Variational Markov Chain Monte Carlo for Bayesian smoothing of non-linear niffusions

In this paper we develop set of novel Markov chain Monte Carlo algorithms for Bayesian smoothing of partially observed non-linear diffusion processes. The sampling algorithms developed herein use a deterministic approximation to the posterior distribution over paths as the proposal distribution for a mixture of an independence and a random walk sampler. The approximating distribution is sampled by simulating an optimized time-dependent linear diffusion process derived from the recently developed variational Gaussian process approximation method. Flexible blocking strategies are introduced to further improve mixing, and thus the efficiency, of the sampling algorithms. The algorithms are tested on two diffusion processes: one with double-well potential drift and another with SINE drift. The new algorithm's accuracy and efficiency is compared with state-of-the-art hybrid Monte Carlo based path sampling. It is shown that in practical, finite sample, applications the algorithm is accurate except in the presence of large observation errors and low observation densities, which lead to a multi-modal structure in the posterior distribution over paths. More importantly, the variational approximation assisted sampling algorithm outperforms hybrid Monte Carlo in terms of computational efficiency, except when the diffusion process is densely observed with small errors in which case both algorithms are equally efficient.

Variational Markov chain Monte Carlo for Bayesian smoothing of non-linear diffusions

In this paper we develop set of novel Markov Chain Monte Carlo algorithms for Bayesian smoothing of partially observed non-linear diffusion processes. The sampling algorithms developed herein use a deterministic approximation to the posterior distribution over paths as the proposal distribution for a mixture of an independence and a random walk sampler. The approximating distribution is sampled by simulating an optimized time-dependent linear diffusion process derived from the recently developed variational Gaussian process approximation method. The novel diffusion bridge proposal derived from the variational approximation allows the use of a flexible blocking strategy that further improves mixing, and thus the efficiency, of the sampling algorithms. The algorithms are tested on two diffusion processes: one with double-well potential drift and another with SINE drift. The new algorithm's accuracy and efficiency is compared with state-of-the-art hybrid Monte Carlo based path sampling. It is shown that in practical, finite sample applications the algorithm is accurate except in the presence of large observation errors and low to a multi-modal structure in the posterior distribution over paths. More importantly, the variational approximation assisted sampling algorithm outperforms hybrid Monte Carlo in terms of computational efficiency, except when the diffusion process is densely observed with small errors in which case both algorithms are equally efficient. © 2011 Springer-Verlag.

Tillage and the environment in sub-tropical Australia-Tradeoffs and challenges

Tillage is defined here in a broad sense, including disturbance of the soil and crop residues, wheel traffic and sowing opportunities. In sub-tropical, semi-arid cropping areas in Australia, tillage systems have evolved from intensively tilled bare fallow systems, with high soil losses, to reduced and no tillage systems. In recent years, the use of controlled traffic has also increased. These conservation tillage systems are successful in reducing water erosion of soil and sediment-bound chemicals. Control of runoff of dissolved nutrients and weakly sorbed chemicals is less certain. Adoption of new practices appears to have been related to practical and economic considerations, and proved to be more profitable after a considerable period of research and development. However there are still challenges. One challenge is to ensure that systems that reduce soil erosion, which may involve greater use of chemicals, do not degrade water quality in streams. Another challenge is to ensure that systems that improve water entry do not increase drainage below the crop root zone, which would increase the risk of salinity. Better understanding of how tillage practices influence soil hydrology, runoff and erosion processes should lead to better tillage systems and enable better management of risks to water quality and soil health. Finally, the need to determine the effectiveness of in-field management practices in achieving stream water quality targets in large, multi-land use catchments will challenge our current knowledge base and the tools available.

