In recent decades, China has experienced rapid urbanization and industrialization, which has led to significant changes in land use/cover (Qiao et al. 2016). As one of the most important indicators for studying urbanization, impervious surface area (ISA) directly reflected ecological and environmental issues from local to global (Ayalew et al. 2022). The research on the spatiotemporal dynamic distribution of ISA contributed to understanding the impact of human activities on the natural world (Seto et al. 2012). In general, ISA refers to man-made surface materials that impeded water infiltration into underground, such as asphalt, concrete roads, roofs, or car parks (Chwster et al. 1996). The Chinese government attached great importance to the sustainable development of the Yellow River Delta (YRD) region due to its unique geographical location and significant strategic position. For example, a specialized regional plan was approved by Chinese State Council for the reasonable development, called "Development Plan for the Yellow River Delta Efficient Ecological Economic Zone". Therefore, it is of great significance to study the spatiotemporal evolution of ISA in the YRD region for deeply understanding the urbanization process of the region, planning sustainable urban development, protecting the ecological environment as well as promoting harmonious co-existence between human and nature (Ma et al. 2014).
With the continuous development of computer technique and sensor, numerous approaches have been developed for the ISA extraction from remotely sensed data with several spatial resolution over the past decades (Wu et al. 2015). For example, Wang et al. generated time series ISA data over Huai'an central urban area and subsequently explored the impact of ISA spatiotemporal changes on urban heat islands (Wang et al. 2022); Wang et al. adopted a neighborhood-based spatiotemporal filter for extracting ISA continuous change information from Landsat images over the Qinhuai River basin from 1988 to 2017, aiming to provide input for future urban planning in this basin (Wang et al. 2021). Different from the traditional field surveys, remote sensing has the advantages of wide coverage, short revisit and low cost, which has been considered as the primary means to obtain ISA distribution (Duan et al. 2022). The main differences among researches on ISA using remote sensing images included: (i) selection and extraction of classification features, such as different combinations of spectral and texture features (Fu et al. 2021; Liu et al. 2016; Li et al. 2021; Zhao et al. 2012); (ii) selection and design of classifiers, the used classification methods for extracting ISA mainly involved: random forest (Liu et al., 2020), support vector machine method (Xue et al. 2009), decision tree (Li et al. 2002) and neural network (R. Cots-Folch et al. 2007), etc. Gu et al. compared and analyzed the accuracy of the four classification methods mentioned above, and the results showed that the random forest algorithm had great advantages over others (Gu et al. 2019). The random forest classification method has the advantages of higher accuracy, less parameters to set, and more convenient operation and estimation of feature variables’ importance (Breiman 2001), leading to its wide application in remote sensing image classification for the extraction of land cover/utilization information
For the ISA of the YRD, some scholars have conducted relevant studies. Song et al. took Binzhou City, the hinterland of the YRD, as the study area and investigated the impact of ISA on rainwater resources (Song 2018); Shen et al. extracted ISA distribution information of Dongying City in time series based on MNF transform, PPI index and LSMA, and used root mean square error to verify the accuracy (Shen et al. 2019); Liu et al. used random forest method based on Sentinel-1/2 image method to extract the impervious layer of Dongying City in 2019, and the final overall accuracy reached 93.37% (Liu et al. 2021). Although the previous process of ISA extraction based on the traditional desktop side can obtain highly accurate impervious surface information, the image data pre-processing process is tedious and at the same time, high-performance processing equipment is also essential to obtain the resultant data quickly and accurately, which will undoubtedly increase the cost expenditure.
In the 21st century, China has entered a new stage in the field of Earth observation, demanding higher performance of the automated processing and analysis of massive remote sensing data, and there is an urgent need to make innovative breakthroughs in intelligent, efficient and low-cost processing of massive remote sensing data. With the rapid development of space information technology, diverse and integrated remote sensing cloud computing platforms have emerged. During the development of the platforms, the Google Earth Engine (GEE) cloud computing platform was born, presenting the remote sensing data processing and application to the general public (Tamiminia et al. 2020). GEE is a cloud platform provided by Google to achieve efficient visualization and processing of massive remote sensing data (Tamiminia et al. 2020). The GEE platform, which has been applied in many fields, simplifies the process of downloading and processing of remote sensing data by using globally open-source datasets capable of processing large amounts of remote sensing data efficiently, such as Landsat and Sentinel (Arman et al. 2021; Yang et al. 2021).
In this study, we aim to generate a time series of consistent ISA data in the YRD from 1992, 1998, 2004, 2010, 2016 and 2021 using a random forest classifier and GEE, and subsequently analyze the spatiotemporal changes of ISA in the last three decades. More specifically, this study aimed (i) to determine whether ISA mapping accuracy was improved when spectral and GLCM texture features were synergistically used, (ii) to understand what the ISA spatial-temporal evolution characteristics were in the YRD spanning almost three decades, (iii) to attempt to provide beneficial suggestions for the protection of ecological environment and sustainable development of the YRD.