为了探究随机森林算法在春玉米氮营养指标预测精度，于2021-2022年在东北地区进行氮肥梯度试验，以迪卡159为试验材料，设置10个施氮水平，利用无人机搭载多光谱分别在小喇叭口期(V9)、大喇叭口期(V12)和抽雄期(VT)获取遥感数据。利用19个植被指数分别构建地上部生物量(AGB)、植株吸氮量(PNU)、叶面积指数(LAI)和比叶氮(SLN)模型。结果表明，随机森林算法在预测AGB和LAI有较高的精度，R²分别是0.83和0.9。通过相关分析，LAI、AGB、PNU及SLN与结构不敏感色素指数(SIPI)相关性最高，相关系数分别为-0.75、-0.70、-0.84和0.63，SIPI在4个模型中重要性均最高。研究结果表明，随机森林算法在春玉米氮素监测中具有一定的发展潜力，SIPI在氮素监测中有重要作用，研究结果可为春玉米氮素营养监测提供参考依据。

To investigate the accuracy of the random forest algorithm for nitrogen nutrition prediction in spring maize, this study conducted a nitrogen fertilization gradient trial in northeast China in 2021-2022. 10 nitrogen application levels were set with Deka 159 as the test material, and remote sensing data were acquired at the small trumpet stage(V9), large trumpet stage(V12) and male tapping stage(VT) using a UAV with multispectral. Using 19 vegetation indices, four models were constructed, including aboveground biomass(AGB), nitrogen uptake by plants (PNU), leaf area index(LAI) and specific leaf nitrogen(SLN). The results showed that the random forest algorithm had high accuracy in predicting AGB and LAI, and R2 is 0.83 and 0.9. The correlation analysis of LAI, AGB, PNU and SLN had the highest correlation with the structurally insensitive pigment index(SIPI), with correlation coefficients of -0.75, -0.7, -0.84 and 0.63, respectively; Structurally SIPIis the highest importance among the four models. The results of the study indicate that the random forest algorithm has a certain development potential in spring maize nitrogen monitoring, and theSIPI has an important role in nitrogen monitoring, and the results of the study can provide a reference basis for monitoring nitrogen nutrients in spring maize.

S513.062

内蒙古科技重大专项(2021ZD0003)、国家重点研发计划项目(2016YFD0300605)、中国现代农业产业技术研究体系(CARS-02-25)、财政部和农业部中国农业研究体系以及农业科技创新计划项目(CAAS-ZDRW202004)