Analysis of Slug Tests in Formations of High Hydraulic Conductivity by James J. Butler Jn, Elizabeth J. Garnett, and John M, Healey! Abstract A new procedure is presented for the analysis of sug tests performed in partially penetrating wells in formations of high hydraulic conductivity. ‘This approach is a simple, spreadsheet-based implementation of existing models that in be used for analysis of tests from confined or unconfined aquifers, Field examples of tests exhibiting oscillatory and nonoscillatory behavior are used to illustrate the procedure and to compare results with estimates obtained using alternative approaches. Tae procedure is considerably simpler than recently proposed methods for this hydrogeologic setting. Although the simplifications required by the approach can introduce error into lydraulic-conductivity esti ‘mates, this additional error becomes negligible when appropriate measures are taken in the field. These measures are summarized in a set of practical field guidelines for slug tests in highly permeable aquifers. Introduction Slug tess in formations of high hydraulic conductivity (K) ae often affected by mechanisms that are not consid cred in models for tests in less permeable formations (Hvorslev 1951; Cooper et al. 1967; Bouwer and Rice 1976), Although a number of specialized models have been developed for slug tests in tis hydrogeologic setting (Van der Kamp 1976; Kipp 1985; Springer and Gelhar 1991; McElwee and Zenner 1998; Zurbuchen etal. 2002) there is still no consensus regarding the analysis of tests in partially penetrating wells, the most common condition faced in practice. The result is that inappropriate methods are often used for the analysis of such tests, thereby introducing error into K estimates. In addition, published field guidelines for slug tests in highly permeable formations are incomplete, so ‘many tests are performed in a manner that, at best, signifi- cantly increases the complexity of the analysis process and, at worst, yields K estimates of low reliability. The purpose of this paper isto present new analysis and field procedures that should help resolve these problems. ‘The major focus of this paper will be on the presenta tion of a new procedure for the analysis of slug tests per- "Kansas Geological Survey, 1930 Constant Ave, Campus West, University of Kansas, Lawrence, KS 66047 Department of Gealagy, University of Kansas, Lawrence, KS e085, corresponding author: {785) 864-2116; butler@kasku.eds Received March 2001, accented September 2002. 620 Vol. 41, No, S~GROUND WATEF formed in partially penetrating wells in highly permeable formations, This procedure is considerably simpler than recently proposed alternatives, and can be used for tests in both confined and unconfined aquifers. However, the effec liveness of the method strongly depends on the use of appropriate measures in the field. An additional focus of this paper, therefore, is the definition of field guidelines for slug tests in highly permeable systems. If these guidelines are used in conjunction with the more general guidelines previously proposed by Butler et al. (1996) and Butler (1998), the analysis method presented here will yield hydraulie-conductivity estimates tha, forall practical pur poses, are indistinguishable from those obtained using con siderably more involved approaches “The procedure described in this paper isa spreadsheet implementation of a modification of the analysis approach proposed by Butler (1998) for slug tests in highly perme- able aquifers, This procedure is based on models previously described by Springer and Gethar (1991) and Butler (1998) for tests in unconfined and confined formations, respec- tively. These two models ate straightforward extensions of those of Bouwer and Rice (1976) and Hvorsiev (1951) to slug tests in highly permeable aguifers and are therefore designated in this article as the high-K Bouwer and Rice and high-K Hyorslev models, respectively, to emphasize their relationship to those earlier models. The spreadsheet ‘method described here was devised to make the proposed analysis procedure more accessible o the field practitioner, Although spreadsheet implementations of the Van der —September-October 2003 {pages 620-630)Kamp and Kipp methods for slug tests in fully penetrating ‘wells are available (Wylie and Magnuson 1995; Weight and ‘Wittman 1999), simple spreadsheet approaches for analysis of tests in partially penetrating wells have not been pre- sented previously. ‘This paper begins with a general overview ofthe analy sis procedure, followed by a description of the spreadsheet implementation of the approach. Two field examples are then presented to demonstrat the method for the analysis of ‘both oscillatory and nonoseilatory response data. Field measures required to ensure that the underlying assump- tions of the approach can be justified are then described, ‘The paper concludes with a brief summary of the major points, All of the Tiles needed (o implement the