在淮河流域蚌埠站以上区域构建人工神经网络(ANN)、HBV-D 模型和SWIM模型等水文模型,评估不同时间、空间尺度和数据基础下水文模型的适宜性。得出:①时间尺度上三个模型对数据要求不同, ANN模型需要月尺度数据即可建立降水-径流关系且能取得较好的模拟效果,HBV-D 和SWIM模型为日尺度水文模型,需逐日降水、温度和径流量等数据,SWIM模型还需作物管理、营养盐和土壤侵蚀等数据;②空间尺度上,ANN模型适应于大尺度,HBV-D 模型适用于1×10⁴ km²及以上流域,SWIM模型更适合于1×10⁴ km²以下小流域降水-径流关系建立;③模拟效果分析,月尺度统计上ANN模型对水文模拟的整体效果较好,但不适合用于气候变化背景下水文水资源等研究,而有物理基础的HBV-D和SWIM模型虽模拟的纳希效率系数不及ANN模型,但在气候变化背景下仍是较好的工具。

In this paper,the applicability of three hydrological models, including artificial neural network (ANN) model, Hydrologiska Byråns Vattenbalansavdelning-D (HBV-D) model and Soil and Water Integrated Model (SWIM), are examined at different temporalspatial scales and databases in the Huaihe River basin which is above the Bengbu hydrological gauging station. It is found that ANN model only needs monthly data to build rainfall-runoff relationship and can obtain well simulation results, but HBV-D and SWIM models require data on daily scale such as daily precipitation, daily temperature and daily runoff. SWIM model even requires crop management data, nutrient data, soil erosion data etc. In addition, on spatial scale, the applicability of ANN model is adequate to large-scale basin, SWIM model may only be suitable for small-scale basin with an area of less than 10000 km², and HBV-D model can apply to a basin of about 10000 km². Furthermore, according to simulation results, ANN model can get better result for overall hydrological simulation, but it is not suitable for the hydrological and water resources research under climate change. Although their Nash-Sutcliffe coefficients are less than ANN model, the physically based distributed hydrological model, HBV-D model and the SWIM model are good tools to study impacts of climate change, which is significantly controlled by model structure.

利用1998 -2012年重庆市34个站点逐时降水观测资料, 分析了重庆地区高时空分辨率卫星产品（Tropical Rainfall Measuring Mission, TRMM; Climate Prediction Center morphing technique, CMORPH）和再分析资料（ECMWF Re-Analysis Interim, ERAIN; Modern-Era Retrospective Analysis for Research and Applications, MERRA; Climate Forecast System Reanalysis, CFSR; Japanese 55-year Reanalysis, JRA55）的可靠性, 并进一步对卫星和再分析资料降水日变化的准确性进行了评估。结果表明, 再分析数据总体上高估了重庆的日降水量。在重庆西南和东南部, TRMM比较接近观测数据。CMORPH在东北部表现最好, 低估了其他地区的降水量。就降水量而言, 卫星资料的结果好于再分析资料。对于日平均降水强度和降水频率, 卫星资料的分布形态与观测资料差异较大, 再分析资料与观测则较为一致。观测资料中降水峰值多出现在午夜, 沿着西南至东北方向有滞后现象。卫星资料各降水要素（PA， PI和PF）也表现出由西南至东北方向峰值时间滞后的现象。对于不同地形特征上的降水日变化, 卫星和再分析资料的表现有明显差距, 尤其是在夏季山地地区, 再分析资料高估了12:00 -17:00（北京时）的降水。

The hourly precipitation data observed at 34 gauge stations in Chongqing during 1998-2012 is compared with the satellite products [Tropical Rainfall Measuring Mission (TRMM), Climate Prediction Center morphing technique (CMORPH)] and reanalysis data [ECMWF Re-Analysis Interim (ERAIN), Modern-Era Retrospective Analysis for Research and Applications (MERRA), Climate Forecast System Reanalysis (CFSR), Japanese 55-year Reanalysis (JRA55)].The reliability of the precipitation products in high spatiotemporal resolution is analyzed, and the accuracy in diurnal variation is also evaluated.The results show the daily precipitation in Chongqing is generally overestimated by the reanalysis data while the satellite products show agreement with the observations.TRMM is accurate and close to the observational data in the southwest and southeast.CMORPH works best in the northeast while underestimates the rainfall in the other areas.There are evident differences in precipitation intensity and frequency between the satellite and observational data, while the reanalysis data appears to be in agreement with the observations.The peak time of precipitation diurnal variation is mostly from midnight to dawn and displays a hysteresis from the southwest to northeast.The precipitation amount (<i>PA</i>), intensity (<i>PI</i>) and frequency (<i>PF</i>) in the satellite products show the hysteresis from the southwest to northeast, too.There is significant difference in diurnal cycle of precipitation between the reanalysis and satellite data over complex terrain.The reanalysis data overestimates the precipitation from 12:00 to 17:00(Beijing time), especially in the mountainous areas in summer.

