Sandra M. Eberts

Comparison of particle-tracking and lumped-parameter age-distribution models for evaluating vulnerability of production wells to contamination

Environmental age tracers have been used in various ways to help assess vulnerability of drinking-water production wells to contamination. The most appropriate approach will depend on the information that is available and that which is desired. To understand how the well will respond to changing nonpoint-source contaminant inputs at the water table, some representation of the distribution of groundwater ages in the well is needed. Such information for production wells is sparse and difficult to obtain, especially in areas lacking detailed field studies. In this study, age distributions derived from detailed groundwater-flow models with advective particle tracking were compared with those generated from lumped-parameter models to examine conditions in which estimates from simpler, less resource-intensive lumped-parameter models could be used in place of estimates from particle-tracking models. In each of four contrasting hydrogeologic settings in the USA, particle-tracking and lumped-parameter models yielded roughly similar age distributions and largely indistinguishable contaminant trends when based on similar conceptual models and calibrated to similar tracer data. Although model calibrations and predictions were variably affected by tracer limitations and conceptual ambiguities, results illustrated the importance of full age distributions, rather than apparent tracer ages or model mean ages, for trend analysis and forecasting.RésuméDes traceurs environnementaux ont été utilisés de différentes façons pour aider à évaluer la vulnérabilité de puits de production d’eau potable à la pollution. L’approche la plus appropriée dépendra de l’information disponible et de celle qui est recherchée. Pour comprendre comment le puits répond à des sources de polluants non localisées et variables, il faut nécessairement une représentation de la distribution des âges de l’eau dans le puits. Une telle information sur des puits productifs est rare et difficile à obtenir, particulièrement dans les zones où manquent des études de terrain détaillées. Dans cette étude, des distributions d’âge déduites de modèles de flux souterrain détaillés avec suivi de particules advectives ont été comparés avec ceux générés par des modèles paramétriques globaux pour examiner les conditions dans lesquelles des estimations tirées de modèles plus simples, de modèles paramétrés moins liés à la ressource pourraient être utilisés à la place d’estimations issues de modèles de suivi de particule. Dans chacun des quatre sites hydrogéologiques contrastées aux USA, des modèles de suivi de particule et des modèles à paramétrage global basées sur des schémas conceptuels similaires et paramétrés pour des traçage similaires ont fourni en gros des distributions d’âge et des tendances de contamination largement semblables. Bien que les paramétrages de modèle et prévisions aient été affectés de façon variable par des limitations du traçage et par des ambiguités conceptuelles, les résultats ont illustré l’importance des distributions de l’âge total, comparativement aux âges apparents de traçage ou âges moyens du modèle, pour l’analyse de tendance et pour la prévision.ResumenLos trazadores de edad ambiental han sido usados de varias maneras para ayudar a evaluar de vulnerabilidad de los pozos de producción de agua potable a la contaminación. El enfoque más apropiado depende de la información que está disponible y que la que es necesaria. Para entender como el pozo responde a la entrada variable de la fuente no puntual del contaminante en la capa freática se necesita alguna representación de la distribución de las edades de agua subterránea en el pozo. Tal información para los pozos de producción está dispersa y es difícil de obtener, especialmente en áreas carentes de un estudio detallado de campo. En este estudio se compararon las distribuciones de edades derivadas de modelos detallados de flujo de agua subterránea con el seguimiento advectivo de partículas para examinar las condiciones en las cuales las estimaciones a partir de modelos de parámetros concentrados más simples y menos intensivamente dependientes de los recursos podrían ser usados en lugar de la estimación de los modelos de seguimiento de partículas. En cada una de las cuatro configuraciones hidrogeológicas contrastantes en EEUU, el seguimiento de partículas y los modelos de parámetros concentrados brindaron a grandes rasgos las distribuciones de edad y las tendencias de contaminación resultan grandemente indistinguibles cuando estaban basadas en modelos conceptuales similares y calibrados con datos de similares trazadores. A pesar que las calibraciones del modelo y las predicciones estuvieron afectadas variablemente por las limitaciones de los trazadores y las ambigüedades conceptuales, los resultados ilustraron la importancia de las distribuciones de edad completa, más bien que las edades aparentes de trazadores o de modelos de edades medias, para los análisis de tendencias y predicciones.摘要用于评价饮用水生产井污染脆弱性的环境年龄示踪已被广泛应用。最适合的方法取决于可利用的信息及所需数据。为了解井对水面上非点源污染物输入变化的响应,必须知道地下水年龄的分布特征。生产井的这些信息较少,且难获得,尤其是在缺少详细野外调查的地区。本次研究,年龄分布来源于详细的地下水流动模型以及平流溶质运移,并与来源于集中参数模拟的年龄分布进行比较,以确定简单少源的加强集中参数模型估计值可用于替代颗粒示踪模型估计值的条件 。在美国的四个对比水文地质设置点,颗粒示踪及集中参数模拟法基于简单概念模型,对简单示踪数据进行验证,得出了大致的简单年龄分布及主要的难分辨的污染物趋势。尽管模型验证与预测受示踪物局限性及概念的不准确性的变化影响,结果表明对趋势分析及预测而言,完整的年龄分布比表观示踪年龄或模型平均年龄更有意义。ResumoTraçadores ambientais de idade têm sido usados de várias maneiras para ajudar a avaliar a vulnerabilidade das captações de produção de água potável à contaminação. A abordagem mais adequada dependerá da informação que está disponível e do que é desejado. Para entender como o poço vai responder às variações das entradas de contaminantes difusos no nível freático, é necessária alguma representação da distribuição das idades das águas subterrâneas no poço. Estas informações são escassas e difíceis de obter para furos de produção, especialmente em áreas carentes de estudos de campo pormenorizados. Neste estudo, as distribuições de idade provenientes de modelos de fluxo de águas subterrâneas detalhados com rastreio de partículas advetivas foram comparadas com aqueles gerados a partir de modelos de parâmetros agregados, para examinar as condições em que as estimativas obtidas a partir de modelos mais simples, com menos recursos, poderiam ser usadas em vez das estimativas a partir de modelos de partículas de rastreio. Uma em cada quatro das configurações hidrogeológicas nos EUA, obtidas por modelos de partículas de rastreio e por modelos de parâmetros agregados, produziram distribuições de idade mais ou menos semelhantes e tendências de contaminação em grande parte indistinguíveis quando baseadas em modelos conceptuais semelhantes e calibrados para idênticos dados de traçadores. Embora a calibração e as previsões do modelo tenham sido afetadas de forma variável por limitações dos traçadores e por ambiguidades conceptuais, os resultados demonstraram a importância das distribuições de idade total, ao invés de modelos de idades aparentes de traçadores ou de modelos de média de idades, para análises de tendências e previsões.