procedure are included in a Kansas Geological Survey (KGS) report (Butler and Garnett 2000) that can be readily obtained from the KGS Web site Grwukgs.ducedi) General Overview of Analysis Procedure Regardless of whether a slug test has been performed in an unconfined or confined aquifer, the general approach for the processing und analysis of data trom a test in a hhighly permeable formation is the same. This approach cos sists of the following four steps (see Figure I for a sehe- matic ofthe setup fora test performed in a highly permeable aquifer) wa RELEASE ~ ine By toseer} anion coumune ' dq ‘sora ree pion at ne eS lateral change) PRESSURE ‘TRANSDUCER aso Figure 1. Schematic of a well in which the pneumatic method Is being used for slug-testiitation (after MeLane et a. 1990; utler 1998; notation explained intext; not o scale) 1. Test data are processed into the form required for analysis. The processing bepins by plotting pressure transducer readings venus the time sinee some arbiteary starting point (Figure 2a). The time at which the test began and the pressure head corresponding to static conditions are estimated from this plot. The static pressure head and the fest start time are then subtracted from the head and time records, respectively, to obtain plot ofthe deviation of the pressure head from static conditions (H(0) versus the time since test inition (Figure 26). The deviation data are vided by the change in water level that initiated the test (Hip) 16 obtain the normalized head required for the analysis (Figure 2c), For tests performed using the poeuratie method (Figure 1), this inital water-level change is bes estimated using the air-pressure transducer 2. A graph of theoretical type curves is prepared. “These ype curves are in te form of plots othe normalized deviation of the water level from static conditions versus dimensionless time (Figure 3), The same set of type curves is used for analysis of tests in unconfined and confined ‘aquifers. The type curves are generated using the damped spring solution of classical physics (Kreyszig 1979): “costs + © sin(onas) |scye2 walt) 20, a walty) = EL + ty). Cy = 2 @) walla) where C, = dimensionless damping parameter g) = gravitational acceleration Hy = change in water level initiating a slug test (initial dis placement) 1, = effective length of water colurmn in well 1g. = dimensionless ime parameter (git)! w= deviation of water level from static Tevel in well normalized water-level deviation (fH) dimensionless trequeney parameter (=l-(Cy/2)") Cp D = 2 — op, = —P + wg 7 eh = ~F 3. ‘The Cy type curves are superimposed on a plot Of the test data. The dimensionless time axis is expanded or contracted until a reasonable match is obtained beoveen aacurve for a particular Cy, (Cjy*) and the test data (Figures 4a through 4c). Time match points are then determined by reading the corresponding values from the actual (¢*) and dimensionless (1,*) time axes (Figure 4c). Note that a near- ‘unique match can be obtained because the Cj, value con- trols the shape of the response (degree of oscillatory bbchavior or curvature), while the f, value controls the dura tion (period). Different users can obiain different manual curve matches, but these differences will be quite small due to the sensitivity of the match (0 variations in fy (change of JJ, Butler Jr etal. GROUND WATER 41,no. §:620-630 621Gems4S - Multilevel Slug Test #5 17.68 m Below TOC - 7/19/99 ' 2 © aad : Pressure Head at Koc al oe “me se) : Gems4S - Multilevel Slug Test #5 17.68 m Below TOC - 7/19/99 . (oo sesnd Tins ata 5 i 3 Tie (oo) 8 Gems4S - Multilevel Slug Test #5 10) 17.68 m Below TOC - 7/19/99 Eo (7 Normalized Bata ] : Time (sec) c "igure2. (A) Example plot of pressure head versus ime since some arbitrary starting point (in this case, 12:00:00 am); (B) ‘example plot of deviation of pressure head from statie condi- tons (H1(0) versus time since test initation; (C) example plot ‘of normalized head (H(2/H1,) versus time since test initation. 622 JJ. Butler Jr. et al. GROUND WATE! 41, no. 8: 620-630 ‘Type Curves for High-K Media Normal eee Dimensiontess Time dimensionless time type curves (C,, and dimensionless time defined in text). 18% between Figures 4b and 4e) and Cy (type curves for (Cy? and Cy##20% are shown in Figure 4c), 4, Hydraulic conductivity is estimated from the type- curve mateh. The radial hydraulic conductivity (K,) is esti- ‘mated by substituting values for the well-construction parameters, Cy*,and the time match-point ratio (4%) into the equation appropriate for (est conditions: Unconfined — High-K Bouwer and Rice Model (Springer and Gelhar 1991) ti inRe/r © 26C5 Confined — High-K Hvorslev Model (Butler 1998) _ Pinib/ (25) + (1+ YI) os) aa 26C5 where b= screen length; R, = effective radius parameter of Bouwer and Rice (1976) r= effective radius of well casing (corrected for radius of transducer cable) radius of well screen or borehole in isotropic iquifers, for anisotropic aquifers r, should be replaced