Identification of physical processes inherent in artificial neural network rainfall runoff models

The emergence of artificial neural network (ANN) technology has provided many promising results in the field of hydrology and water resources simulation. However, one of the major criticisms of ANN hydrologic models is that they do not consider/explain the underlying physical processes in a watershed, resulting in them being labelled as black‐box models. This paper discusses a research study conducted in order to examine whether or not the physical processes in a watershed are inherent in a trained ANN rainfall‐runoff model. The investigation is based on analysing definite statistical measures of strength of relationship between the disintegrated hidden neuron responses of an ANN model and its input variables, as well as various deterministic components of a conceptual rainfall‐runoff model. The approach is illustrated by presenting a case study for the Kentucky River watershed. The results suggest that the distributed structure of the ANN is able to capture certain physical behaviour of the rainfall‐runoff process. The results demonstrate that the hidden neurons in the ANN rainfall‐runoff model approximate various components of the hydrologic system, such as infiltration, base flow, and delayed and quick surface flow, etc., and represent the rising limb and different portions of the falling limb of a flow hydrograph. Copyright © 2004 John Wiley & Sons, Ltd.     

An integrated hydrological model for rain‐induced landslide prediction

This paper describes an extension to the Combined Hydrology And Stability Model (CHASM) to fully include the effects of vegetation and slope plan topography on slope stability. The resultant physically based numerical model is designed to be applied to site‐specific slopes in which a detailed assessment of unsaturated and saturated hydrology is required in relation to vegetation, topography and slope stability. Applications are made to the Hawke's Bay region in New Zealand where shallow‐seated instability is strongly associated with spatial and temporal trends in vegetation cover types, and the Mid‐Levels region in Hong Kong, an area subject to a variety of landslide mechanisms, some of which may be subject to strong topographic control. An improved understanding of process mechanism, afforded by the model, is critical for reliable and appropriate design of slope stabilization and remedial measures. Copyright © 2002 John Wiley & Sons, Ltd.     

Momentum transport of wave zero during March: A possible predictor for the Indian summer monsoon

Analysis of monthly momentum transport of zonal waves at 850 hPa for the period 1979 to 1993, between ‡S and ‡N for January to April, using zonal (u) and meridional (v) components of wind taken from the ECMWF reanalysis field, shows a positive correlation (.1% level of significance) between the Indian summer monsoon rainfall (June through September) and the momentum transport of wave zero TM(0) over latitudinal belt between 25‡S and 5‡N (LB) during March. Northward (Southward) TM(0) observed in March over LB subsequently leads to a good (drought) monsoon season over India which is found to be true even when the year is marked with the El-Nino event. Similarly a strong westerly zone in the Indian Ocean during March, indicates a good monsoon season for the country, even if the year is marked with El-Nino. The study thus suggests two predictors, TM(0) over LB and the strength of westerly zone in the Indian Ocean during March.     

月降水量的神经网络混合预报模型研究

以均生函数表征预报量自身周期变化，结合500hPa月平均高度场和月平均海温场预报因子，采用神经网络方法建立了一种新的短期气候预报模型。分别以广西桂北、桂中和桂南6月降水量作为预报对象进行预报试验，结果表明，这种新的预报方法比均生函数回归预报模型及高度场、海温场预报因子的回归预报模型，具有更好的物理基础和预报能力。     

山西省主要河流流域面雨量预报业务流程

以T213、HLAFS模式、MM5中尺度模式输出的格点资料以及日本降水量格点资料为基础，将影响山西降水的天气动力模型归纳为诊断模型，从中引出多个能够全面反映降水模型特征的综合物理因子；根据各种数值模式输出的降水量预报性能和质量优劣特点，依据数值模式的形势场预报优于要素场预报的现实，构造在不同环流形势背景下，启动不同预报方程的面雨量预报业务流程，有效地遏止了在环流形势调整时预报输出不能快速响应的弱点，提高了点和面雨量预报的准确度。     

TRMM卫星微波成像仪分级产品及其反演降水算法

文章叙述了获取定量降水信息的意义，简要介绍了对热带测雨卫星TRMM(Tropical Rainfall Measurement Mission)的仪器、美国国家宇航局提供的微波成像仪TMI(TRMM Microwave Imager)分级产品。对比了物理方法和经验方法反演降水的特点，并对一些经验方法以及倾斜对流系统对反演降水的影响、动态聚类分析、神经网络反演方法的研究成果进行了介绍。     

四川盆地及东部山区暖季降水的日变化特征

利用国家气象信息中心1998—2018年分辨率为0.1°×0.1°的降水融合资料和ERA5小时再分析数据,分析了四川盆地及东部山区6—9月降水的日变化特征及成因。研究表明,四川盆地(简称盆地)降水量和频次有相似的日变化特征,盆地降水主要集中在夜间至清晨,降水高值区东传到达盆地东部山区后,降水量和频次都减少,东部山区没有明显的降水东传特征。夜间至清晨,盆地东南山地背风坡形成了较强的下沉气流,促进盆地低层形成质量堆积,同时盆地气旋性涡旋、水汽输入和大气层结不稳定的增强,造成盆地降水增强。盆地东南地区水汽通量方向自东南转为偏南以及盆地低层水汽辐合区向东扩展为盆地降水高值区向东传播提供水汽条件。盆地东部山区降水量峰值时间呈早晨和午后的双峰型分布,与地形触发局地性降水有关。受大气层结稳定性、山地气旋环流和山地—平原螺线管环流日变化的影响,盆地东部山区降水频次峰值时间呈午后的单峰型分布。     

