基于随机森林模型研究孑遗植物银杏的历史分布和迁移趋势

【作者】 刘静；

【导师】 卫海燕；

【作者基本信息】 陕西师范大学 ， 地图学与地理信息系统， 2019， 硕士

【摘要】 气候变化尤其是全球变暖已成为不争的事实,气候变暖将导致冰川融化、海平面上升、生态退化等,物种分布区也随气候条件的改变发生迁移,进而影响物种的生态多样性,因此基于物种分布模型(SDMs)研究物种对气候变化的响应是目前的主流趋势。孑遗植物,被称为“活化石”,经过了地质和气候环境的剧烈变动一直存活至今,因其特殊的地位,决定了它在全球生物多样性保护中具有不可替代的作用。本研究以孑遗植物银杏(Ginkgo biloba L.)为研究对象,依据122个采样点数据和筛选出的17个环境数据(气候因子、地形因子和人类影响指数),基于随机森林(RF)模型预测银杏在过去(末次冰盛期、全新世中期)、当前和未来(2050s、2070s)四种浓度路径(RCP2.6、RCP4.5、RCP6.0和RCP8.5)适生区分布变化,分别用3种评价指标(AUC、TSS和Kappa)对所构建模型进行评价;根据各因子对建模结果的贡献程度,筛选影响银杏分布的主要因子,绘制因子响应曲线进一步确定适宜银杏生长的因子阈值,分析环境因子的范围对银杏适宜性分布的影响;通过叠加银杏在末次冰盛期和当前时期适生区分布,推断银杏在冰期的潜在避难所并分析形成的原因;对比当前和未来不同浓度路径下银杏的适宜区分布,判断未来气候变化对银杏的扩张或收缩趋势、重心转移轨迹以及空间分布格局;分析人类影响指数参与建模,探讨人类活动对银杏分布和传播的影响。本文主要的研究结论:(1)当前银杏的总适宜生境基本涵盖我国秦岭—淮河线以南到横断山脉以东的大部分地区,包括重庆、湖北、湖南、江西、安徽、浙江、江苏和福建大部分地区,地形区包括长江中下游平原、四川盆地、云贵高原、江南丘陵和浙闽丘陵,为我国亚热带季风气候区。(2)3种评价指标对每个时期模型构建的模型预测精度均成功,按评价结果值从大到小依次为 AUC(0.92～0.96)>TSS(0.7～0.85)>Kappa(0.61～0.79)。(3)影响银杏生长的主导环境变量有6个,分别为Bioll(最冷季平均温度)2℃～10℃、Bio12(年降水量)900 mm～1800 mm、Bio2(平均日较差)7℃～9℃、Bio17(最干季降水量)190 mm、Biol(年平均温度)14℃～20℃,Bio14(最干月降水量)大于32 mm,都会促进银杏的生长,表明银杏的适生环境特征为夏季高温多雨,冬季温和少雨。(4)银杏在冰期的潜在避难所分布在横断山脉南段、华中和华东地区。这些区域地势平坦,受周围高大山脉保护,阻挡了冷空气的侵袭,使得银杏得以保存并延续至今。(5)未来气候变化下,银杏在长江中下游平原沿岸,尤其两湖平原、鄱阳湖平原和沿江平原地区的分布处于收缩状态。从2050s到2070s,随着时间的推移银杏在空间上的分布沿“东—西”方向越来越分散,重心逐渐向高纬度地区转移。(6)银杏的存在概率随着人类影响指数的增加而升高,说明人类活动对银杏的生长起促进作用,主要体现在对银杏古树资源的保护和资源的文化传播。本研究合理预测了气候变化对孑遗植物银杏在长时间尺度下潜在适生区分布,也预测了其在冰期可能的避难所以及在未来气候变化下的空间分布格局变化,为孑遗植物的就地保护和迁地保护提供了一定的理论依据。更多还原

