In My Experience: The Lessons from Dispersion—Don't Believe Everything You Read

IF 1.8 4区 环境科学与生态学 Q3 WATER RESOURCES
John Cherry
{"title":"In My Experience: The Lessons from Dispersion—Don't Believe Everything You Read","authors":"John Cherry","doi":"10.1111/gwmr.12603","DOIUrl":null,"url":null,"abstract":"<p>This note is about my experience investigating and thinking about hydrodynamic dispersion, simply known as dispersion. In keeping with the tried and true practice of academic reductionism, I will constrain the discussion to the relatively simple and idealized case of point-source groundwater plumes in granular deposits with the aim of illustrating some persistent myths and insights about this process, which has attracted so much interest over the past several decades. Dispersion is the most enigmatic topic in hydrogeology and for good reason.</p><p>To most hydrogeologists, dispersion refers to mixing of solute concentration during transport of the solutes. It occurs at the plume periphery and, in some cases, internally within the plume. The partial differential equation (the advection-dispersion equation… the ADE), on which nearly all mathematical models (analytical and numerical) for representing solute transport and fate in groundwater are based, does not represent fundamentally the way in which advection and mixing happen in the field. This equation can be useful when simulating bulk plume spreading at a specified point in time but is deeply flawed for the intended purpose of representing plume mixing with background groundwater at the field scale. Although, when used carefully, these models can capture the nature of a contaminant plume for particular purposes at a moment in time, they cannot correctly represent the actual processes that govern the evolution of plumes, in the past or future.</p><p>The derivation of the ADE is founded on assumptions that cannot be expected to be met for heterogeneous geologic media and all such media in this context are substantially heterogeneous in their hydraulic conductivity distribution. Solute transport is dependent on the velocity field at a fine spatial scale that can be as small as millimeters or centimeters where molecular diffusion, caused by the local concentration gradients, is the process driving mixing with groundwater at lower solute concentration. This scale is too small for measuring the full velocity field and would require frequent temporal measurements for realistic assessment. Hence, the dispersivity values used in advection-dispersion models are bulk <i>black-box</i> parameters intended to capture the overall effects of the media heterogeneity. But, when viewed in the context of the spatial and temporal scale diffusive mixing, there are important deviations from the assumptions inherent in the derivation of the ADE. This is a serious matter because a common reason to use mathematical models is for realistic representation of the past as a basis for predicting the future. Moreover, the failures of the ADE to accurately represent the processes, limit model usefulness in exploring parameter interdependencies and sensitivities. In essence, what is published in textbooks about these for “transport and fate” models is, at best, misleading and at worst simply wrong. This has been perpetuated from textbook to textbook, starting with the Freeze &amp; Cherry GROUNDWATER, 1979.</p><p>A key parameter in the ADE is the dispersion coefficient that contains a few parameters, including one known as dispersivity with the dimension of length. In the derivation of the ADE, dispersivity is assumed to be a fixed property of the geologic medium, like hydraulic conductivity and porosity. Values of dispersivity are obtained by fitting ADE models to tracer test data or actual plumes. However, these fitting values increase with the time and distance traveled by the solute cloud, implying ever increasing rates of mixing—recall that in the simple case under consideration here, accelerated mixing with time and distance traveled is beyond justification. The reason for this outcome derives from the way in which the diffusion coefficient is represented in the ADE. In other words, diffusion is incorrectly incorporated in the ADE for heterogeneous granular media, even for minimal spatial variability of grain size layers and lenses for which there are only weak variations in the hydraulic conductivity distribution.</p><p>I regret that it has taken me 43 years to get around to writing this warning about the established nonsense written about dispersion in the literature; unfortunately, this nonsense only became evident to me with the great advantage of hindsight. I am thankful that this journal has provided me with the opportunity to unburden myself by correcting this long-standing erroneous thinking that became so embedded in the literature. However, as I try to set the record straight concerning the problems of the ADE, I wish to be clear that I do not advocate its abandonment. The ADE models can be useful and even necessary given lack of practical alternatives, if used with judgment and care. Ironically, I believe that they likely should be used much more than they are. But that tangent is the subject for another essay.