The accuracy of historical land use/land cover change (LUCC) reconstruction on the regional scale is fundamental for global change research. In this study, based on gazetteers, we developed a methodological framework. It follows the steps including 1) collecting the land tax payment records which had information on cropland and calculating the total registered cropland data in each county; 2) validating the registered land data based on the understanding of the traditional Chinese principles of land assessment; 3) selecting the time slices which obtain as many counties with cropland data as possible and 4) interpolating the missing data based on their temporal and spatial distribution, to produce the cropland dataset at the county level in 1677, 1755 and 1884 in Zhili province (including the current administrative territory of Beijing municipality, Tianjin municipality, most of Hebei province and a few parts of Henan province, Shangdong province, Liaoning Province and Inner Mongolia Autonomous Region). In comparison, our cropland data is close to most of the other researchers' and its spatial resolution is improved. The developed methods were reliable and could be applied to improve the precision of the existing LUCC datasets and could be used to study regional climate and environment simulations.