为了探讨ERA5再分析降水数据在中国区的适用性,以全国728个站点的日降水数据为参考,使用Pearson相关系数、均方根误差、平均绝对误差、探测率、误报率以及公正先兆评分分析了ERA5再分析降水数据在不同时间尺度（月、季）、不同气候区、不同海拔梯度下的精度以及ERA5再分析数据对暴雨和干旱事件的刻画能力。结果表明：ERA5降水数据对日降水事件的识别能力在空间和时间上均存在差异,整体来看：在北温带的精度最高;在夏、秋两季较冬、春两季的精度低;海拔&gt;500 m地区的精度低于海拔&#x02264;500 m地区的精度。ERA5数据在进行暴雨识别时,与站点观测数据的偏差较大,且阈值越大（即暴雨越强）偏差越大;基于ERA5计算的不同尺度的标准化降水指数的精度不同,其中3个月尺度的标准化降水指数的精度最高。在进行干旱事件的识别时,阈值越低（即干旱越严重）误差越大。本研究可为ERA5降水数据的使用范围和使用方法提供参考,有助于分析相关研究的不确定性。

ERA5 is the latest generation of ECMWF reanalysis product. Its precipitation data can be used to analyze the spatial and temporal characteristics of precipitation, extracting rainstorm and flood events, driving crop models, hydrological models, and land surface models. Studies on the effects of using reanalysis data to drive various models to simulate climate change are becoming more common in the context of climate change. However, as atmospheric reanalysis data, the data quality of ERA5 precipitation data will be affected by the errors of forecast products, observation data, and assimilation methods. The precision of reanalysis data will influence the uncertainty of related study results. As a result, verifying and evaluating the accuracy of reanalysis data is both necessary and urgent. To investigate the applicability of ERA5 precipitation data in China, daily precipitation data from 728 Chinese sites were used as a reference to examine the accuracy of ERA5 reanalysis precipitation data at various time scales (month, quarter), different climate zones and different elevation gradients using the Pearson correlation coefficient (r), root mean square error (RMSE), mean absolute error (MAE), probability of detection (POD), false alarm rate, and equitable threat score. Furthermore, the capability of ERA5 to depict heavy rain and drought events was investigated. The results demonstrate that there are spatial and temporal differences in the ability of ERA5 precipitation data to identify daily precipitation events. Among the eight climatic zones, the north temperate zone has the highest accuracy. The r, RMSE, and MAE of ERA5 daily precipitation and station observation are 0.587, 4.040 mm·d^-1, and 1.472 mm·d^-1, respectively, in the north temperate zone. Precipitation data from the ERA5 are less accurate in the summer and autumn than in the winter and spring. The accuracy of ERA5 precipitation data in areas with elevations greater than >500 m is lower than in areas with elevations of less than ≤500 m. When the ERA5 data is used to identify the heavy rain, there is a large deviation from observations of the stations, and the larger the threshold (i.e., the stronger the heave rain), the larger the deviation is. The accuracy of standardized drought index (SPI) of different time scales calculated using ERA5 is different, varies with SPI on a three-month time scale having the highest accuracy. When using ERA5 data to identify drought events, the lower the threshold (that is, the greater the severity of the drought), the greater the error. Generally speaking, the drought capture ability of the north temperate zone, south temperate zone, and north subtropical zone is better. This research broadens the scope of validation of climate data sets and serves as a model for future research. These findings serve as a reference for related studies considering whether to use ERA5 data, as well as a tool for analyzing the uncertainties of related studies.

