基于MaxEnt的青藏高原区光核桃适宜分布区预测  

Simulation of Suitable Habitat Distribution of Prunus mira in Qinghai-Tibet Plateau based on MaxEnt Species Distribution Model

作　　者：李小熊 李佶芸 喻武[1] 杨东升[1] LI Xiao-xiong;LI Ji-yun;Yu Wu;YANG Dong-sheng(Tibet Agricultural and Animal Husbandry University,Nyingchi Tibet 860000,P.R.China)

机构地区：[1]西藏农牧学院,西藏林芝860000

出　　处：《西部林业科学》2020年第5期142-148,154,共8页Journal of West China Forestry Science

基　　金：西藏高原生态安全实验室开放基金(STX2019-09);水利部沙棘开发管理中心项目(2017-zg-zx-57)。

摘　　要：光核桃是藏东南地区分布较为广泛的一种乔木,其主要分布于海拔2600~4500 m的河谷区域,外观美丽,特别是在春季光核桃花盛开时具有良好的景观效果。为弄清其在青藏高原的分布格局,本文基于光核桃实际空间分布点及23个环境因子,以MaxEnt模型为基准,建立组合物种分布模型预测光核桃在青藏高原区的潜在地理分布范围。结果表明:最冷月年均最低温、土壤酸碱度、最暖季年均降水量这3个环境因子对其潜在适宜生长区的影响最大。经PCA降维后,基于10种环境因子的组合物种分布模型比单独使用MaxEnt模型可以更精确地描绘光核桃在青藏高原区的基本分布特征,所有的光核桃实际分布点均落在模型所预测出的高及中高适宜分布区,其面积为6.94×104 km 2。以ANN、RF、MARS为代表的复杂机器学习模型可以起到较好的预测效果,对其进行建立组合物种分布模型后的预测效果好于使用单一模型。Prunus mira is a widely distributed tree in Southeast Tibetan region.It is mainly distributed in the valley area with an altitude of 2600-4500 meters,and has an impressive landscape,especially in spring bloom season.In order to understand its distribution pattern in the Qinghai-Tibet Plateau,based on the actual spatial distribution points of Prunus mira and 23 environmental factors by the MaxEnt model,a combined species distribution model was established to predict the potential geographic distribution of Prunus mira in the Qinghai-Tibet Plateau.The results showed that:(1)The average annual minimum temperature in the coldest month,the soil pH,and the average precipitation in the warmest season,have the greatest impact on their potential suitable growth areas.(2)After PCA dimensionality reduction,the combined species distribution model based on 10 environmental factors could more accurately describes the basic distribution characteristics of Prunus mira in the Qinghai-Tibet Plateau than the MaxEnt model alone.Almost all the actual distribution points of Prunus mira fall in the model.The predicted suitable distribution area for high and middle heights is 6.94×104 km 2.(3)A complex machine learning model represented by ANN,RF,and MARS could play a good prediction effect,and the prediction effect after establishing a combined species distribution model is better than using a single model.

关 键 词：MAXENT 光核桃 青藏高原 组合物种分布模型 适宜分布区 

分 类 号：S664.1[农业科学—果树学]