by 1,VK,/K,, where K, is vertical Inydraulie conductivity (Zlotik 1994), Although the unconfined model (Equation 4) applies for wells screened either away from or up against an imper- ‘meable lower boundary, the confined model (Equation 5) does not apply fo wells in which the sereen abuts an imper- meable boundary. In that case, a modified version of the Hyorslev model should be used in which the 2r, terms inGems4S - Multilevel Slug Tests Analysis of Slug Test #5 17.68 m Below TOC - 7/19/99 Gems4S - Multileve! Slug Tests Analysis of Slug Test #5 17.68 m Below TOC - 7/19/99 Dimensoniess Time Dimensionless Time get le 34 E og) oo 3 i g E 5° Bos : roe | 7 ares | in © Time (800) “Time (see) a. Gems4S - Multilevel Slug Tests Analysis of Slug Test #5 17.68 m Below TOC - 7/19/99 Dimensionless Time 02 a2 ea os a7 ‘K ‘ ¢ i i E 2 co -096 | oe 1 NBimalaes Data aa : 19 1] oe c Time (800) Figure 4, (A) Superposition af type curve and normalized data plots (type eurve plots reference upper x-axis wi data plot references lower x-axis; every second data point plotted; type curves not relevant fora ‘and normalized data plots (maximum value for upper x-axis reduced f inal match between type curve and n value for upper x-aXis reduced from Figure 4b; starred (*) 4 lysis are hidden from view): m Figure 4a In an alized data plots (maximum values for x-axis used as tities are match parameters). the numerator of Equation $ are replaced by ry. Note that Botler (2002) has proposed a correction for slug tests in small diameter wells in highly permeable formations that can be readily incorporated into the proposed procedure, ‘The definition of dimensionless time following Equa- tion 3 includes a parameter (Z, the effective length of the water column) that arises from the derivation of the momentum balance for the wellbore (Kipp 1985). Zur- ‘buchen ot al. (2002) and others have analyzed slug-test data assuming that [, ean be computed from well-construction information and considered as a known quamtity for the analysis, That approach, however, does not generally pro- duce a good match between the type curve and the response data, indicating that the relevant physical processes are not J. Butler Ir etal, GROUND WATER 41, no, 5: 620-680 623‘completely described with existing definitions of L,. Give the complexity of slug-induced flow in a well in highly permeable aguifer, itis difficult to account forall contribu~ tions to the £, term. Thus, in the approach described here, L, is calculated as part of the analysis: ye (i)e © In most cases, the analysis-caleulated 1, value will be somewhat larger than the nominal value computed from the well-construction parameters (Kipp 1985). McElwee and Zenner (1998) discuss possible mechanisms that could be responsible for this difference. ‘The type curves of Figure 3 represent the theoretical deviation of the water level in the well from the static posi fon during a slug tet, The general upprouch outlined in the preceding paragraphs is based on the assumption that a pressure transducer in the water column will provide an accurate recont of the water-level position. However, MeEwee (2001) and Zurbuchen etal. (2002) have recently pointed out that a transducer may not provide an accurate recon of water-level position in conditions of high water- column aevelerations. The accuracy of the apparent water= level record obiained from a pressure transducer in an accelerating water column is a function of, among other things, the length ofthe water column above the transducer. ‘Zusbuchen et al (2002) proposed the following type-curve ‘correction to account for the position of the transducer in an accelerating water column L Hltadon = 4+ (Plan + Host) *(WidLe) where Htcaox= theoretical normalized head corrected for trans- ducer postion Playin = depth of pressure transducer below static “i second derivative of w, with respect to ime ‘This correction requires an estimate of L,, which, as dis- ‘cussed in the previous paragraph, is usually not known prior to the analysis, so error can be introduced into the correc tion process. Fortunately, the need for this correction can generally be avoided by placing the transducer close to the static level as discussed in a Tater section. Spreadsheet Implementation ‘The general procedure outlined in the previous section is implemented using two Excel 97 (Microsoft, Redmond, Washington) spreadsheets. Spreadsheet “Type Curve Gen- erator" is used to generate type curves with Equations 1 through 3, and spreadsheet “High K Estimator” is used to ‘process and analyze the slug-test data, Both spreadsheets are included in the file “High K Slug Tests.xs,” which ean be ‘obiained from Butler and Garnett (2000). The workings of cach ofthese spreadshects will be briefly described in this section. Note thatthe processing and analysis of individual tests can be readily completed within afew minutes, so no attempt has been made here to fully automate the procedure. 