基于MaxEnt的秦艽物种空间分布预测模型的不确定性分析

严胡勇 ¹何运媚 ²张婧月 ³谭蕾 ²唐思萌⁴

扫码查看

作者信息

1. 重庆理工职业学院大数据学院,重庆 401320;重庆工商大学人工智能学院,重庆 400067
2. 重庆工商大学数学与统计学院,重庆 400067
3. 重庆理工职业学院大数据学院,重庆 401320
4. 重庆大学计算机学院,重庆 400030
折叠

摘要

由于参数设置等不确定性因素的变化,同一模型预测的物种生态分布可能会有所不同.因此,量化不同不确定性因素的贡献对于减少生态预测的变化至关重要.然而,很少有研究分析特定模型的建模不确定性.该研究以秦艽为例,探讨其分布预测的不确定性,重点关注参数设置.首先,采用主成分分析法(PCA)和生态变量组法(EVGM)筛选环境因子.参数设置使用 25%的存在点数据和 2种缺失点数据方法作为测试方法,建立 6套模型,探讨存在点测试集比例对模型性能的影响,综合分析训练、测试AUC值和分布面积,确定物种的最佳模型参数,发现 20%的随机测试抽样比例是最佳的.该模型不仅可以为秦艽等野生药材的保护和生态规划提供指导,也能为确定物种空间分布的最优模型提供理论参考.

Abstract

Changes in uncertain factors such as parameter setting can lead to variations in the ecological distribution predicted by the same model.Therefore,quantifying the contributions of different uncertainty factors is crucial for reducing variability in ecological predictions.However,there is limited research analyzing the modeling uncertainty of specific models.This study,using Gentiana macrophylla as an example,explores the uncertainty in its distribution prediction,with a specific focus on parameter settings.Initially,principal component analysis(PCA)and ecological variable grouping method(EVGM)were employed to select environmental factors.Six sets of models were established using 25%of presence point data and two methods for handling missing point data as test methods.The study investigates the impact of the presence point test set proportion on model performance,conducting a comprehensive analysis of training,testing AUC values,and spatial distribution area.The optimal model parameters for species were determined,revealing that a 20%random testing sampling proportion was optimal.This model not only provides guidance for the conservation and ecological planning of Gentiana macrophylla and other medicinal herbs but also serves as a theoretical reference for determining the optimal model for species spatial distribution.

关键词

参数设置/不确定性分析/MaxEnt/秦艽/存在点数据

Key words

parameter setting/uncertainty analysis/MaxEnt/Gentiana macrophylla/presence point data

引用本文复制引用

基金项目

重庆市教委科学技术研究项目(KJQN202215901)

出版年

2024

智慧农业导刊

ISSN：

段落导航