利用1982&#x02014;2018年内蒙古地区115个气象站点的月降水观测资料评估了FLDAS、ERA5、CRA40/Land和GLDAS 4种陆面同化及再分析降水资料在内蒙古地区的可靠性。结果表明：（1） 4种降水资料均能较好地表征降水量在内蒙古地区从东北向西部递减和冬季降水少、夏季降水多的时空变化特征。（2） 相关系数、绝对平均误差、均方根误差和纳什效率系数的计算结果表明,4种降水资料在夏季表现最优、冬季表现最差,在东部半湿润区和半干旱区好于西部干旱区和极干旱区。（3） ERA5资料在绝大多数时间对内蒙古的降水有高估现象,GLDAS却对降水存在低估,尤其是在冬季,GLDAS对固态降水几乎没有观测能力,而新发布的FLDAS资料表现较好。总体来看,相对于ERA5和GLDAS,FLDAS和CRA40/Land降水资料与观测值之间的差别最低,有着最优的统计特征。

Monthly precipitation data observed at 115 weather stations in Inner Mongolia during 1982-2018 was compared with land data assimilation and reanalysis precipitation datasets [Famine Early Warning Systems Network (FEWS NET) Land Data Assimilation System (FLDAS), the Fifth Generation of ECMWF Reanalysis (ERA5), China Meteorological Administration, Global Land surface Reanalysis (CRA40/Land), and Global Land Data Assimilation System (GLDAS)]. The reliability of the four precipitation datasets was analyzed and accuracy of monthly variation was also evaluated. The results showed that the four precipitation datasets captured features of precipitation distribution. Precipitation levels decreased from the northeast to west and were lower in winter and higher in summer. By analyzing correlation coefficients, mean absolute errors, mean bias errors, root mean square errors, and Nash-Sutcliffe efficiency coefficients, the four precipitation datasets showed the best performance in summer and the worst performance in winter, and better values in eastern semi-humid and semiarid areas than western arid and extremely arid areas. Compared with ERA5 and GLDAS, FLDAS, and CRA40/Land showed the lowest differences from observed values and the best statistical characteristics. ERA5 data mostly overestimated precipitation in Inner Mongolia during most of the study period, while GLDAS underestimated it. Particularly in winter, GLDAS showed almost no observation capability for snow precipitation, while the newly released FLDAS data performed better. Overall, FLDAS and CRA40/Land precipitation datasets performed best with better statistical power.

根据长江源区直门达水文站1957—2021年径流资料,使用长、短期径流旱涝急转指数(LRDFAI和SRDFAI),分析长江源区旱涝急转事件发生变化情况,采用希尔伯特-黄变换方法,分析长江源区旱涝急转强度周期变化。结果表明:长江源区直门达水文站短期径流旱涝急转事件主要发生在3—9月份,“旱转涝”和“涝转旱”事件发生频次相当。发生频次在不同年代波动较大,总体上呈现“多-少-多-中-多-多”的趋势。短期旱涝急转事件中“旱转涝”事件强度呈现微弱的增长趋势,而“涝转旱”事件强度呈现微弱的递减趋势,分别存在7.1 a和12.3 a的主周期。共发生长期径流旱涝急转事件20次,“旱转涝”和“涝转旱”事件发生频次分别为8次和12次,存在8.3 a和14.2 a的周期。

According to runoff data from the Zhimenda hydrological station from 1957 to 2021, we analyzed the abrupt alteration between drought and flood in the Changjiang River source area by using the Long Runoff Drought-Flood Area Index (LRDFAI) and the Short Runoff Drought-Flood Amplitude Index (SRDFAI). We also investigated the periodical change of the intensity of the alteration events by the Hilbert-Huang Transform method. Results reveal that short-term alteration events in the study area mainly occur from March to September with comparable frequencies of “drought to flood” and “flood to drought” events. The frequencies of these events fluctuate considerably in different years and exhibit an overall trend of more-less-more-medium-more-more. The intensity of short-term “drought-to-flood” events displays a weak increasing trend, while the intensity of “flood-to-drought” events shows a weak decreasing trend. Both types of events exhibit main cycles of 7.1 years and 12.3 years. The area witnessed 20 long-term runoff alteration events with 8 and 12 occurrences of “drought to flood” and “flood to drought” events, respectively, and cycles of 8.3 and 14.2 years.