Effect analysis of road structure on multidimensional fragmentation of cultivated land:A case study in the international metropolis of Shanghai City*

LI Shiyao,CAI Yinying

(College of Public Administration,Huazhong Agricultural University,Wuhan 430070,China)

Abstract: Cultivated land fragmentation (CLF)is an important factor restricting agricultural scale production and modernization development.Recent studies have revealed that CLF is a multidimensional problem.A systematic study of the spatial distribution and driving factors of multidimensional cultivated land fragmentation (MCLF)has practical meaning for preserving high-quality farmland in suburbs and restraining urban land expansion.Taking the suburbs of the international metropolis of Shanghai City as an example,this article used canonical correlation analysis to calculate the effects of road structure on the MCLF in urban fringe,rural fringe and rural areas.The results showed that there was spatial heterogeneity in the degree of MCLF in the Shanghai suburbs in 2015.The size and distribution characteristics of CLF mainly occurred near the center of the main city,while the shape characteristics of CLF mainly occurred on the southeast coast.In addition,there was no positive correlation between road density and the degree of MCLF.Without the highest road density,the rural fringe had the highest level of distribution characteristics of CLF.Furthermore,the MCLF was affected by the road structure in heterogeneous zones.The higher density of the fourth-class road density represented a lower degree of the size characteristics of CLF in the urban fringe,and a higher density of the third-class road represented a higher degree of the distribution characteristics of CLF in the rural fringe.In rural areas,the area with a lower density of higher roads and a higher density of lower roads showed a lower CLF in terms of size characteristics and a higher CLF in terms of shape characteristics,while the area with a higher density of higher roads and a lower density of lower roads showed a higher CLF in terms of distribution characteristics.These discoveries enlightened us that to relieve the stress of cultivated land protection and achieve the goal of sustainable land use,we need to be explicit about the main challenges of regional cultivated land protection and consider the potential risks of road construction to the MCLF in the process of road planning and design.

Keywords: Road structure; Land fragmentation; Cultivated land protection; Canonical correlation analysis; Shanghai suburbs

The rapid progress of urbanization has led to the explosive growth of roads and other infrastructure.By 2050,there will be at least 25 million kilometers of newly constructed roads in the world,90% of which will be concentrated in developing countries,involving many important areas with biodiversity and high ecological value (Laurance et al.,2014).Existing studies have shown that the construction of roads has two sides.On the one hand,road expansion will promote economic growth,social integration and the flow of elements between urban and rural areas; on the other hand,the uncertain risks caused by road construction will damage the local ecological environment,result in loss of habitat and lead to landscape fragmentation (Laurance and Arrea,2017).As an important part of the ecosystem,the changes in quantity,quality and landscape structure of cultivated land will profoundly affect species diversity,ecosystem service values,and soil carbon sequestration(Jiang et al.,2019; Bian et al.,2022).Thus,studying the characteristics and causes of cultivated land change is an important prerequisite for socioeconomic sustainability and the coordinated development between agriculture and the environment.

