首页 > , Vol. , Issue () : -
Crop diseases have a severe impact on food security, and the excessive use of pesticides in crop disease prevention and control is a common issue. The evaluation of disease habitat suitability is able to provide important information for disease forecasting and control. The occurrence of crop diseases is closely associated with factors such as the growth status of host and environmental conditions. While the disease habitat conditions vary significantly due to cultivation practices and microclimate variations in the field. At present, disease habitat monitoring and evaluation are generally coarse, mainly relying on meteorological information and lacking detailed descriptions of spatially heterogeneous factors such as crop growth status and environmental conditions among fields. In this study, rice sheath blight (RSB), a major disease widespread in rice cultivation, was selected as the research object, the disease surveys were conducted at a county level in 2018 and 2019. Multi-source remote sensing data including optical, microwave and thermal infrared images were used for monitoring the key disease habitat factors. Multi-temporal Sentinel-2 optical images were utilized to extract the planting area of the host crop, which solved the problem of confusing the host with other vegetation in single phase images; the growth status of host was indicated by the tasseled cap products of Sentinel-2 optical images; the status of water layer in rice field was extracted by combining Sentinel-1 microwave images and Sentinel-2 optical images, the optical image of rice region was segmented by object-oriented analysis method to obtain the rice plot boundary in order to eliminate the noise of microwave image; and the MODIS land surface temperature (LST) products were utilized to reflect evapotranspiration and respiration status of rice plants. Based on these remote sensing habitat features of the RSB and a spatial gridding analysis, the habitat suitability evaluation model was established using the partial least squares (PLS) regression method. By validating against the disease survey data, the results showed that the remote sensing information can effectively characterize the disease habitat features. The R2 of the habitat suitability evaluation model was 0.60-0.65, and the RMSE was 0.72 and 0.56, respectively, and the output of the model was in good agreement with the actual spatial pattern of the disease. Besides, the hot and cold spots of the disease habitat suitability map were highly consistent with the actual pattern of disease occurrence in the region. Moreover, the rate of habitat suitability under each disease grade was also analyzed, and the results further confirmed the rationality of the evaluation. Therefore, this study demonstrates the feasibility of utilizing multi-source remote sensing data in evaluating the disease habitat suitability. The disease habitat evaluation map can be integrated into some disease epidemic models to develop spatio-temporal dynamic disease forecasting models at regional scale, and multi-source data, such as meteorological data, remote sensing data, and ground sensor networks, can be incorporated to establish a more comprehensive habitat suitability evaluation model, which is expected to be beneficial for large-scale disease control.