626 JJ. Butler Jr. et al. GROUND WATER 41, no. 5: 620-630 Figure Sais a view of a portion of spreadsheet “Type Curve Generator." The Cy value for which a type curve ito bbe generated is entered in cell B12. The theoretical normal- ized responses are then calculated in column H using the “appropriate equation (Equations | through 3, depending on C_, value). A plot of dimensionless time versus the theoret- ical normalized response (type curve) is automatically gen- crated as Chart | (Figure 3). Figures Sb and Se are views of spreadsheet “High K Estimator." The time and pressure records are entered in columns C and D and then automatically plotted as Chart 2 (Figure 2a), The start time forthe test andthe static level are determined from these records and entered in cells C12 and 9, respectively. The time since test initiation and the devi- ation feom static are then calculated in columns F and G. “Tho inital change in water level is entered in cell CLI and the normalized deviation from static (normalized head) is calculated in column J The time since test initiation and the normalized head are then auromaticaly plotted in Chart 3 along with the type curve generated in spreadsheet “Type Curve Generator” (Figure 4). The appropriate type curve (Cp is determined through iterative curve generation and comparison of theoretical and actual normalized head plots ‘The focus of the comparison ison the peak and trough val- ues for Cjy< 2, whereas the emphasis is on plot curvature for Cy> 2. Even when trough and peak values correspond, the type curve and response data plots will usually be offset (Figure 4a). To obtain a match, the dimensionless times are ‘adjusted using a modulation factor. This fetor is entered in cell N13 and the adjusted dimensionless times are caleu- Jated in column O. Char 3 is then automatically reploted using the new adjusted dimensionless time. The time adjustment continues in an iterative fashion until an aecept- able match is obtained (Figure 4e). The ratio of the time ‘match points (,°/*) is automatically computed in cell L6, and the K, estimate is then calculated using the values for this rao, Cp, and the well-construction parameters. The estimate obtained with the high-K Horslov model (Equ- tion 5) is given in various units in cells $911 and UIO, while that obtained with the high-K' Bouwer and Rice model (Equation 4) is given in cells $28~30 and U29. The effective length ofthe water columm i ealeulated in N8 and ‘compared 10 the nominal value computed from well con- struction parameters (Equation 6 of Kipp 1985) in cell NI. ‘The spreadshoct “High K Estimator” shown in Figures ‘5b and Seis for slug tests in wells in which the sereen does not abut an impermeable boundary. Ifthe well is sereened up to an impermeable boundary, the version of “High K Estimator" in the file “High K Slug Tests Boundary.xls” should be used. English-unit equivalents of "High K Slug ‘Tests.als” and “High K Shug Tests Boundary.xls” are also provided in Butler and Garnett (2000) Example Field Applications ‘This spreadsheet-based procedure is demonstrated using ficld data from a research site of the Kansas Geolog- ical Survey. To illustrate the range of conditions that can bee observed in tests in highly permeable aquifers, both oscillatory and nonoscillatry response data are considRe a a 7 Sproadshoots and Charis z prepared by z Garett ands Buller for | Bute, J Jr and. Garnet 2000, Simple procedure or analysis f slug tats informations of | _highnydraule conductivity using epreadehet and sclentitegreptos sofware, Kansas Geological © |_Surmy Open Rept 200040, Lawrence, Ks. 1 Type Curve Generator Spreadanaat | n q co ermso lad Dimensionless | Gp= avis 0918 oss -12004 Tine | 077s 3 0000 ct ses rT cs Gseee Ca case cs caete cs oe cy cere ce are as Gees + oss th bsreo 12 sire 13 base ta so80 1s sss ie ares ir obec View ofa portion of sprendsheet Type Curve Generator. A [8 [65 EF [6.7 Z K 7 |Rligh K Estimator Spreadshest i Specs -"a" not used In confined case 2 Metric Units Depth to Botton of Scrven (rom toc) Teal a [Screen Length (5p z25|m 4 Jbepn to Static Water Level (Fam toc: B274[m 5 |Site Location: GEMS, Hop of Seroen to Wator Table (): 3fim 6 [ates #27100 JRacus of Well Soreen (yh con 7 [rimo: Nominal Radius of Well Casing (ink ‘019, @ [Test Designation: — [a8425 055] Radius of Transducer Cable (a 0.008 3 [Static Level: ‘o45)mn Etfecive Casing Radius (= (r2r BPO). | a.0talm 10 [intial Water Level Modiied Screen Radius () ‘2.017 Ti] Change (Ha oem [Aspect Ratio (br) T3779] 72 [Slat Time for Test | S50525]s0e Formation Thicknoss 6] To.a7|m 7 18 5 Tine | Pressure 16 in Head Test | Daviation Tost | Novmalized 7 seconds | in meters Time _|from Static Time _| Head 7 esesz7 | _o3ea oz | 0.082 02 0.869 19 35082.0_| 0307 4 | 0.085, a 0.808 20 s5993.1_ | 0.308 06 | 0.048 06. 