Abandoned rural residential land: Using machine learning techniques to identify rural residential land vulnerable to be abandoned in mountainous areas

Feng Xu, Hung Chak Ho, Guangqing Chi, Zhanqi Wang

Research output: Contribution to journalArticle

Abstract

Rural residential land has been increasingly abandoned in China given the rapid and massive rural-to-urban migration. From the aspect of land-use planning and policy making, it is important to understand the determinants of residential land abandonment across rural areas as well as to know what rural residential land is vulnerable to be abandoned. However, neither of these can be known via qualitative evaluation of residential land abandonment or via remote sensing applications and land-use modelling. In this study, we develop an approach of combining machine learning techniques (Random Forest, Supported Vector Machine, and Naive Bayes) and land-as-an-object analysis to identify the rural residential land that has a high possibility of being abandoned in mountainous areas. We applied this approach to Fang County, Central China. The results indicate a reasonable and reliable prediction of rural residential land abandonment based on our approach, particularly in estimating the potential occurrence of local land abandonment. The geographic characteristics of the land and the living conditions of the land user were found to have relatively significant impacts on rural residential land use. Our approach also provides a pathway to evaluate specific land use and identify its potential change. This approach can be useful for the development of a standardized protocol for the evaluation of residential land abandonment across other rural areas and may be applicable for the investigation of other land types that may be vulnerable to abandonment. Quantitatively assessing the influential factors of residential land use can also provide alternative insights for the development of planning protocols in order to ultimately improve the quality of life and living environments in rural areas.

Original languageEnglish (US)
Pages (from-to)43-56
Number of pages14
JournalHabitat International
Volume84
DOIs
StatePublished - Feb 1 2019

Fingerprint

learning
land use
rural area
machine learning
land
mountainous area
land type
China
planning
land use planning
quality of life
policy making
living conditions
evaluation
remote sensing
migration
determinants
prediction
modeling
protocol

All Science Journal Classification (ASJC) codes

  • Urban Studies

Cite this

@article{c9a3a421ef79460eb1f9173650345d45,
title = "Abandoned rural residential land: Using machine learning techniques to identify rural residential land vulnerable to be abandoned in mountainous areas",
abstract = "Rural residential land has been increasingly abandoned in China given the rapid and massive rural-to-urban migration. From the aspect of land-use planning and policy making, it is important to understand the determinants of residential land abandonment across rural areas as well as to know what rural residential land is vulnerable to be abandoned. However, neither of these can be known via qualitative evaluation of residential land abandonment or via remote sensing applications and land-use modelling. In this study, we develop an approach of combining machine learning techniques (Random Forest, Supported Vector Machine, and Naive Bayes) and land-as-an-object analysis to identify the rural residential land that has a high possibility of being abandoned in mountainous areas. We applied this approach to Fang County, Central China. The results indicate a reasonable and reliable prediction of rural residential land abandonment based on our approach, particularly in estimating the potential occurrence of local land abandonment. The geographic characteristics of the land and the living conditions of the land user were found to have relatively significant impacts on rural residential land use. Our approach also provides a pathway to evaluate specific land use and identify its potential change. This approach can be useful for the development of a standardized protocol for the evaluation of residential land abandonment across other rural areas and may be applicable for the investigation of other land types that may be vulnerable to abandonment. Quantitatively assessing the influential factors of residential land use can also provide alternative insights for the development of planning protocols in order to ultimately improve the quality of life and living environments in rural areas.",
author = "Feng Xu and Ho, {Hung Chak} and Guangqing Chi and Zhanqi Wang",
year = "2019",
month = "2",
day = "1",
doi = "10.1016/j.habitatint.2018.12.006",
language = "English (US)",
volume = "84",
pages = "43--56",
journal = "Habitat International",
issn = "0197-3975",
publisher = "Elsevier Limited",

}

Abandoned rural residential land : Using machine learning techniques to identify rural residential land vulnerable to be abandoned in mountainous areas. / Xu, Feng; Ho, Hung Chak; Chi, Guangqing; Wang, Zhanqi.

In: Habitat International, Vol. 84, 01.02.2019, p. 43-56.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Abandoned rural residential land

T2 - Using machine learning techniques to identify rural residential land vulnerable to be abandoned in mountainous areas

AU - Xu, Feng

AU - Ho, Hung Chak

AU - Chi, Guangqing

AU - Wang, Zhanqi

PY - 2019/2/1

Y1 - 2019/2/1

N2 - Rural residential land has been increasingly abandoned in China given the rapid and massive rural-to-urban migration. From the aspect of land-use planning and policy making, it is important to understand the determinants of residential land abandonment across rural areas as well as to know what rural residential land is vulnerable to be abandoned. However, neither of these can be known via qualitative evaluation of residential land abandonment or via remote sensing applications and land-use modelling. In this study, we develop an approach of combining machine learning techniques (Random Forest, Supported Vector Machine, and Naive Bayes) and land-as-an-object analysis to identify the rural residential land that has a high possibility of being abandoned in mountainous areas. We applied this approach to Fang County, Central China. The results indicate a reasonable and reliable prediction of rural residential land abandonment based on our approach, particularly in estimating the potential occurrence of local land abandonment. The geographic characteristics of the land and the living conditions of the land user were found to have relatively significant impacts on rural residential land use. Our approach also provides a pathway to evaluate specific land use and identify its potential change. This approach can be useful for the development of a standardized protocol for the evaluation of residential land abandonment across other rural areas and may be applicable for the investigation of other land types that may be vulnerable to abandonment. Quantitatively assessing the influential factors of residential land use can also provide alternative insights for the development of planning protocols in order to ultimately improve the quality of life and living environments in rural areas.

AB - Rural residential land has been increasingly abandoned in China given the rapid and massive rural-to-urban migration. From the aspect of land-use planning and policy making, it is important to understand the determinants of residential land abandonment across rural areas as well as to know what rural residential land is vulnerable to be abandoned. However, neither of these can be known via qualitative evaluation of residential land abandonment or via remote sensing applications and land-use modelling. In this study, we develop an approach of combining machine learning techniques (Random Forest, Supported Vector Machine, and Naive Bayes) and land-as-an-object analysis to identify the rural residential land that has a high possibility of being abandoned in mountainous areas. We applied this approach to Fang County, Central China. The results indicate a reasonable and reliable prediction of rural residential land abandonment based on our approach, particularly in estimating the potential occurrence of local land abandonment. The geographic characteristics of the land and the living conditions of the land user were found to have relatively significant impacts on rural residential land use. Our approach also provides a pathway to evaluate specific land use and identify its potential change. This approach can be useful for the development of a standardized protocol for the evaluation of residential land abandonment across other rural areas and may be applicable for the investigation of other land types that may be vulnerable to abandonment. Quantitatively assessing the influential factors of residential land use can also provide alternative insights for the development of planning protocols in order to ultimately improve the quality of life and living environments in rural areas.

UR - http://www.scopus.com/inward/record.url?scp=85059309207&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85059309207&partnerID=8YFLogxK

U2 - 10.1016/j.habitatint.2018.12.006

DO - 10.1016/j.habitatint.2018.12.006

M3 - Article

VL - 84

SP - 43

EP - 56

JO - Habitat International

JF - Habitat International

SN - 0197-3975

ER -