Cultivated land fragmentation (CLF)is a common phenomenon in the process of socioeconomic transformation and is closely related to issues of food security and social stability.As early as 2008,scholars proposed using CLF as a system indicator for sustainable development monitoring (Jaeger et al.,2008),while the main method to deal with CLF is land consolidation (Bruschi et al.,2015; Hartvigsen,2015; Du et al.,2018; Jiang et al.,2019; Liu et al.,2019).Initially,the understanding of CLF was mainly limited to the dispersed land plots,considering this phenomenon to be rural basic space issues due to lack of planning (King and Burton,1982; McPherson,1982).At present,the most commonly used definition of CLF comes from the FAO’s interpretation modified by Sundqvist and Lisa in 2006,which views land fragmentation as a problem with spatial attributes that can be described in terms of farm size,the number of plots owned by a single ownership,the size and shape of a plot and the spatial distribution of the land parcels(Sundqvist and Lisa,2006; de Vries,2016).As the research continues,the understanding of CLF has been extended to the natural endowment,spatial characteristics and utilization aspects (Liu et al.,2019; Ntihinyurwa et al.,2019).Simultaneously,measurement methods related to CLF have also been enriched.From using the number of plots and the average cultivated area to measure the CLF,indices such as the Simmons index,Januszewski index and Igbozurike index were constructed for quantitative measurement (Demetriou et al.,2013),and then the spatial distribution and shape attributes were applied to describe CLF,such as the Schmook index and shape index (Schmook,1976; Chen et al.,2016).Most of these qualitative and quantitative analyses show that CLF leads to a series of negative impacts,including increasing production costs,reducing food production,and having adverse effects on collective action and governance of the common areas (Latruffe and Piet,2014; Ciaian et al.,2018; Laurance,2018; Looga et al.,2018; Qi and Dang,2018; Pili et al.,2019; Zang et al.,2019).To avoid and reduce the negative impacts of CLF,research on the driving factors of CLF is of great importance.Generally,the factors for CLF can be classified into two types.One is human activities,and the other is natural conditions.The former mainly includes the factors of social culture,land system and economic location,while the latter mainly includes the factors of soil quality,topographic conditions and geographic environment (Hartvigsen,2014; Gomes et al.,2019; Liu et al.,2019; Ntihinyurwa et al.,2019).

However,little attention has been given to the impact of roads on CLF.Studies show that approximately 80% of Earth’ s terrestrial surface remains roadless,but this area is fragmented into 600 000 patches,more than half of which are less than 1 km,resulting in adverse effects such as the loss of biological habitats,destruction of species diversity and fragmentation of landscape patterns (Ibisch et al.,2016; Assis et al.,2019; Dissanayake et al.,2019).Roads encroach on cultivated land resources by directly occupying cultivated land and indirectly attracting construction projects to occupy the cultivated land surrounding the roads,resulting in CLF,which is manifested in changes in the size,distribution and shape of cultivated land (Keken et al.,2014; Alphan,2017; Xiao et al.,2018; Fiedeń,2019).At present,research on the impacts of roads on CLF is mainly divided into three aspects.One is to use the method of overlay analysis to measure the segmentation effect of liner roads on the cultivated land (Ibisch et al.,2016).Another is to rely on aerial photographs to describe the changes in land use and cover (Fiedeń,2019).The third is to apply regression analysis to explore the causal relationships between road construction and CLF (Xiao et al.,2018;Mann et al,2021).Most of these studies have homogenized roads of various types,and few studies have focused on the relationship between road structure and CLF.In the meantime,CLF is a multidimensional problem; when referring to the analysis of the relationship between roads and CLF,different dimensions of CLF should also be taken into account.

Thus,this paper proposed a new idea to study the multidimensional relationship between road structure and CLF by using the canonical correlation analysis method.The specific objectives of this study were to 1)briefly describe the spatial differentiation characteristics of the multidimensional cultivated land fragmentation (MCLF)through representative indicators; 2)observe the relationship between road density and the degree of the MCLF through regional comparison; and 3)apply canonical correlation analysis to explore the effect of road structure on the MCLF in heterogeneous areas.The findings of this study can provide a basis for road construction and cultivated land protection.

1 Materials and methods

1.1 Study area and data sources

Shanghai was one of the first pilot cities to be opened up along the coast in 1984; since then,the acceleration of the urbanization process has led to the loss of a large amount of arable land (Xiao et al.,2018).After more than 30 years of development,Shanghai has become the center of economic,trade,shipping and technological innovation in China.showed that the GDP of Shanghai City increased from 39.09 billion Yuan in 1984 to 3815.53 billion Yuan in 2019,with an average annual growth rate of 13.98%,which created a total economic of 3.85% with a national land area of less than 0.07% in 2019.The rapid economic and urbanization progress has resulted in the urgent growth of demand for roads and other infrastructure.From 1984 to 2019,official statistics showed that the length of road transportation in Shanghai increased from 2136 km to 13 045 km,with an average annual growth rate of 5.31%.

