*3.2. Sediment Calibration*

Sediment concentration and sediment flux results are shown in Table 5 and Figure 5. Overall, the sediment concentration and sediment flux show good agreement with measurements at the outlet, indicating that the model can well reproduce the sediment erosion and transport processes on a watershed scale. Compared with streamflow, sediment concentration and sediment flux are less well predicted, with the NSE value of 0.25 and 0.28, respectively. For both sediment concentration and sediment flux, the simulated rising limbs fit well with the measured ones, while the peak values are underestimated. The falling limbs recess earlier but last longer. The long recession should inherit from that of the streamflow, and essentially the overland flow. The total sediment yield simulated during this event is around 9600 tons, which is 40 percent lower than the measurement. This error is close to the one reported by [13], which is considered to be acceptable in hydrological engineering.

**Table 5.** Performance of the simulation of sediment concentration and sediment flux for calibration event.


We notice that the model simulated only one peak for sediment flux and sediment concentration, while the observation exhibited multiple peaks (Figure 5). Such a mismatch can be attributed to the uncertainty of observation or limitation of model algorithm. Due to the large fluctuation of sediment concentration during a rainfall event, the point-wise observation data might not be able to detect the real condition of sediment concentration. Thus, the multiple peaks shown in Figure 5 might not be the real case. If we assume that the observed multiple peaks are the real condition, the multiple peaks or rapid fluctuation of the sediment concentration and sediment flux might be attributed to local erosion of channel or bank collapse, which are currently not represented in the model.
