Large Scale Monitoring And Forecasting Of Wheat Diseases And Pests Based On Multi-source Remote Sensing Data

With the rapid development of remote sensing technology, remote sensing data has the characteristics of diversity, real-time and dynamic etc. Therefore remote sensing data plays a significant role in monitoring and forecasting occurrence and development of crop diseases and insect pests owe to its advantages of fast, nondestructive, and large scale. However, how to choose and adopt appropriate methods, effective integration of multi-source remote sensing data, to maximize the mining data of information benefits and construct a simple and universal monitoring and forecasting methods is an important issue. Study on the model and method of monitoring and forecasting of wheat diseases and pests using multi-temporal satellite optical data and thermal infrared data at regional scale firstly. At the end of study, we use the Moderate Resolution Imaging Spectroradiometer (MODIS) data of USA observation planning system and National Meteorological grid data of China Meteorological Administration to monitor the main diseases and pests of the wheat producing areas at the national scale, such as wheat stripe rust, powdery mildew and aphids. The specific research contents and results are as follows:1. In the monitoring the area of wheat powdery mildew at regional scale, the study using multi-temporal HJ-CCD data, extract commonly used spectral features which reflect the leaf area index (LAI), chlorophyll (Chl) and canopy structure change mechanism and construct single variable monitoring model and multi-variable monitoring model to monitor wheat powdery mildew occurrence area respectively. Verification result showed that the overall accuracy range of monitoring results based on single phase and single spectrum feature is 68.4%～75.4%. Among them, the overall accuracy of modified soil adjusted vegetation index (MSAVI) is the highest, which may be due to MSAVI can not only reduce the disturbance of soil and vegetation canopy background, but also can better reflect the dynamic changes of leaf area index LAI. In the four original bands, the overall accuracy of near infrared (RNIR) is the highest, and the monitoring accuracy of the other three bands is not good enough. Compared with single phase monitoring model, the accuracy of MSF model based on multi temporal phase was significantly improved, which showed that the change of spectral characteristics of multi temporal phase could help to eliminate the influence of other field stress. The overall accuracy of the MSF-AdaBoost model was the highest, and the omission error and commission error of the disease sample points were 9.5% and 7.3%, respectively, commission error of the healthy sample points was 25%, which were lower than that of the other models.2. Forecasting the occurrence of crop diseases and pests can alert the stakeholders to take preventive measures more effective in real time and reduce yield losses compare to monitoring. Crops growth status and environmental conditions of field has a significant impact on diseases and pests occurrence, development, and dispersal. In this study, we retrieve two parts of information from multi-temporal HJ-CCD optical data and HJ-IRS thermal infrared data, including Ratio Vegetation Index (RVI) and Normalized Difference Vegetation Index (NDVI) monitoring grow status of crop, Perpendicular Drought Index (PDI) and Land Surface Temperature (LST) monitoring the field habitat information. Using relevance vector machine (RVM), support vector machine (SVM) and logistic regression (LR) method to establish forecasting model predict the wheat aphid distribution of filling stage in the Beijing suburbs. To evaluate the difference of performances of the three types of models, we obtain accuracy of models respectively with ground validation samples. The results showed that:prediction accuracy as high as 87.5% by using RVM algorithm which indicate that it has the best performance among the three methods. Additionally, the RVM model needs less computation work and faster prediction speed. These results revealed that the RVM model with high accuracy level and less computing time is much more favorable in practical application.3. Monitoring the occurrence and severity of wheat diseases and insect pests on the national scale. Considering the differences in phenological period of different vegetation types at same year make normalized difference vegetation index (NDVI) variation curve different of different vegetation types within a year, we can extract wheat growing regions of provinces in the Chinese Mainland by using MODIS-NDVI time series products and analysis of vegetation types, main types of crops, cropping systems and the corresponding NDVI curve characteristics of different provinces in China. The occurrence and development of wheat diseases and insect pests are closely related to meteorological conditions, habitat conditions and crop growth. Based on the occurrence mechanism of wheat diseases and pests, this paper combined the remote sensing and meteorological data, and established the integrated warning monitoring model to break through the previous model which simply use the meteorological data or simply use the remote sensing data to monitor. Firstly, the suitable range of occurrence of plant diseases and pests was preliminarily determined by meteorological data. On this basis, combined with the sensitive spectral index of different diseases and insect pests, MODIS-LST products and MODIS-NDVI were used to monitor the surface temperature and wheat growth respectively, the large area monitoring and severity evaluation of wheat diseases and insect pests are realized.