Results

4.1 Evaluation of multi-precipitation products

4.1.1 Evaluation at basin-scale

According to Fig. 5 , except TRMM, the data from other precipitation products show a decreasing trend from south-east to north-west, which is consistent with the results of Hu et al(2011). Compared with GO, the precipitation data of IMERG and IMERG_T are the closest, while the precipitation data of TRMM and CFSR are significantly overestimated, and the precipitation data of CMADS are significantly underestimated. Several literatures (Ghodichore et al ., 2018; Graham et al ., 2019; Saha et al ., 2014) found that reanalysis precipitation products obviously overestimated or underestimated observed precipitation.
The further to reflect the difference between the precipitation products and the GO, the PBIAS, CC, and RMES of the precipitation products and the GO were counted on a monthly time-scale. Based on Fig. 4 , the PBIAS values of TRMM, IMERG_T, IMERG, CMADS, and CFSR were characterized by low warm season precipitation and high cold season precipitation. TRMM precipitation data were underestimated in January and February, and overestimated at other times, especially from October to December. IMERG_T precipitation data were underestimated in the rainy season (MayNovember) and overestimated in the dry season (DecemberApril). IMERG precipitation data were underestimated in the dry season (DecemberApril), but IMERG performed best in observing precipitation in the rainy season (average PBIAS = -2.26%). CMADS precipitation data were underestimated in other months except December. The precipitation data of CFSR overestimated the precipitation in all months. Except IMERG, the CC values of other precipitation products also show characteristics of being lower in the warm season and higher in the cold season, among which CFSR has the best correlation with GO (average CC = 0.73), while CMADS, TRMM, IMERG_T, and IMERG perform poorly, with mean CC values of 0.23, 0.01, -0.01, and -0.28, respectively. However, the RMSE values of five types of precipitation products show seasonal characteristics related to the greater precipitation in the warm season and lower precipitation in the cold season in the YRSR (Hu et al ., 2011). IMERG precipitation products have the smallest deviation, with RMSE average of 13.71 mm, followed by CMADS (17.35 mm), CFSR (21.32 mm), IMERG_T (32.42 mm), and TRMM (47.89 mm).
To reveal whether different precipitation products can capture precipitation events within various precipitation intensity groups, we use the probability density function approach to evaluate the daily precipitation intensity (PI), divided PI into nine bins (0 ≤ PI < 0.1, 0.1 ≤ PI < 1, 1 ≤ PI < 5, 5 ≤ PI < 10, 10 ≤ PI < 15, 15 ≤ PI < 20, 20 ≤ PI < 30, 30 ≤ PI < 40, and PI ≥ 40). Based onFig. 5 , IMERG, IMERG_T, CMADS, and CFSR can correctly capture precipitation classifications, but TRMM overestimates high rainfall of > 10 mm/day. IMERG and CFSR overestimate the intensity of all precipitation events, especially CFSR, which significantly overestimates moderate precipitation events of 110 mm/d. The precipitation underestimation by CMADS is mainly concentrated within the range of 120 mm/d, while events within the range of 0.11 mm/d are overestimated.

4.1.2 Evaluation at grid-scale

According to Fig. 6 , the qualities of the TRMM, IMERG_T, IMERG, CMADS, and CFSR were generally better in the south-east than in the north-west. The north-western areas are covered with snow all year round, owing to their high altitude and higher latitude. This leads to poor-quality precipitation observations in this area (Mark et al ., 2016; Noh et al ., 2009). The overestimation of TRMM is the largest with PBIAS of 33.11% to 59.74%, and this gradually increases from downstream to upstream. The precipitation data of CFSR were overestimated except for the station at Dari, while CMADS precipitation data were underestimated except for the station at Maqu. IMERG precipitation data were overestimated in the downstream area and underestimated upstream. Compared with satellite precipitation products (CC of 0.090.40), the reanalysis precipitation products (CC of 0.340.58) have a better correlation with GO. The RMSE values of five precipitation products were large in the south-east and small in the north-west. According to the statistical indicators pertaining to various precipitation products, the overall performance of CMADS precipitation products is the best, with PBIAS of 27.22%2.48%, CC of 0.430.58, and RMSE of 2.684.96 (mm/d), followed by IMERG, CFSR, IMERG_T, and TRMM.
IMERG_T and TRMM have the same detection index value [Figs. 7(a) and (b) ], and the specific reason for this is given inSection 2.2.1 , so here we only analyzed TRMM. According toFig. 7 , the four precipitation products have high detection rates (POD ≥ 0.60), of which CFSR performs best (POD ≥ 0.90), followed by IMERG (0.67 ≤ POD ≤ 0.82), CMADS (0.63 ≤ POD) ≤ 0.84), and TRMM (0.60 ≤ POD ≤ 0.70). FAR values of four precipitation products increase with latitude. Among the four precipitation products, TRMM shows the highest false alarm ratio (0.40 ≤ FAR ≤ 0.57), followed by IMERG (0.40 ≤ FAR ≤ 0.57), CFSR (0.29 ≤ FAR ≤ 0.57) and CMADS (0.30 ≤ FAR ≤ 0.48). CFSR has the highest comprehensive forecasting ability, with a CSI of 0.480.69, followed by CMADS and IMERG, and TRMM exhibits the worst comprehensive forecasting ability. According to the detection indicators of various precipitation products, the overall performance of CFSR precipitation products is the best, with a POD of 0.900.98, FAR of 0.290.51, and CSI of 0.48-0.69, followed by CMADS, IMERG, and TRMM.

