SinoLC-1: the first 1-meter resolution national-scale land-cover map of China created with the deep learning framework and open-access data

Abstract

Abstract. In China, the demand for a more precise perception of the national land surface has become most urgent given the pace of development and urbanization. Constructing a very-high-resolution (VHR) land-cover dataset for China with national coverage, however, is a non-trivial task and thus, an active area of research impeded by the challenges of image acquisition, manual annotation, and computational complexity. To fill this gap, the first 1-meter resolution national-scale land-cover map of China, SinoLC-1, was established using a deep learning-based framework and open-access data including global land-cover (GLC) products, open street map (OSM), and Google Earth imagery. Reliable training labels were generated by combining three 10-meter GLC products and OSM data. These training labels and 1-meter resolution images derived from Google Earth were used to train the proposed framework. This framework resolved the label noise stemming from a resolution mismatch between images and labels by combining a resolution-preserving backbone, a weakly supervised module, and a self-supervised loss function, to refine the VHR land-cover results automatically without any manual annotation requirement. Based on large storage and computing servers, processing the 73.25 TB dataset to obtain a final SinoLC-1 land-cover product covering the entire land surface of China, ~9,600,000 km2, took about 10 months. The SinoLC-1 product was validated using a visually interpreted validation set including 106,852 random samples and a statistical validation set collected from the official land survey report provided by the Chinese government. The validation results showed SinoLC-1 achieved an overall accuracy of 73.61 % and a kappa coefficient of 0.6595. Validations for every provincial region further indicated the credible accuracy of this dataset across whole China. Furthermore, the statistical validation results indicated SinoLC-1 conformed closely to the official survey reports. In addition, SinoLC-1 was qualitatively compared with five other widely used GLC products. These results indicated SinoLC-1 had the highest spatial resolution, the most accurate land-cover edges, and the finest landscape details. In conclusion, as the first 1-meter resolution national-scale land-cover map of China, SinoLC-1 delivered accuracy and provided primal support for related research and applications throughout China. The SinoLC-1 land-cover product is freely accessible at https://doi.org/10.5281/zenodo.7707461 (Li et al., 2023).

Local knowledge of homegarden plants in Miao ethnic communities in Laershan region, Xiangxi area, China

“…Barren and sparse vegetation accounts for 3.22%, while grassland (1.03%) and water resources (0.32%) are relatively scarce (see Fig. 1 ) [ 27 ]. The main crops in this region include corn, rice, sweet potatoes, and soybeans, which are well-suited to the subtropical monsoon humid climate with hot summers and mild winters.…”

“…Spatial map visualization and data processing were performed using ArcGIS 10.8 [ 35 ]. The geographic data utilized in this study were obtained from various sources including National Geomatics Center of China ( https://www.ngcc.cn/ngcc/ ), Geospatial Data Cloud ( https://www.gscloud.cn/ ), Standard Map Service ( http://bzdt.ch.mnr.gov.cn/ ), and the research conducted by Li et al [ 27 ].…”

Spatial Distribution, Source Analysis, and Health Risk Assessment of Heavy Metals in the Farmland of Tangwang Village, Huainan City, China

“…The 1 m resolution national land cover data of China were downloaded from the Earth System Science Data journal website. The land cover data (SinoLC-1) were obtained using vector boundary cropping of the study area [36] and imported into ArcGIS 10.8 software with a 500 m × 500 m grid placement to present the sampling points for cropland type (Figure 1). Soil samples were collected in late November of 2022, and the geographical coordinates of the sampling points were determined on site using a handheld GPS.…”

A high‐performance cellular automata model for urban expansion simulation based on convolution and graphic processing unit

“…The cell size of urban CA model is usually the same as the resolution of land-cover data used. As high-resolution spatial data are employed in urban CA models, the cell size becomes smaller, enabling the model to produce more detailed urban morphology (Li et al, 2023;Liu et al, 2023). However, in the same study area, a higher resolution means an increase in the number of pixels, which leads to a significant increase in the computational intensity of the model and occupies more memory space.…”