Flood runoff was calculated from continuous surface velocity data by applying SIV (Surface Image Velocimeter) to various flood runoff events in ephemeral streams of Jeju Province. In the process of analyzing surface velocity of collected runoff images, a method of improving analytical accuracy of nighttime runoffimages, the biggest weakness of SIV, was proposed to improve the limit and compare accuracy of daytime and nighttime runoff images. In addition, error rate was analyzed by calculating variables for far distance and near distance from the collected runoff images during conversion of image coordinate system. Cross-correlation coefficient of SIV was analyzed to present a post-processing method for improved accuracy of runoff calculation. Runoffimages of disaster monitoring CCTVs installed in streams of Jeju Province were used to analyze runoff of measured and non-measured watersheds. Advanced technology for runoff observation was studied by considering rainfall-runoff characteristics of ephemeral streams. Based on comparison of surface velocity analysis on nighttime runoff images with observed surface velocity of Kalesto, coefficient of determination was 0.848. Daytime runoff images had of 0.891. Lighting device of 1,000W installed to make up for weakness of nighttime runoff images was extremely effective in increasing accuracy of analysis on nighttime runoff images of streams in Jeju Province, which have narrow stream width.
Runoff, calculated by rate of velocity change (16.77%∼317.69%) from input error of far distance reference point in the process of setting reference point for SIV, had arange of 0.82%~338.63%. If runoff is calculated using SIV in inland streams with wide stream width, minor reference point input error can lead to greater velocity error. Error rate must be reduced in the process of setting reference point. Cross-correlation coefficient of continuous image analyzed using SIV was interpreted to propose an appropriate filtering value for the collected runoff images. For daytime runoff images, 61.0% of 59 velocity vectors corresponded to correlation coefficient range of 0.9∼1.0. For nighttime runoff images, the ratio of vectors in correlation coefficient range of 0.9~1.0 was 17% lower. This indicates that quality of nighttime runoff images is lower than daytime runoff images. Therefore, reasonable filtering of correlation coefficient is necessary on runoff images obtained during calculation of runoff using SIV. A runoff observation system that accounts for peculiar runoff characteristics of ephemeral streams was proposed based on field observation experience and the results of using SIV on various flood runoff images in ephemeral streams caused by heavy rains and typhoons. To improve accuracy of observed runoff data, observation angle of cameras and effective installation of 1,000W lighting device were presented. This increased accuracy of nighttime runoff image analysis, which is a weakness of the conventional SIV. In addition, ephemeral streams of Jeju Province that have extremely narrow stream width compared to inland streams showed advantage of using CCTV. Water bubbles formed on water surface during flood runoff functioned as tracer particles appropriate for surface velocity analysis. Future studies shall be conducted on continued runoff observation and creation of an optimal observation system that accounts for peculiar hydrological characteristics and runoff phenomena in ephemeral streams of Jeju Province.