基于最大信息熵模型的中药材品质区划方法研究

From the beginning of the 1990s, the relevant theories, principles and methods of Chinese Materia Medica regionalization has developed more and more comprehensive. But there are still some difficult problems which need to be solved, such as samples’ representativeness are not enough, ecological facor data accuracy and wideth are not exact, the relationship between herb quality and ecological factor is not clear, etc..The Maximum Entropy Model is compatible with different types of ecological facters, its result only relies on the objective sampling datum, and it’s a machine – learning model based on ecology. The model has wide applications in terrestrial / marine flora and fauna distribution areas. Atractylodes is an important medicinal herb plant group, which is recorded by the South Korean, North Korean, Janpanese and Chinese Pharmacopoeia. Duing to the long-term excavation, the natural reserve of A. lancea（Thunb.）DC. is currently endangered. A. chinensis (Bunge) Koidz. is widely disributed, it has a large amount distribution and exports international. A. coreana (Nakai) Kitam. and A. japonica Koidz.are used as A. lancea（Thunb.）DC. in some areas of China, both of them are the potential substitutes of the endangered herbs. This research makes the 4 different Atractylodes for examples, focuses on the quality regionalization method of Chinese Materia Medica, the flow is stated below: first of all, construct a fine mountain ecological factors database by geographic information system and meteorological data, and then make habitual suitability index as a link, based on the complete sampling datum, to explore the relationship between herb quality and ecological environmental factors by Maximum Entropy Model. At last, draw the quality regionalization map based on the the similarity of environmental factors. This data – abundant, objective method will make a significant guide to the traditional Chinese herb planting.

中药资源区划方法经过20多年的发展逐渐完善，但仍存在着样本代表性不够、生态因子数据的精度和广度较低、药材品质与生态因子关系不明确等问题。最大信息熵模型是兼容不同类型生态因子、只依据客观采样结果的机器学习型生态模型，在研究陆地/海洋动植物分布领域有广泛应用。苍术类药材是重要的药用植物资源，被中日朝韩等国家药典收载，南苍术由于长期采挖，目前已近濒危；北苍术分布广泛，用量大且出口国际；朝鲜苍术和关苍术在中国部分地区有作为苍术用的地方习惯，有作为濒危药材的替补潜力。本研究选择以上4种不同特点的苍术类药材，研究中药材品质的区划方法体系：先利用地理信息系统和气象观测数据构建精细的山区生态因子数据库，然后以生态适宜度为纽带，基于完备的采样数据，利用最大信息熵探寻药材品质和生态环境之间的关系，使用环境相似度进行等级划分，得到品质区划结果。该方法数据详尽、客观性强、突出药材品质，对中药实际生产具有指导意义。

本研究针对目前中药资源区划中存在的样本代表性不够、生态因子数据的精度和广度较低、药材品质与生态因子关系不明确等问题，以精细的生态因子数据为基础，利用生境适宜度建立生态因子与中药品质之间的联系，提出中药材品质区划的新方法。主要内容有：.（1）精细生态因子数据构建方法研究。山地地形多变导致山地气候是一个复杂的气候类型，本研究把GIS 方法和气象原理结合起来构建高分辨率的气象生态因子数据（太阳直接辐射、气温和降雨量），同时对现势的遥感影像进行目视解译获取地表植被类型数据。.（2）以苍术为研究对象，依据2010版《药典》对苍术指标性药用成分苍术素含量的规定（干燥品中不得少于0.3 %）来考察苍术的品质。基于实地采样数据和丰富的生态因子数据，首先通过分析生态因子与苍术素（C13H10O）含量的相关性筛选所需生态因子，再依据2010年版中国《药典》中的指标成分含量将采样点分为苍术素低于《药典》标准与苍术素高于《药典》标准两类，然后分别利用最大信息熵模型计算出苍术的生态适宜度，最后整合两类生态适宜度结果，得到了全国范围内苍术素含量分布图。并用课题组以前的采样数据来检验该方法的准确性。结论有：.①苍术的品质评价：苍术素含量（y）与苍术素低于《药典》标准的苍术生境适宜度（x）呈负相关（y = -2.70x + 3.16, R2=0.64）；苍术素含量与苍术素高于《药典》标准的苍术生境适宜度呈正相关（y = 4.28x + 4.04, R2=0.62）。.②分析苍术素低于《药典》标准的苍术分布结果，发现土壤类型（贡献率39.7%）、土壤粘土量（26.7%）、12月平均气温（22.3%）、土壤阳离子交换能力（6%）等对苍术素的积累有抑制作用。分析苍术素高于《药典》标准的苍术分布结果，发现土壤类型（50%）、土壤粘土量（34.8%）、12月平均气温（10%）等对苍术素的积累有促进作用。.本研究实现了任意坐标点的苍术品质查询，满足了“根据品质定地点”的实际种植需求。在苍术素含量分布较高的地区种植苍术，用科学的田间管理技术调控田间生态因子，可提高种植苍术的经济效益。