基于祁连山西段老虎沟流域断面的径流数据与大本营气象站的气象资料，对冰川区径流与气象要素之间的相关性进行了分析，并建立了多元指数非线性回归方程对径流进行重建以弥补缺失径流，此外，对冰川区径流的年际、季节、日尺度的变化特征进行了分析。结果表明：（1） 冰川区径流量与气温的相关性最高0.86，其次是水汽压（0.81）、相对湿度（0.46）、降水量（0.27），径流受气温影响最大。（2） 21世纪观测资料显示日平均径流量为2.10 m³·s^-1，较20世纪50年代末的1.65 m³·s^-1有所增加，主要因消融季气温上升0.75 ℃所致，强消融期（7、8月）径流量的年际变化较大，消融期初（5、6月）和末（9月）年际变化较小。消融季5—9月产流占比分别为5.3%、16.1%、37.3%、35.1%、6.2%。（3） 多元指数非线性回归方程可以较好地模拟日径流量（纳什效率系数平均为0.70），补充缺失径流后，对于径流的日变化，消融期初和末日变化较小，而强消融期径流的日变化较大。对于径流的时滞效应，老虎沟流域消融季各月径流呈现“谷-峰”的日变化特征，消融期初和末气温达到最大值与径流达到最大值的时间间距较大，而强消融期较小，在消融期初的6月两者相差最大为3 h。明晰冰川融水变化特征对干旱区水资源管理，生态保护及社会经济可持续发展具有重要意义。

Based on the runoff data and meteorological data of Laohugou watershed in the western of Qilian Mountains, northwest China, the research analyzed the relationship between runoff and meteorological variables in glacier area, and established a multivariate exponential nonlinear regression to reconstructe the runoff. In addition, the characteristics of interannual, seasonal, diurnal variations of runoff in the glacial region were analyzed. The results showed that: (1) The correlation between runoff and temperature is the highest (0.86), followed by water vapor pressure (0.81), relative humidity (0.46), and precipitation (0.27). Runoff is most affected by temperature. (2) The mean daily runoff is 2.10 m³·s^-1 in this century, which is higher than 1.65 m³·s^-1 in the late 1950s, the main reason is that the temperature increased by 0.75 ℃ in the ablation season. The interannual variation of runoff is large in the strong ablation period, and the interannual variation is small in the beginning (May-June) and end (September) of the ablation period. The proportion of runoff generation from May to September was 5.3%, 16.1%, 37.3%, 35.1% and 6.2%, respectively. (3) Multivariate exponential nonlinear regression equation can better simulate the daily runoff (the mean Nash efficiency coefficient is 0.70). After the supplement of the missing runoff, the diurnal variation of runoff is small in the beginning and end of the ablation period, but the diurnal variation of runoff is large in the strong ablation period. For the time-lag effect of runoff, monthly runoff in Laohugou watershed showed a characteristic of diurnal variation of “valley-peak” during the ablation period. The time interval between the maximum temperature and the maximum runoff was long in the beginning and end of the ablation period, while the time interval was short between the maximum temperature and the maximum runoff during the strong ablation period, and the maximum difference between the maximum temperature and the maximum runoff was 3 hours in June.