0.804 2 55033.3 | 0.01 Ds | 0.045: og 0.836 |] 2 55893.5 | 0.405 1 “041 1 0.763 23 556957 [0408 123.087 iz 0.695 24 5593.9 | 414 14 | 0.032 14 0.803 B 350341 | 0.418 is | 008 is. Osi7 28 55ea4.3 | 0.423 ta | 007 ie 0.425 27 55994,5 | 0.426, 2 0.018 2 0.335 28 559947 | 0.899 23 [ 0.013 22 0.248 23 350349 | 0437 240.000, 2a D164 30 55835.1 | 0.842 26 [0.004 26. (0.083 Figure Sb, View of a portion of spreadsheet High K Estimator. JJ. Butler Jr. et al. GROUND WATER 41, no, §:620-630 625E i B ae = T w v w 7 z Bost Fit (Confined High-K Hvorsiey Model 3 Tine — Itype Curve cameeselad | Correlation Ratio Co, Kes Ur h2 nfirar (re ie270.5) 5 nt ‘OTs ag BCs] 5 ‘760 I I I 7 [Bracketied quanti 13.786] 1 [oomputed Fomrato —_Le= Fee4in 2 [nominal te 13.449) Ke | 1.98609|misee To] aiteronce 26% 1.88E+02|miday | SS2E*02| day 1 1956.01 emisec 2 13 | Modulation Factor= Tasal (Gaconfined -High-K Souwer and Rice Medal 74 T To 16 Ke 2) 16] Dimensionless | _Co= ahusied Pa 7 Time 077s Time 78 ° 7 c (A ETSI a= i879] 19 0] TeesTay Tas B=) 0307 20 02 0.981087 02632 lrstterm TAI) 2 oS 0.95560 03047, T_o19i] 22 oa 0.528003) 0.5260 [second erm (AB Ceo No 25 5 ‘oa5Te14) ‘08579 T oz 24 08: 0.840386) 0.7805 ined) 5.788 25 oF, 80863] Tai [cannot excasds 25 08) ‘O.750175] 1.0526, [See Gute (1998) =p.105, 2 oa 0.695129} 18> 2a 4 0.637505) 13165 T= | 1608-03) nisee 23) ia 0.576049] Taare 7.39+02|miday | A5E902| 0) 301 12 0.51749 5780, “{80E-0t|emisee ‘Figure Se, View of a portion of spreadsheet High K Estimator. cred. Note that in both cases the hydraulie-conduectivity estimates agreed well with estimates obtained using the dipole flow probe (Zlotnik and Zurbuchen 1998) in the same (nonoscillatory example) oF adjacent (oscillatory example) wells Oscillatory Example. Figure 6a displays the normal- ized head versus time plot for three ofa series of slug tests performed in a direct-push installation screened in the con- fined coarse sand and gravel aquifer underlying the Kansas River Moodplain near Lawrence, Kansas (Butler et al 2002). The coincidence of the normalized plots indicates that dynamic-skin effects and any dependence on the mag- ‘nitude ofthe intial displacement can be neglected for these tests (Butler et al. 1996; Butler 1998). Figure 6b shows the type-curve fit determined using the spreadsheet provedure. In this case, a C,* value of 0.775 and time match points of 16.72 (1,8) and 22.0 (I) are obtained. Substituting these values and those for the well-construetion parameters (6 0.229 m, r,= 0.019 m, r, = 0.017 m) into Equation 5 yields #8 K, value of 168 m/day. To verify the appropriateness of this estimate, a joint analysis ofall three tests of Figure 6a ‘was performed with the McElwee and Zenner (1998) mode! coupled to an optimization routine. This analysis yielded K, estimate ~4% higher, a negligible difference given that a ‘visual match was used in the spreadsheet procedure. Note ‘that cests in partially penetrating wells in high-K aquifers are often analyzed with the Van der Kamp (1976) method {or tests in fully penetrating wells or the conventional forms of the Hvorslev (1951) and Bouwer and Rice (1976) meth- ods. In this case, the Van der Kamp method yields a very 626 J. Butler Jr. et al. GROUND WATER 41, no. 5: 620-630 significant overestimation of K, (factor of 3.9), as would be expected when a method developed for fully penetrating wells is used to analyze atest in a partially penetrating well (Butler 1998), Aa analysis with the conventional form of the Hvorslev or Bouwer and Rice method using the initial falling limb of the test data will also lead to a significant overestimation of K, (overestimation by a factor of 1.9 using the Horsley method). This degree of overestimation is commonly observed when oscillatory response data are analyzed with the conventional forms of the Horsley and Bouwer and Rice method Nonascillatory Example, Figure 7a displays the nor- ‘malized head versus time plot for three of a series of slug {ests performed ina packer-isolated interval ina monitoring well in the same aquifer as in the previous example. The nnear-coincidence of the normalized piots indicates that ddynamic-skin effects and any dependence on the magnitude of te initia displacement can again be neglected, Figure 7b shows the type-curve fit determined using the spreadsheet based procedure. In this case, a Cy value of 3.0 and time _match points of 17.39 (1,*) and 16.0 (8) are obtained. Sub- stituting these values and those for the well-construction parameters (b = 0,610 m, 7, = 0.025 m, 7, = 0.057 m) into Equation $ yields a K, value of 38.7 mda. A joint analysi of ull three tests of Figure Tu performed with the McElwee and Zenner (1998) model coupled to an optimization rou tine yielded a K, estimate ~1% lower. Analysis with the conventional form of the Hvorslev method yielded a K,esti= mate ~13% higher. Additional work has shown that the conventional forms of the Hvorslev and Bouwer and RiceBB426 - Test Series D3 4/27/00 rae asf las Baas Boa Bos i So i — i A rine 8) Analysis of Slug Test D3-6 at BB426 4/27/00 _ 4 ¢ ore : + Normalized Gata-H-odstm | uve 52-097 Normalized Head B Time (sec) iguce 6. (A) Normalized head versus time plo for three sug tests performed in