Modeling nitrate at domestic and public-supply well depths in the Central Valley, California

Aquifer vulnerability models were developed to map groundwater nitrate concentration at domestic and public-supply well depths in the Central Valley, California. We compared three modeling methods for ability to predict nitrate concentration >4 mg/L: logistic regression (LR), random forest classification (RFC), and random forest regression (RFR). All three models indicated processes of nitrogen fertilizer input at the land surface, transmission through coarse-textured, well-drained soils, and transport in the aquifer to the well screen. The total percent correct predictions were similar among the three models (69-82%), but RFR had greater sensitivity (84% for shallow wells and 51% for deep wells). The results suggest that RFR can better identify areas with high nitrate concentration but that LR and RFC may better describe bulk conditions in the aquifer. A unique aspect of the modeling approach was inclusion of outputs from previous, physically based hydrologic and textural models as predictor variables, which were important to the models. Vertical water fluxes in the aquifer and percent coarse material above the well screen were ranked moderately high-to-high in the RFR models, and the average vertical water flux during the irrigation season was highly significant (p < 0.0001) in logistic regression.

Vulnerability of a Public Supply Well in a Karstic Aquifer to Contamination

To assess the vulnerability of ground water to contamination in the karstic Upper Floridan aquifer (UFA), age-dating tracers and selected anthropogenic and naturally occurring compounds were analyzed in multiple water samples from a public supply well (PSW) near Tampa, Florida. Samples also were collected from 28 monitoring wells in the UFA and the overlying surficial aquifer system (SAS) and intermediate confining unit located within the contributing recharge area to the PSW. Age tracer and geochemical data from the earlier stage of the study (2003 through 2005) were combined with new data (2006) on concentrations of sulfur hexafluoride (SF(6)), tritium ((3)H), and helium-3, which were consistent with binary mixtures of water for the PSW dominated by young water (less than 7 years). Water samples from the SAS also indicated mostly young water (less than 7 years); however, most water samples from monitoring wells in the UFA had lower SF(6) and (3)H concentrations than the PSW and SAS, indicating mixtures containing high proportions of older water (more than 60 years). Vulnerability of the PSW to contamination was indicated by predominantly young water and elevated nitrate-N and volatile organic compound concentrations that were similar to those in the SAS. Elevated arsenic (As) concentrations (3 to 19 microg/L) and higher As(V)/As(III) ratios in the PSW than in water from UFA monitoring wells indicate that oxic water from the SAS likely mobilizes As from pyrite in the UFA matrix. Young water found in the PSW also was present in UFA monitoring wells that tap a highly transmissive zone (43- to 53-m depth) in the UFA.