Due to the advantages of geographical location and climatic conditions,Shanghai City possesses good conditions for agricultural production.Therefore,cultivated land resources had become one of the ultimate victims of social development in Shanghai City,where a large number of agricultural resources were occupied.According to the results of the land use change survey in 2015,there was 191 200 hmof cultivated land in Shanghai City,most of which is located in the undeveloped southern area.Compared with 1984,cultivated land decreased by 44.69%,which meant that nearly half of the cultivated land resources were lost.Under the background of the sharp contradiction between land demand and land supply in the new stage of social progress,the trade-off of cultivated land protection to road infrastructure construction and economic development must be reconsidered.Hence,the suburbs of Shanghai City,which had been undergoing a great impact of urbanization,were selected here as the typical research areas to explore the relationship between road infrastructure construction and CLF,hoping to provide suggestions for urban construction and cultivated land protection.

Because data accuracy varied from different data sources,this study calculated the data to the town level,which was the most basic administrative unit in Shanghai City.The specific data sources and related information are provided in Table 1.

Table 1 Data sources and descriptions

1.2 Multidimensional measurement of CLF

It is difficult to calculate CLF using metrics of a certain dimension.However,comprehensive metrics of measuring will result in the loss of information.Thus,to provide a more objective description of the situation from different angles,a multidimensional approach was adopted to measure CLF.Based on previous studies,the effective mesh size,Schmook index and areaweighted mean shape index were selected to measure CLF from the three dimensions of cultivated land size,plot distribution and parcel shape (Jaeger,2000;Moser et al.,2002; Moser et al.,2007; Demetriou et al.,2013).The calculation methods are presented in Table 2.

1.3 Road level classification

According to the traffic volume and usage attributes,roads can be divided into five levels:motorways,first-class roads,second-class roads,third-class roads,and fourth-class roads.The scale and service range of roads at all levels fall in order (Table 3).Because of the better management and service provision,the motorway is relatively independent of the outside,and the main impact on CLF lies in its spatial division.With the decline in the road level,the division effect weakens gradually.For the fourth-class roads,due to greater openness,the service targets of the fourth-class roads are mainly aimed at towns with rich arable land resources.Therefore,local residents can build houses around the road to improve the convenience of going out,which will lead to the occupation of the surrounding cultivated land resources.Thus,the main impact of the fourth-class roads on CLF lies in the dimension of shape.As the road level increases,the shape fragmentation of cultivated land weakens gradually.

1.4 Correlation analysis of road structure and MCLF

To study the relationship between road structure andCLF,other leading factors must be excluded as much as possible (Fig.1).Existing research has shown that there are two main reasons for CLF.One is human activities,and the other is natural conditions.Considering that there were relatively few differences in land inheritance and social systems in Shanghai,population density,GDP per unit area,distance from the main center of the city and distance from the subcenter of the city were selected to measure the impacts of human activities.In terms of natural conditions,the proportion of cultivated land area to town area,natural quality of cultivated land,slope and water network density were chosen as the main natural influencing factors of CLF (Hartvigsen,2014; Gomes et al.,2019; Liu et al.,2019; Ntihinyurwa et al.,2019).Definite descriptions are shown in Table 4.

Table 2 Indicators for assessing multidimensional cultivated land fragmentation (MCLF)

Table 3 Brief introduction of roads at all levels

Fig.1 Analysis framework of the impact of road structure on multidimensional cultivated land fragmentation(MCLF)

Table 4 Main factors influencing cultivated land fragmentation (CLF)

Performing the least-variance cluster analysis on the above mentioned influencing factors,zones with similar human activities and natural conditions could be obtained apart from road structures.Therefore,it was reasonable to believe that the analysis of the relationship between the road structure and MCLF was reliable when these external conditions were similar.A qualitative relationship could be gained if we conducted a descriptive statistical analysis between road structure and MCLF.Then,through quantitative canonical correlation analysis,the effect of the road structure on the MCLF could be discovered.

