随机森林
机器学习
人工智能
特征选择
超参数
栖息地
计算机科学
特征(语言学)
环境科学
草原
气候变化
生态学
物种分布
变量(数学)
小气候
概率分布
植被(病理学)
分布(数学)
支持向量机
降水
环境生态位模型
可解释性
草原
植被分类
自然地理学
空间分布
生态系统
联轴节(管道)
遥感
集成学习
作者
Xiaolong Ding,Youchen Zhen,Kuanyan Tang,Narisu,Fangzheng Yue,Namuriga,Ning Wang,Guangjun Wang
标识
DOI:10.1134/s1995425526700113
摘要
Abstract This study systematically identified the key drivers of Myrmeleotettix palpalis habitat distribution using integrated machine learning models and ecological mechanism analysis. Through Pearson correlation matrix-based feature selection and hyperparameter optimization, we retained 15 environmental variables across topographic, climatic, and surface parameter dimensions. Random Forest and MaxEnt models were employed to quantify variable importance and stability, revealing a “tripartite regulatory framework”: (1) Topographic factors (Aspect, DEM) mediate microclimate through solar radiation redistribution; (2) Climatic factors (Bio18: warmest quarter precipitation, Bio4: temperature seasonality) serve as core energy-water suppliers; (3) Surface parameters (soil sand content, land use) constitute the resource foundation. The optimized MaxEnt model exhibited excellent performance (AUC = 0.966), identifying optimal habitats concentrated in the central-eastern Inner Mongolia steppe (3.67 × 105 km2) with a critical precipitation threshold (Bio18 > 320 mm). Our composite feature evaluation framework (importance-weighted 0.6 + stability-weighted 0.4) reliably prioritized ecologically robust drivers (Aspect composite score = 0.94, Bio18 = 0.81). These findings provide scientific targets for precision locust control and ecological intervention in grassland ecosystems.
科研通智能强力驱动
Strongly Powered by AbleSci AI