| 引用本文: | 李 磊,齐 识,舒和平,等.黄河干流降水时序分布模型构建及预测研究[J].灌溉排水学报,2026,45(6):147-156. |
| LI Lei,QI Shi,SHU Heping,et al.黄河干流降水时序分布模型构建及预测研究[J].灌溉排水学报,2026,45(6):147-156. |
|
| 摘要: |
| 【目的】解析黄河干流兰州至包头段降水时序变化规律及概率分布特征，并对未来区域降水量及变化趋势进行预测。【方法】基于1956—2016年兰州、喜集水、银川、磴口、临河、五原、乌前旗、包头8个站点的月降水量，采用Mann-Kendall检验和小波分析法揭示降水量时空演变规律，通过Normal、Gamma和Exponential函数建立降水量时序分布模型，并应用SARIMA模型和随机森林（RF）模型进行降水量预测。【结果】①研究区降水突变年份主要集中在1956—1990年及2010年前后，呈约10 a的突变周期；各站点均存在10~20 a的年代际周期、5~9 a的年际周期以及50~60 a的显著长周期。②Gamma分布在超过60%的站点上拟合效果最优，表明降水样本的经验频率呈偏态分布特征。③2024—2036年，区内降水整体呈“总体缓升、局地下滑、年内单峰型”的特征，其中兰州站和五原站降水量轻微下降，而其余站点则出现小幅增长，最大增幅出现在包头站（+1.8%）；兰州、银川、乌前旗、包头站第二季度降水量可达140~210 mm，磴口站和乌前旗站降水波动较高（极差为53~77 mm），银川站和五原站则波动较低（标准差低于5 mm）。【结论】研究区历史降水量呈频发突变、多周期交替及偏态分布特征；未来降水量总体缓升，降水时段更加集中且局地波动增强，建议重点防范降水集中期及局部突发暴雨引发的洪涝风险。 |
| 关键词: 降水量；分布函数；SARIMA模型；随机森林；黄河 |
| DOI：10.13522/j.cnki.ggps.2025343 |
| 分类号: |
| 基金项目: |
|
| Construction and prediction of spatiotemporal variation in precipitation in the main stem of the Yellow River |
|
LI Lei, QI Shi, SHU Heping, WANG Xingkun, CUI Weidong, LI Kuijing
|
|
1. College of water conservancy and Hydropower Engineering, Gansu Agricultural University, Lanzhou 730070, China;
2. Key Laboratory of Cryospheric Science and Frozen Soil Engineering, Northwest Institute of Eco-environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China; 3. Gansu Provincial Agricultural Smart Water-saving
Technology Innovation Center, Lanzhou 730070, China
|
| Abstract: |
| 【Objective】Understanding and predicting precipitation dynamics under climate change is important for water resource management in arid and semi-arid regions. Taking the Lanzhou-Baotou reach of the Yellow River as an example, we analyzed the temporal variation and probability distribution of precipitation and predicted the future changes in precipitation.【Method】Monthly precipitation data collected from 1956 to 2016 from eight stations (Lanzhou, Xijishui, Yinchuan, Dengkou, Linhe, Wuyuan, Wuchuanqi and Baotou) were used in the analysis. Spatiotemporal variations in precipitation in the reach were analyzed using the Mann-Kendall test and wavelet analysis. The precipitation time series distribution was fitted with the Normal, Gamma, and Exponential functions. Future precipitation changes were predicted using the Seasonal Autoregressive Integrated Moving Average (SARIMA) model and the Random Forest (RF) model.【Result】①Abrupt changes in precipitation mainly occurred between 1956 and 1990 and around 2010, with an approximate 10-year cycle. At all stations, we identified interdecadal (10-20 years), interannual (5-9 years), and long-term (50-60 years) cycles. ②The Gamma distribution provided the best fit for more than 60% of the eight stations, indicating that the precipitation distribution is positively skewed. ③The results showed that the precipitation will exhibit a fluctuation but increasing trend from 2024 to 2036, with a single annual peak. Lanzhou and Wuyuan stations are expected to experience a slight decrease in precipitation, while other stations show modest increases, with the largest increase (+1.8%) expected at Baotou. Precipitation from April to June is projected to reach 140-210 mm at Lanzhou, Yinchuan, Wuchuanqi and Baotou. High precipitation variability (53-77 mm) is projected at Dengkou and Wuchuanqi, while lower variability (＜5 mm) is expected at Yinchuan and Wuyuan.【Conclusion】Historical precipitation exhibited frequent abrupt changes, multi-scale periodic oscillations, and skewed distribution. Prediction reveals that precipitation is expected to show a fluctuating increase. These results highlight the possibility of flood risks due to the changes in precipitation pattern. |
| Key words: precipitation; distribution function; SARIMA model; random forest; the Yellow River |
