The land use transfer matrix approach obtains a two-dimensional matrix through the generation of a land use transfer path, and explains this path concerning the structural characteristics of land use change and the direction of transfer between types (Li C et al., 2021). In particular, the different land use types are identified by codes in a fixed order (1, 2 and 3). For example, the change path of cropland to “forest and grassland” and “forest and grassland” can be expressed as 1-2-2. Finally, a Sankey diagram was created to visualize different land transfer paths from 1980 to 2020.

The Space Time cube is a three-dimensional cube proposed by Hagerstrand (1968). A two-dimensional plane is used to express the spatial location and region, and another dimensional time axis expresses the temporal change in the location of the two-dimensional plane (Figure 2) (https://desktop.arcgis.com/). The analysis process includes the identification of land use cold spots and hot spots in two dimensions, the creation of a Space Time Cube, and a statistical test for cold and hot spots, as follows:

1) Land Use Hotspot Identification. The accuracy of the land use data, the size of the study area and the visualization of the data were all taken into account in this study. After a preliminary test to identify the most suitable scale (data not presented), a 600 m * 600 m grid was selected as the study unit. The area of land use change at different times within each unit was first obtained, and then the Getis-Ord Gi* tool was used to calculate and compare the local sum of adjacent elements within each grid with the sum of all elements, thus allowing the analysis of the local spatial aggregation of area change at different times.

All trend analysis operations were completed in ArcGis 10.7.

The Geodetector method (http://www.geodetector.org/) is based on statistical principles for spatial variance analysis. It represents a set of statistical methods to detect spatial variance and its mechanisms. This method is mainly applied to detect the spatial variance and to what extent factor explains the spatial variance of attribute. q = 1 − 1 N σ 2 ∑ h = 1 L N h σ h 2 Where q is the spatial variance of a factor indicator, which is used to detect the factor’s influence on land use activities. N is the sample size of the whole region. σ 2 is the variance of land change in the whole region. h is the identification partition, h = 1, 2, ..., L, where L denotes the number of partitions. 0 ≤ q≤ 1, the larger the value of q, the stronger the factor’s influence on land activities. Conversely, the smaller the value of q, the weaker the influence. A q of 0 indicates no spatial variance in the study area and the spatial distribution of land use is random.

In this paper, 18 townships were used as detection areas. Based on the existing studies and data accessibility, the accumulated area of land change in each township between 2000 and 2020 as a proportion of its administrative area was Y. The socioeconomic factors (change in GDP X1, change in the agricultural output value X2, change in the gross industrial output value X3, and change in the road network density X4) and natural environment (DEM X5, slope X6, average annual temperature change X7, and average annual precipitation change X8), for a total of eight factors. Since the statistical objects of socioeconomic data changed considerably from 1980 to 2000, this study also took the local development history into account to provide a context for the identified land activity mechanisms.

Baiyun District is located in the transition zone between the low mountains of central Guangdong Province and the Pearl River Delta Plain. The north and northeast of Baiyun District are characterized by large hills and mountains, preserving large areas of pristine ecological forest. In the west and south, there are flat, broad alluvial plains with fertile soil and various soil types and genera, providing favorable conditions for a number of plantations (rice, vegetables, drought-tolerant cereals, peanuts, forestry). Therefore, since the reform and opening up, industrialization and urbanization were delayed in Baiyun District compared with the main urban area of Guangzhou. Traditional agricultural activities still dominate the study area, forming a stable spatial pattern of natural spaces and large-scale agricultural production. In the early stage, under the premise of not changing land use, the farmer families contracted the land to the collective and enjoyed full management autonomy as the primary production unit. As a result, the rural economy grew rapidly and provided sufficient labor and land for subsequent industrialization (Baiyun District Information Database in Guangzhou, China, 2021).