</p><p>So how did the ADE become so predominant in groundwater education? When groundwater contamination was first recognized as a topic worthy of attention in the 1960's, the research community conducted laboratory tracer tests in columns and boxes packed with homogeneous sand or glass beads. A mathematical model was needed to simulate the laboratory results, otherwise the experimental results could have no generality. The ADE was imported from the chemical engineering literature and it was found that analytical solutions fitted the laboratory data well enough to support adoption of the ADE in contaminant hydrogeology. There were no competing choices. Soon, the literature provided many analytical solutions to the ADE in 1, 2 and 3 dimensions.</p><p>Dispersion in point-source plumes is typically considered in three directions, first in the direction of flow, known as longitudinal dispersion, and in the two principal directions transverse to the longitudinal flow, known fittingly as transverse dispersion. Advection refers to the transport of the solute with the bulk motion of the water. For clarity and ease of discussion, we will limit this note to the circumstances wherein there is a contaminant plume migrating primarily in the horizontal direction so that longitudinal dispersion refers to dispersion in the horizontal direction. Vertical transverse dispersion occurs at the top and bottom of the plume and horizontal transverse dispersion occurs at the sides of the plume. Dispersion operates most importantly at the periphery of plumes but becomes increasingly important in the interior of plumes as the simple homogeneous case (but unrealistic in the field) gives way to plumes in “real” heterogeneous aquifers.</p><p>My research on contaminant transport and fate began in 1968, focusing on field investigations mostly in sandy deposits. First, I investigated landfill and radionuclide plumes where high-resolution plume characterization methods were first developed and applied. This was early in the transport game. The first paper in a peer review journal about groundwater plumes appeared in 1966 (Water Resources Research, WRR). In1977, colleagues and I began to conduct natural gradient field tracer tests in the Borden aquifer (Sudicky et al., 1978), which is a weakly heterogeneous sand deposit. These tests were found to be essential in the development of understanding solute transport and especially dispersion. They are labor and sample analysis intensive and require much time for the tracer to travel far enough for diffusion to result in meaningful results, which is why such tests had not been done earlier. We initiated more of these tracer tests (e.g., Mackay et al. 1985, WRR) because of claims in the literature that showed a surprisingly large range of transverse dispersivity values. Also, questions arose because the early numerical models used to simulate plumes, without exception, showed strong transverse spreading when actual plumes characterized by high resolution measurements showed the opposite. Hence, the published record for the effects of transverse dispersion had contradictions between field observations and numerical simulations.</p><p>While the first Borden experiment was in progress in 1977-1978 (Figure 1), I was writing the chapter on groundwater contamination in the Freeze &amp; Cherry book. In retrospect, the dispersion discussion in the transport chapter is scientifically schizophrenic in that two conflicting lines of evidence are presented in the chapter's figures without recognition that any conflict is there. One, based on high resolution sampling, shows the weak strength of transverse vertical dispersion (small vertical spreading) of the landfill plume at the Borden site. On the other hand, there is a figure also in this chapter, produced by numerical simulations, for a longitudinal cross section through a hypothetical sand aquifer that shows strong vertical transverse dispersion (large vertical spreading entirely unlike the actual Borden plume; Figure 1). This simulation was done using one of the earliest numerical models for solute transport. Like all the early numerical models for transport, it suffered from strong numerical dispersion. Revisiting these figures, it is apparent to me now that both represented plumes misleadingly because both suggested what appeared to be strong transverse vertical dispersion. As I was writing the Freeze-Cherry chapter, the results of the first Borden tracer experiment arrived. They showed weak transverse dispersion, as was the case for the nearby landfill plume, but the significance of this did not adequately register in my mind. I continued to think of plumes as they were depicted in the first numerical models of solute transport, such as one that appeared in a paper in 1973 (Pinder, WRR) in which the simulations were made to fit a chromium plume mapped using only low-resolution methods (i.e., conventional monitoring wells). Ironically, in this case the erroneous simulations based on numerical dispersion fitted the erroneous field plume based on the blended samples typical of conventional monitoring wells where the water intake interval is too long. These two wrongs resulted in an even bigger wrong in conceptualizing plumes.