In some areas of the world, treated or treatable water is being injected and stored beneath the earth’s surface in aquifers (i.e., aquifer storage programs) to preserve current water resources for future uses, plan for future droughts, retain instead of releasing surplus water during times of plenty, recharge wellfields and/or store water for use at a later point in time to sustain development [1] . As a result, most aquifer storage programs are found in arid areas (with consistent water deficits) or coastal areas (subjected to salt water intrusion) or areas with pronounced seasonal wet and dry periods (mismatched demands and supplies). The beneficial use of water that is not needed at the present is the main driver for the various aquifer storage programs. To date, all aquifer storage strategies are proven, comparatively cost-effective programs that are capable of successfully managing existing water resources to increase reliability for long-term water supply. But that does not mean all such programs are or will be successful to the client.

For the purpose of this discussion, AWWA Manual M21 on Groundwater [1] divides aquifer storage programs into four categories: Artificial Aquifer Creation, Aquifer Recharge, Aquifer Reclamation, and Aquifer Storage and Recovery (ASR). All of these approaches are used by the water supply industry to insure that reliable water resources are available for agricultural, environmental and urban uses. The focus of this paper is on ASR projects which are different from the other aquifer storage strategies because the associated injection and recovery wells are generally the same wells.

ASR has been a viable approach in the management of both potable and non-potable water supplies. In general, ASR technology can be used to increase the efficiency of water system operations because during the wet, low-demand portion of the year, some or all of the unused water treatment plant capacity can be used to treat water and store it in an aquifer for later retrieval and usage. Common applications are the injection of potable or raw water into an aquifer, with the intention to provide future withdrawal for augmentation of water supplies at a later time. The storage period can be over multiple months to allow the stored water to meet the next high demand season or an emergency (e.g., a severe drought or an extreme flood) or during an interruption of water withdrawal (e.g., due to equipment breakdown or water right dispute). Figure 1 shows the concept behind ASR graphically [2] .

A benefit of ASR technology in areas with significant fluctuations in raw water supplies and/or demands during the year is that it may allow the water utility to size its treatment plants for average conditions rather than seasonal high demands; thereby saving capital infrastructure costs. Once treated, potable water is often stored and recovered to deal with unmet demands. The use of potable water in ASR projects often allow the water utility to withdraw the injected water, disinfect it, and pump it directly to the distribution system without much further treatment, although this objective may not be met for various reasons. Other sources may require treatment after recovery like untreated raw water surface and ground sources, stormwater runoff and reclaimed water (Englewood, FL, Las Vegas Valley and several Arizona wells do this among others). All have been captured, injected, stored, and recovered.

In the United States, ASR wells are covered under U.S. Environmental Protection Agency (EPA)’s Underground Injection Control (UIC) program that was promulgated under the Safe Drinking Water Act (SDWA) Sections 1421 through 1426. U.S. EPA’s authority to govern underground injection programs is codified at 40 Code of Federal Regulation (CFR) 144 through 148. The purpose of the rules is to protect underground sources of drinking water (USDWs¹). The UIC program requirements were developed to ensure that emplacement of fluids via injection wells did not endanger current and future USDWs. Much of the UIC regulations were aimed at regulating disposal of waste fluids via underground injection, and focused on the injection of hazardous wastes (Bloetscher, et al., 2002). As a part of delegation of federal programs to the states, 40 states and territories, along with 2 tribes, now administer all or a portion of the UIC programs in their jurisdiction [1] . Under the auspices of protecting public health, the regulatory agencies have the authority to require, accept, and reject permits for underground injection wells on the basis of their successes or failures to protect USDWs.

Based on a survey conducted in 2013, there are 204 ASR sites in the United States [3] . Data were collected for each site via the survey and subsequent follow-up phones and email communication. As a part of this survey, the status of the various ASR systems was obtained and reviewed. Valuable information can be drawn from the data and analyses regarding lessons learned on the various ASR projects. Figure 2 shows the status of the ASR wells by site. Approximately 37% of the ASR sites are considered to be operational, while 24% are under various stages of testing. In addition, feasibility studies were conducted at 12% of the surveyed sites but these projects were not pursued further (as of the date of the survey); while 26% of the sites were associated with the cessation of ASR activities (see Figure 2). It is these inactive projects that are the focus of this paper and their

failures would help provide insight into the challenges associated with the ASR technology. For instance, inactive and discontinued project sites comprise 26% of the 204 sites, and a total of 220 wells (see Table 1), indicating that ASR projects are not always successful. While a lot can be learned from successful projects, reviewing the failed systems provides insights into some of the criteria and process associated with the development of ASR programs.