Since 1990, with the second wave of global industrial transfer and the promotion of the second round of the contract management responsibility system, low-end industries have become the target of foreign investment, and a large amount of rural collective land has been expropriated or illegally occupied (Long, 2012; Gong et al., 2019). At the same time, industrial production has also created a large number of employment opportunities. Local farmers have entered the workforce, and migrants from the surrounding cities and counties around Guangzhou have moved in to work (Yang, 2019). This has led to a boom in the informal expansion of self-housing and external leasing in remote areas. The real estate planned economy (1990-1995) and the real estate market economy (1996-2000) directly contributed to the large-scale construction of commercial housing in the southern part of Baiyun District (Su et al., 2005). Therefore, the 1990-2000 presented in Figure 4 also verifies the construction development in the north and south of Baiyun District. In addition, The “rural land capitalization” of Baiyun District has led to the rapid shrinkage of cropland area. However, the shortage of cropland in the urban area of Guangzhou and the timely open policy of purchasing and selling agricultural products caused Baiyun District to become the main grain supply source area for urban Guangzhou. At the same time, Baiyun District also became prominent in the northern transportation and export of agricultural products. As a result, the agricultural production space gradually shifted from the suburbs to the middle and outer suburban agricultural areas (Baiyun District Information Database in Guangzhou, China, 2021). The land space dominated by traditional agriculture in Baiyun District reduced rapidly, and the fragmentation of land development intensified. Spatially, there was a large area of built-up land in the south and fragmentation of land in the center and north.

Since the turn of the 21st century, the global economy has been developing in an integrated way, with industrial adjustment and upgrading and international transfer occurring in parallel (Yang et al., 2020). Baiyun District, which used to rely on traditional agriculture and low-end industries, was facing unprecedented challenges. Under the government’s leadership, there was a massive transfer of agricultural production space to urban residential and public built-up land in 2000–2010 relative to previous periods. The built-up land expanded from the south to the center of the district, and the urban villages formed in a low-cost way, outside of the modern urban management, and the land uses were highly mixed (Yang, 2019; Yao and Tian, 2020). Driven by the “Strong Industrial Zone,” “Modern Logistics Development Plan” and “Optimized Northern Development” polices, a large number of professional markets, logistics parks or industrial agglomeration areas were built the central and western regions and the northern part of remote areas. Thus, as shown in Figure 4, the expansion of other construction land (industrial land) from 2000 to 2010 was particularly significant relative to other periods. This also validates the significant role of gross industrial output as the second influencing factor on land use change. This construction was reliant on the transfer of rural collective land, which had the advantage of cheaper prices than urban areas (Baiyun District Information Database in Guangzhou, China, 2021). As a result, a large number of laborers from the surrounding inland provinces of Guangdong Province were absorbed. Since 2000, in response to the requirements of the tenth Five-Year Plan for developing the national economy, the north of Renhe Town in Baiyun District developed large-scale infrastructure land to construct the new Baiyun International Airport (Yang et al., 2020). In general, under the government’s coordination, the land activities during this period showed a significant trend of spatially pushed development from the middle to the north.

The rapid population growth and urban expansion in the first decade of the 21st century led to the shortage of new built-up land and reserve land for further development (Gong et al., 2018a). In 2009, Baiyun District government mapped idle land, low-utility land, old factories and hillsides to supplement cropland (Baiyun District Information Database in Guangzhou, China, 2021). Under the premise of maintaining the current land use, the southern part of Baiyun District carried out the reconstruction of urban villages, evacuation of villages to build houses or revitalization of the stock of built-up land (Li et al., 2019). Thanks to better transport facilities and lower land rents, the central and northern parts took over the urban capital from the south (Gong et al., 2018b). In central Baiyun District, the old factories and historical sites were rebuilt into a fine chemical industrial zone, innovation industrial zone or science and technology park in the process of the “Exit the secondary industry and enter the tertiary industry” policy, and develop an “industrial clustering” regional model (Lin et al., 2019). The northern part focused on implementation as a land reserve planning region. Since 2010, Baiyun District has seen a decrease in the overall land-use changes under the land-intensive approach of limiting incremental land and excavating stock land. However, under the guidance of local and national planning divisions, the land functions have diversified, leading to a spatial pattern of ecological tourism in the east, Baiyun New City in the south, a creative industry gathering area in the middle, an industrial logistics area in the west, and an airport economic zone in the north of the district (People’s Government of Baiyun District, Guangzhou City, 2022a; People’s Government of Baiyun District, Guangzhou City, 2022b).