2 Results

2.1 Spatial differences in the MCLF

In the suburbs of Shanghai,there was 13.12 hmof cultivated land according to the land use change survey in 2015,which was mainly distributed in the southern region (Fig.2a).Using the methods mentioned in Table 3,we obtained the characteristics of size,distribution and shape of CLF in Shanghai suburbs and 107 towns under its jurisdiction,which still possessed cultivated land in 2015 (Fig.2,Table 5).The results showed that the effective mesh size,Schmook index and area-weighted mean shape index in Shanghai suburbs were 5800 m,0.22 and 1.47,respectively,which significantly showed a high degree of MCLF,especially in the size characteristics of CLF.On the township scale,approximately half of the towns were more fragmented in CLF size and distribution characteristics,and most of these towns were located in areas close to the main center or subcenter of the city,indicating that these kinds of fragmentation were closely related to the level of regional economic development.For the CLF shape characteristics,one-fifth of towns showed a higher degree of CLF shape characteristics,many of which were situated in the southeast coastal area.

Fig.2 Cultivated land resources and distribution of multidimensional cultivated land fragmentation (MCLF)in the Shanghai suburbs

Table 5 Spatial differences in the multidimensional cultivated land fragmentation (MCLF)at the town level in the Shanghai suburbs

2.2 Leading factors of CLF and zoning results

The main factors influencing CLF were divided into four grades by using the quartile method (Fig.3).It is easy to determine that human activities had similar dispersion around the main center and along the east coast area.The natural conditions caused distinct differences in distribution.Cultivated land resources were relatively abundant in the southern region of Shanghai,and the slope was much flatter in the southeastern region.However,the quality of cultivated land and the distribution of water network density were relatively scattered.Carrying out the least variance clustering analysis on these eight factors,three types of zones could be obtained,which contained the smaller square sum of the deviations in the same type of zones and larger square sum of the deviations among the different types of zones.We called these types of zones urban fringe,rural fringe and rural area (Fig.4,Table 6).The urban fringe was the closest zone to the city center,showing higher values in GDP per unit area and population density,and lower values in the proportion of cultivated land and water network density.Rural areas were quite different from what we called the urban fringe,which possessed the largest proportion of cultivated land and the gentlest sloped land,relatively lower values in GDP per unit area and population density,and the furthest distance from the city center.For the rural fringe,which was located between the urban fringe and rural area,the values of indices such as the proportion of cultivated land,GDP per unit area of land,population density and slope,were also basically between those of the urban fringe and rural area.

Fig.3 Calculation results of the main influencing factors of cultivated land fragmentation (CLF)in the Shanghai suburbs

Fig.4 Zoning result of cultivated land fragmentation(CLF)in the Shanghai suburbs according to the main influencing factors

Table 6 Features of cultivated land in different zones of cultivated land fragmentation (CLF)in the Shanghai suburbs

2.3 Impact of road structure on MCLF

Based on the zoning results,the statistics of the road structure and MCLF indices in different zones are illustrated in Fig.5 and Fig.6.From Fig.5,it can be seen that the road structure differed among these three zones.From the urban fringe to the rural fringe and the rural area,the road density decreased from 4.35 to 3.60 and 2.68 km·km,respectively.The density of first-class roads,second-class roads and third-class roads decreased in turn,while the density of fourth-class roads increased.The density of third-class roads occupied the largest proportion in the urban fringe,which accounted for 48.16% of the road density in this area,followed by fourth-class roads,accounting for 20.65%,and first-class roads,accounting for 15.34%.The density of fourth-class roads occupied the largest proportion in the rural areas,accounting for 55.39% of the road density in this area,followed by third-class roads,accounting for 21.25%.In the rural fringe,the density proportions of third-class roads and fourth-class roads were equivalent,accounting for 37.02% and 35.27%,respectively,followed by that of second-class roads with a proportion of 12.16%.The statistics in Fig.6 show that the rural area obtained the best evaluation results,as the effective mesh size,Schmook index and area-weighted mean shape index values in the rural area were 7951.86 m,0.31 and 1.48,respectively.However,not all indices that measured the degree of MCLF improved in the order of urban fringe,rural fringe and rural area.Unsurprisingly,the urban fringe had the worst evaluation in terms of the size and shape characteristics of CLF,the values of effective mesh size and area-weighted mean shape index of which were 298.53 mand 1.72,respectively.However,the distribution characteristics of CLF in the rural fringe received the worst evaluation,with a Schmook index of 0.12.