Prediction of Ground Water Levels in the Uplands of a Tropical Coastal Riparian Wetland using Artificial Neural Networks

Artificial Neural Networks (ANNs) have been found to be a robust tool to model many non-linear hydrological processes. The present study aims at evaluating the performance of ANN in simulating and predicting ground water levels in the uplands of a tropical coastal riparian wetland. The study involves comparison of two network architectures, Feed Forward Neural Network (FFNN) and Recurrent Neural Network (RNN) trained under five algorithms namely Levenberg Marquardt algorithm, Resilient Back propagation algorithm, BFGS Quasi Newton algorithm, Scaled Conjugate Gradient algorithm, and Fletcher Reeves Conjugate Gradient algorithm by simulating the water levels in a well in the study area. The study is analyzed in two cases-one with four inputs to the networks and two with eight inputs to the networks. The two networks-five algorithms in both the cases are compared to determine the best performing combination that could simulate and predict the process satisfactorily. Ad Hoc (Trial and Error) method is followed in optimizing network structure in all cases. On the whole, it is noticed from the results that the Artificial Neural Networks have simulated and predicted the water levels in the well with fair accuracy. This is evident from low values of Normalized Root Mean Square Error and Relative Root Mean Square Error and high values of Nash-Sutcliffe Efficiency Index and Correlation Coefficient (which are taken as the performance measures to calibrate the networks) calculated after the analysis. On comparison of ground water levels predicted with those at the observation well, FFNN trained with Fletcher Reeves Conjugate Gradient algorithm taken four inputs has outperformed all other combinations.

黄土坡面下坡位土壤侵蚀过程的模拟试验

在中国科学院水利部水土保持研究所,采用2个不同坡长小区室内模拟降雨试验方法,对黄土坡面下坡位侵蚀过程进行研究。结果表明:1)坡面下坡位侵蚀模数随降雨过程、降雨强度及坡度的变化均具有大小交错,上下波动的特征,侵蚀模数随降雨过程的变化总体呈先上升后趋于稳定的态势,随降雨强度的增大而增加,随坡度的增大而先增大后减小,再增大再减小;2)坡面上、下坡位侵蚀模数随降雨过程、坡度及降雨强度的变化均具有明显差异,下坡位明显不如上坡位;3)坡面上坡位汇流和下坡位产流与坡面上坡位输沙对坡面下坡位侵蚀模数的影响可用二元线性方程很好地描述,前者的贡献率为47.8%,后者为20.4%;只通过观测分析小区平均侵蚀特征得出的坡面侵蚀过程,掩盖了坡面下坡位的真实侵蚀过程,采取水土保持措施减少坡面上坡位汇流及增加降雨就地入渗,可以有效地治理坡面下坡位的水土流失。

近地表土壤水分条件对坡面农业非点源污染物运移的影响

采用人工模拟降雨实验,研究了近地表土壤水分条件,尤其是土壤水分饱和条件对土壤侵蚀过程中农业非点源污染物运移的影响.结果表明,前期近地表土壤水分条件对土壤侵蚀过程中农业非点源污染物的迁移有着很大影响.饱和含水量时径流及泥沙中非点源污染物的流失浓度和流失量大于非饱和含水量,且前期近地表土壤含水量越大,径流及泥沙中农业非点源污染物的流失浓度和流失量越大.土壤氮素的主要流失途径是降雨所产生的径流,约占总流失量的90.4%～99.8%;土壤磷素的流失途径是降雨径流和侵蚀产沙,分别占总流失量的2.67%～23.5%和76.5%～97.3%.同时,土壤质地对磷素养分的流失有很大影响,杨凌土随泥沙流失的DP浓度和流失量均大于安塞黄绵土.最后,提出了采取最佳管理措施等控制农业非点源污染的针对性建议.

高精度GPS,三维激光扫描和测针板三种测量技术监测沟蚀过程的对比研究

采用连续模拟降雨试验,对坡沟系统概化模型进行坡面沟蚀发育过程模拟,再现坡面片蚀—细沟侵蚀—切沟侵蚀演变过程。结合3种测量技术从测量精度、测量人员要求、数据处理、数据通用性、配套软件使用、前期投入、测量条件等几个方面入手,对比分析高精度GPS(Trimble 5700)、三维激光扫描仪(Leica HDS 3000)和测针板的3种测量方法的优缺点,同时对3种测量方法在沟蚀过程监测和侵蚀量估算方面进行对比研究。结果表明,激光扫描仪能很好地监测沟蚀演变过程,且对侵蚀量估算精度较高,误差仅为4.5%。高精度GPS也能很好地监测沟蚀演变过程,对侵蚀量估算精度误差为7.38%。测针板法不能很好的反映沟蚀演变过程,但是对于侵蚀量的估算可以满足日常要求,误差为-12.78%。