</p><p>I now recognize that, at the time, there were extreme differences between high-resolution field plumes and model plumes and that the field evidence overwhelmingly supported weak transverse dispersion in sandy aquifers of weak to moderate heterogeneity. I recall that euphoria abounded in contaminant hydrogeology in that era due to the great possibilities of the relatively powerful computers of the time. This apparently clouded my perception of what the conflicting data should have been telling me. It was not until 1994 that Hoekanan and Fried (WRR) properly simulated the Borden landfill plume in 3D without appreciable numerical dispersion. Misunderstandings about dispersion were promulgated further in the 1980-1990 era when graphs appeared in the literature indicating that dispersivity values showed longitudinal dispersivity directly related to the distance traveled by the plume and most importantly these papers also claimed a clear relationship between the magnitude of longitudinal dispersivity and the transverse dispersivity values. These graphs are most reasonably attributed to artifacts arising from the blending of water drawn into monitoring wells that connect zones with different solute concentrations (low-resolution sampling) and the inclusion of results from different hydrogeologic circumstances (apples and oranges). Unfortunately, this relationship between the longitudinal and transverse dispersivity values has been reproduced in many textbooks. It represents the ultimate misleading conceptualization about dispersion. Unfortunately, this perception is widely held by practitioners. Instead, reasonable dispersivity values can be useful in some cases for contaminated site simulations when selected based on what has been learned cumulatively based on plumes characterized since the late 1950's using high-resolution methods and, after 1978, based on natural gradient tracer tests.</p><p>Not long after the publication of the Freeze &amp; Cherry book (1979), we pointed out (Gillham and Cherry, GSA Special Vol., 1982) that the ADE representation of transport was flawed. We presented an alternative conceptual model, referred to as the advection-diffusion model. In this conceptualization, if we were to know the velocity distribution in enough detail, what passes for dispersion (i.e., mixing) in the ADE model is explained for mobile solutes based on advection combined only with solute diffusion. This thinking was to be the start of what could have been a paradigm shift for dispersion but the shift did not evolve to maturity. The literature supporting the advection-diffusion view grew by a few papers but was largely ignored, perhaps because there are disadvantages associated with such a paradigm shift. For example, acceptance of it requires site characterization at spatial scales much finer than what has become “standard practice.” Also, there is continuing convenience to the belief in strong transverse dispersion in the use of ADE models because this makes numerical simulations easier. Large dispersivity values stabilize the model numerics and the simulations produce wide, fan-shaped rather than narrow pencil-shaped plumes. Furthermore, the perceived fan-shaped plumes are much less expensive to monitor because they do not require closely spaced monitoring points for plume detection. For example, for a landfill with a leaky patch of liner that creates a 5-m wide contaminant input to a sand aquifer with steady flow, the simulated plume width at a distance of 100 m downgradient is only about 6-8 m wide when reasonably small lateral transverse dispersivity values are used. However, the lateral plume spread (width) is 20 m or even much more when the much larger values in common in use in plume modeling are assigned. Conveniently, plumes that go under-detected are less expensive to deal with in the regulatory realm, at least in the short term. In this case and many others, longitudinal dispersion is irrelevant. It is the transverse dispersion that matters.</p><p>In summary, in the conceptual and mathematical representations of plumes, dispersion should not be ignored when a particular understanding of contaminant plumes is needed. An example is when monitored natural attention (MNA) is proposed because mixing can govern rates of biodegradation. But as a parameter in models for predicting the behavior of plumes into the future, dispersivity is problematic because the values cannot be measured in any practical way at the appropriate spatial and temporal scales. The values must be assigned based on judgment derived from comprehensive knowledge of the site in the context of relevant dispersion literature.</p><p>When new or altered numerical models for transport are to be used, verification with analytical solutions that incorporate the appropriate processes and/or field data are essential to avoid excessive misrepresentations of reality. A practical challenge in the assessment of contamination plumes is the decision to apply ADE models or simply rely on assessment of the groundwater flow system, which may be all that is needed for the circumstance.</p><p>The overall lesson I have taken away from this, and a few other experiences during my career, is that there are concepts published in books that, although they appear to be valid at face value, are nonetheless wrong. We must maintain inquiring minds and be vigilant to recognize when simulations deviate from the field evidence and be willing to adjust the written record and academic training accordingly. Concerning dispersion, more thinking is still needed.</p>","PeriodicalId":55081,"journal":{"name":"Ground Water Monitoring and Remediation","volume":"43 3","pages":"145-147"},"PeriodicalIF":1.8000,"publicationDate":"2023-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/gwmr.12603","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Ground Water Monitoring and Remediation","FirstCategoryId":"93","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1111/gwmr.12603","RegionNum":4,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"WATER RESOURCES","Score":null,"Total":0}
引用次数: 0

