Introduction

Soil organic carbon (SOC), as a critical component of soil fertility, not only affects soil quality and crop production (Vågen & Winowiecki, 2013; Jiang et al., 2014; Zhao et al., 2018), but also plays an important role in the global carbon cycle (Wei et al., 2011; Borrelli et al., 2018). The SOC levels are influenced by natural factors and human activities (Bolinder et al., 1997; Wang et al., 2015a; Fujisaki et al., 2018; Fan et al., 2019), while management practices usually have more profound influences on the cropland SOC levels (Nandwa, 2001; Li et al., 2016; Sandeep et al., 2016). For example, unreasonable management practices (i.e., overuse of chemical fertilizers and straw burning) may cause soil nutrient imbalances, and further accelerate cropland SOC loss in a short time (de Blécourt et al., 2019; Keel et al., 2019; Yang et al., 2019). Consequently, information and knowledge on the spatiotemporal changes in cropland SOC and their driving factors are extremely important for developing effective management practice recommendations and addressing climate change.
Multiple factors affect the spatiotemporal changes in cropland SOC, including natural factors such as climate, soil types, and topography (Meersmans et al., 2011; Ou et al., 2017; Yang et al., 2019), and anthropogenic factors such as agricultural management practices (i.e., tillage, fertilization) (VandenBygaart et al., 2004; Han et al., 2018; Zhou et al., 2019) and land use change (Guo & Gifford, 2002; Luo et al., 2019). The systematic review of existing SOC studies (Schillaci et al., 2018; Ramesh et al., 2019; Wiesmeier et al., 2019) shows that natural factors are generally recognized as important contributors to spatial variability of SOC in cropland. For example, Li et al., (2018b) found that topographic factors such as slope and aspect significantly influenced the spatial distributions of cropland SOC in central Iowa and explained more than 62% of the spatial variability of SOC in that area. The spatial prediction of topsoil SOC with the random forest model in an intensively cultivated area in eastern China demonstrated that soil properties, climate, and geographic locations were the most important predictors for mapping SOC distributions (Deng et al., 2018). However, numerous studies have indicated that anthropogenic factors, in particular management practices and land use changes, have even more profound influences on the spatiotemporal changes in cropland SOC (Bas et al., 2010; Leifeld et al., 2013; Zhao et al., 2018; Shi et al., 2019), because anthropogenic activities usually alter the balance between soil carbon inputs and decomposition over a relatively short time period (Jiang et al., 2014). For example, a significant difference in the annual SOC change rates across different agricultural regions of China (-2.0% to 0.6% yr-1) was mainly caused by regional differences in management practices, such as cultivation, fertilization, and straw/residue return rate (Yan et al., 2011). The CENTURY model analysis of the SOC changes in the agricultural soils of northeastern Spain also revealed that the soil carbon inputs associated with management practices were the primary reasons for SOC increases (rate of 0.27 Mg C ha-1 yr-1) over the period of 1977-2007(Álvaro-Fuentes et al., 2011). Qiu et al., (2013) identified spatiotemporal changes in SOC in eastern China over the period of 1979-2006, and found that the changes in SOC were closely related to the conversion from croplands to urban lands due to rapid urban expansion.
China has been experiencing rapid economic growth since the 1980s (Bao et al., 2019). Meanwhile, in the rapidly urbanizing area, the urban expansion and the industry-dominated economic growth have caused a significant shrinkage of cropland area and a decrease of economic investment in cropland management (Bren d’Amour et al., 2017; van Vliet et al., 2017; Zhong et al., 2020). However, the impacts of these changes on the spatiotemporal changes of SOC and their implications remain unclear, in particular over different time periods. Therefore, the specific objectives of this study were (1) to map the spatial distribution of cropland SOC in a representative area that has been experiencing rapid urbanization; (2) to quantify the changes in SOC spatial distributions over different time periods while considering the associated mapping uncertainties; and (3) to identify the primary driving factors of the SOC spatiotemporal changes.