基于四种常用机器学习的梭梭(Haloxylon ammod-endron(C.A.Mey)Bunge)适宜生境分布研究  

The suitable habitat distribution of Haloxylon Ammodendron(C.A.M.)Bunge based on four machine learning algorithms

作　　者：朱妮[1,2] 郭彦龙 ZHU Ni;GUO Yanlong(Academician and Expert Workstation of Shaanxi,Ankang University,Ankang 725000,China;Engineering Technology Research Center for Water Resource Protection and Utilization of Hanjiang River,Ankang 725000,China;National Tibetan Plateau Data Center,Institute of Tibetan Plateau Research,Chinese Academy of Sciences,Beijing 100101,China)

机构地区：[1]安康学院,陕西省院士专家工作站,安康725000 [2]安康市汉江水资源保护与利用工程技术研究中心,安康725000 [3]中国科学院青藏高原研究所国家青藏高原科学数据中心,北京100101

出　　处：《生态科学》2023年第4期190-199,共10页Ecological Science

基　　金：国家自然科学基金重点项目(41230852);陕西省教育厅项目(20JK0009)。

摘　　要：梭梭Haloxylon ammodendron (C.A.Mey) Bunge是适应中亚荒漠生境的耐寒、抗旱、耐盐碱C4植物,预测该物种地理分布范围以及未来不同气候变化情景下该物种适宜生境分布的变化,对于我国荒漠区生态环境保护具有重要意义。利用收集的梭梭分布数据与33个环境因子,基于四种常用的机器学习算法组成的组合模型对该物种适宜生境在我国西北干旱区的分布进行了预测。结果表明:(1)不同的模型算法的结果在空间上具有较高的一致性,但其分布细节具有一定的差异性。(2)梭梭在我国的适宜生境面积约为0.91×10^(6) km^(2),主要分布在新疆的准噶尔盆地和塔里木盆地边缘,甘肃河西走廊地区,内蒙古巴丹吉林、乌兰布和、腾格里沙漠和库布齐沙漠的西部,青海省柴达木盆地等地区。(3)未来气候变化对该物种适宜生境分布的影响取决于升温幅度,但是在大部分的增温情景中,该物种的适宜生境变化较小。(4)高适宜生境面积为0.19×10^(6) km^(2),应该作为梭梭原生产地保护以及人工种植的重点区域。(5)集成多模型结果的组合物种分布模型能够在一定程度上减少模型的不确定性,增加模型的精度。在环境条件限制性较高的干旱区,加入土壤因素能够有效提高物种分布模型的建模科学性和合理性。Haloxylon ammodendron(C.A.M.)Bunge is a C4 plant,perennial xero-halophytic shrubs,has unique physiological and morphological traits and strong resistance to high temperature,drought,salinity and other stresses,which adapted to the desert habitat in Central Asia.It is a great significance to predict the geographical distribution of this species for the ecological environment protection in the desert areas of China.In this study,based on the distribution data of H.ammodendron and 33 environmental variables using four popular machine learning algorithms(Backpropagation neuronal network(BP_ANN),Boosted regression trees(BRT),Random forest(RF)and MaxEnt)we built an ensemble model(EM)to simulate the present and future potential suitable habitats of H.ammodendron,We found that:(1)The four model algorithms results showed high consistency for the suitable habitat distribution,but the details were different.(2)The suitable habitat area of H.ammodendron in China was 0.91×10^(6) km^(2) that mainly located in the Junggar basin and the edge of the Tarim Basin in Xinjiang;the Hexi Corridor in Gansu;Badain Jaran Desert,Ulan Buh desert,Tengger desert and west of Hobq Desert in Inner Mongolia,and the Qaidam Basin in Qinghai province.(3)Effects of future climate change on the suitable habitat distribution of this species depended on the amplitude of temperature rise.However,the suitable habitat of this species changed little in most warming scenarios.(4)The high suitable habitat area was 0.19×10^(6) km^(2),which could be used as the important area for the protection and artificial cultivation for H.ammodendron.(5)The EM strategy could reduce the uncertainty caused by single model algorithm and improve model accuracy.In arid regions where environmental conditions are more restrictive,adding soil factors can effectively improve the scientific and rational of species distribution simulation.

关 键 词：梭梭(Haloxylon ammodendron(C.A.Mey)Bunge 物种分布 机器学习 组合模型 荒漠物种 

分 类 号：P935.1[天文地球—自然地理学]