Drone-based photogrammetry for riverbed characteristics extraction and flood discharge modeling in taiwan’s mountainous rivers

Abstract

Flooding is a natural hazard that poses considerable risks to the mountainous regions of Taiwan. Accurate flood modeling can be used for flood prevention, but it requires detailed knowledge of components of riverbed characteristics, which can be challenging to investigate due to the complex terrain in Taiwan's mountains. To overcome these challenges, this study employed an automated grain sizing (AGS) photogrammetry technique using unmanned aerial vehicle (UAV) based orthophotos to extract riverbed characteristics. The AGS model was developed and calibrated in the Linbien River in Western Taiwan, and then the calibrated AGS model was used to extract riverbed characteristics for calculating Manning's roughness coefficient (n value) for flood modeling in the Beinan River in Eastern Taiwan. The study assessed the applicability of AGS in flood modeling using two scenarios: (i) traditional one sample (S1) and (ii) three UAV-based samples (S2) within each cross-section. The model's accuracy was evaluated using the coefficient of determination (R2), root-mean-squared error (RMSE), and mean absolute error (MAE). The results indicated only 3.7% to 9.9% differences of mean representative grain diameters between AGS and PCS. In flood discharge modeling, the R2, RMSE, and MAE of S1 were 0.9948, 106.40 m3/s, and 60.32 m3/s, respectively, while for S2, these values were 0.9994, 34.94 m3/s, and 19.08 m3/s, respectively. The results demonstrate that UAV-based AGS has the potential to enhance flood discharge modeling performance in Taiwan's mountainous rivers, leading to future research focused on refining sampling techniques and incorporating additional data sources for improved AGS-based applications.