水流作用下疏松土壤材料中细沟的再生及其临界剪应力的实验研究

动态土壤侵蚀包括依赖于空间与时间的土壤剥离、细沟的形成、水流中过量泥沙的沉积、沉积泥沙中细沟的再生等物理过程。与泥沙从沉积区域再剥离相关联的细沟再生是细沟变窄加深过程中的一个重要的现象。该文介绍一种实验装置 ,用于测量类似于沉积土壤的疏松土壤材料细沟再生的力学参数。用一种沙壤土进行实验 ,测定了不同团聚体大小和坡度下细沟再生开始和停止所对应的临界水流量。计算出了与沟坡相关的临界剪切应力。这些结果有助于理解细沟的再生及其数量关系 ,以及临界剪切应力与坡度间的定量关系 ;对于发展由沉积土壤再生的细沟演化过程模型特别有用。

大气散落核素复合示踪在土壤侵蚀科学中的应用

介绍了大气散落核素 7Be、2 1 0 Pbex、1 37Cs复合在土壤侵蚀和泥沙来源领域中的研究进展 ;阐述了核素比率应用的原理、特点及其能更为敏感地反映土壤侵蚀速率和过程等优点 ;说明多元素复合示踪在推断土壤不同侵蚀过程及估算土壤侵蚀速率方面的作用 ;论述了建立两核素相关深度曲线确定泥沙起源深度的应用原理和特点及其应用范围 ,并对各种方法进行了简单的评价。

细沟侵蚀过程与细沟水流水力学参数的关系研究

利用供沙土槽和试验土槽的双土槽径流小区 ,定量研究了在不同降雨强度下上方来水来沙对陡坡地细沟侵蚀产沙过程和细沟水流水力学参数的影响及其细沟水流水力学特征参数与细沟侵蚀产沙量的关系。结果表明 :坡面细沟侵蚀过程以侵蚀—搬运过程为主 ,坡上方来沙不仅被径流全部搬运 ,且上方来水在坡下方引起了另外的侵蚀产沙量 ,其值随上方来水含沙量的减少和降雨强度的增加而增大。上方来水的汇入或降雨强度的增大可使细沟水流流态由层流转化为紊流。上方来水对细沟水流水力学参数 (流速、水力半径、雷诺数、弗劳德数和阻力系数 )有重要影响。定量分析了细沟水流水力学特征参数 (流速、雷诺数和阻力系数 )与上坡来水引起坡下方净侵蚀产沙量的关系 ,建立了净侵蚀产沙量与细沟水流流速、雷诺数和阻力系数统计模型。

Release of β-galactosidase from poloxamine/α-cyclodextrin hydrogels

All mammals lose their ability to produce lactase (β-galactosidase), the enzyme that cleaves lactose into galactose and glucose, after weaning. The prevalence of lactase deficiency (LD) spans from 2 to 15% among northern Europeans, to nearly 100% among Asians. Following lactose consumption, people with LD often experience gastrointestinal symptoms such as abdominal pain, bowel distension, cramps and flatulence, or even systemic problems such as headache, loss of concentration and muscle pain. These symptoms vary depending on the amount of lactose ingested, type of food and degree of intolerance. Although those affected can avoid the uptake of dairy products, in doing so, they lose a readily available source of calcium and protein. In this work, gels obtained by complexation of Tetronic 90R4 with α-cyclodextrin loaded with β-galactosidase are proposed as a way to administer the enzyme immediately before or with the lactose-containing meal. Both molecules are biocompatible, can form gels in situ, and show sustained erosion kinetics in aqueous media. The complex was characterized by FTIR that evidenced an inclusion complex between the polyethylene oxide block and α-cyclodextrin. The release profiles of β-galactosidase from two different matrices (gels and tablets) of the in situ hydrogels have been obtained. The influence of the percentage of Tetronic in media of different pH was evaluated. No differences were observed regarding the release rate from the gel matrices at pH 6 (t50 = 105 min). However, in the case of the tablets, the kinetics were faster and they released a greater amount of 90R4 (25%, t50 = 40–50 min). Also, the amount of enzyme released was higher for mixtures with 25% Tetronic. Using suitable mathematical models, the corresponding kinetic parameters have been calculated. In all cases, the release data fit quite well to the Peppas–Sahlin model equation, indicating that the release of β-galactosidase is governed by a combination of diffusion and erosion processes. It has been observed that the diffusion mechanism prevails over erosion during the first 50 minutes, followed by continued release of the enzyme due to the disintegration of the matrix.

