentration Pathway (RCP) 2.6, 4.5, 6.0, 8.5) from 2024 to 2049 and to analyze the temporal and spatial distribution of bamboo forests in Anji County. Our results provide four outcomes. (1) The transition probability matrices indicate that the area of bamboo forests shows an expansion trend, and the largest contribution to the expansion of bamboo forests is the cultivated land. The Markov chain composed of the average transition probability matrix could perform excellently, with only small errors when simulating the areas of different land-use types. (2) Based on the optimized BPNN, which had a strong generalization ability, a high prediction accuracy, and area under the curve (AUC) values above 0.9, we could obtain highly reliable land suitability probabilities. After introducing more driving factors related to bamboo forests, the prediction of bamboo forest changes will be more accurate. (3) The BPNN_CA_Markov coupling model could achieve high-precision simulation of LUCC at different times, with an overall accuracy greater than 70%, and the consistency of the LUCC simulation from one time to another also had good performance, with a figure of merit (FOM) of approximately 40%. (4) Under the future four RCP scenarios, bamboo forest evolution had similar spatial characteristics; that is, bamboo forests were projected to expand in the northeast, south, and southwest mountainous areas of Anji County, while bamboo forests were projected to decline mainly around the junction of the central and mountainous areas of Anji County. Comparing the simulation results of different scenarios demonstrates that 74% of the spatiotemporal evolution of bamboo forests will be influenced by the interactions and competition among different land-use types and other driving factors, and 26% will come from different climate scenarios, among which the RCP8.5 scenario will have the greatest impact on the bamboo forest area and spatiotemporal evolution, while the RCP2.6 scenario will have the smallest impact. In short, this study proposes effective methods and ideas for LUCC simulation in the context of climate change and provides accurate data support for analyzing the impact of LUCC on the carbon cycle of bamboo forests.

show abstract

Section: Discussionmentioning

confidence: 99%

Spatiotemporal LUCC Simulation under Different RCP Scenarios Based on the BPNN_CA_Markov Model: A Case Study of Bamboo Forest in Anji County

Huang

et al. 2020

IJGI

View full text Add to dashboard Cite

show abstract

“…Traditional CNN models perform well in image classification (He et al, 2016;Krizhevsky, Sutskever, & Hinton, 2012;Simonyan & Zisserman, 2014). In the field of urban study, CNNs have been used for the semantic segmentation of urban traffic scenes and land-use change analysis (Deng et al, 2017;Zhai et al, 2020). As shown in Fig.…”

Section: Cnn-based Urban Perception Modelmentioning

confidence: 99%

A comparison of two deep-learning-based urban perception models: which one is better?

et al. 2021

Self Cite

View full text Add to dashboard Cite

Urban perception is a hot topic in current urban study and plays a positive role in urban planning and design. At present, there are two methods to calculate urban perception. 1) Using a model to learn image features directly automatically; 2) Coupling machine learning and feature extraction based on expert knowledge (e.g. object proportion) method. With two typical streets in Wuhan as the study area, video data were recorded and used as the model input. In this study, two representative methods are selected: 1) End to end convolution neural network (CNN-based model); 2) Based on full convolution neural network and random forest (FCN + RF-based model). By comparing the accuracy of two models, we analyze the adaptability of the model in different urban scenes. We also analyze the relationship between CNN-based model and urban function based on POI data and OSM data, and verify its interpretability. The results show that the CNN-based model is more accurate than FCN + RF-based model. Because the CNN-based model considers the topological characteristics of the ground objects, its perception results have a stronger nonlinear correlation with urban functions. In addition, we also find that the CNN-based model is more suitable for scenes with weak spatial heterogeneity (such as small and medium-sized urban environments), while the FCN + RF-based model is applicable to scenes with strong spatial heterogeneity (such as the downtown areas of China’s megacities). The results of this study can be used as a reference to provide decision support for urban perception model selection in urban planning.

show abstract

“…In essence, each CNN layer identifies features at a particular level of detail. In combination, CNN can capture features at multiple scales in image recognition, street‐front assessment based on Google street views (Law et al ., 2020), traffic flow prediction (Ren et al ., 2019a), and land‐use change simulation (Zhai et al ., 2020). As such, scale becomes a new means to recognize the spatial organization.…”

Section: Four Kinds Of Tension Triggering Big Geographic Questionsmentioning

confidence: 99%

GIS research to address tensions in geography

Yuan

2020

Singap J Trop Geogr

View full text Add to dashboard Cite

The rise of Geographic Information Systems (GIS) technologies has been producing powerful tools for spatial data processing, management, analysis, modeling, and visualization. While supporting many tasks, GIS technologies have evoked new thinking and advanced intellectual inquiries in geography. Technological advances in other fields often stimulate new research questions and lead to revolutionary discoveries. The Hubble telescope revolutionizes our understanding of the universe, and 3D digital microscopes transform our knowledge of the coordination among biological, neurological, and physiological systems in living organisms. Can GIS claim similar revolutionary effects on geography? The answer is much up for debate. With GIS technologies, geographic studies can explore a broader extent across multiple scales in space and time and tackle problems through increasingly complex spatial statistics, visual analytics, computation, simulation, machine learning, and artificial intelligence. Both the Hubble telescope and 3D digital microscope were built based on scientific research that subserves the technological advances for inquiries into space and life systems. Likewise, GIScience research innovates GIS methods that enable novel geographic investigations and therefore contribute to geographic knowledge production. For the sake of simplicity, the term GIS used here represents both GIScience and GIS technologies. This essay attempts to clarify the intellectual contributions of GIS to geography on the following two questions: (1) What novel geographic thinking is driven by GIS? (2) How may GIS provoke new geographic inquiries and knowledge? Building on Nystuen's notion of four tensions that trigger geographic questions, the essay discusses how GIS innovations mediate historical tension, space‐time tension, dimensional tension and scale tensions.

show abstract

Simulating urban land use change by integrating a convolutional neural network with vector-based cellular automata

Cited by 97 publications

References 53 publications

Spatiotemporal LUCC Simulation under Different RCP Scenarios Based on the BPNN_CA_Markov Model: A Case Study of Bamboo Forest in Anji County

Spatiotemporal LUCC Simulation under Different RCP Scenarios Based on the BPNN_CA_Markov Model: A Case Study of Bamboo Forest in Anji County

A comparison of two deep-learning-based urban perception models: which one is better?

GIS research to address tensions in geography

Contact Info

Product

Resources

About