Abstract

This note is about my experience investigating and thinking about hydrodynamic dispersion, simply known as dispersion. In keeping with the tried and true practice of academic reductionism, I will constrain the discussion to the relatively simple and idealized case of point-source groundwater plumes in granular deposits with the aim of illustrating some persistent myths and insights about this process, which has attracted so much interest over the past several decades. Dispersion is the most enigmatic topic in hydrogeology and for good reason.

To most hydrogeologists, dispersion refers to mixing of solute concentration during transport of the solutes. It occurs at the plume periphery and, in some cases, internally within the plume. The partial differential equation (the advection-dispersion equation… the ADE), on which nearly all mathematical models (analytical and numerical) for representing solute transport and fate in groundwater are based, does not represent fundamentally the way in which advection and mixing happen in the field. This equation can be useful when simulating bulk plume spreading at a specified point in time but is deeply flawed for the intended purpose of representing plume mixing with background groundwater at the field scale. Although, when used carefully, these models can capture the nature of a contaminant plume for particular purposes at a moment in time, they cannot correctly represent the actual processes that govern the evolution of plumes, in the past or future.

The derivation of the ADE is founded on assumptions that cannot be expected to be met for heterogeneous geologic media and all such media in this context are substantially heterogeneous in their hydraulic conductivity distribution. Solute transport is dependent on the velocity field at a fine spatial scale that can be as small as millimeters or centimeters where molecular diffusion, caused by the local concentration gradients, is the process driving mixing with groundwater at lower solute concentration. This scale is too small for measuring the full velocity field and would require frequent temporal measurements for realistic assessment. Hence, the dispersivity values used in advection-dispersion models are bulk black-box parameters intended to capture the overall effects of the media heterogeneity. But, when viewed in the context of the spatial and temporal scale diffusive mixing, there are important deviations from the assumptions inherent in the derivation of the ADE. This is a serious matter because a common reason to use mathematical models is for realistic representation of the past as a basis for predicting the future. Moreover, the failures of the ADE to accurately represent the processes, limit model usefulness in exploring parameter interdependencies and sensitivities. In essence, what is published in textbooks about these for “transport and fate” models is, at best, misleading and at worst simply wrong. This has been perpetuated from textbook to textbook, starting with the Freeze & Cherry GROUNDWATER, 1979.

A key parameter in the ADE is the dispersion coefficient that contains a few parameters, including one known as dispersivity with the dimension of length. In the derivation of the ADE, dispersivity is assumed to be a fixed property of the geologic medium, like hydraulic conductivity and porosity. Values of dispersivity are obtained by fitting ADE models to tracer test data or actual plumes. However, these fitting values increase with the time and distance traveled by the solute cloud, implying ever increasing rates of mixing—recall that in the simple case under consideration here, accelerated mixing with time and distance traveled is beyond justification. The reason for this outcome derives from the way in which the diffusion coefficient is represented in the ADE. In other words, diffusion is incorrectly incorporated in the ADE for heterogeneous granular media, even for minimal spatial variability of grain size layers and lenses for which there are only weak variations in the hydraulic conductivity distribution.

I regret that it has taken me 43 years to get around to writing this warning about the established nonsense written about dispersion in the literature; unfortunately, this nonsense only became evident to me with the great advantage of hindsight. I am thankful that this journal has provided me with the opportunity to unburden myself by correcting this long-standing erroneous thinking that became so embedded in the literature. However, as I try to set the record straight concerning the problems of the ADE, I wish to be clear that I do not advocate its abandonment. The ADE models can be useful and even necessary given lack of practical alternatives, if used with judgment and care. Ironically, I believe that they likely should be used much more than they are. But that tangent is the subject for another essay.