direet-push equipment at the Geohydro- logic Experimental and Monitoring Site (GEMS) near Lawrence, Kansas (all tests Initiated with the pneumatic method; 72) determined from air-pressure transducer); (B) superposition of normalized data and best-fit type curve for test 6 (every second data point plotted; starred (°) quantities are match parameters; calculated L, (16.98 m) is approxi- rately 26% larger than nominal value). ‘methods will yield reasonable parameter estimates when C, >3 Recommended Field Guidelines ‘The procedure described in the preceding sections is ‘considerably simpler than recently proposed approaches for analysis of slug tests in highly permeable aquifers (MCE! wee and Zenner 1998; Zurbuchen et al. 2002). Although the simplifications required for this procedure can introduce Gems4S - Multilevel Slug Tests 14.63 m Below TOC - 7/20/99 Tes H 0189 Tost3 “50.285 m [os Test 6 -H0-138 m Normalized Head 00 o 2 4 6 6 © Rw MW Time (sec) A Gems4S - Multilevel Slug Tests Analysis of Slug Test #3 14.63 m Below TOC - 7/20/99 Dimensionless Time 900 498 70 r04 78 Nonnaized Bata F J Typo cure Gy2e88 Normalized Head Time (seo) B Figure 7. (A) Normalized head versus time plat for three slug tests performed in a monitoring well at GEMS (all tests init ‘ated with the pneumatic method; H, determined from pressure transducer); (B) superposition of normalized data and best-fit type curve for test 3 (every second data point plotted; starred (*) quantities are match parameters; caleu- ated L, (8.31 m) is ‘ jn 14% of nominal value) ertor into K, estimates, the results of the comparisons pre- sented in the previous section show that this error is negli gible when appropriate measures are taken in the field ‘These field measures can be encapsulated in the following, three guidelines for slug tests in formations of high hydraulic conductivity 1. Slug tests should be initiated very rapidly rela tive to the formation response, so that details of the ini- JJ. Butler Jr. et al. GROUND WATER 41, no. 5:620-630 627Dependence on Transducer Position GP43C - 4/4/00 - Hy<0.26 m = Pha 20381 | foam Phan ae Normalized Head 10 rs Time (sec) Figure 8. Normalized head versus time plot dependence on the depth (2,4) ofthe transducer below the statle water level (7, = 0.007 iy ry = 0.013 m, b = 0.156 ms all tests initiated with’ the pneumatic method; H, dependence sand dynamie-skin effects (Butler 1998) negligible: Hd ‘mined from air-pressure transducer; tests performed in dlirect-push pipe in the same aquifer as in Figures 6 an tiation process can be ignored in the analysis. The analy sis procedure described here is based on the assumption that the initiation process has a negligible impact on the response data. As recommended by Butler (1998) and Zur- buchen et al. (2002), the pneumatic method shoutd be used Tor test initiation whenever possible (ie in wells not screened across the water table) in order to avoid the error introduced into the K, estimate by shifts in phase and mag- nitude of the response data produced by slower initiation methods. Although Butler (1998) and others have sug- ‘gested using truncated data sets when test initiation is rela- lively slow, Zurbuchen et al, (2002) have recently demonstrated that this approach may often be inappropriate in highly permeable systems and can lead {0 an undespre- diction in hydraulic conductivity as large as 30%. 2. A.series of tests should be performed at each well using a range of initial displacements to demonstrate that any dependence on H, can be justifiably neglected in the analysis. Butler et al. (1996) and others have noted that response data can be dependent on the magnitude of the initial water-level change (H,). The analysis method described here is based on the assumption that any depen- dence on Hy is negligible. To ascertain the range of Hy for \which that assumption is valid, Butler et al. (1996) recom- ‘mend that a series of slug tests be performed in which Hy is changed in a systematic fashion between tests (¢.2., 0.06, 0.12, 0.18, 0.12, and 0.06 m). Ifthe plot of a series of tests demonstrates that dependence on His negligible fora cer~ tain range of Hy, then the method presented here can be used with confidence, When such conditions cannot be demonstrated, more involved approaches must be consid 628 JJ. Butler Jr et ak GROUND WATER 41, no. §: 620-630 Dependence on Transducer Position 48GP2900 - Lv2 - 2/10/00 Test 15 HO 2EM Pl OSE Test 20 Hpe0.143 m= Plyeen 827 Normalized Head Time (sec) A Dependence on Transducer Position 4SGP2900 - Lv2 - 2/10/00 Test 15-H,=O1ZEM-Plygg DBE | 695 Uncorrected coracted - Lg=14.96 m Normalized Head Time (se) B igure 9, (A) Normalized head versus time plot displaying dependence on the depth (pf) ofthe transducer below the static water level (r,=0.007 mr, = 0.013 m, = 0.076 my; both, method; Hy dependence -skin effets (Butler 1998) negligible; Hy deter- Jrspressure transducer; tests performed in ures 6 and 7); GB) superposition of normalized head plot and