Figure 1
In order to investigate the sensitivity of the number of climate observations on the large-scale hydrological modeling, the following research tasks are proposed. For the precipitation, three cases in the data network as shown in Figure 1 can be defined. First, 100 precipitation stations in the East river basin will be used to generate rainfall Thiessen polygons as the input climate forcing to the hydrological model. Second, 11 out of 100 stations which has multi-year rainfall measurements will be used to generate polygons and force the model. This simulation will be followed by the simulation using the basin-average precipitation (i.e., arithmetic average from the 100-station precipitation observations). For the evaporation, two cases similar to the precipitation as also shown in Figure 1 can be defined base on the 8-station potential evaporation measurements in the East River basin.
A set of optimal model parameters will be identified based on the simulations forced by the largest numbers of climate observations (i.e., case 1 in Figure 1). This set of parameters will be used in the subsequent simulations in which the numbers of climate stations gradually decrease. The objective is to examine the degree to which the model performance (judging from the comparisons with the observed hydrographs in multiple locations of the basin) would be impacted by the availability of the climate forcing measurements, mainly precipitation and potential evaporation.
Figure 1
In order to investigate the influence of the number of calibration targets (mainly streamflow measurements in this study) on the calibrated effective parameters and large-scale hydrological model performance, the following research tasks are proposed. Four levels of complexity in the calibration targets can be identified, as elaborated in Figure 2. It starts from the model calibration from using only single station of streamflow measurement at the outlet of the basin, as usually encountered for the hydrological modeling in the data-scarce regions. The next is to calibrate the effective parameters from 4-5 streamflow stations along the mainstream. Then, another 4-5 different streamflow observations in the tributaries of the basin will be added into the calibration targets to further improve the model performance. Notice that each of the above three cases will yield a set of optimal parameters, whereas no spatial variability of the calibrated parameters has yet to be account for. The last (highest) level of complexity is to calibrate the parameters individually from sub-catchment to sub-catchment based their respective streamflow condition in order to characterize the spatial heterogeneity of the calibration parameters. This last complexity will allow studying the perhaps most important issue for the PUBs, namely the transferability of model parameters from basin to basin.
Figure 2
2003 Pat Yeh and Lincoln Fok