Comparison of the Positional Accuracy of Stream Mapping Methods: Considering the Effects of Minimum Contributing Area and Spectral Data

Numerous remotely sensed stream-mapping methods have been developed in the last several decades that use an array of remote sensing data in order to improve small stream management. One of the currently prevailing methods is the Digital Elevation Model (DEM)-based D8 method. Since the D8 method only uses the DEM-topographic data without using any information in relation to water existence, the Local Difference Algorithm (LDA) is suggested to utilize spectral information of riparian zones as well as DEM-topographic information.;While the positional accuracy of the two stream maps needs to be evaluated and compared to determine map quality, the estimation or comparison of the positional accuracy of small stream map products has not been well studied, and rarely attempted by remotely sensed stream map developers.;Therefore, the objectives in this study are (1) to compare the positional accuracy of the stream maps from D8 and LDA, (2) to better understand the limitation of LDA and (3) to contribute to the establishment of an efficient and appropriate way to measure and compare the positional accuracy of stream maps.;I suggested new methodologies in sampling and statistical analysis. For sampling, I suggested theoretically, application of a line-transect method and practically, a road crossing stream points-based method. The Producer's Accuracy, User's Accuracy and KHAT were proven to be appropriate positional accuracy estimates for stream maps. In addition, a modified McNemar test was developed to compare the positional accuracy estimates of two different stream maps. Using the three accuracy estimates and statistical comparison methods, the LDA and D8 maps were compared in consideration of the nine spectral bands and seven Minimum Contributing Areas (MCAs).;Through the study, I found that, first, the LDA maps were not statistically significantly better than the D8 maps. However, there is an indication that the LDA may outperform the D8 if higher resolution of spectral data is used. Second, the MCAs were found to significantly affect the positional accuracy of stream maps, while the spectral bands had little effects on it. Finally, I demonstrated the feasibility of the methodologies and procedures in estimations and comparisons of stream maps.