为筛选出具有良好降噪效能的城市道路绿化带植物并构建最优配置的植物群落,以广州市8条主干道绿化带为研究对象,分析不同时间段道路绿化带物种类型、结构特征以及环境因子对降噪效能的影响。结果表明:广州道路绿化带植物有42科68属81种（含种下等级，下同）。高峰期道路绿化带的最小声级噪声衰减率(R_Lmin)、最大声级噪声衰减率(R_Lmax)、等效声级噪声衰减率(R_Leq)、声暴露级噪声衰减率(R_SEL)、综合噪声衰减率(R_com)的均值均高于平峰期,整体降噪效能较优。“乔+灌+草”模式道路绿化带的R_Lmin、R_Lmax、R_Leq、R_SEL、R_com值最大,整体降噪效能最优,“乔+灌”和“灌+草”模式次之,“灌木”模式最差。高峰期道路绿化带R_com值的关键影响因子为枝下高、叶面积指数、湿度；平峰期道路绿化带R_com值的关键影响因子为Simpson多样性指数、叶面积指数、Shannon多样性指数和Pielou均匀度指数。高峰期,R_Lmax、R_Leq、R_SEL和R_com值与绿化带宽度、叶面积指数、Pielou均匀度指数、湿度呈极显著(P＜0.01)或显著(P＜0.05)正相关,而与枝下高呈极显著或显著负相关；平峰期,R_Lmin、R_Lmax、R_Leq、R_SEL和R_com值与绿化带宽度、叶面积指数、物种丰富度指数、Simpson多样性指数、Shannon多样性指数和Pielou均匀度指数呈极显著或显著正相关。综合研究结果显示:在绿化带植物群落构建上,应推行“乔+灌+草”模式,其中乔木层应选择小叶榄仁(Terminalia neotaliala Capuron)等冠幅和胸高断面积较大、枝叶繁茂的物种；灌木层应选择朱蕉〔Cordyline fruticosa (Linn.) A. Chev.〕等叶片数量多、叶表面被毛、质地厚且分枝低的物种；草本层应选择蓝花草(Ruellia brittoniana Leonard)等叶片狭长且生长密集的物种。

Taking the green belts along eight major roads of Guangzhou City as research objects, the effects of plant species types, structural characteristics, and environmental factors on noise reduction efficiency of roadside green belts at different time periods were analyzed, and the urban roadside green belt plants with superior noise reduction efficiency were screened and the optimal plant community configuration pattern was constructed. The results show that there are 81 species (including infraspecies, the same below) of plants belonging to 68 genera of 42 families in roadside green belts of Guangzhou City. The averages of noise attenuation rate of the minimum sound level (R_Lmin), noise attenuation rate of the maximum sound level (R_Lmax), noise attenuation rate of the equivalent sound level (R_Leq), noise attenuation rate of the sound exposure level (R_SEL), and the comprehensive noise attenuation rate (R_com) of the roadside green belt at peak period are higher than those at off-peak period, and the overall noise reduction efficiency is better. RLmin, R_Lmax, R_Leq, R_SEL and R_com values of the “tree+shrub+grass” pattern roadside green belt are the largest, the overall noise reduction efficiency is the best, followed by the “tree+shrub” and “shrub+grass” patterns, and the “shrub” pattern is the worst. The key influence factors for Rcom value of the roadside green belt at peak period are height under branch, leaf area index, and humidity; those of the roadside green belt at off-peak period are the Simpson diversity index, leaf area index, Shannon diversity index, and Pielou evenness index. At peak period, R_Lmax, R_Leq, R_SEL, and R_com values show highly significant (P＜0.01) or significant (P＜0.05) positive correlations with green belt width, leaf area index, Pielou evenness index, and humidity, while show highly significant or significant negative correlations with height under branch; at off-peak period, R_Lmin, R_Lmax, R_Leq, R_SEL, and Rcom values show highly significant or significant positive correlations with green belt width, leaf area index, species richness index, Simpson diversity index, Shannon diversity index, and Pielou evenness index. The comprehensive research results indicate that for green belt plant community construction, the “tree+shrub+herb” pattern should be carried out, among which the tree layer should select species such as Terminalia neotaliala Capuron with large crown width and basal area at breast height, and dense foliage; the shrub layer should select species such as Cordyline fruticosa (Linn.) A. Chev. with numerous leaves, hairy leaf surface, thick texture, and low branching; the herb layer should select species such as Ruellia brittoniana Leonard with narrow and elongated leaves and dense growth pattern.