A ternary age-mixing model to explain contaminant occurrence in a deep supply well

The age distribution of water from a public-supply well in a deep alluvial aquifer was estimated and used to help explain arsenic variability in the water. The age distribution was computed using a ternary mixing model that combines three lumped parameter models of advection-dispersion transport of environmental tracers, which represent relatively recent recharge (post-1950s) containing volatile organic compounds (VOCs), old intermediate depth groundwater (about 6500 years) that was free of drinking-water contaminants, and very old, deep groundwater (more than 21,000 years) containing arsenic above the USEPA maximum contaminant level of 10 µg/L. The ternary mixing model was calibrated to tritium, chloroflorocarbon-113, and carbon-14 (14C) concentrations that were measured in water samples collected on multiple occasions. Variability in atmospheric 14C over the past 50,000 years was accounted for in the interpretation of 14C as a tracer. Calibrated ternary models indicate the fraction of deep, very old groundwater entering the well varies substantially throughout the year and was highest following long periods of nonoperation or infrequent operation, which occured during the winter season when water demand was low. The fraction of young water entering the well was about 11% during the summer when pumping peaked to meet water demand and about 3% to 6% during the winter months. This paper demonstrates how collection of multiple tracers can be used in combination with simplified models of fluid flow to estimate the age distribution and thus fraction of contaminated groundwater reaching a supply well under different pumping conditions.

A hybrid machine learning model to predict and visualize nitrate concentration throughout the Central Valley aquifer, California, USA

Intense demand for water in the Central Valley of California and related increases in groundwater nitrate concentration threaten the sustainability of the groundwater resource. To assess contamination risk in the region, we developed a hybrid, non-linear, machine learning model within a statistical learning framework to predict nitrate contamination of groundwater to depths of approximately 500m below ground surface. A database of 145 predictor variables representing well characteristics, historical and current field and landscape-scale nitrogen mass balances, historical and current land use, oxidation/reduction conditions, groundwater flow, climate, soil characteristics, depth to groundwater, and groundwater age were assigned to over 6000 private supply and public supply wells measured previously for nitrate and located throughout the study area. The boosted regression tree (BRT) method was used to screen and rank variables to predict nitrate concentration at the depths of domestic and public well supplies. The novel approach included as predictor variables outputs from existing physically based models of the Central Valley. The top five most important predictor variables included two oxidation/reduction variables (probability of manganese concentration to exceed 50ppb and probability of dissolved oxygen concentration to be below 0.5ppm), field-scale adjusted unsaturated zone nitrogen input for the 1975 time period, average difference between precipitation and evapotranspiration during the years 1971-2000, and 1992 total landscape nitrogen input. Twenty-five variables were selected for the final model for log-transformed nitrate. In general, increasing probability of anoxic conditions and increasing precipitation relative to potential evapotranspiration had a corresponding decrease in nitrate concentration predictions. Conversely, increasing 1975 unsaturated zone nitrogen leaching flux and 1992 total landscape nitrogen input had an increasing relative impact on nitrate predictions. Three-dimensional visualization indicates that nitrate predictions depend on the probability of anoxic conditions and other factors, and that nitrate predictions generally decreased with increasing groundwater age.

Educational webtool illustrating groundwater age effects on contaminant trends in wells

Trends in concentrations of nonpoint-source contaminants in wells, springs, and streams are related to the history of contamination in groundwater recharge and the age distribution in the groundwater discharge. The age distribution in discharge depends on the groundwater age distribution in the aquifer and the subset of flowpaths that are sampled by the discharge. Groundwater travel times from recharge to discharge are variable; consequently, responses at discharge locations to changing contaminant loading in recharge can include delayed initial responses, dilution of peak concentrations, and prolonged flushing times. These effects are well understood in principle and have important consequences for water resource management (Eberts et al. 2013), but their implications may not be easy to visualize or communicate.