The project comprised two phases, data collection and data analysis. Data were collected in two steps. States and U.S. EPA were contacted by phone or email to track down the regulations that govern ASR programs from state

regulatory agencies and to determine whether the state had such programs in place or not. Note that there is a lack of ASR specific regulations at the federal level because the states are only required to submit inventory information on Class V wells and ensure non-endangerment of USDWs. While prior inventories were prepared by regulatory agencies and consultants, none were complete, comprehensive, and inclusive of all inactive projects. The regulatory investigations undertaken for this paper yielded a comprehensive list of ASR sites in the U.S., which was a critical component of the project. For the ASR projects included in the survey, variables of interest were used to account for operations, physical construction, and localized differences. These variables were extracted from the data sets. During the data extraction process, a number of wells were found to be older and/or not in service. In addition, data for some ASR sites were lost or missing. A confounding problem is that some states (e.g., California) have no centralized data sources, so there is a lack of consistency in the parameters collected.

The dataset collected for this project is the latest and most comprehensive representation of the ASR site inventory as of July 1, 2013. Data collected included:

・ State (where the ASR site is located);

・ Date the program was initiated or first well drilled;

・ Stage of development/status, categorized as study, testing, operational or abandoned;

・ Number of wells drilled: the number of individual ASR wells onsite to accommodate the designed injection capacity;

・ Number of abandoned wells, the number of individual ASR wells onsite no longer in service;

・ Source of water, categorized as ground, surface, reclaimed of industrial water;

・ Use of recovered water, categorized by irrigation, potable water supplies, raw water supplies, and surface water augmentation;

・ Number of storage cycles (estimated), which is indicative of age of the system;

・ Injection rate per well as a measure of injection capacity (converted to millions of gallons per day (MGD));

・ Withdrawal rate per well as a measure of withdrawal capacity (converted to MGD);

・ Peak flow as a measure of total available capacity on the site (converted to MGD);

・ Total water stored (converted to millions of gallons);

・ Depth of well casing below the surface: A measure of depth in feet of the most interior and deepest well casing that is installed at the final construction stage;

・ Depth of well below the surface: A measure of depth in feet of the deepest point of the well;

・ Casing material which is the final casing categorized as steel, PVC, fiberglass or stainless steel, separating stainless steel screens form stainless steel casings;

・ Casing diameter which is a measure of diameter in inches of the most interior and deepest well casing that is installed at the final construction stage;

・ Whether tubing and packer wells were used;

・ Injection zone formation category, categorized into formation type such as limestone, sand, sandstone, basalt, and alluvial formation;

・ Injection zone transmissivity or hydraulic conductivity (converted to transmissivity); as the measure the ability of water to maneuver through a porous media (the rate of flow per unit time per unit cross-sectional area, which is converted to gallons per day/ft);

・ Total dissolved solids (TDS) of the water in the injection formation fluid;

・ Type of confinement (formation type) categorized into formation type such as clay, dolomite, silt, shale, sandstone, basalt, or no confinement;

・ Number of monitoring wells if known; and

・ Operational issues reported. This is the area that provides much information on the inactive wells.

It should be noted that not all of the data listed above were available for all wells in the survey, especially the aquifer parameters associated with the older wells (e.g., TDS and hydraulic conductivity).

This study examines a set of 24 ASR data elements, including 15 continuous and 9 categorical variables. Continuous variables are estimated start date, number of active wells, number of abandoned wells, injection capacity, withdrawal capacity, in/out ratio, peak flow on site, amount of water stored, depth of well, depth of casing, injection horizon, diameter of casing, transmissivity, total dissolved solids (TDS), and number of monitoring wells. Descriptive statistics for the 15 continuous variables are summarized in Table 2(a). Categorical variables are geographic region, issues arisen for an ASR program, water source, water use, number of storage cycles (there were grouped because of the range of numbers), casing material, tubing and packer (T&P) code, injection formation code and confinement unit formation code. Descriptive statistics for the 9 categorical variables are summarized in Table 2(b). Geographic regions are summarized in Table 3. Table 4 shows the summary of logistic regression statistics predicting ASR success in the United States by region. Statistically significant coefficients are highlighted in bold and include issues or the lack thereof predominate by region with clogging, metals migration and recovery as the major issues (95% confidence).