So how did the ADE become so predominant in groundwater education? When groundwater contamination was first recognized as a topic worthy of attention in the 1960's, the research community conducted laboratory tracer tests in columns and boxes packed with homogeneous sand or glass beads. A mathematical model was needed to simulate the laboratory results, otherwise the experimental results could have no generality. The ADE was imported from the chemical engineering literature and it was found that analytical solutions fitted the laboratory data well enough to support adoption of the ADE in contaminant hydrogeology. There were no competing choices. Soon, the literature provided many analytical solutions to the ADE in 1, 2 and 3 dimensions.

Dispersion in point-source plumes is typically considered in three directions, first in the direction of flow, known as longitudinal dispersion, and in the two principal directions transverse to the longitudinal flow, known fittingly as transverse dispersion. Advection refers to the transport of the solute with the bulk motion of the water. For clarity and ease of discussion, we will limit this note to the circumstances wherein there is a contaminant plume migrating primarily in the horizontal direction so that longitudinal dispersion refers to dispersion in the horizontal direction. Vertical transverse dispersion occurs at the top and bottom of the plume and horizontal transverse dispersion occurs at the sides of the plume. Dispersion operates most importantly at the periphery of plumes but becomes increasingly important in the interior of plumes as the simple homogeneous case (but unrealistic in the field) gives way to plumes in “real” heterogeneous aquifers.

My research on contaminant transport and fate began in 1968, focusing on field investigations mostly in sandy deposits. First, I investigated landfill and radionuclide plumes where high-resolution plume characterization methods were first developed and applied. This was early in the transport game. The first paper in a peer review journal about groundwater plumes appeared in 1966 (Water Resources Research, WRR). In1977, colleagues and I began to conduct natural gradient field tracer tests in the Borden aquifer (Sudicky et al., 1978), which is a weakly heterogeneous sand deposit. These tests were found to be essential in the development of understanding solute transport and especially dispersion. They are labor and sample analysis intensive and require much time for the tracer to travel far enough for diffusion to result in meaningful results, which is why such tests had not been done earlier. We initiated more of these tracer tests (e.g., Mackay et al. 1985, WRR) because of claims in the literature that showed a surprisingly large range of transverse dispersivity values. Also, questions arose because the early numerical models used to simulate plumes, without exception, showed strong transverse spreading when actual plumes characterized by high resolution measurements showed the opposite. Hence, the published record for the effects of transverse dispersion had contradictions between field observations and numerical simulations.

While the first Borden experiment was in progress in 1977-1978 (Figure 1), I was writing the chapter on groundwater contamination in the Freeze & Cherry book. In retrospect, the dispersion discussion in the transport chapter is scientifically schizophrenic in that two conflicting lines of evidence are presented in the chapter's figures without recognition that any conflict is there. One, based on high resolution sampling, shows the weak strength of transverse vertical dispersion (small vertical spreading) of the landfill plume at the Borden site. On the other hand, there is a figure also in this chapter, produced by numerical simulations, for a longitudinal cross section through a hypothetical sand aquifer that shows strong vertical transverse dispersion (large vertical spreading entirely unlike the actual Borden plume; Figure 1). This simulation was done using one of the earliest numerical models for solute transport. Like all the early numerical models for transport, it suffered from strong numerical dispersion. Revisiting these figures, it is apparent to me now that both represented plumes misleadingly because both suggested what appeared to be strong transverse vertical dispersion. As I was writing the Freeze-Cherry chapter, the results of the first Borden tracer experiment arrived. They showed weak transverse dispersion, as was the case for the nearby landfill plume, but the significance of this did not adequately register in my mind. I continued to think of plumes as they were depicted in the first numerical models of solute transport, such as one that appeared in a paper in 1973 (Pinder, WRR) in which the simulations were made to fit a chromium plume mapped using only low-resolution methods (i.e., conventional monitoring wells). Ironically, in this case the erroneous simulations based on numerical dispersion fitted the erroneous field plume based on the blended samples typical of conventional monitoring wells where the water intake interval is too long. These two wrongs resulted in an even bigger wrong in conceptualizing plumes.