4.2 Evaluation of hydrological simulations

4.2.1 Results of streamflow simulation using different precipitation datasets

According to Fig. 8 , the runoff simulation results of Scenario S1 are the best overall, with R 2 and NSE values of 0.85/0.75, 0.84/0.51 in the calibration/validation periods at TNH and 0.81/0.57, 0.80/0.39 in the calibration/validation periods at JM. Scenario S6 performed second best, and in the validation periods (R 2 = 0.78, NSE = 0.53 at TNH;R 2 = 0.64, NSE = 0.53 at JM) yielded the satisfactory performance and outperformed Scenario S1, but it performed poorly in the calibration periods. Scenario S6 underestimates the runoff during the dry season, owing to the CMADS precipitation data being underestimated (Fig. 4 ). The runoff simulation results of Scenarios S3 and S7 were significantly overestimated, and neither TNH nor JM reached a satisfactory performance, especially with respect to Scenario S3 at JM. The reason for this is that the precipitation data of TRMM and CFSR were overestimated (Fig. 4 ), and the precipitation data of TRMM overestimate the upstream precipitation [Figs 3(c) and 6(b) ].
Based on Figs 8 and 9 , the runoff simulation results of Scenario S5 were significantly better than those of Scenario S3, but slightly worse than in Scenario S2. In calibration periods, scenario S2 (R 2 = 0.76, NSE = 0.75 at TNH;R 2 = 0.77, NSE = 0.70 at JM) and S5 (R 2 = 0.70, NSE = 0.65 at TNH;R 2 = 0.66, NSE = 0.66 at JM), the runoff simulation results yielded a satisfactory performance, but the performance of the two in the validation periods was extremely poor (NSE ≤ 0.26). This may be due to the short time-series of precipitation data in Scenarios S2 and S5, and the limited number of calibration times of parameters, which leads to significant differences in the performance of simulation results in the calibration and validation periods. In summary, the runoff simulation results based on GO performed best overall, followed by IMERG, CMADS, CFSR, IMERG_T, and TRMM. IMERG and CMADS precipitation products can be used in this data-scarce alpine region.

4.2.2 Results of streamflow simulation using corrected precipitation datasets

As mentioned in Section 4.1 , the GO and the reanalyzed precipitation products have a high correlation at basin and grid-scales, but the correlation with the satellite precipitation products is poor (Figs 4 and 6 ). Therefore, we only corrected the precipitation data of CMADS and CFSR. We used GO to perform daily-scale regression analysis on CMADS and CFSR precipitation data at basin-scale, owing to scarcity of data in the YRSR. Comparing the fitting effects of different functions, it is found that R 2 of the resulting cubic polynomial is the highest. According to cubic polynomial fitting,R 2 of CMADS is 0.827, andR 2 of CFSR is 0.934 (Fig. 10 ).
Fig. 11 shows that the corrected CFSR precipitation data has improved the simulation results at TNH. The simulation results have changed from unsatisfactory to satisfactory, and theR 2 (NSE) value during the calibration and validation periods increased (increased) by 0.28 (0.34) and 0.22 (0.27), respectively. However, the overall performance of CMADS after correction remains unsatisfactory because the correlation between GO and CFSR precipitation data is better than that of CMADS (Fig. 4 ). Compared with TNH, the corrected CMADS and CFSR precipitation data generate no improvements in runoff model results of JM, and the simulated results remain unsatisfactory.

4.2.3 Results of streamflow simulation using combined precipitation datasets

By using R 2 and NSE indicators, it is found that the simulated results of IMERG and CMADS precipitation data are close to, or even better than, the GO in calibration or validation periods (Figs 8 and 9 ). The performance of CFSR precipitation data after correction is better (Fig. 11 ). Therefore, we choose the combination of CMADS, CFSR_C, and IMERG precipitation data and GO, corresponding to Scenarios S10, S11, and S12. The spatial distribution of precipitation stations is shown in Fig. 1(b) .
According to Table 2 , the overall performance of Scenario S10 combining GO and CMADS is the best, and the simulation results at TNH resulted in good performance (R 2 = 0.77, NSE = 0.72), which is superior to Scenario S1 (R 2 = 0.80, NSE = 0.68) and Scenario S8 (R 2 = 0.59, NSE = 0.50). Although the simulation results at JM yielded unsatisfactory performance, they were close to being deemed satisfactory (calibration periods: R 2 = 0.50, NSE = 0.48; validation periods: R 2 = 0.55, NSE = 0.47). The runoff simulation results of Scenarios S11 and S12 are not as good as those of Scenarios S1 and S2, but slightly better than those of Scenarios S5 and S9.