Application of integrated back‐propagation network and self organizing map for flood forecasting

Much of the nonlinearity and uncertainty regarding the flood process is because hydrologic data required for estimation are often tremendously difficult to obtain. This study employed a back‐propagation network (BPN) as the main structure in flood forecasting to learn and to demonstrate the sophisticated nonlinear mapping relationship. However, a deterministic BPN model implies high uncertainty and poor consistency for verification work even when the learning performance is satisfactory for flood forecasting. Therefore, a novel procedure was proposed in this investigation which integrates linear transfer function (LTF) and self‐organizing map (SOM) to efficiently determine the intervals of weights and biases of a flood forecasting neural network to avoid the above problems. A SOM network with classification ability was applied to the solutions and parameters of the BPN model in the learning stage, to classify the network parameter rules and to obtain the winning parameters. The outcomes from the previous stage were then used as the ranges of the parameters in the recall stage. Finally, a case study was carried out in Wu‐Shi basin to demonstrate the effectiveness of the proposal. Copyright © 2009 John Wiley & Sons, Ltd.     

Hydrologic characteristics of lake‐ and stream‐side riparian wetted margins in the McMurdo Dry Valleys,Antarctica

Water is a limiting factor for life in the McMurdo Dry Valleys (MDV), Antarctica. The active layer (seasonally thawed soil overlying permafrost) accommodates dynamic hydrological and biological processes for 10–16 weeks per year. Wetted margins (visually wetted areas with high moisture content) adjacent to lakes and streams are potential locations of great importance in the MDV because of the regular presence of liquid water, compared with the rest of the landscape where liquid water is rare. At 11 plots (four adjacent to lakes, seven adjacent to streams), soil particle size distribution, soil electrical conductivity, soil water content and isotopic signature, width of the wetted margin, and active layer thaw depth were characterised to determine how these gradients influence physicochemical properties that determine microbial habitat and biogeochemical cycling. Sediments were generally coarse‐grained in wetted margins adjacent to both lakes and streams. Wetted margins ranged from 1·04 to 11·01 m in average length and were found to be longer at lakeside sites than streamside. Average thaw depths ranged from 0·12 to 0·85 m, and were found to be deepest under lake margins. Lake margins also had much higher soil electrical conductivity, steeper topographic gradients, but more gradual soil moisture gradients than stream margins. Patterns of soil water δ¹⁸O and δD distribution indicate capillary action and evaporation from wetted margins; margin pore waters generally demonstrated isotopic enrichment with distance from the shore, indicating evaporation of soil water. Lake margin pore waters were significantly more negative in D_XS (D_XS = δD‐8δ¹⁸O) than streamside pore waters, indicating a longer history of evaporation there. Differences between lake and stream margins can be explained by the more consistent availability of water to lake margins than stream margins. Differences in margin characteristics between lakes and streams have important consequences for the microbial habitat of these margins and their functional role in biogeochemical cycling at these terrestrial–aquatic interfaces. Copyright © 2009 John Wiley & Sons, Ltd.     

Coupled surface water–groundwater model and its application in the arid Shiyang River basin,China

The planning and management of water resources in the Shiyang River basin, China require a tool for assessing the impact of groundwater and stream use on water supply reliabilities and improving many environment‐related problems such as soil desertification induced by recent water‐related human activities. A coupled model, integrating rule‐based lumped surface water model and distributed three‐dimensional groundwater flow model, has been established to investigate surface water and groundwater management scenarios that may be designed to restore the deteriorated ecological environment of the downstream portion of the Shiyang River basin. More than 66% of the water level among 24 observation wells have simulation error less than 1·0 m. The overall trend of the temporal changes of simulated and observed surface runoff at the Caiqi gauging station remains almost the same. The calibration was considered satisfactory. Initial frameworks for water allocation, including agricultural water‐saving projects, water diversion within the basin and inter‐basin water transfer, reducing agricultural irrigation area and surface water use instead of groundwater exploitation at the downstream were figured out that would provide a rational use of water resources throughout the whole basin. Sixteen scenarios were modelled to find out the most appropriate management strategies. The results showed that in the two selected management options, the groundwater budget at the Minqin basin was about 1·4 × 10⁸ m³/a and the ecological environment would be improved significantly, but the deficit existed at the Wuwei basin and the number was about 0·8 × 10⁸ m³/a. Water demand for domestic, industry and urban green area would be met in the next 30 years, but the water shortage for meeting the demand of agricultural water use in the Shiyang River basin was about 2·2 × 10⁸ m³/a. It is suggested that more inter‐basin water transfer should be required to obtain sustainable water resource use in the Shiyang River basin. Copyright © 2009 John Wiley & Sons, Ltd.     