Modeling the Potential Impact of Seasonal and Inactive Multi-Aquifer Wells on Contaminant Movement to Public Water-Supply Wells

Wells screened across multiple aquifers can provide pathways for the movement of surprisingly large volumes of groundwater to confined aquifers used for public water supply (PWS). Using a simple numerical model, we examine the impact of several pumping scenarios on leakage from an unconfined aquifer to a confined aquifer and conclude that a single inactive multi-aquifer well can contribute nearly 10% of total PWS well flow over a wide range of pumping rates. This leakage can occur even when the multi-aquifer well is more than a kilometer from the PWS well. The contribution from multi-aquifer wells may be greater under conditions where seasonal pumping (e.g., irrigation) creates large, widespread downward hydraulic gradients between aquifers. Under those conditions, water can continue to leak down a multi-aquifer well from an unconfined aquifer to a confined aquifer even when those multi-aquifer wells are actively pumped. An important implication is that, if an unconfined aquifer is contaminated, multi-aquifer wells can increase the vulnerability of a confined-aquifer PWS well.

Assessing the vulnerability of public-supply wells to contamination: Rio Grande aquifer system in Albuquerque, New Mexico

The Rio Grande aquifer system in New Mexico consists of a series of basins located along a generally north-south-trending zone called the Rio Grande Rift. Here, tectonic forces in the past forced the crust to begin to pull apart, creating downwardfaulted basins flanked by mountains. Over time, the basins gradually filled with alluvial sediments. Albuquerque is within the Middle Rio Grande Basin in central New Mexico. Sediments consisting of sand, silt, gravel, and clay in varying amounts have filled the basin to a maximum thickness of about 15,000 feet. The public-supply well selected for study is screened from 351 to 1,179 feet below land surface. Like most public-supply and commercial wells in the Albuquerque area, it draws water from the part of the alluvial fill known as the Santa Fe Group.

Assessing the vulnerability of public-supply wells to contamination--Glacial aquifer system in Woodbury, Connecticut

Printed on recycled paper The glacial aquifer system in the Pomperaug River Basin is made up of glacial deposits and fractured bedrock. Stratified (layered) sediments deposited by glacial meltwater fill in the valley underlying the Town of Woodbury. Such deposits are generally the most productive sources of groundwater in New England. Fractured bedrock lies beneath the valley and the adjacent uplands. Glacial till (a dense mixture of different-sized rock particles in a fine-grained matrix) blankets the bedrock surface in the uplands and in some areas beneath the stratified glacial deposits in the valley. Together, stratified glacial deposits and till are referred to as “glacial deposits” in this fact sheet. Bedrock beneath the valley is composed of basalt and sedimentary rocks (shale, siltstone, conglomerate, and organic-rich limestone), whereas bedrock in the uplands includes basalt and metamorphic and granitic rocks. Recharge to the aquifer system originates as rainfall or as stream water that infiltrates to the water table. In the uplands, infiltrating water reaches the water table and then flows laterally through till to upland streams or downward into bedrock fractures. Water eventually flows from the bedrock through the sides and bottom of the filled valley into the glacial deposits. Groundwater in the valley discharges to streams, ponds, or wells in the area. The public-supply well selected for study is open to the glacial deposits from 45 to 60 feet below land surface. It supplies water to about 2,000 Woodbury residents who live in the valley bottom. Most of the groundwater withdrawn in the study area is returned to the subsurface as flow through septic-tank drainfields, because all residences and businesses in Woodbury are unsewered and use onsite septic systems to treat human waste. Glacial aquifer system

Assessing the Vulnerability of Public-Supply Wells to Contamination: Central Valley Aquifer System near Modesto, California

Near Modesto, the primary water-bearing units of the Central Valley aquifer system are a sequence of unconsolidated, interlayered lenses of sand, gravel, silt, and clay that originated as alluvial deposits. The sediments consist largely of granitic detritus and weathering products from the Sierra Nevada. The aquifer in the study area is at least 400 feet thick and is unconfined in the shallow part of the system, meaning that it can freely receive water percolating down from land surface. With depth, however, the aquifer becomes partially isolated from sources of vertical recharge by discontinuous lenses of fine-grained silt and clay. The public-supply well selected for study is screened from 91 to 366 feet below land surface. This interval spans the zone typically tapped by public-supply wells in the region, although the majority of public-supply wells have a shorter screened interval and do not extend to the depth of the selected well. Groundwater flow near Modesto has been greatly altered by human activities. Most notably, large amounts of groundwater withdrawals and application of irrigation water have created a significant downward component of flow. These water-use practices also control local horizontal movement of groundwater. Irrigation has created a recharge rate of 23.6 to 27.6 inches per year in the agricultural areas northeast of Modesto. Groundwater flows from these agricultural areas towards the center of the city to the southwest, where large groundwater withdrawals paired with low recharge (estimated at 11.8 inches per year) have created a regional water-level depression (Burow and others, 2004).