I now recognize that, at the time, there were extreme differences between high-resolution field plumes and model plumes and that the field evidence overwhelmingly supported weak transverse dispersion in sandy aquifers of weak to moderate heterogeneity. I recall that euphoria abounded in contaminant hydrogeology in that era due to the great possibilities of the relatively powerful computers of the time. This apparently clouded my perception of what the conflicting data should have been telling me. It was not until 1994 that Hoekanan and Fried (WRR) properly simulated the Borden landfill plume in 3D without appreciable numerical dispersion. Misunderstandings about dispersion were promulgated further in the 1980-1990 era when graphs appeared in the literature indicating that dispersivity values showed longitudinal dispersivity directly related to the distance traveled by the plume and most importantly these papers also claimed a clear relationship between the magnitude of longitudinal dispersivity and the transverse dispersivity values. These graphs are most reasonably attributed to artifacts arising from the blending of water drawn into monitoring wells that connect zones with different solute concentrations (low-resolution sampling) and the inclusion of results from different hydrogeologic circumstances (apples and oranges). Unfortunately, this relationship between the longitudinal and transverse dispersivity values has been reproduced in many textbooks. It represents the ultimate misleading conceptualization about dispersion. Unfortunately, this perception is widely held by practitioners. Instead, reasonable dispersivity values can be useful in some cases for contaminated site simulations when selected based on what has been learned cumulatively based on plumes characterized since the late 1950's using high-resolution methods and, after 1978, based on natural gradient tracer tests.

Not long after the publication of the Freeze & Cherry book (1979), we pointed out (Gillham and Cherry, GSA Special Vol., 1982) that the ADE representation of transport was flawed. We presented an alternative conceptual model, referred to as the advection-diffusion model. In this conceptualization, if we were to know the velocity distribution in enough detail, what passes for dispersion (i.e., mixing) in the ADE model is explained for mobile solutes based on advection combined only with solute diffusion. This thinking was to be the start of what could have been a paradigm shift for dispersion but the shift did not evolve to maturity. The literature supporting the advection-diffusion view grew by a few papers but was largely ignored, perhaps because there are disadvantages associated with such a paradigm shift. For example, acceptance of it requires site characterization at spatial scales much finer than what has become “standard practice.” Also, there is continuing convenience to the belief in strong transverse dispersion in the use of ADE models because this makes numerical simulations easier. Large dispersivity values stabilize the model numerics and the simulations produce wide, fan-shaped rather than narrow pencil-shaped plumes. Furthermore, the perceived fan-shaped plumes are much less expensive to monitor because they do not require closely spaced monitoring points for plume detection. For example, for a landfill with a leaky patch of liner that creates a 5-m wide contaminant input to a sand aquifer with steady flow, the simulated plume width at a distance of 100 m downgradient is only about 6-8 m wide when reasonably small lateral transverse dispersivity values are used. However, the lateral plume spread (width) is 20 m or even much more when the much larger values in common in use in plume modeling are assigned. Conveniently, plumes that go under-detected are less expensive to deal with in the regulatory realm, at least in the short term. In this case and many others, longitudinal dispersion is irrelevant. It is the transverse dispersion that matters.

In summary, in the conceptual and mathematical representations of plumes, dispersion should not be ignored when a particular understanding of contaminant plumes is needed. An example is when monitored natural attention (MNA) is proposed because mixing can govern rates of biodegradation. But as a parameter in models for predicting the behavior of plumes into the future, dispersivity is problematic because the values cannot be measured in any practical way at the appropriate spatial and temporal scales. The values must be assigned based on judgment derived from comprehensive knowledge of the site in the context of relevant dispersion literature.

When new or altered numerical models for transport are to be used, verification with analytical solutions that incorporate the appropriate processes and/or field data are essential to avoid excessive misrepresentations of reality. A practical challenge in the assessment of contamination plumes is the decision to apply ADE models or simply rely on assessment of the groundwater flow system, which may be all that is needed for the circumstance.

The overall lesson I have taken away from this, and a few other experiences during my career, is that there are concepts published in books that, although they appear to be valid at face value, are nonetheless wrong. We must maintain inquiring minds and be vigilant to recognize when simulations deviate from the field evidence and be willing to adjust the written record and academic training accordingly. Concerning dispersion, more thinking is still needed.