Simultaneous identification of parameter,initial condition,and boundary condition in groundwater modelling

In this paper, we perform an inverse method to simultaneously estimate aquifer parameters, initial condition, and boundary conditions in groundwater modelling. The parameter estimation is extended to a complete inverse problem that makes the calibrated groundwater flow model more realistic. The adjoint state method, the gradient search method, and the least square error algorithm are combined to build the optimization procedure. Horizontal two‐dimensional groundwater flow in a confined aquifer is exemplified to demonstrate the correlation between unknowns, the contribution of observation, as well as the suitability of applying the inverse method. The correlation analysis shows the connection between storage coefficient and initial condition. Besides, transmissivity and boundary conditions are also highly correlated. More observations at different location and time are necessary to provide sufficient information. A time series of unsteady head is requested for estimation of storage coefficient and initial condition. Observation near boundary is very effective for boundary condition estimation. The observation at pumping well mostly contributes to the estimation of transmissivity. According to all observations, it is possible to identify parameters, initial condition, and boundary condition simultaneously. Furthermore, the results not only illustrate the traditional assumption of known boundary condition but also initial condition, which may cause an incorrect estimation. Copyright © 2009 John Wiley & Sons, Ltd.     

Bayesian analysis of stage–fall–discharge models for gauging stations affected by variable backwater

Parsimonious stage–fall–discharge rating curve models for gauging stations subject to backwater complications are developed from simple hydraulic theory. The rating curve models are compounded in order to allow for possible shifts in the hydraulics when variable backwater becomes effective. The models provide a prior scientific understanding through the relationship between the rating curve parameters and the hydraulic properties of the channel section under study. This characteristic enables prior distributions for the rating curve parameters to be easily elicited according to site‐specific information and the magnitude of well‐known hydraulic quantities. Posterior results from three Norwegian and one American twin‐gauge stations affected by variable backwater are obtained using Markov chain Monte Carlo simulation techniques. The case studies demonstrate that the proposed Bayesian rating curve assessment is appropriate for developing rating procedures for gauging stations that are subject to variable backwater. Copyright © 2009 John Wiley & Sons, Ltd.     

The thermal conductivity for granite with various water contents

The thermal conductivities of granite were measured under different conditions of porosity and water content to investigate the effects of the porosity and water content on the thermal conductivity. For the dry samples, the thermal conductivities range from 2.12 W/mK for the rocks with a high porosity to 3.12 W/mK for the ones with a low porosity. Water-sorbed samples have greater thermal conductivities than dry samples of the same granite. The thermal conductivities range from 2.99 W/mK for granites with a high porosity to 3.62 W/mK for ones with a low porosity under saturated condition.     

Retrieval of outgoing longwave radiation from COMS narrowband infrared imagery

Hourly outgoing longwave radiation(OLR) from the geostationary satellite Communication Oceanography Meteorological Satellite(COMS) has been retrieved since June 2010. The COMS OLR retrieval algorithms are based on regression analyses of radiative transfer simulations for spectral functions of COMS infrared channels. This study documents the accuracies of OLRs for future climate applications by making an intercomparison of four OLRs from one single-channel algorithm(OLR12.0using the 12.0 μm channel) and three multiple-channel algorithms(OLR10.8+12.0using the 10.8 and 12.0 μm channels; OLR6.7+10.8using the 6.7 and 10.8 μm channels; and OLR All using the 6.7, 10.8, and 12.0 μm channels). The COMS OLRs from these algorithms were validated with direct measurements of OLR from a broadband radiometer of the Clouds and Earth's Radiant Energy System(CERES) over the full COMS field of view [roughly(50°S–50°N, 70°–170°E)] during April 2011.Validation results show that the root-mean-square errors of COMS OLRs are 5–7 W m-2, which indicates good agreement with CERES OLR over the vast domain. OLR6.7+10.8and OLR All have much smaller errors(～ 6 W m-2) than OLR12.0and OLR10.8+12.0(～ 8 W m-2). Moreover, the small errors of OLR6.7+10.8and OLR All are systematic and can be readily reduced through additional mean bias correction and/or radiance calibration. These results indicate a noteworthy role of the6.7 μm water vapor absorption channel in improving the accuracy of the OLRs. The dependence of the accuracy of COMS OLRs on various surface, atmospheric, and observational conditions is also discussed.     

