Remote sensing analysis of rice disease stresses for farm pest management using wide-band airborne data

On-farm pest management and crop protection strongly depend on diagnosis of crop disease stress in fields. In this paper, we first examine the applicability of broadband high-spatial-resolution ADAR (airborne data acquisition and registration) remote sensing data in visible and near infrared regions for rice disease detection and then develop an approach to explore their applicability. Experiments were conducted in 1999 on a large-scale rice field with sheath blight infection in central Arkansas, USA. Based on the measured symptoms, a comprehensive ground disease index (DI) was constructed to indicate the infection severity. Correlations between ground data and image data were analyzed with attempt to develop an applicable method for remote sensing of the rice disease. The results indicate that the broadband remote sensing imagery has valuable capability of application. Some image indices such as the RI/sub 14/, SDI/sub 14/ and SDI/sub 24/ have a correlation of above 0.62, hence are valuable for rice disease identification. A method based on the indices has been developed. Validation with the ground data indicates that the method has an average accuracy of above 70% for classification. The standard estimate error of the method is /spl sim/13%. In spite of this encouraging result, we also realize that it is really difficult to discriminate the healthy plants from light infection ones (DI<20%) because of their heavily overlapping in the estimated image indices. Identification is much more accurate when infection reaches to mediate-to-severe levels (DI>35%). High spectral resolution remote sensing imagery with more bands and narrower bandwidth is required for remote sensing diagnosis of crop disease stress.