A Chi-square test was performed to determine if there is a statistically difference between program status (i.e., active versus inactive) and each of the categorical variables (Table 5). A Chi-square test is a statistical test to determine if there is a statistically relationship between two categorical variables. When the association between two categorical variables is statistically significant, a contingency table of the residual values would reveal if these values make sense based on a priori knowledge of the two phenomena being compared (Table 5). Residual values are calculated by subtracting the observed value (O_ij) by the expected value (E_ij) in each cell of a

contingency table between the two categorical variables of interest, where the expected values are defined as:

where R_i is the row total corresponding to row i, C_i is the column total corresponding to column i and n is the total number of observations. The association between two categorical variables can be assessed using the Chi-Square statistic, which can be defined as:

In addition to Chi-Square tests, logistic regression was performed to identify explanatory variables of ASR success in the United States. Logistic regression is a probabilistic statistical technique that uses a logistic function to predict the outcome of a dichotomous variable based on one or more predictor variables. A logistic function can be defined as:

where t is a linear function of at least one or more explanatory variables. Equation (1) can also be rewritten as:

where is the intercept from the linear regression equation and is a regression coefficient for one of the explanatory variables. Equation (2) can be interpreted as the probability of equaling a value of 1 (rather than 0). One can also define the inverse of the logistic function, , which is also called the logit, log-odds or natural logarithm of the odds function:

The β coefficients in logistic regression are estimated using a maximum likelihood estimator.

We used success (1) or failure (0) of ASR sites in the United States, as the dichotomous dependent variable. A number of different logistic regression models were tested by varying the sets of continuous (Table 2(a)) and categorical (Table 2(b)) explanatory variables; the model sets were driven by the exploratory questions and missing data. Variables shown to be measuring similar phenomena (i.e., based on physical characteristics and correlation analysis) such as well depth and casing depth, were not used together in the tested models to avoid multi-co-linearity and other artifacts. Seven continuous explanatory variables were used in the logistic regression analyses: estimated start date, withdrawal capacity, in/out water volume ratio, depth of well, injection horizon, casing diameter and transmissivity. Six categorical variables were transformed into dummy variables and then used as explanatory variables in the logistic regression analyses: region, issues with ASR program, water source, casing material, injection formation code and confinement unit formation code.

The state with the largest numbers of ASR programs are Florida (54), followed by California, New Jersey, Arizona and Oregon (see Figure 3). However, the presence of ASR sites is not necessarily indicative for success of ASR projects. For example, in Florida, over half the sites are no longer active or have wells that are no longer used (see Figure 4). Other states that have large numbers of inactive sites are California and Arizona. With the elimination of inactive and test sites, there are only 22 active ASR sites (as compared to the total of 54 ASR sites) in Florida. Figure 5 shows the number of sites in the test phase by state. It should be noted that most of the inactive ASR sites have only one well, although a few outliers with many wells skews the average number of wells per site in Table 1. There also appears to be a slowing of ASR development after a marked rise in the 1990s and early 2000s (see Figure 6 [3] ).

With 95% confidence, chi-square tests results indicate that there are no statistical differences between ASR program status (active & inactive) and the following categorical variables: water source, water use, casing material, T&P code and confinement unit formation code. With 95% confidence, regression findings suggest that success rate is higher for ASR sites located in California, the Pacific Northwest, and Rocky Mountain/Plains.

Chi-square tests results indicate that there are statistically differences between ASR program status (active and inactive) and the following categorical variables: state, region, operational issues, storage cycles and injection formation code. Success rate is lower for ASR sites with deeper wells, clogging problems and water quality/ arsenic issues; clogging and water quality are operational issues. Based on these observations, the authors re-examined the data in order to discern the causes of operational issues with ASR systems, focusing primarily on the inactive wells, and those that have indicated issues with operations.