【Abstract】 It is an indisputable fact that climate change,especially global warming,will lead to glacier melting,sea level rising,ecological degradation,and so on.Species habitats will migrate with the change of climate conditions,which further affect ecological diversity.Therefore,the research on species response to climate change based on species distribution models(SDMs)is the mainstream trend.Relict plants,known as"living fossil" plants,have survived the drastic changes of geological and climatic environment until now.Because of their special status,they play an irreplaceable role in global biodiversity conservation.In this study,Ginkgo biloba L.was taken as the study object.Based on 122 sampling sites of G.biloba in China and 17 selected environmental variables including climate,topographic and human influence index,using Random Forest model to predict the suitable distribution of G.biloba in the past(Last Glacial Maximum(LGM),Mid-Holocene(MH)),current,and future(2050s,2070s)under four representative concentration pathways(RCP2.6,RCP4.5,RCP6.0,and RCP8.5).Three evaluation indexes(AUC,Kappa,and TSS)were used to evaluate model accuracy.According to the factor contribution,the main factors affecting the distribution of G.biloba were screened,the suitable threshold and environment condition for the growth of G.biloba was further obtained by factor response curve.By overlaying the suitable habitats of G.biloba in LGM and current,to identify the potential refugia during the glacial period and analyze the formation causes.In addition,we compared the suitable distribution of G.biloba under different RCPs in the future to judge the spatial distribution pattern,the transfer trajectory of the centroid,and the expansion or contraction trend under climate change.Finally,we analyzed the results of human influence index participating in the modeling,and discussed the influence of human activities on the distribution and transmission of G.biloba.The main conclusions of this study are as follows:(1)At present,the suitable habitats of G.biloba basically cover most areas from the south of Qinling-Huaihe Line to the east of Hengduan Mountains,which are mainly distributed in Chongqing,Hubei,Hunan,Jiangxi,Anhui,Zhejiang,Jiangsu,and Fujian Province.The topographic areas include the plains of the Middle-Lower Yangtze plains,Sichuan Basin,Yunnan-Kweichow Plateau,Southeast Hills,which are dominated by subtropical monsoon climate.(2)The evaluation index values constructed by the model in each period are all of the high accuracies,indicating that all models are successful in predicting accuracy.The order of evaluation results is AUC(0.92～0.96)>TSS(0.7～0.85)>Kappa(0.61～0.79).(3)There are six major environmental variables affecting the growth of G.biloba.Bio 11(mean temperature of coldest quarter)is 2℃～10 ℃,Bio 12(annual precipitation)is 900 mm～1800 mm,Bio2(mean diurnal temperature range)is 7 ℃～9 ℃,Bio 17(precipitation of wettest quarter)is 190 mm,Biol(annual mean air temperature)is 14 0C一20 0C,and Bio 14(precipitation of driest month)is more than 32 mm will promote the growth of G.biloba,indicating that the suitable environment is characterized by high temperature and abundant rainfall in summer,mild and less rain in winter.(4)Results show that the potential refuges of G.biloba during the glacial period are located in the southern section of Hengduan Mountains,central and Eastern China.The flat terrain together with the protection of the surrounding high mountains prevented the invasion of cold air so that G.biloba can be preserved and continued to this day.(5)The habitat expansion or contraction of G.biloba in the future is mainly manifested by the shrinkage of the middle and lower reaches of the Yangtze River,especially in the Plain of Hubei and Hunan,Poyang Plain and the Hanjiang Plain.From 2050s to 2070s,the spatial distribution of G.biloba will become more and more dispersed along the "east-west" direction,and the centroid gradually will shift to the northwest high latitudes with the passage of time.(6)The presence probability of G.biloba increases with the increase of human influence index,which indicates that human activities have played a promoting role in the growth of G.biloba,mainly reflected in the protection of ancient tree resources and the cultural communication.This study reasonably predicts the potential habitat distribution of G.biloba in long-term scale under climate change.It also predicts the possible refugia of G.biloba in the glacial period and the change of spatial distribution pattern under future climate change,which will provide a theoretical basis for the in-situ and ex-situ conservation of relic plants.更多还原