I.M. Breckels, * SPE, Koninklijke/Shell Exploratie en Produktie Laboratorium H. A. M. van Eekelen, SPE. Koninklijke/Shell Exploratie en Produktie LaborJtorium Summary Previous work on establishing a relationship between in- situ stress and depth is reviewed. On the basis of data from literature, supplemented by available field data, equations are proposed to describe the trend of horizontal stress with depth for nonnally pressured fonnations in the U.S. gulf coast region. A separate correlation be- tween horizontal stress and pore pressure allows this trend to be extended to fonnations containing abnonnal pore pressures. Having achieved a good correlation for data from this well-documented region, we repeat the process for other parts of the world, using data from hydraulic fracture treatments and fonnation integrity (leakoff) tests. Similar relationships between horizontal stress and depth are found to hold for nonnally pressured areas that are geologically and tectonically similar to the U. S. gulf coast. Coupled with regional correlations between horizontal stress and pore pressure, these relationships enable horizontal stress levels to be estimated for a given depth, provided the pore pressure also is known. Introduction The instantaneous shut-in pressure (lSIP) recorded dur- ing or after a fracturing job or during a mini fracture test provides a good approximation to the minimum principal in-situ total stress component S Hillin' The vertical total stress or overburden stress S v (nonnally the maximum principal in-situ stress) can be derived by integration of the fonnation density, compensated (FDC) log . In tec- tonically inactive areas, the intemlediate principal stress, S Hlllax' can be assumed approximately equal to the minimum principal stress: SHmin ==SHmax <sv. In the more general case, when S Hmax > S Hmin' the value of S Hillax in principle can be derived from the fonnation breakdown pressure Ph as measured at the start of a frac- . Now Wllh Shell Winni ng NV. Cairo. 01 19.2136182/0091.0336$00.25 Copyright 1982 Society of Petroleum Engineers 01 AIMt: SEPTEMBER 1982 turing job. For a fully plastering fluid, for instance, P /, = 3s Hmin - S Hmax - P c where Pc is the pore pressure. In practice, however, S Hmax cannot be detennined ac- curately because the breakdown value will be influenced by hole geometry, hole integrity, mud-cake properties, and the extent to which the fluid penetrates into and pressurizes the pore space around the borehole. Also, hydraulic fracturing data are often lacking in many areas of the world. The only remairung way, then, to obtain in- fomlation relating to in-situ stresses is to analyze the results of fonnation integrity tests (leakoff tests and casing-seat tests). In these tests, the pressure at which the fonnation starts taking fluid may range from S Hmin to 2s IImin - Pc. Therefore. it is the lower end of the range of values from fomlation-integrity tests that is of interest for in-situ stress detennination. Review of Previous Work The majority of hydraulic fracturing/fOlmation-integrity test data quoted in literature are from the U.S. gulf coast and Santa Barbara Channel (CA). From these data, a number of authors have derived fracture gradient correla- tions to be used in planning drilling programs. All correlation methods begin with the writing of the fracture gradient as follows. p/ D=Pc ID+k(sv -Pc)/D =PcID+kavID , .................. . ... (1) where P, fracturing pressure, P (' fomlation pore pressure, s v vertical or overburden stress, and a v vertical effective stress. The parameter k then is correlated with depth. D, over- hurden gradient. .\' vi D. vertical effective stress, a v' or Pd", :;it), (fer sh::llcs) The (:'.'rrc!atif'n mll<Jlly i ~ h ~ ~ e 2191 >- ... 0 0 e :z: >- a.. UJ 0 ...J co: 0 >- <r: UJ > 0 2 4 6 8 iO i2 14 . \\ 0 16 -. 1. MATTHEWS 8 KELLEY ---- 2. ADJUSTED M8KCURVE i8 00000000 3. PENNEBAKER --- 4. EATON 5. CHRISTMAN :1 20 0.3 0.4 0.5 0.6 0.7 0.8 0.9 i.O EFFECTIVE STRESS RATIO Fig. l-Effective stress ratio curves from varioLlS authors. on measured or estimated values of s v' plus measured fracture gradients or lost circulation data. The only ex- ception is Pennebaker, I who uses ISIP's to derive k(D) but still applies the result for fracture gradient prediction, thereby ignoring the difference between fracture initia- tion and propagation pressures (hoop stress effect). This difference also is ignored implicitly by other authors, who use fracture gradients to derive a correlation for k but interpret k as the horizontal to vertical effective stress ratio, whereas in fact it lies somewhere above this ratio. If P j is the pressure at which the formation starts fluid (Sllmin <Pj<2S llmin -Pc), the parameter kIn Eq. I equals the stress ratio aH/aV multiplied by a factor which, depending on the strength of the hoop stress ef- fect and on the difference between S IImin and 5 Hmax' may take on any value between one and two. Matthews and Kelly 2 assume that the overburden gra- dient s vlD equals 1.00 psilft. The parameter k is called the matrix stress coefficient and is correlated with ver- tical effective stress, a v: k =k(a v). For normally pressured formations, one then obtains k=k(D) In prin- ciple it is recognized that k(D) may differ for different areas and should be determined from fracture data for each specific area. Example correlations of k(D) are given for the Texas and Louisiana gulf coasts. Matthews and Kelly recognize that fracture extension pressures may be lower than Pr and therefore do not interpret k as horizontal to verticai effective stress ratio. Costley 3 also assumes s vlD = 1.00 psi/ft. He uses lost circulation data in normally pressured fonnations to determine k(D) for the Louisiana gul f coast. It is as- 2192 17500 6000 20000 1m) (II) Doplh 400 600 7!500 800 1000 10000 12500 teooo 17500 (pIn Iroclurino doto o _Ioninlogrlfy _I dolO o Fig. 2-Minimum horizontal stress vs. depth-U.S. gulf