融合SIF和反射光谱的小麦条锈病遥感监测

日光诱导叶绿素荧光(SIF)能够敏感反映作物病害胁迫信息，然而冠层几何结构等因素严重影响了SIF对植被光合功能变化及其受胁迫状况的捕捉能力。为此，将能够敏感反映作物群体生物量的归一化差值植被指数（NDVI）和MERIS陆地叶绿素指数(MTCI)与SIF_P相融合(SIF_P-NDVI，SIF_P-MTCI，SIF_P-NDVI*MTCI)，对比分析融合前后SIF对小麦条锈病的遥感监测精度。结果表明：（1）融合反射率光谱指数的SIF_P-NDVI，SIF_P-MTCI和SIF_P-NDVI*MTCI较融合前的SIF_P与病情指数(DI)相关性均有不同程度的提高，其中O₂-B波段提高最为明显，分别提高了23.48%，33.61%和36.49%，O₂-A波段提高量最小，分别提高了2.39%，2.14%和1.51%；（2）以SIF_P-NDVI和SIF_P-MTCI为自变量，基于随机森林回归(RFR)算法构建的小麦条锈病遥感监测模型预测DI值和实测DI值间的R2较SIF_P分别平均提高了1.15%和4.02%，RMSE分别平均降低了2.7%和14.41%；（3）综合利用NDVI和MTCI处理后的SIF_P-NDVI*MTCI为自变量构建的小麦条锈病遥感监测模型精度最优，其预测DI值和实测DI值间的R2较SIF_P平均提高了5.74%，RMSE平均降低了22.52%。研究结果对提高小麦条锈病遥感监测精度具有重要意义，同时亦对其他作物的病害监测具有一定的参考价值。

Monitoring of Wheat Stripe Rust Based on Integration of SIF and Reflectance Spectrum

Solar-induced chlorophyll fluorescence (SIF) can sensitively reflect crop disease stress information, but the geometric structure of canopy and other factors seriously affected the ability of SIF to capture changes in photosynthetic function and stress status of vegetation. Therefore, in this paper, the normalized difference vegetation index (NDVI) and MERIS terrestrial chlorophyll index (MTCI), which can sensitively reflect crop biomass, were integrated with SIF_P (SIF_P-NDVI，SIF_P-MTCI，SIF_P-NDVI*MTCI), and the remote sensing monitoring accuracy of SIF on wheat stripe rust before and after the integration was compared and analyzed. The results show that: (1) at the O₂-B, O₂-A and H₂O absorption at 719 nm bands, integrated reflectance spectral indices of SIF_P-NDVI, SIF_P-MTCI and SIF_P-NDVI*MTCI showed different improvements in correlation with disease index (DI) than SIFP. The O₂-B band increased the most significantly, by 23.48%, 33.61% and 36.49% respectively, while the O₂-A band increased the least by 2.39%, 2.14% and 1.51%, respectively. (2) If SIF_P-NDVI and SIF_P-MTCI were regarded as independent variables respectively, the averaged R2 value of the prediction model based on random forest regression (RFR) algorithm were increased by 1.15% and 4.02%, and the averaged RMSE value were decreased by 2.7% and 14.41%, respectively, compared to those with SIFP as the independent variable. (3) The prediction model based on SIF_P-NDVI*MTCI gave the best performance with an R2 value 5.74% higher than that of SIF_P, and an RMSE value 22.52% lower than that of SIF_P. The results of this paper are of great significance to improve the accuracy of remote sensing monitoring of wheat stripe rust and have a certain reference value for disease monitoring of other crops.