Abstract:

Global historical land use and land cover change (LUCC) datasets are of great significance for understanding the long-term global or regional environmental change. However, uncertainties in the cropland data of the global LUCC datasets have been revealed by historical record-based reconstructed regional data. Therefore, assessing the reliability of global historical LUCC datasets combined with regional cropland reconstruction results and the land reclamation history is the basis and prerequisite for improving the quality of LUCC datasets. In this study, based on the regional historical record-based cropland datasets in Xinjiang area, including the cropland datasets from the Qitai-Jimusa region covering 1776-1909 (Yang-dataset), the Manas River Basin region covering 1909-1944 (Zhang-dataset), and the Xinjiang province covering 1960-2000 (Chen-dataset), and the historical data of land reclamation points of migrants for the Kangxi and Qianlong periods (AD1716-1778) of the Qing Dynasty in Xinjiang area, the reliability of cropland data in Xinjiang area included in these global datasets (the HYDE 3.2 and SAGE) were evaluated through comparative analyses from three aspects, including the changing tendency of cropland area, the area of cropland and the differences at spatial pattern. The comparisons show that the quantity of cropland area has a large difference in different reconstruction regions, even though all of the cropland areas in different reconstruction regions from multiple datasets increased. Specifically, the cropland area of the Qitai-Jimusa region for 1776-1909 in HYDE 3.2 dataset was less than that in the Yang-dataset and the relative difference ratios between the two datasets were around 0.04-0.56. However, the cropland area of the Manas River Basin region for 1909-1944 in HYDE 3.2 and SAGE datasets was notably greater than that in the Zhang-dataset and the relative difference ratios between the two global datasets and the Zhang-dataset were around 3.08-8.95 and 4.80-5.12, respectively. In contrast, the cropland area of the Xinjiang area for 1960-2000 in the two global datasets were similar to that in the Chen-dataset. The relative difference ratios between the two global datasets and the Chen-dataset were around 0.67-0.75 and 0.71-1.05, respectively. Spatially, both the expansion process of land reclamation points for migrants and the land reclamation history indicated that the land reclamation in Xinjiang area was under the expansion from south to north and from east to west in the early and middle Qing Dynasty. Moreover, the northern Xinjiang area was still dominated by nomadic pastoralism in early Qing Dynasty. However, according to the two global datasets, most of cropland had been allocated to the northern area of Xinjiang in the early Qing Dynasty, accounting for 42.84% and 61.26% of the total regional cropland, respectively. Therefore, the cropland data in the global datasets cannot reflect the process and spatial pattern characteristics of the cropland change in Xinjiang area objectively because of the differences in data sources and reconstruction methods.

Key words: land use and land cover change, historical cropland, global datasets, accuracy evaluation, Xinjiang area

摘要：

全球历史土地利用/覆被变化数据是开展长时段气候与生态效应模拟研究的科学基础,然而,现有全球LUCC数据集在区域尺度上仍存在较大不确定性。结合区域土地垦殖史实评估全球数据集的可靠性,是改进数据集质量的基础和前提。文章以中国学者利用区域历史文献和土地资源调查数据重建的清中后期（1776—1909年）奇台-吉木萨尔地区、清末民国时期（1909—1944年）玛纳斯河流域和建国以来（1960—2000年）新疆地区耕地数据为基准,结合清康乾时期（1716—1778年）区域屯田点数据,采用趋势对比、数量对比和空间格局对比相结合的方法,对全球数据集中新疆地区耕地数据的可靠性进行了评估。结果表明,HYDE 3.2和SAGE数据集中清代以来新疆地区的耕地数据与区域重建结果相比,虽然耕地变化趋势均呈增加态势,但数量上差异较大。其中,HYDE 3.2数据集在奇台-吉木萨尔地区的耕地数据存在明显低估,仅为区域数据的0.04—0.56;而在玛纳斯河流域地区,HYDE 3.2和SAGE中的耕地数据又存在明显高估,分别为区域数据的3.08—8.95和4.80—5.12;与之相比,1960年以来全球数据集中新疆地区的耕地数据与区域数据差异有减小趋势,分别为陈氏数据的0.67—0.75和0.71—1.05。从耕地空间分布格局看,清康乾时期新疆地区屯田点扩散过程及区域垦殖史实均表明该区土地垦殖经历了“从南到北,由东向西”逐步推进的历史过程,且在清前期北疆地区仍以游牧为主,而该时期全球数据集中北疆地区已有较多耕地分布,且分别占到区域耕地总量的42.84%和61.26%。可见,全球数据集中新疆地区耕地数据未能客观反映该区域耕地变化的过程与格局特征,而数据来源和重建方法的差异是造成不同全球与区域性数据集结果差异的主要原因。

关键词: 土地利用/覆被变化, 历史耕地, 全球数据集, 可靠性评估, 新疆地区