Our research results confirm that the trend turning of land use morphology can integrate land use morphology, use function, land use effect and the mechanism of land use change, and systematically sort out land use transition. Meanwhile, this study innovatively combines the history, policy measures and socio-economic development of the study area for a time-phased refinement analysis, which further confirms that land use transition in urban suburbs is a spatiotemporal evolution process of a complex man-land relationship. It is also the most direct result of multiple forces such as physical geography and socioeconomic conditions on land use morphology (Gong et al., 2019; Yao and Tian, 2020; Cao et al., 2021). The unique natural environment has shaped the original farming style, and the continuous improvement of the agricultural business system after the founding of New China has injected continuous momentum into agricultural production (Li H et al., 2021; Zhou et al., 2021). Since the reform and opening up, the original dual/dualistic structure has collapsed/subsequently disintegrated, and the urban suburbs have gradually developed on the periphery of the traditional rural core (Yang, 2019). The combination of foreign capital and local resource liberalization has contributed to the conversion of land use morphology from hybridization to large-scale development. The limited land resources are gradually formed into the regional space of specific function positioning types under the rational allocation planning (Gong et al., 2018a). However, urban suburbs continue to be faced with the problem of limited land resources. Even under the multiple constraints of the “Controlling growth and forcing storage” policy, the problems of mixed tenure, spatial fragmentation, poor living environment and industrial inefficiency continue to persist (Gong et al., 2019; Yang, 2019; Yao and Tian, 2020; Cao et al., 2021). The recent construction of the Guangdong-Hong Kong-Macao Bay Area and the national strategy of rural revitalization have brought new opportunities and challenges to the urban suburbs of the Pearl River Delta (Gong, et al., 2019). However, it remains unclear how the new circumstances could help to solve the problem of land use transition and stimulate the vitality of land use development in the urban suburbs. In addition, the reshaping of the position of urban suburbs in the urban-rural relationship and consideration of how the urban suburbs should act as a bridge between urban and rural area need to be topics of further research.

With the accumulation of spatiotemporal data and the advent of the big data era, feature mining of spatiotemporal data with geology as the object of analysis with statistical methodology has brought a new opportunity for land use research. For example, it can solve the problems of massive data analysis and visualization on different spatial units and in different periods (Wang et al., 2014). Moreover, it can also meet the needs of attribute diversification, function differentiation and mechanism diversification, and realize long time series of evolutionary laws, interactive dynamics and prediction research. The spatiotemporal feature mining method selected in the current study is an effective coupling of spatiotemporal statistics and land system science. The method focuses on the breakthrough of land use research from traditional two-dimensional plane space to the realization of three-dimensional spatial visualization, which is a fundamental attempt to study the mono-temporal land use morphology to long-temporal land use transition (Esri, 2021; Wang et al., 2010). As in this study, the cold hotspot analysis of land activities breaks through the traditional two-dimensional land transfer, effectively visualizing the multidimensional land use morphology and revealing the dominant land use morphology. Based on this approach, we deeply explored the changes of recessive land use morphology, for example, to reveal the process of land substitution caused by the changes of socioeconomic development stages (such as industrial transfer, infrastructure projects) and the real problems derived from it. However, this study considered a combination of existing spatiotemporal data analysis methods, which continue to have associated problems such as the scale disunity of the source data, the lack of data attributes and the required improvement to the detection factor system. In addition, the uneven distribution of data acquisition at large spatial scales leads to a lack of spatiotemporal slice information, and the type data generation is biased (Wang et al., 2014; Zhou et al., 2020). Therefore, it remains necessary to conduct considerable spatiotemporal data analysis practice alongside the further innovation and development of spatiotemporal statistical methods and theoretical systems.