SIMULATION OF BOUNDARY LAYER EFFECTS ON A HEAVY RAINFALL EVENT CAUSED BY A MESOSCALE CONVECTIVE SYSTEM OVER THE YELLOW RIVER MIDSTREAM AREA

A heavy rainfall event caused by a mesoscale convective system (MCS), which occurred over the Yellow River midstream area during 7–9 July 2016, was analyzed using observational, high-resolution satellite, NCEP/NCAR reanalysis, and numerical simulation data. This heavy rainfall event was caused by one mesoscale convective complex (MCC) and five MCSs successively. The MCC rainstorm occurred when southwesterly winds strengthened into a jet. The MCS rainstorms occurred when low-level wind fields weakened, but their easterly components in the lower and boundary layers increased continuously. Numerical analysis revealed that there were obvious differences between the MCC and MCS rainstorms, including their three-dimensional airflow structure, disturbances in wind fields and vapor distributions, and characteristics of energy conversion and propagation. Formation of the MCC was related to southerly conveyed water vapor and energy to the north, with obvious water vapor exchange between the free atmosphere and the boundary layer. Continuous regeneration and development of the MCSs mainly relied on maintenance of an upward extension of a positive water vapor disturbance. The MCC rainstorm was triggered by large range of convergent ascending motion caused by a southerly jet, and easterly disturbance within the boundary layer. While a southerly fluctuation and easterly disturbance in the boundary layer were important triggers of the MCS rainstorms. Maintenance and development of the MCC and MCSs were linked to secondary circulation, resulting from convergence of Ekman non-equilibrium flow in the boundary layer. Both intensity and motion of the convergence centers in MCC and MCS cases were different. Clearly, sub-synoptic scale systems in the middle troposphere played a leading role in determining precipitation distribution during this event. Although mesoscale systems triggered by the sub-synoptic scale system induced the heavy rainfall, small-scale disturbances within the boundary layer determined its intensity and location.     

ENSO事件对云南及临近地区春末初夏降水的影响

本文采用合成及相关分析的方法,应用55年中国降水资料、美国NOAA海表温度资料以及NCEP/NCAR再分析资料,研究了ENSO事件对我国云南及其邻近地区春末初夏降水的影响及物理机理.研究结果表明:(1)在El Ni(n)o (La Ni(n)a)年,云南大部分地区4～5月降水偏少(多),东部地区相关信号尤其明显;(2)...     

A CLOUD-RESOLVING MODELING STUDY OF SURFACE RAINFALL PROCESSES ASSOCIATED WITH LANDFALLING TYPHOON KAEMI(2006)

The detailed surface rainfall processes associated with landfalling typhoon Kaemi(2006) are investigated based on hourly data from a two-dimensional cloud-resolving model simulation. The model is integrated for 6 days with imposed large-scale vertical velocity, zonal wind, horizontal temperature and vapor advection from National Center for Environmental Prediction (NCEP) / Global Data Assimilation System (GDAS) data. The simulation data are validated with observations in terms of surface rain rate. The Root-Mean-Squared (RMS) difference in surface rain rate between the simulation and the gauge observations is 0.660 mm h^-1, which is smaller than the standard deviations of both the simulated rain rate (0.753 mm h^-1) and the observed rain rate (0.833 mm h^-1). The simulation data are then used to study the physical causes associated with the detailed surface rainfall processes during the landfall. The results show that time averaged and model domain-mean P_s mainly comes from large-scale convergence (Q_WVF) and local vapor loss (positive Q_WVT). Large underestimation (about 15%) of P_s will occur if Q_WVT and Q_CM (cloud source/sink) are not considered as contributors to P_s. Q_WVF accounts for the variation of P_s during most of the integration time, while it is not always a contributor to P_s. Sometimes surface rainfall could occur when divergence is dominant with local vapor loss to be a contributor to P_s. Surface rainfall is a result of multi-timescale interactions. Q_WVE possesses the longest time scale and the lowest frequency of variation with time and may exert impact on P_s in longer time scales. Q_WVF possesses the second longest time scale and lowest frequency and can explain most of the variation of P_s. Q_WVT and Q_CM possess shorter time scales and higher frequencies, which can explain more detailed variations in P_s. Partitioning analysis shows that stratiform rainfall is dominant from the morning of 26 July till the late night of 27 July. After that, convective rainfall dominates till about 1000 LST 28 July. Before 28 July, the variations of in rainfall-free regions contribute less to that of the domain-mean Q_WVT while after that they contribute much, which is consistent to the corresponding variations in their fractional coverage. The variations of Q_WVF in rainfall regions are the main contributors to that of the domain-mean Q_WVF, then the main contributors to the surface rain rate before the afternoon of 28 July.