Table 6 outlines the information related to wells that are inactive. Table 7 outlines issues that have been raised with respect to active wells. Among the wells that are inactive, 10 wells were affected by water quality issues. Five of those are related to arsenic in Florida [4] [5] , and four are associated with arsenic, manganese, iron or a combination of metals (e.g., in Wisconsin [6] ). Two additional wells are related to radon in the west [7] [8] . The arsenic releases are all from limestone aquifers, and are associated with a range of injected water quality. Nevertheless, change in the geochemistry induced by injected water may lead to the release of arsenic from the geologic formations. Among the wells that are still active but not operational, a total of 6 wells in Florida are all releasing arsenic from limestone formations. One of these wells (i.e., Bradenton), is actively looking at deploying ion exchange as a means to deal with arsenic in the recovered water. Recently, U.S. EPA and the state of Florida have entered into a memorandum of understanding on permitting issues associated with arsenic in recovered water [9] .

Clogging is another ongoing problem at many sites. Eleven sites were identified as being abandoned as a result of clogging. Although 18 other sites are still active, they are associated with ongoing clogging issues. Among the inactive wells, 9 of them are injecting surface water into alluvial formations. The mechanism for clogging was not explicitly identified in these cases, but biofouling, particulates and geochemistry are all potential factors because of the organics in the water sources. Gin [10] [11] and Walker [12] reported that the clogging of the Phoenix wells (one using surface water and two injecting reclaimed water) got so bad that clearing the wells was required every three days, which led to the abandonment of the systems. Clogging is an issue at 18 active sites for which 13 of them are alluvial formations with a majority of them injecting surface water. Most of these sites are located in the west and they all have steel casings which may contribute to the potential for biofouling [13] .

Low recovery of injected water has derailed ASR programs at 17 sites. The majority of these sites are in Colorado or Florida. Boomgaard [14] indicated that recovery was low in the abandoned Colorado wells. In addition, it was easier and cheaper to obtain new water rights from west of the Continental divide than to keep the ASR program going.

As for the Florida sites that are no longer active, low recovery is mostly associated with sites that attempted to inject potable water into limestone formations containing brackish groundwater. Many are in southeast Florida where only the Boynton Beach wells have been successful to date [5] . Bloetscher and Muniz [15] suggest that the percent recovery would improve with lesser recovery from the well, but the clients were discouraged by the low recovery and hence abandoned the project. The CERP projects partially fall into this category of low recovery (although arsenic was a bigger issue). The Dalton, Georgia well was a private project that injected into granite [16] [17] . It is the only granite well and relies of fissure for success. Given north Georgia is known for its tight geologic formations and limited fissures, it is not surprising that the project was abandoned due to a lack of recovery.

Table 6 and Table 7 indicate that well problems leading to inactivity or abandonment can be broken into three groups: 1) physical and chemical clogging or biological fouling, 2) inability to adequately recover injected water, 3) water quality changes which include arsenic, metals mobilization and trihalomethanes [5] [6] [18] . Each of these issues are outlined in the following sections.

This paper results from a nationwide analysis of ASR systems through a data gathering effort as of July 1, 2013. Unlike the previous efforts, every utility that was identified as having investigated ASR was contacted, whether they proceeded with the development of ASR projects or not, and whether the system was still active or not. Nine months of data collection, literature reviews, online searches, phone calls, emails, file searches and permit collection yielded information on operations of the ASR systems, stratigraphy, ASR challenges and information were gathered and compiled for every system we knew of in the U.S. Copies of reports and rules have been posted at www.h2o-pe.com. One of the issues that arise from the inventory of the 204 systems in the U.S. is that 26% percent of the ASR projects have been functionally abandoned.

This paper focuses on the causes of failures associated with ASR projects and provides solutions to avoid some of the common pitfalls. While the wells at over 50 sites were technically abandoned, only one (Broward County) was reported as permanently plugged and abandoned, and three systems in California were converted to potable supply wells, which was not surprising because many of the California ASR wells started as supply wells. Geochemistry, clogging and recovery are the most common issues associated with failed efforts. Both physical and chemical properties of the injected water may change after injection as a function of chemical reactions between injected water and the native groundwater/aquifer matrix. The stored water has the potential to change as a result of increased aquifer pressure, temperature (increases or decreases), and the potential for micro-organisms activities. The changes in water and the formation are directly related to recovery potential.

The combined challenges of meeting requirements for drinking water storage, groundwater injection, water quality and disinfection byproducts rose in several states. The need to establish ownership of the stored water in several states has created additional challenges that must be addressed by the regulatory agencies within those states where ASR is used and water rights are at issue.