Micro-abrasion wear behaviour of TiAlCrSiN nanostructured coatings

The injection process of glass fibres reinforced plastics promotes the moulds surface degradation by erosion. In order to improve its wear resistance, several kinds of PVD thin hard coatings were used. It is well-known that nanostructures present a better compromise between hardness and toughness. Indeed, when the coating is constituted by a large number of ultra-thin different layers, cracks and interface troubles tend to decrease. However, it is not clear that these nanostructures present a better wear behaviour in erosion processes. In order to study its wear behaviour, a sputtered PVD nanostructured TiAlCrSiN coating was used. The substrate and film surfaces topography were analyzed by profilometry and atomic force microscopy techniques. Film adhesion to the substrate was evaluated by scratch tests. The surface hardness was measured with a Vickers micro-hardness tester. The wear resistance was evaluated by micro-abrasion with a rotating ball tribometer tests. Slurry of SiC particles in distilled water was used in order to provoke the surface abrasion. Different duration tests were performed in order to analyze the wear evolution. After these tests, the wear mechanisms developed were analyzed by scanning electron microscopy. Wear craters were measured and the wear rate was calculated and discussed. With the same purpose, coated inserts were mounted in an injection mould working with a 30% glass fibres reinforced polypropylene. After 45 000 cycles no relevant wear was registered.

Investigación geofísica de los paleovalles pleistocenos del NE de Cataluña: el ejemplo de la Riera de Tossa (Costa Brava meridional)

The geometry of Riera de Tossa (Costa Brava, Catalonia) Pleistocene paleovalley has been studied by two geophysical prospecting methods: vertical electric sounding (VES-DC) and seismic refraction. The results show the existence of a sudden slope change of the thalweg that permits to differentiate two sections where erosion processes operated in unequal intensity. Based on geological and geomorphological available data, it can be inferred that this slope rupture is caused by the litological control imposed by the alineation of resistant materials, which limited the headward erosion extend of the creek during the Upper Pleistocene

Developing a desertification indicator system for a small Mediterranean catchment: a case study from the Serra de Rodes, Alt Empordà, Catalunya, NE Spain

La desertificació és un problema de degradació de sòls de gran importància en regions àrides, semi-àrides i sub-humides, amb serioses conseqüències ambientals, socials i econòmiques com a resultat de l'impacte d'activitats humanes en combinació amb condicions físiques i medi ambientals desfavorables (UNEP, 1994). L'objectiu principal d'aquesta tesi va ser el desenvolupament d'una metodologia simple per tal de poder avaluar de forma precisa l'estat i l'evolució de la desertificació a escala local, a través de la creació d'un model anomenat sistema d'indicators de desertificació (DIS). En aquest mateix context, un dels dos objectius específics d'aquesta recerca es va centrar en l'estudi dels factors més importants de degradació de sòls a escala de parcel.la, comportant un extens treball de camp, analisi de laboratori i la corresponent interpretació i discussió dels resultats obtinguts. El segon objectiu específic es va basar en el desenvolupament i aplicació del DIS. L'àrea d'estudi seleccionada va ser la conca de la Serra de Rodes, un ambient típic Mediterràni inclòs en el Parc Natural del Cap de Creus, NE Espanya, el qual ha estat progressivament abandonat pels agricultors durant el segle passat. Actualment, els incendis forestals així com el canvi d'ús del sòl i especialment l'abandonament de terres són considerats els problemes ambientals més importants a l'àrea d'estudi (Dunjó et al., 2003). En primer lloc, es va realitzar l'estudi dels processos i causes de la degradació dels sòls a l'àrea d'interés. En base a aquest coneixement, es va dur a terme la identificació i selecció dels indicadors de desertificació més rellevants. Finalment, els indicadors de desertificació seleccionats a escala de conca, incloent l'erosió del sòl i l'escolament superficial, es van integrar en un model espaial de procés. Ja que el sòl és considerat el principal indicador dels processos d'erosió, segons la FAO/UNEP/UNESCO (1979), tant el paisatge original així com els dos escenaris d'ús del sòl desenvolupats, un centrat en el cas hipotétic del pas d'un incendi forestal, i l'altre un paisatge completament cultivat, poden ser ambients classificats sota baixa o moderada degradació. En comparació amb l'escenari original, els dos escenaris creats van revelar uns valors més elevats d'erosió i escolament superficial, i en particular l'escenari cultivat. Per tant, aquests dos hipotètic escenaris no semblen ser una alternativa sostenible vàlida als processos de degradació que es donen a l'àrea d'estudi. No obstant, un ampli ventall d'escenaris alternatius poden ser desenvolupats amb el DIS, tinguent en compte les polítiques d'especial interés per la regió de manera que puguin contribuir a determinar les conseqüències potencials de desertificació derivades d'aquestes polítiques aplicades en aquest escenari tan complexe espaialment. En conclusió, el model desenvolupat sembla ser un sistema força acurat per la identificació de riscs presents i futurs, així com per programar efectivament mesures per combatre la desertificació a escala de conca. No obstant, aquesta primera versió del model presenta varies limitacions i la necessitat de realitzar més recerca en cas de voler desenvolupar una versió futura i millor del DIS.

