中文摘要：

在中国耕地保护形势日益严峻的情况下,如何在青藏高原区有效地开发耕地后备资源备受关注,开展自然质量适宜性评价研究是进行耕地后备资源开发利用的重要前提。该文以西藏一江两河流域为典型研究区,选取地形地貌、气候、土壤和水资源状况等4个方面的8个因子,引入知识与数据驱动一体化的模糊证据权模型,得出全域耕地后备资源自然质量适宜性等级,并与综合指数模型评价结果进行对比分析。结果表明：1）模糊证据权模型的评价结果显示,99.38%的耕地＂参照物＂处于适宜性三等以上,说明该模型能有效评价耕地后备资源自然质量适宜性;2）对比综合指数模型评价结果中90.83%的耕地＂参照物＂处于三等以上,说明模糊证据权模型较综合指数模型在某种程度上能够更好地克服评价因子权重确定的主观性和自然条件多样化所带来的计算复杂性等问题;3）研究区11 434.1 km~2耕地后备资源在自然质量上具有不同程度的适宜性,其中一等占9.6%,二等占5.5%,三等占38.6%,四等占22.8%,五等占23.4%;耕地后备资源开发可以重点关注日喀则市、贡嘎、扎囊、拉孜、乃东、拉萨市和白朗等县市。该文可为耕地后备资源适宜性评价提供方法借鉴,并为西藏一江两河流域耕地后备资源的合理开发提供依据。

英文摘要：

With the rapid development of industrialization and urbanization in China, arable land resources are still continuing to be occupied, leading to the intensification of conflicts between an increasing population and decreasing arable land resources. Thus, how to exploit and use the reserve resources of arable land in the Qinghai-Tibet Plateau has attracted much more attention under the serious situation of protecting arable land. But it is the necessary prerequisite to carry out a study on natural suitability evaluation before the exploitation and utilization of reserved arable land. This paper took the reserved arable land of 18 counties in the ＂Three Rivers＂ Area region of Tibet as an example. Eight factors of landform, climate, soil and water, that is, slope, elevation, slope direction, temperature, rainfall, soil depth, organic matter content, and distance to water area, were adopted as evaluation factors. The fuzzy weight of evidence model, which is both knowledge and data driven, was employed to get the grade of natural suitability evaluation of reserve resources of arable land, which was then compared with the result evaluated by a comprehensive index model. Evaluation results showed as follows： first, 99.38% of ＂reference objects＂ of the arable land are above the third grade on the basis of five suitability grades in total, which means that the fuzzy weight of evidence model could evaluate the natural suitability of reserved arable land effectively. Second, compared with the evaluation results of a comprehensive index model, 90.83% of the arable land resources are above the third grade, which shows the fuzzy weight of evidence model is more objective and reasonable than the comprehensive index model in this case. And it also overcomes the subjective arbitrariness in weight determination and the computation complexity by various natural conditions. Third, 11 434.1 km2 of reserved arable land had suitable natural quality, among which grade 1, 2, 3, 4, and 5 accounted for 9.6%, 5.5%, 38.6%, 22.8%, a