Early solar system aqueous activity: K isotope evidence from Allende

The alkali element K is moderately volatile and fluid mobile; thus, it can be influenced by both primary processes (evaporation and recondensation) in the solar nebula and secondary processes (thermal and aqueous alteration) in the parent body. Since these primary and secondary processes would induce different isotopic fractionations, K isotopes could become a potential tracer to distinguish them. Using recently developed methods with improved precision (0.05‰, 95% confidence interval), we systematically measured the K isotopic compositions and major/trace elemental compositions of chondritic components (18 chondrules, 3 CAIs, 2 matrices, and 5 bulks) in the carbonaceous chondrite fall Allende. Among all the components analyzed in this study, CAIs, which formed initially under high‐temperature conditions in the solar nebula and were dominated by nominally K‐free refractory minerals, have the highest K₂O content (average 0.53 wt%) and have K isotope compositions most enriched in heavy isotopes (δ⁴¹K: ?0.30 to ?0.25‰). Such an observation is consistent with previous petrologic studies that show CAIs in Allende have undergone alkali enrichment during metasomatism. In contrast, chondrules contain lower K₂O content (0.003–0.17 wt%) and generally lighter K isotope compositions (δ⁴¹K: ?0.87‰ to ?0.24‰). The matrix and bulks are nearly identical in K₂O content and K isotope compositions (0.02–0.05 wt%; δ⁴¹K: ?0.62 to ? 0.46‰), which are, as expected, right in the middle of CAIs and chondrules. This strongly indicates that most of the chondritic components of Allende suffered aqueous alteration and their K isotopic compositions are the ramification of Allende parent‐body processing instead of primary nebular signatures. Nevertheless, we propose the small K isotope fractionations observed (< 1‰) among Allende components are likely similar to the overall range of K isotopic fractionation that occurred in nebular environment. Furthermore, the K isotope compositions seen in the components of Allende in this study are consistent with MC‐ICP‐MS analyses of the components in ordinary chondrites, which also show an absence of large (10‰) isotope fractionations. This is not expected as evaporation experiments in nebular conditions suggest there should be large K isotopic fractionations. Nevertheless, possible nebular processes such as chondrules back exchanging with ambient gas when they formed could explain this lack of large K isotopic variation.     

The evolution of the z distribution of normal neutron stars in the Galaxy

Under the two initial 1‐D one parameter velocity distribution forms (one is normal, the other is exponential), the z direction scale height evolution of normal neutron stars in the Galaxy is studied by numerical simulation. We do statistics for the cases at different time segments, also do statistics for the cumulative cases made of each time segment. The results show in the cumulative cases the evolution curves of the scale heights are smoother than in the each time segment, i.e., the cumulation improve the signal‐to‐noise ratio. Certainly the evolution cases are different at different Galactic disk locations, which also have very large difference from the average cases in the whole disk. In the initial stages of z evolution of normal neutron stars, after the beginning transient states, the cumulative scale heights increase linearly with time, and the cumulative scale height increasing rates have linear relationship with the initial velocity distribution parameters, which have larger fluctuation in the vicinity of the Sun than in the whole disk. We utilize the linear relationship of the cumulative scale height increasing rates vs. the initial velocity distribution parameters in the vicinity of the Sun to make comparison with the observation near the Sun. The results show if there is no magnetic decay, then the deserved initial velocity parameters are obvious lower than the present well known results from some authors; whereas if introducing magnetic decay, for the 1‐D normal case we can make consistence among concerning results using magnetic decay time values which are supported by some authors, while for the 1‐D exponential case the results show the lackness of young pulsar samples in the larger z in the vicinity of the Sun (© 2010 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)