Machine learning for pore-water pressure time-series prediction: Application of recurrent neural networks

Knowledge of pore-water pressure(PWP)variation is fundamental for slope stability.A precise prediction of PWP is difficult due to complex physical mechanisms and in situ natural variability.To explore the applicability and advantages of recurrent neural networks(RNNs)on PWP prediction,three variants of RNNs,i.e.,standard RNN,long short-term memory(LSTM)and gated recurrent unit(GRU)are adopted and compared with a traditional static artificial neural network(ANN),i.e.,multi-layer perceptron(MLP).Measurements of rainfall and PWP of representative piezometers from a fully instrumented natural slope in Hong Kong are used to establish the prediction models.The coefficient of determination(R^2)and root mean square error(RMSE)are used for model evaluations.The influence of input time series length on the model performance is investigated.The results reveal that MLP can provide acceptable performance but is not robust.The uncertainty bounds of RMSE of the MLP model range from 0.24 kPa to 1.12 k Pa for the selected two piezometers.The standard RNN can perform better but the robustness is slightly affected when there are significant time lags between PWP changes and rainfall.The GRU and LSTM models can provide more precise and robust predictions than the standard RNN.The effects of the hidden layer structure and the dropout technique are investigated.The single-layer GRU is accurate enough for PWP prediction,whereas a double-layer GRU brings extra time cost with little accuracy improvement.The dropout technique is essential to overfitting prevention and improvement of accuracy.     

Analysis of the contributions of topographic,soil, and vegetation features on the spatial distributions of surface soil moisture in a steep natural forested headwater catchment

Surface soil moisture has been extensively studied for various land uses and landforms. Although many studies have reported potential factors that control surface soil moisture over space or time, the findings have not always been consistent, indicating a need for identification of the main factors. This study focused on the static controls of topographic, soil, and vegetation features on surface soil moisture in a steep natural forested headwater catchment consisting of three hillslope units of a gully area, side slope, and valley‐head slope. Using a simple correlation analysis to investigate the effects of the static factors on surface soil moisture at depths of 0–20 cm at 470 points in 13 surveys, we addressed the characteristics of surface soil moisture and its main controlling factors. The results indicated that the mean of surface soil moisture was in the decreasing order of gully area > valley‐head slope > side slope. The relationship between the mean and standard deviation of surface soil moisture showed a convex‐upward shape in the headwater catchment, a negative curvilinear shape in the gully area, and positive curvilinear shapes at the side and valley‐head slopes. At the headwater catchment and valley‐head slope, positive contributions of soil porosity and negative contributions of slope gradient and saturated hydraulic conductivity were the main controlling factors of surface soil moisture under wetter conditions, whereas positive contributions of topographic wetness index and negative contributions of vegetation density were the main controlling factors of surface soil moisture under drier conditions. At the side slope underlain by fractured bedrocks, only saturated hydraulic conductivity and vegetation density were observed to be the controlling factors. Surface soil moisture in the gully area was mainly affected by runoff rather than were static features. Thus, using hillslope units is effective for approximately estimating the hydrological behaviours of surface moisture on a larger scale, whereas dependency between the main static factors and moisture conditions is helpful for estimating the spatial distributions of surface moisture on a smaller scale.     

Impact of precipitation characteristics on soil hydrology in tile‐drained landscapes

Understanding the variables regulating tile‐flow response to precipitation in the US Midwest is critical for water quality management. This study (1) investigates the relationship between precipitation characteristics, antecedent water table depth and tile‐flow response at a high temporal resolution during storms; and (2) determines the relative importance of macropore flow versus matrix flow in tile flow in a tile‐drained soya bean field in Indiana. In spring, although variations in antecedent water table depth imparted some variation in tile‐flow response to precipitation, bulk precipitation was the best predictor of mean tile flow, maximum tile flow, time to peak, and run‐off ratio. The contribution of macropore flow to total flow significantly increased with precipitation amount, and macropore flow represented between 11 and 50% of total drain flow, with peak contributions between 15 and 74% of flow. For large storms (>6 cm bulk precipitation), cations data indicated a dilution of groundwater with new water as discharge peaked. Although no clear dilution or concentration patterns for Mg²⁺ or K⁺ were observed for smaller tile flow generating events (<3 cm bulk precipitation), macropore flow still contributed between 11 and 17% of the total flow for these moderate size storms. Inter‐drain comparison stressed the need to use triplicate or duplicate tile drain experiments when investigating tile drainage impact on water and N losses at the plot scale. These results significantly increase our understanding of the hydrological functioning of tile‐drained fields in spring, when most N losses to streams occur in the US Midwest. Copyright © 2010 John Wiley & Sons, Ltd.     

Reservoir rainfall‐runoff geomorphological model. II: analysis,calibration and validation

The model presented in the complementary document entitled, Reservoir rainfall‐runoff geomorphological model I: parameter application and analysis is analysed, calibrated and validated in this paper. The accuracy of simulated hydrographs is analysed by means of the efficiency defined by Nash and Sutcliffe. The sensitivity of the influence of five parameters on the behaviour of the model developed is analysed. Two different calibration and validation processes of Reservoir rainfall‐runoff geomorphological model are performed in Aixola watershed. Twelve events have been selected for calibrations and 25 for validations. With the first calibration and validation process, the model parameters are set by assigning the medians' values of the distributions obtained by means of the optimum results. The second process is performed by calibrating the most determinant parameter in the adjustment, which is the one that indicates the proportion of infiltrated water that is retained and does not flow; this is done with an empirical formulation depending on the event characteristics. Subsequently, the obtained results are validated. This last process has achieved very good adjustments in both calibrated and validated events. Copyright © 2012 John Wiley & Sons, Ltd.     