Fig.5 Road structure in different zones of cultivated land fragmentation (CLF)in the Shanghai suburbs

Fig.6 Indicators in different zones of cultivated land fragmentation (CLF)in the Shanghai suburbs

Furthermore,canonical correlation analysis was utilized to explore the effect of road structure on MCLF(Table 7),and we obtained three pairs of canonical coefficients in each zone.In the urban fringe and rural fringe,only the first pair of canonical coefficients passed the significance test,while all three pairs of canonical coefficients passed the significance test in the rural areas.According to the criterion that the cumulative contribution rate should be between 70% and 85%,the first pair of typical variables was selected in the urban fringe and the rural fringe,and the first and the second pairs of typical variables were selected in the rural area.The standardized canonical coefficients for road structure and MCLF are shown in Table 8 and Table 9.From the standardized canonical coefficients,we could see that there were differences in the effect of road structure on MCLF in heterogeneous areas.In the urban fringe,the density of fourth-class roads was related to the CLF size character-istics.The higher the density of fourth-class roads is,the lower the degree of CLF in size characteristics.In the rural fringe,the density of third-class roads was related to the CLF distribution characteristics.The higher the density of third-class roads is,the higher the CLF in distribution characteristics.In rural areas,the relationship between the road structure and the MCLF is much more complicated.The area with a lower density of motorways,first-class roads and second-class roads and a higher density of third-class roads and fourth-class roads showed a lower CLF size characteristics and a higher CLF shape characteristics,while the area with a higher density of second-class roads and third-class roads and a lower density of fourth-class roads showed a higher CLF distribution characteristics.

Table 7 Canonical correlation and significance test

Table 8 Standardized canonical coefficients for road structure

Table 9 Standardized canonical coefficients for MCLF

3 Discussion

3.1 Relationship between road structure and MCLF in Shanghai suburbs

As the precedent and succeeding infrastructural activity in the urbanization process,road construction brings about landscape and ecological risks such as CLF(Liu et al,2008; Mann et al,2021).The study aimed to determine the effect of the road structure on the MCLF and provided a new idea to explore the connection between them.This study systematically summarized external factors of human activities and natural conditions that affect CLF on the basis of existing research and used the method of cluster analysis to divide the suburbs of Shanghai City into urban fringe,rural fringe and rural areas with similar external conditions.Then,the degree of CLF was calculated in three dimensions of size,distribution and shape on the parcel scale.Finally,a canonical correlation analysis was performed between the MCLF and road structure in heterogeneous zones,so the effect of road structure on the MCLF could be obtained.

Most existing studies suggest that road construction worsened the situation of CLF,but few have studied the relationship between road structure and CLF in heterogeneous areas (Alphan,2017; Fiedeń,2019; Karlson and M?rtberg,2015; Keken et al.,2014; Laurance and Balmford,2013; Mehdipour et al.,2019).Compared with previous studies,we found that road density did not always facilitate CLF in all forms,which might not have been given enough attention in past research.From the urban fringe to the rural fringe and then to the rural area,the CLF size and shape characteristics gradually improved with decreasing road density,while the CLF distribution characteristics in the rural fringe were more dispersed than those in the urban fringe.Another discovery was that the road structure in heterogeneous zones showed different effects on the MCLF.There was a negative correlation between the density of fourth-class roads and the CLF size characteristics in the urban fringe,while a positive correlation existed between the density of third-class roads and the CLF distribution characteristics in the rural fringe.In rural areas,areas with a lower density of motorways,first-class roads and second-class roads and a higher density of third-class roads and fourthclass roads showed lower CLF size characteristics and higher CLF shape characteristics,while areas with a higher density of second-class roads and third-class roads and a lower density of fourth-class roads showed higher CLF distribution characteristics.