best-fit uncor- rected and corrected type curves (type curve corrected according to Equation 7). cred (eg., MeBlwee and Zenner 1998; MeEtwee 2001) Zorbuchen et al. (2002) have recently demonstrated that neglect of the dependence on Hand initiation of tests with relatively large displacements (7.5 m) can lead to an under- jon in K, approaching a factor of two, Note that the radius of influence ofa slug testis independent of the mag- nitude ofthe initial displacement, o there is little reason to use a large displacement if a reasonable signal-to-noise ratio ccan be obsained with smaller displacements (Figure 6a). estima3. The pressure transducer in the water column should be placed close tothe static water level, so that & type-curve correction for water-column acceleration is not necessary. As stated earlier, the record of pressure head versus time obtained from a transducer in a well ina highly permeable aquifer is a function of the depth of the trans- ducer below static. Figures 8 and 9a present the results of {0 series of slug tests in which the depth ofthe transducer ‘was systematically varied between tests to reveal a very pronounced dependence on transducer position. These fig- "ures indicate thatthe correction defined in Equation 7 is not necessary when the pressure transducer i close to static ‘This was verified by analyzing test 15 of Figure 9a and using the analysis-calculated 2, (14.96 m) to compute a type curve corrected for water-column acceleration. Figure 9b demonstrates that the correction is not of practical sig- nificance when the transducer is close to static, The defini- tion of “close” will be a function of the well-formation configuration, For example, “close” would be within 0.7 m of static for the configuration of Figure 8. Results ofa lange ‘number of tests indicate that placing the transducer within (0.5 mof static is sufficiently “close” for the vast majority of practical applications. In all eases, the need for the type- ‘curve correction can be readily assessed by considering the ‘maximum normalized head observed during atest. A maxi mum normalized head greater than 0.9 (Figures 6a, 7a, 8 [fay = O41 m), and 9a [pty yq = 0.936 m)) isan indi tion that the transducer is close enough to statie to make type-curve comection unnecessary. However, if the maxi mum normalized head is below 0.9 (Figures 8 [Pty > 0.341 mi] and 98 [pigq, = 4827 m), the tests should be repeated with the transducer eloxer tothe static level or the type curves must be corrected using Equation 7. If he trans- ducer is not repositioned othe type-curve correction is not applied, the underestimation in hydraulic conductivity can exceed factor of two when the transducer is at a consider- able distance below static (¢.., factor of 2.03 for test with transducer ata depth of 6.116 m in Figure 8). Note that the ‘maximum normalized head criterion of 0.9 proposed here is conservative threshold determined from field experiments (Figure 8) and simulations of hypothetical tests. When the ‘maximum normalized head isabove this threshold, the error Inroduced into the K, estimate through neglect ofthe type- curve correction will be much less than 10% ‘Summary and Conclusions A simple, spreadsheet-based procedure was presented forthe analysis of slug tests performed in partially penetrat- ing wells in formations of high hydraulic conductivity. In this procedure, theoretical type curves ate graphically ft (0 normalized plots of slug-test response data (0 oblain esti- mates of the hydraulic conductivity of the near-well por- tions of the formation. Field examples were used to illustrat the approach for the range of conditions common in highly permeable aquifers. Although the procedure is considerably simpler than previously proposed methoxl, example analyses demonstrated that the error introduced into K, estimates by the simplicity of the approach is negl- gible when appropriate measures are taken in the field. ‘These measures can be summarized in the following set of practical field guidelines. First, slg tests in highly perme- ble aquifers should be initiated very rapidly relative to the formation response to ensure that details of the test initia- tion process can be justifiably ignored in the analysis, Except in wells screened across the water table, the pneu- ‘matic method is the preferred approach for test initiation, Second, a series of tests should be performed at each well using a range of inital displacements (o demonstrate that any dependence on H, can be assumed negligible, Finally, the pressure transducer in the water column should be placed close tothe static level (< 0.5 m), so tha type curves do not have to be corrected for water-column avceleration. ‘When these guidelines are followed, the spreadsheet proce- dure presented here should yield K, estimates that are rea sonable representations ofthe hydraulic conductivity of the formation in the vicinity ofthe test interval. Acknowledgments “This research was supported in part by external funding provided by the Kansas Water Resources Institute under grant HQ96GRO2671 Modif. 