Accounting for sparsely observed rainfall space—time variability in a rainfall—runoff model of a semiarid Tunisian basin/Prise en compte d'observations peu denses de la variabilité spatiotemporelle de la pluie dans une modélisation pluie—débit d'un bassin semi-aride Tunisien

Abstract

The management of water excesses and deficits is a major task in semiarid Mediterranean regions, where the variability of rainfall inputs is high at different time and space scales. Thus intense hydrometeorological events, which generate both potential resource and hazards, are of major interest. A simple method is proposed, with the example of the Skhira basin (192 km²) in central Tunisia, to account for the event space–time variability of rainfall in a rainfall–runoff model, in order to check its influence on the shape, magnitude and timing of resulting hydrographs. The transfer function used is a geomorphology-based unit hydrograph with an explicit territorial significance. Simulations made for highly variable events show the relevance of this method, seen as the first step of a downward approach, and its robustness with respect to the quality and the density of rainfall data.     

Groundwater–surface water interactions in a lowland watershed: source contribution to stream flow

The lower coastal plain of the Southeast USA is undergoing rapid urbanisation as a result of population growth. Land use change has been shown to affect watershed hydrology by altering stream flow and, ultimately, impairing water quality and ecologic health. However, because few long‐term studies have focused on groundwater–surface water interactions in lowland watersheds, it is difficult to establish what the effect of development might be in the coastal plain region. The objective of this study was to use an innovative improvement to end‐member mixing analysis (EMMA) to identify time sequences of hydrologic processes affecting storm flow. Hydrologic and major ion chemical data from groundwater, soil water, precipitation and stream sites were collected over a 2‐year period at a watershed located in USDA Forest Service's Santee Experimental Forest near Charleston, South Carolina, USA. Stream flow was ephemeral and highly dependent on evapotranspiration rates and rainfall amount and intensity. Hydrograph separation for a series of storm events using EMMA allowed us to identify precipitation, riparian groundwater and streambed groundwater as main sources to stream flow, although source contribution varied as a function of antecedent soil moisture condition. Precipitation, as runoff, dominated stream flow during all storm events while riparian and streambed groundwater contributions varied and were mainly dependent on antecedent soil moisture condition. Sensitivity analyses examined the influence of 10% and 50% increases in analyte concentration on EMMA calculations and found that contribution estimates were very sensitive to changes in chemistry. This study has implications on the type of methodology used in traditional forms of EMMA research, particularly in the recognition and use of median end‐member water chemistry in hydrograph separation techniques. Potential effects of urban development on important hydrologic processes (groundwater recharge, interflow, runoff, etc.) that influence stream flow in these lowland watersheds were qualitatively examined. Copyright © 2011 John Wiley & Sons, Ltd.     

Understanding and modeling basin hydrology: interpreting the hydrogeological signature

Basin landscapes possess an identifiable spatial structure, fashioned by climate, geology and land use, that affects their hydrologic response. This structure defines a basin's hydrogeological signature and corresponding patterns of runoff and stream chemistry. Interpreting this signature expresses a fundamental understanding of basin hydrology in terms of the dominant hydrologic components: surface, interflow and groundwater runoff. Using spatial analysis techniques, spatially distributed watershed characteristics and measurements of rainfall and runoff, we present an approach for modelling basin hydrology that integrates hydrogeological interpretation and hydrologic response unit concepts, applicable to both new and existing rainfall‐runoff models. The benefits of our modelling approach are a clearly defined distribution of dominant runoff form and behaviour, which is useful for interpreting functions of runoff in the recruitment and transport of sediment and other contaminants, and limited over‐parameterization. Our methods are illustrated in a case study focused on four watersheds (24 to 50 km²) draining the southern coast of California for the period October 1988 though to September 2002. Based on our hydrogeological interpretation, we present a new rainfall‐runoff model developed to simulate both surface and subsurface runoff, where surface runoff is from either urban or rural surfaces and subsurface runoff is either interflow from steep shallow soils or groundwater from bedrock and coarse‐textured fan deposits. Our assertions and model results are supported using streamflow data from seven US Geological Survey stream gauges and measured stream silica concentrations from two Santa Barbara Channel–Long Term Ecological Research Project sampling sites. Copyright © 2004 John Wiley & Sons, Ltd.     

陕西金堆城花岗岩类主要造岩矿物特征及其岩石学意义

本文依据大量测试数据,对金堆城地区花岗岩类主要造岩矿物学特征同钼迁移富集成矿关系进行了探讨。研究结果表明:异常晶胞条纹长石和正长石以及富重稀土和具明显负铕异常的黑云母是含钼花岗岩类标型矿物,并且可作为花岗岩类岩石发育地区寻找隐伏钼矿体的指示性矿物。钾钠长石和斜长石的结构状态和成分不仅可以提供有关花岗岩类成因演化的信息,其矿物本身亦可成为钼的载体矿物。岩浆期后,强烈的热液蚀变作用可造成钾钠长石和斜长石分解,并且引起钼的释放,从而形成高品位钼矿石。     

Asymptotic Rayleigh instantaneous unit hydrograph

The instantaneous unit hydrograph for a channel network under general linear routing and conditioned on the network magnitude,N, tends asymptotically, asN grows large, to a Rayleigh probability density function. This behavior is identical to that of the width function of the network, and is proven under the assumption that the network link configuration is topologically random and the link hydraulic and geometric properties are independent and identically distributed random variables. The asymptotic distribution depends only on a scale factor, , where is a mean link wave travel time.     

中国水质站网规划中几个问题的探讨

在完成中国水质站网规划工作的基础上,对下列问题作了进一步探讨:(1)水质站网规划单元的划分;(2)聚类分析在水质站网规划中的应用;(3)站网密度与人口和经济发展的关系;(4)中国水质站网的分区特征。