China's first 1-meter resolution national-scale land-cover map SinoLC-1 was built using a framework based on deep learning and open data, including Global Land Cover (GLC) products, OpenStreetMap (OSM) and Google Earth Images. Reliable training tags were generated by combining three 10-meter GLC products and OSM data. These training tags and 1-meter-resolution images derived from Google Earth were used to train the proposed framework. The framework solves the problem of label noise caused by mismatch between image and label resolutions by combining a resolution-maintaining backbone network, a weak supervision module, and a self-supervision loss function, allowing automatic optimization of high-resolution land cover results without manual labeling. Based on a large storage and computing server, it took about 10 months to process a 73.25 TB data set to obtain SinoLC-1 covering the entire territory of China (9.6 million square kilometers).

The SinoLC-1 product was verified through a visual interpretation verification set containing more than 100,000 random samples and a statistical verification set collected from official land survey reports provided by the China government. The verification results showed that the overall accuracy of SinoLC-1 was 73.61%, and the κ coefficient was 0.6595. Validation of each provincial region further demonstrated the accuracy of the dataset across China. In addition, statistical verification results show that SinoLC-1 is consistent with the official survey report, with an overall error rate of 6.4%. In addition, SinoLC-1 was compared to five other widely used GLC products. These results show that SinoLC-1 has the highest spatial resolution and the finest landscape detail. In short, as China's first national-scale land cover map with 1-meter resolution, SinoLC-1 provides accuracy and basic support for relevant research and applications across the country.