模糊证据权法在西藏一江两河流域耕地适宜性评价中的应用

开展耕地适宜性评价是编制农业发展规划、优化耕地资源配置的基础性工作。文章以西藏一江两河流域18 个县的耕地资源为评价对象，将模糊证据权模型引入到耕地适宜性评价中，通过该方法筛选出地形地貌、气候、土壤和水资源状况等4 个方面的9 个因子作为证据图层，并计算因子图层的模糊证据权重，最后综合各因子图层的证据权重参数得出后验概率并修正形成耕地适宜性分布图。结果表明：①80.83%的适宜样本点处于修正后验概率分级图的中等适宜级别及以上，即已知信息体现出较高的吻合性，表明模糊证据权模型能够科学、有效地评价耕地资源的适宜性；②集知识驱动与数据驱动一体化的模糊证据权模型既能够充分发挥先验知识、客观规律的作用，也可以通过隶属度函数增加知识干预，有效利用专家的丰富经验，在一定程度上克服了知识驱动方法中因子筛选和权重确定主观性强或纯粹数据驱动方法的不足；③研究区21.32×10⁴hm²的耕地资源中，高度适宜、中度适宜、一般适宜和不适宜的耕地资源分别占32.03%、31.94%、27.19%和8.84%，对应其空间分布情况，可以优化耕地资源规模与布局，指导高原耕地区农业生产实践，促进耕地资源的合理利用。

Suitable Evaluation on Cultivated Land Based on Fuzzy Weights of Evidence Method in the Yarlung Zangbo River,Nyangqu River and Lhasa River Region,Tibet

Cultivated land suitability evaluation of the plateau is an important prerequisite to promote the planning of regional agricultural development and the construction of food security system. This paper takes cultivated land of 18 counties in the Yarlung Zangbo River, Nyangqu River and Lhasa River (YNL) region of Tibet as the object of evaluation. The fuzzy weight of evidence model is introduced to screen out nine factors included in the four aspects of topography, climate, soil and water conditions as evidence layers, calculate the fuzzy weight of evidence of factor layer, and finally synthesize parameters of all factor layers to calculate posterior probability weight and revise it to format a distribution map of cultivated land suitability. The study shows that: 1) 80.83% of the known suitable points are above moderately suitable level in the grading figure of modified posterior probability, reflecting a high coherency, which indicates that the fuzzy weight of evidence connecting with modified posterior probability can evaluate the suitability of cultivated land scientifically and effectively. 2) Fuzzy weight of evidence model, which is both data-driven and knowledge-driven, can give full play to the prior knowledge and the role of objective laws. It can also increase the intervention of knowledge by the membership function, effectively using the experience of experts, which, to some extent, avoids arbitrariness in selecting indicators and determining the weight in the knowledgedriven method and overcomes the weakness of purely data-driven method. 3) Highly suitable, moderately suitable, generally suitable and unsuitable culti-vated land in the study area cover 213200 hm² of cultivated land, accounting for 32.03%, 31.94%, 27.19% and 8.84% respectively, which corresponds to the spatial distribution. This can optimize the scale and layout of cultivated land and also can be used in the practice of cultivated area on the plateau to promote the rational utilization and management of cultivated land.