Spatiotemporal scale and integrative methods matter for quantifying the driving forces of land cover change

Highlights

• Structural equation and mixed-effects model are effective land-cover change attribution.

• Driving factors on land-cover transitions tend to be spatiotemporal scale dependent.

• Fine-tuned analysis on driving forces is critical for landscape planning and management.

Abstract

Land cover change (LCC) is a major part of environmental change. Exploring the spatiotemporal differences in LCC and the driving factors is the basis for comprehensive research on landscape planning, and it is of great significance for future effective and sustainable landscape management. In this respect, cross-scale research with integrated methods is worthy of more attention, although some studies have discussed the driving forces of LCCs at either regional or local scale. We combined a structural equation model and a mixed-effects model for quantifying the driving forces of LCCs across different scales in the Loess Plateau (China), which is a typical region that has experienced significant LCCs over recent decades. The impacts of biophysical and socioeconomic factors on different change trajectories (agricultural intensification, urbanization and ecological restoration) were found to be inconsistent at different temporal and spatial scales. We found that topography had a negative effect on agricultural intensification during 1990–2010 and on urbanization during 1990–2000, but it had a positive effect on ecological restoration during 2000–2015 at the regional scale. Moreover, although there was no significant impact from economic development on any type of LCCs at the regional scale, its important influence could be seen in some of the township categories. Therefore, the path and scale dependence of driving forces is an important consideration in landscape planning and management to accommodate local conditions and fine-tuned analysis as decision-making supports.

Introduction

Land cover patterns have been altered at an unprecedented pace, magnitude and spatial scale during recent decades (Gao et al., 2016). A large amount of natural land (such as forests, grassland and wetland) has been converted into developed land (such as urban and cropland), thus pressures on the environment and ecosystems are increasing (Liang et al., 2017), including atmospheric composition, energy and water balance, degradation of soil and water and declines in biodiversity (Foley et al., 2005). Land cover change (LCC) thus constitutes one of the nine “planetary boundaries” (Rockstrom et al., 2009), and land degradation neutrality was proposed as a UN 2030 global sustainable development goal. To achieve this objective, understanding the processes, trends, causes and consequences of LCCs is a key issue in global change research (Pielke, 2005; Verburg et al., 2009), and its focus has gradually transitioned from ‘globe’ to ‘region’ and from ‘nature’ to ‘humanity’ (Liu et al., 2017).

