中文摘要：

土地覆被变化是气候与生态效应模拟研究的重要参量。SAGE和HYDE两个全球历史土地利用数据集在相关研究中得到广泛应用，但在区域尺度上的应用，其可靠性如何，至今少有论及。以我国学者重建的传统农区历史耕地数据集（CHCD）为基础，从全区、省区和网格（60km×60km）三个空间尺度，对SAGE（2010）和HYDE3．1数据集中有关中围传统农区历史耕地重建结果进行对比分析，结果表明：（1）SAGE（2010）数据集对中国传统农区耕地数量重建是以单一线性插补而得，其中1700-1950年是以0．51％的年均增长率线性递增，1950年后是以0．34％年均速率线性递减，这种“标准化”变化趋势不能客观反映传统农区土地垦殖的真实历史，耕地面积也明显高估，与CHCD数据集不具有可比性；（2）HYDE3．1数据集吸纳了区域性研究成果，使其在总量上与CHCD数据集较为接近，具有较好的可比性，但其在省区和网格尺度上与CHCD存在显著差异，其中相对差异率超过70％（〈-70％或〉70％）的网格占比高达56％-63％，超过90％（〈-90％或〉90％）的网格占比也高达40％-45％；而相对差异率介于-10％-10％的网格占比仅为5％-6％，介于-30％-30％的网格占比也仅为17％左右；（3）充分利用我国丰富的历史文献，建立更高精度的中围区域历史土地利用数据集，是提高区域气候与生态效应模拟研究质量的重要保障。

英文摘要：

Land cover change is an important boundary condition in climate and ecological simulations. Although they have been widely used in researches, SAGE and HYDE, the two global scale historical land use datasets, have been rarely discussed about their rationality and accuracy in regional scale. On the basis of CHCD, the historical cropland dataset of traditional cultivated region of China （TCRC, hereafter） which was established by Chinese scholars, we analyzed the rationality and accuracy of TCRC part of SAGE （2010） and HYDE3.1 by comparing them with CHCD in the whole study area scale, provincial scale and 60 km ~ 60 km grid cell scale. The results indicated that （1） TCRC part of SAGE （2010） dataset was established by linear interpolation and the average annual growth rate was 0.51% from 1700 to 1950 and-0.34% after 1950. The normalized trend could not reflect TCRC＇s history of cultivation, and the estimation of cropland area was obviously high compared with CHCD dataset. （2） In comparison of more regional research results, HYDE3.1 dataset was very close to CHCD dataset in the whole study area scale. However, there were significant differences between HYDE3.1 and CHCD in provincial and 60 km~ 60 km grid cell scales. The percentages of grid cells whose relative differences are greater than 70% （〈-70% or 〉 70%） and 90% （〈-90% or 〉90%） were 56%-63% and 40%-45% respectively while those whose relative difference ranges from-10% to 10% and from-30% to 30% were only 5%- 6% and about 17%, respectively. （3） Making full use of abundant historical archives in China and establishing historical land use dataset of China with high accuracy are significant ways of improving the quality of climate and ecological simulation results.