*2.5. Model Calibration*

In this study, the streamflow, sediment concentration, sediment flux and sediment yield were selected as the major variables to calibrate and validate the hydro- and sediment model. A stepwise calibration was performed where first the hydro-parameters and then the sedimentation parameters were calibrated. The streamflow was calibrated first manually through trial and error, and then automatically using the NCAR developed calibration toolbox, which is based on the Dynamically Dimensioned Search (DDS) calibration methodology [30]. Once finished, the calibrated hydro-parameters were used to drive the sediment module and the sediment parameters were then calibrated manually.

#### 2.5.1. Streamflow Calibration

Streamflow was calibrated for a 3-year rainfall event on 17 October 1981 (calibration event hereafter), which started at 21:19 and lasted for approximately five hours. The average rainfall intensity is 14.7 mm/h and total rainfall is 74.4 mm. Calibrated hydro-parameters were selected based on sensitivity analysis and previous studies [23,31,32].

The calibration was carried out through two ways: automated calibration using the NCAR developed calibration toolbox and manual calibration based on trial and error. The reason for calibrating the model in two ways is that the automated calibration tools, which are usually based on standard objective metrics, may weigh more on timing error while weighing less on amplitude error, or the other way around, and thus may result in unreasonable results [33]. A manual evaluation, if executed in a rational way, can take the advantage of both visual inspection and standard metrics. While laborious and highly dependent on researchers' experience, the manual calibration may produce a better result.

Before calibration, a two-year run was performed starting from 1 January 1981 to let the model reach an equilibrium state before the calibration rainfall event. The results from each calibration were statistically evaluated using the correlation coefficient, Root Mean Square Error (RMSE), Nash-Sutcliffe coefficient (NSE) and Kling-Gupta efficiency (KGE). A detailed description of relevant equations is provided in the Supplementary Materials.

The Dynamically Dimensioned Search tool [30] was used for automated calibration. The objective function is the weighted *NSE* and *logNSE*:

$$ObjFn = 0 - \left(\frac{NSE}{2} + \frac{\log NSE}{2}\right) \tag{11}$$

$$NSE = 1 - \frac{\sum\_{t=1}^{T} \left(O\_t - P\_t\right)^2}{\sum\_{t=1}^{T} \left(O\_t - \overline{O\_t}\right)^2} \tag{12}$$

where *NSE* is the Nash-Sutcliffe coefficient, *Ot* is the observed streamflow at time *t*, *Pt* is the modeled streamflow at time *t*, *Ot* is the average of observed streamflow.

Table 1 summarizes the hydro-parameters calibrated by automated calibration with their lower and upper limits and default values. These parameters are mostly related to the land surface model. The hydro-parameters were adjusted within a reasonable bound of values through a 300-iteration automated run. With 32 processors, it took 10 h to finish the automated calibration.

For manual calibration, the refkdt and RETDEPRTFAC in Table 1 were selected as they were identified as the most sensitive parameters by the automated calibration and previous studies [23,31,32]. In addition, channel parameters including channel bottom width (Bw) in meters, channel side slope (Chsslp) and Manning roughness coefficient (MannN) were also calibrated.


**Table 1.** Calibrated hydro-parameters using the NCAR developed calibration tool.

#### 2.5.2. Sediment Calibration

Sediment concentration, sediment flux, and sediment yield were calibrated manually via a series of sensitivity tests following previous studies [9,13,23,34]. Calibrated sediment parameters can be categorized into two groups: soil-type related and land-use-type related. Soil erodibility factor *K* is soil-type related, while cropping-management factor *C* and conservation practice factor *P* are considered to be land-use determined. The calibrated sediment parameters are shown in Table 2.

**Table 2.** Calibrated sediment parameters and values for sediment model.


#### **3. Results**