008 (subaward S104) and bby Geoprobe Systems Inc. Field support provided by Steven P. Loheide II is gratefully acknowledged. ‘This work has benefited from discussions over the years with Wes MeCall, Carl MeEwee, Vitaly Zlotnik, and Brian Zurbuchen, References Bouwer, H., and R.C. Rice. 1976, A slug test for determining Inydraulic conductivity of unconfined aquifers with com- pletely or partially penetrating wells. Water Resources Research 12, 00.3: 433-428. Butler, JJ, Jt. 1998, The Design. Performance, and Analysis of ‘Slug Tests. Boca Raton, Flori: Lewis Publisher. Butler, 1. Jr. 2002. A simple correction for slug tess in small- diameter wells. Ground Warer 40, no, 3: 308-207, Butler, JJ. Je, and EJ. Gamett. 2000. Simple procedures for analysis of slug tests in formations of high hydraulic conductivity using spreadsheet and scientilie graphics sot ware. Kansas Geological Survey Open-File Rep. 2000-40 (available at wwi-kgs Audit Publication OFROO 40/ index-ham), Burler, JJ. t, C.D, McElwee, and WZ, Lix. 1996, Improving the reliability of parameter estimates obtained from s| texts. Ground Water 34, no, 3: 480-490. Butler, J. Jr .M. Healey, G.W. McCall, E.J. Garnett, and S.P, LLoheide I. 2002. Hydraulic tests with direet-push equip. ment. Ground Water'40, 90.1: 25-36, Cooper, HHL, 1D. Bredeboelt, and LS. Papadopules. 1967, Response ofa finite diameter well oan instantaneous charge of water. Wazer Resources Research 3, no, I: 263-269, Horsey. MJ. 1951. Time lag and soil permeability in ground ‘water observations. US. Army Corps of Engineers Water ways Exper. Sta, Bull no. 36 Kipp, KL. Je. 1985. Type curve analysis of inertial effets inthe response of a well toa slug test, Water Resources Research 21, no. 9: 1397-1408, Kreysrig, F. 1979. Advanced Engineering Mathematics, New "York: John Wiley and Sons McElwee. C.D. 2001. Application of a nonl Ground Water 39,00, §: 737-184 MeElvee, C.D, and M.A. Zenner. 1998, A nonlinear model for analysis of slug-test cata. Water Resources Research 34, no. 155-66. a slug test model ‘JJ. Butler Jr et al. GROUND WATER 41, n0. 5: 620-630 629‘MeLane, G.A., D.A. Harty, and K.O. Thomsen, 1990. A pneu matic method for conducting rising an falling head tests in highly permeable aquifers. In Proceedings of the 4th Annual NWWA Outdoor Action Conference, 1219-1231 Springer, RK. and LW. Gelhar. 1991. Characterization of large scale aquifer heterogeneity in glacial outwash by analysis of slug tests with oscillatory response, Cape Cod, Massach- scits, US, Geological Survey Water Resources Investigation Report 91-4034: 36-40, ‘Van dec Kamp, G, 1976, Determining aquifer transmissivity by means of well response tsts: The underdamped exse. Water Resources Research 12, 0. 1: 1-17 Weight, W.D.,and GP. Wittman, 1999, Oscillatory slug-test data sets: A comparison of two methods. Ground Water 37, no. 6: 827-835, Wylie, A. and S. Magnuson, 1995, Spreadsheet modeling of slug tests using the van der Kamp method, Ground Water 3, 20. 2: 326-329, Ziotnik, V.A. 1994. Interpretation of slug and packer tests in anisotropic aquifers. Ground Water 32, no. $: 761-766, Zlotnik, V.A., and B.R, Zurbuchen. 1998. Dipole probe: Design ‘and field applications of a single-borehole device for mes surements of vertical variations in hyéraulic conductivity Ground Water 36, no. 6; 84-893, Zurbuchen, B.R., V.A. Zlotnik, and JJ. Butler J, 2002. Dynamic interpretation of slug tests in highly permeable aquifers, Water Resources Research 38, 50.3: 10-10292001WROOD354, Learn the latest in water well constuction practices! ‘Manual of Water Well Con- struction Practices Edited by Stuart Smith Four years in the making, this book isa revision of document originally ‘writen under contract for the US. EPA, Comprehensive in its scope of current well construction methods, its designed to serve ‘not only contractors but as a guideline for state well inspectors. Referenées T871 NGWA member price: $25 Prospective member price: $31.25 Welland Borehole Seoling: Importance, ‘Materials, Methods, and Recommenda- tions for Decommissioning by Stuart Smith Wisconsin Water Well Association NOWA This illustrated book features discussions of the differences between well abandonment and well sealing, avoiding failures and faults in borehole grout seals, choices in sealing materials, site preparation, emplacement procedures and requirements, and more. ‘An appendix provides excellent ready-reference tables on {grouts and grout use. The 30 illustrations found through= ‘ut the book clearly explain the procedures. ‘and benefits of proper atiandonment and sealing, Referenced 1345 NGWA member price: $15 Prospective member price: $18 For additional information on these tiles, or any of the other titles offered through the NGWA Bookstore, call our Customer Service Department at 1-800-551-7379, or visit us on the web at www.NGWA.org. 630 J. Butler Jr. et al. GROUND WATER 41, no. 5: 620-630