Quantifying Changes of Villages in the Urbanizing Beijing Metropolitan Region: Integrating Remote Sensing and GIS Analysis

Abstract

Abstract:Rapid urbanization has resulted in great changes in rural landscapes globally. Using remote sensing data to quantify the distribution of rural settlements and their changes has received increasing attention in the past three decades, but remains a challenge. Previous studies mostly focused on the residential changes within a grid or administrative boundary, but not at the individual village level. This paper presents a new change detection approach for rural residential settlements, which can identify different types of rural settlement changes at the individual village level by integrating remote sensing and Geographic Information System (GIS) analyses. Using multi-temporal Landsat TM image data, this approach classifies villages into five types: "no change", "totally lost", "shrinking", "expanding", and "merged", in contrast to the commonly used "increase" and "decrease". This approach was tested in the Beijing metropolitan area from 1984 to 2010. Additionally, the drivers of such changes were investigated using multinomial logistic regression models. The results revealed that: (1) 36% of the villages were lost, but the total area of developed lands in existing villages increased by 34%; (2) Changes were dominated by the type of 'expansion' in 1984-1990 (accounted for 43.42%) and 1990-2000 (56.21%). However, from 2000 to 2010, 49.73% of the villages remained unchanged; (3) Both topographical factors and distance factors had significant effects on whether the villages changed or not, but their impacts changed through time. The topographical driving factors showed decreasing effects on the loss of rural settlements, while distance factors had increasing impacts on settlement expansion and merging. This approach provides a useful tool for better understanding the changes in rural residential settlements and their associations with urbanization.

Quantifying Urban Land Sprawl and its Driving Forces in Northeast China from 1990 to 2015

“…The overall accuracy of the 2015 LULC map was 93%, and that for the settlements was 95%. [44,45]. More details about the classification approach can be found in previous studies [44,45].…”

“…The overall accuracies of these three classification maps were 86% for 1990, 86% for 2000, and 88% for 2010, and that of settlements in the three years were 90%, 91% and 93%, respectively. backdating approach with change vector analysis [44,45]. More details about the classification approach can be found in previous studies [44,45].…”

“…[44,45]. More details about the classification approach can be found in previous studies [44,45]. Accuracy assessment was also completed for the LULC classification maps in 1990, 2000, and 2010 using historical field survey points and Google Earth images, as well as visual interpretation of the Landsat TM data as reference data.…”

“…An backdating approach, as a synthesis of the post-classification comparison and pre-classification change detection, typically started with the reference map, based on which the classification and change analysis are conducted [43], which conducted classification only at locations with changes and maintained the consistency of the features with on changes [44,45]. In order to reduce the 'salt and pepper' effect [44], the backdating approach integrated with an object-based method was used for Land Use/Land Cover (LULC) classifications.…”

“…Specifically, urban land was extracted in two steps [45]: (1) The class objects "settlements" were selected from the LULC maps as potential urban land; (2) The urban land for 1990, 2000, 2010 and 2015 were separated from villages using the vector layer of points of cities, counties, towns and villages. …”