Abstract Image

根据我的经验:分散的教训——不要相信你读到的一切
具有讽刺意味的是,在这种情况下,基于数值离散的错误模拟拟合了基于混合样本的错误场羽流,该混合样本是取水间隔过长的传统监测井的典型样本。这两个错误导致了羽流概念化的更大错误。我现在认识到,当时,高分辨率的现场羽流和模型羽流之间存在极端差异,现场证据压倒性地支持弱到中等非均质性的砂质含水层中的弱横向分散。我记得,在那个时代,由于当时相对强大的计算机的巨大可能性,污染水文地质中充满了喜悦。这显然模糊了我对相互矛盾的数据应该告诉我什么的看法。直到1994年,Hoekanan和Fried(WRR)才在没有明显数值离散的情况下正确地模拟了Borden垃圾填埋场的羽流。1980年至1990年,当文献中出现图表表明分散度值显示纵向分散度与羽流行进的距离直接相关时,对分散度的误解进一步加剧,最重要的是,这些论文还声称纵向分散度的大小和横向分散度之间存在明确的关系价值观这些图最合理地归因于将抽取到连接不同溶质浓度区域的监测井中的水混合(低分辨率采样)以及包含不同水文地质环境(苹果和橙子)的结果所产生的伪影。不幸的是,纵向和横向分散性值之间的这种关系在许多教科书中都被复制了。它代表了关于分散的最终误导性概念。不幸的是,从业者普遍持有这种看法。相反,在某些情况下,根据自20世纪50年代末以来使用高分辨率方法和1978年后基于自然梯度示踪剂测试表征的羽流的累积经验选择合理的分散性值,对于污染场地模拟来说可能是有用的。在Freeze&amp;Cherry的书(1979年),我们指出(Gillham和Cherry,GSA Special Vol.1982),ADE对运输的描述是有缺陷的。我们提出了另一种概念模型,称为平流-扩散模型。在这种概念化中,如果我们要足够详细地了解速度分布,那么ADE模型中的分散(即混合)是基于仅与溶质扩散相结合的平流的流动溶质的解释。这种想法是分散模式转变的开始,但这种转变并没有发展到成熟。支持平流-扩散观点的文献在几篇论文中有所增长,但在很大程度上被忽视了,也许是因为这种范式转变存在缺点。例如,接受它需要在空间尺度上对场地进行表征,这比现在的“标准做法”要精细得多。此外,在使用ADE模型时,相信强横向色散仍然很方便,因为这使得数值模拟更容易。较大的分散度值稳定了模型数值,模拟产生了宽的扇形羽流,而不是窄的铅笔状羽流。此外,感知到的扇形羽流的监测成本要低得多,因为它们不需要用于羽流检测的紧密间隔的监测点。例如,对于一个具有泄漏衬垫的垃圾填埋场,该衬垫会产生5米宽的污染物输入到具有稳定流量的含水层中,距离100处的模拟羽流宽度 m降级只有6-8 m宽,当使用相当小的横向横向分散度值时。然而,横向羽流扩散(宽度)为20 m,甚至在羽流建模中常用的更大的值被指定时更大。方便的是,至少在短期内,在监管领域处理探测不足的羽状物的成本更低。在这种情况和许多其他情况下,纵向色散是无关紧要的。重要的是横向色散。总之,在羽流的概念和数学表示中,当需要对污染物羽流进行特殊理解时,不应忽视扩散。一个例子是,当提出监测自然注意(MNA)时,因为混合可以控制生物降解的速率。但是,作为预测未来羽流行为的模型中的一个参数,分散性是有问题的,因为无法在适当的空间和时间尺度上以任何实际的方式测量这些值。必须根据相关分散文献中对现场的全面了解得出的判断来分配值。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
CiteScore
3.30
自引率
10.50%
发文量
60
审稿时长
>36 weeks
期刊介绍: Since its inception in 1981, Groundwater Monitoring & Remediation® has been a resource for researchers and practitioners in the field. It is a quarterly journal that offers the best in application oriented, peer-reviewed papers together with insightful articles from the practitioner''s perspective. Each issue features papers containing cutting-edge information on treatment technology, columns by industry experts, news briefs, and equipment news. GWMR plays a unique role in advancing the practice of the groundwater monitoring and remediation field by providing forward-thinking research with practical solutions.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信