Geração de sulcos em meio coesivo e caracterização hidrossedimentométrica de sua evolução

Na pesquisa em erosão, nas últimas décadas, está se formando um consenso de que é importante entender os processos básicos que regem o fenômeno. Uma alternativas para tentar compreender melhor as etapas do processo erosivo é separá-lo na fase de sulco (fluxo concentrado) e de entressulco. Dentro desse enfoque foi construído no Laboratório de Processos Erosivos e Deposicionais (LaPED) do IPH/UFRGS um canal de declividade para estudar o processo de incisão e o desenvolvimento dos sulcos de erosão. A estrutura experimental projetada e construída permite que seja controlada a vazão através de um medidor eletromagnético e que seja alterada a declividade do canal através de um sistema hidráulico associado a um nível digital. O solo colocado no canal foi um Latossolo Vermelho distrófico típico, as declividades de trabalho foram 3,0; 6,0 e 9,0% e a seqüência de vazões aplicadas foi 10,0; 18,5; 25,5; 38,5 e 51,0L.min-1. A estrutura experimental montada se mostrou de fácil operação e eficiente para permitir o avanço no entendimento dos processos de desagregação e de transporte de partículas sólidas pela ação do escoamento superficial, além de possibilitar a geração de sulco(s) de erosão na superfície do solo. O escoamento passou da condição de difuso para concentrado a partir do momento em que a velocidade superficial do fluxo alcançou 0,26m.s-1, a altura de lâmina atingiu 0,0102m, a velocidade de cisalhamento superou os 0,059m.s-1, a tensão de cisalhamento chegou a 3,50Pa e que a potência do escoamento atingiu pelo menos 0,22N.s-1. O processo de incisão iniciou-se com o canal experimental colocado em baixa declividade e em regime de escoamento sub-crítico e de transição. A velocidade de cisalhamento, no momento da incisão, foi, praticamente, o dobro daquela encontrada na literatura para solos siltosos e arenosos. Entretanto, para as três declividades a fase de sulco definido ocorreu somente em regime de escoamento turbulento. A tensão de cisalhamento foi o parâmetro que melhor descreveu a evolução da perda de solo. A potência do escoamento foi o parâmetro hidráulico que mostrou maior eficiência para separar as fases evolutivas dos sulcos. O desenvolvimento do(s) sulco(s) teve o seu início em uma condição de escoamento difuso (ausência de sulcos) e com a potência do escoamento oscilando entre 0,057 e 0,198N.s-1. O avanço do(s) sulco(s) começou com uma zona de transição (fase de incisão e de aprofundamento) onde a potência do escoamento varia entre 0,220 e 0,278N.s-1 e, logo em seguida, teve início a fase de sulco definido, com a potência do escoamento entre 0,314 e 0,544N.s-1. Na fase de escoamento concentrado foi preponderante o papel do processo de erosão regressiva para aumentar tanto o tamanho como o peso das partículas sólidas em transporte pelo escoamento superficial e assim fazer com que predominasse o transporte via fundo sobre o transporte via suspensão. As cargas de sedimento geradas nos solos de diferentes classes texturais foram separadas em grupos distintos em função da potência unitária do escoamento.