3.2 Policy implications

To reduce the pressure on farmland protection and achieve sustainable use of farmland,it is necessary to adapt to the local environment,comprehensively consider different external conditions,and analyze the potential CLF risks from building new roads in the future.In the urban fringe,road planning and design should try to avoid areas with a high density of fourth-class roads to avoid further increasing CLF size characteristics.In the rural fringe,road planning and design should focus on the problem of CLF in distribution and avoid the increase in fragmentation due to newly built roads.In rural areas,special attention should be given to road planning and design,because the lower density of the higher class of roads and the higher density of the lower class of roads may increase CLF shape characteristics,and the higher density of the higher class of roads and the lower density of the lower class of roads may increase CLF distribution characteristics.Additionally,the rights and responsibilities of road construction authorities and other departments should be enhanced to address the problems that may happen after road construction.

For the point of cultivated land protection,analyzing the relationship between the road structure and MCLF in a heterogeneous area can clarify the key points of regional cultivated land protection,and take preventive measures to protect cultivated land in advance to reduce the probability of land fragmentation.In the urban fringe,attention should be given to CLF size characteristics,and the measures of combining agriculture with tourism,mixing the agricultural industry with villages and interacting agricultural production and management with the three industries can be taken to increase the comparative returns of agricultural production.In the rural fringe,we need to take measures to develop characteristic agriculture and meet the needs of the surrounding market to address the situation of CLF distribution characteristics.In rural areas,we should consider the problem of CLF shape characteristics,and solutions such as land consolidation can be implemented to achieve moderatescale management of agricultural land and satisfy the demand of modern agricultural development.

3.3 Limitations and research prospects

CLF is a complex process that is influenced by human activities,including social culture,land system and economic location,as well as by natural conditions,including soil quality,topographical conditions,and geographical environment (Liu et al.,2019).The zoning indices selected in this article are only applicable at small scales.For a wider range of studies,social culture and even land system factors need to be taken into consideration.At the same time,under the different social and cultural backgrounds and land system forms,there will be differences in the form of CLF,which requires further construction of dimensions in CLF (Ntihinyurwa et al.,2019).In addition,this paper focuses on the relationship between road structure and MCLF,without considering the spatial distribution of roads.To further verify the relationship between them,it is necessary to take the spatial and temporal change data into consideration to reveal these kinds of internal connections.

4 Conclusions

Infrastructure construction in the process of urbanization has become one of the major threats to sustainable development,so it is of great significance to implement advanced road traffic planning (Bovet et al.,2018).We believe that an insufficient focus on the road structure impacts of CLFs on a multidimensional scale results in inadequate protection of cultivated land.Thus,in this study,we used the method of zoning comparison and canonical correlation analysis to explore the correlation between road structure and MCLF.The results showed that the road density did not always facilitate CLF in all forms,and the effect of road structure on MCLF differed in heterogeneous zones.From urban fringe to rural fringe and to rural areas in the Shanghai suburbs,with the increasing density of fourth-class roads and the decreasing density of first-class roads,second-class roads and thirdclass roads,the size and shape characteristics of CLF gradually improved in turn,while the CLF distribution characteristics in rural fringe was more dispersed than that in urban fringe.These findings will help us gain a deeper understanding of the CLF phenomenon and guide the planning and construction of road infrastructure to better implement cultivated land protection policies.