For land systems science, the causes of LCCs have attracted attention at different spatial and temporal scales, and the richness of explanations has greatly increased, often at the expense of generality (Eunice Quintero-Gallego et al., 2018). Current research on the driving forces of LCCs has focused on different scales and methods, but there are still many questions to be solved. First, most previous studies about drivers have focused on the global, country or state scale (Alexander et al., 2015; Amuti and Luo, 2014; Arowolo and Deng, 2018), and a number of studies also have assessed LCC at small scale (local), such as at the county or town level (Abadie et al., 2018; da Silva et al., 2016). The results from these scales have found similar drivers, such as topography, accessibility, urbanization and other biophysical or socioeconomic factors (Abadie et al., 2018; Bansal et al., 2016; Gao et al., 2016; Hoyos et al., 2018). However, the driving forces on differences and similarities across multiple scales is worthy of more attention. Second, LCC exhibits strong path dependence (Celio et al., 2014). To study human impacts on the natural environment, it is necessary to assess the history of LCCs and interpret the spatiotemporal patterns of such changes relevant to other environmental and human factors, so it is not sufficient to assess only the changes in areas (Wang et al., 2013; Zhou et al., 2008). The various components of LCCs, such as total change, net change and swap change, have been detected and estimated in previous studies (Gao et al., 2016; Lu et al., 2004; Pontius et al., 2004; Wang and Wang, 2013). However, most of these studies have focused only on decreases or increases in one certain land type, such as croplands (Alexander et al., 2015; Arowolo and Deng, 2018; Najmuddin et al., 2018; van Vliet et al., 2015; Wang et al., 2015; Wood et al., 2004; Xie et al., 2014), forests (Abadie et al., 2018; Acacio et al., 2017), grasslands (Monteiro et al., 2011), wetlands (Zheng et al., 2017), rural lands and urban lands (Zhou et al., 2008). Recently, the topic of the change trajectory (change from a certain land cover type to another, such as the transition from forest to grassland or cropland) (Calaboni et al., 2018; Lambin and Meyfroidt, 2010; Yackulic et al., 2011) has received increasing attention in studies on landscape change analyses (Hietel et al., 2004; Kayhko and Skanes, 2006; Wang et al., 2013), and it is particularly useful for addressing this challenge. Some studies have focused on multiple trajectories (Wang et al., 2013; Yeh and Liaw, 2016), but typical or dominant paths were not highlighted, especially those characterized by historical changes, such as revegetation. The identification of driving forces that cause dominant LCCs will help managers establish policies that are designed to prevent or minimize the negative effects of LCCs. Third, previous studies have frequently quantified the driving forces linked to LCCs through using descriptive analysis (Patarasuk and Binford, 2012; Teixeira et al., 2014; Zhu et al., 2010), principal component analysis (PCA) (Abadie et al., 2018; Chen et al., 2017; Du et al., 2014; Hoyos et al., 2018; Li et al., 2018; Meneses et al., 2017; Zhao and Liu, 2014), analytic hierarchy process (AHP) (Thapa and Murayama, 2010), redundancy analysis (RDA) (Gao et al., 2016; Parcerisas et al., 2012), canonical correlation analysis (CCA) (Salvati et al., 2016), and regression analysis (Abadie et al., 2018; Acacio et al., 2017; Kindu et al., 2015; Monteiro et al., 2011; Najmuddin et al., 2018; Yu et al., 2017). Machine learning methods have gradually been used in identifying driving forces, such as neural networks (Jalala et al., 2011; Tan et al., 2014), random forests, and gradient boosting trees (Moore and Lin, 2019). These methods have been used to interpret the links between drivers and LCCs at one certain scale, but they could not be used to identify the relationships between measured variables and quantify the differences across scales.

Landscape development usually involves urbanization and industrialization, which can lead to significant LCCs. Therefore, it is especially critical to identify the main LCC processes in economically underdeveloped and ecologically fragile areas because of the relatively rapid pace and extent of LCC and its effects on fragile ecosystems (Gao et al., 2016; Li et al., 2012; Liu et al., 2005; Serra et al., 2008; Sleeter et al., 2013). Over the past several decades, China has been one of the most rapidly developing countries, and it has experienced continuous urbanization, population increases and ecological restoration (Chen et al., 2019). The largest ecological restoration program (GFGP: Grain for Green Program) in China has been implemented since 1999, and it has caused increasing landscape changes that have received widespread attention (Du et al., 2014; Liu et al., 2005; Wang et al., 2012). The Loess Plateau is a representative dryland in China. It was the 1st pilot and demonstration area for GFGP, and it has experienced extensive ecological restoration and significant economic development in the last 20 years. Therefore, the Loess Plateau is a suitable research region with typical LCC processes, including agricultural intensification, urbanization, and ecological restoration.

The studies on these LCC processes and their driving forces need to pay more attention on differences across spatiotemporal scales to provide reasonable suggestions on landscape planning and management. In this paper, we examined the historical dynamics (between 1990 and 2015) of land cover area at both regional and town-cluster scales on the Loess Plateau to deepen our understanding of possible spatiotemporal differences in the driving forces in arid and semiarid regions. The objectives of this paper are to: (1) examine typical land cover transitions at regional and town-cluster scales in three time intervals (1990–2000, 2000–2010, and 2010–2015); (2) quantify the differences in the driving forces (biophysical and socioeconomic factors) of LCCs using integrated methods at two spatial scales, i.e., using structural equation modeling (SEM) and a mixed-effects model (MEM); (3) identify the temporal differences in the impacts of the driving forces; and (4) discuss the differentiated landscape management and planning processes at a small spatial scale for sustainable development in the region.