For better risk management, detailed and quantitative measurement of channel and stream‐bed structure is required to understand and predict water and sediment flow in mountain channels. Our previous research demonstrated good performance of green‐wavelength Terrestrial laser scanning (TLS) for measurement of submerged stream‐bed in a steep mountain channel. This paper examines how the acquisition protocol of TLS affects the accuracy of data collected in the mountain channel. First, it was tested whether varying the scanner height, i.e., incident angle affects the data acquisition in terms of point density and accuracy in the pool unit of step‐pool channel. Then, the effect of varying the minimum point spacing on the derived Digital terrain model (DTM) was examined. It was also analyzed whether a combination of multiple TLS data acquired from different directions would improve the accuracy of data compared to data acquired by a single measurement. Furthermore, TLS data were acquired over a cascade unit of the channel and examined whether TLS is capable of capturing reliable underwater data. All the acquired underwater data by TLS were corrected for water refraction and validated using manual surveyed data. The results showed that the accuracy of derived DTM was improved when the scanner height was increased or data was acquired from multiple directions, however, acquiring denser point cloud with a minimum point spacing of 1 mm did not improve the accuracy of the data. Accuracy of TLS measurement in the cascade unit was considerably lower. Special consideration is required for this area.