coast. sumed that the same correlation may be used in other areas by detemlining the depth in the gulf coast area with an equivalent porosity (by use of a porosity/depth cor- relation by Dickinson) and by use of the stress ratio k associated with that depth. Effectively, this means that parameter k is correlated with porosity, and this correla- tion is supposed to be universal for nomlally pressured shales. Pennebaker I uses actual values of s v' or correlations with depth s v(D), which vary with geological age. He correlates k with depth D (although he suggests that a better correlation may be with overburden gradient s vlD) and assumes the correlation to be the same for all areas. He identifies k with the stress ratio a lila v and derives the correlation k(D) from ISIP's, probably in the V.S. gulf coast area. Hence, his curve k(D) actually is a stress ratio correlation, and its application in Eq. 1 should lead to somewhat conservative estimates of frac- ture gradients since the hoop stress effect is neglected. Eaton 4 correlates both Sv and k with depth D, the cor- relations being different for different areas. The correla- tion s v(D) is obtained from integrated density logs, and k(D) then is back-calculated from fracture gradient data. An example is given for the V.S. gulf coast area on the basis of Costley's lost circulation data and density logs from many V.S. gulf coast wells. Eaton neglects the dif- ference between fracture initiation and extension and in- terprets k as an effective stress ratio. In fact, he writes k = v/ ( I - v) and plots h is results as a correlation of v with depth. This is an unnecessary and somewhat dangerous complication because it might create the wrong impres- JOURNAL OF PETROLEUM TECHNOLOGY sion that k also may be detennined by measuring Poisson's ratio J! on a core. Christman 5 also correlates both s v and k with depth and presents correlations for the Santa Barbara Channel (only down to 5,000 ft below the mudline). He also in- terprets k as an effective stress ratio. As an alternative. Christman presents another correlation for k in the Santa Barbara Channel, with the (local) bulk density of the formation. Pilkington 6 compares the correlations k(D) by Matthews and Kelly, Pennebaker. Eaton. and Christman; the curve by Matthews and Kelly is adjusted for the effect of variable overburden density by applying Eaton's sv(D) correlation for the U.S. gulf coast. The spread in the curves turns out to be small (Fig. I). Pilk- ington then derives an average curve (not plotted), and, by referring to the s v(D) correlations. finally expresses the parameter k in overburden gradient s vi D, as follows. k=3.9 svID-2.88 for and k=3.2 svID-2.224 for svID>0.94, where s v is in pounds per square inch and D is in feet. With this expression for k, Eq. I is assumed valid for both normally and abnormally pressured formations in tectonically relaxed tertiary basins containing plastic shales with interbedded sands, but not for brittle or naturally fractured fonnations, including limestones , dolomites, and brittle shales. Pilkington also interprets k as the horizontal/vertical effective stress ratio. It is somewhat surprising that Pennebaker's correlation (based on ISIP's) agrees as closely as it does with the correlations forkeD) by Matthews and Kelly. Eaton. and Christman (all based on fracture gradients). From this agreement it would seem that the hoop stress effect in fracture gradient data is very small and that the stress ratio a Hl a v and the parameter k in Eq. I scatter around almost identical trends with depth. This result is not sup- ported by the data presented in this paper. Horizontal Stress/Depth Relationships From the previous discussion it can be seen that the ma- jority of previous work has been directed toward establishing values for the fracture gradient by correla- tion of the parameter kin Eg. I with depth. This correla- tion then can be used in conjunction with integrated FDC logs to estimate future fracturing pressures. In this paper basic data are considered from tests car- ried out to determine fracturing pressures or ISIP's, to derive a relationship between minimum horizontal stress S Hmin and depth. No attempt has been made to relale SlImin to the vertical stress s v to obtain a correlation be- tween s Hmin Is v and depth. U.S. Gulf Coast Region The number of data available for the U.S. gulf coast is larger than that for any other single region. For this reason, all available data from hydraulic fracturing or fonnation integrity tests for this area (including data from Refs. 2 and 6 through 14) were considered first: data of a suspect nature (e.g .. where fracturing or fluid lcakoff depth was not known) were ignored, while data (mm kllu", fI Lv be also ...,;ere SEPTEMBER 1982 temporarily removed from the data set. The remaining data points (more than 300) have been plotted against depth in Fig. 2; data from hydraulic fracturing tests have been plotted with open symbols to distinguish them from leakoff test data. As already discussed, these latter tests can give a range of values between 2s Hmin - Pc and S flmin' where s IImin is the minimum in-situ horizontal stress. Because of the difficulty associated with assessing which value applies (which type of failure occurs) in any particular test, a lower-bound curve to the majority of data points has been used to obtain an approximate trend for s flmin as a function of depth . The solid curve in Fig. 2 forms a lower bound to 93 % of the data points. It can be defined by two simple functions: S Hmin =0.197 D 1 145 for D 11,500 ft , and S Hmin = 1.167 D-4,596 for D> 11,500 ft, where S Hmin is in pounds per square inch and D is in feet. This lower-bound curve may be compared with Pilk- ington's average curve (see Fig. 3), which can be inter- preted in tern1S of S Hmin and D as a curve with the equations SHmin =0.24IDI 135 for ft, and SHmin = 1.06ID-I, 986 for D> 7,550 ft. It can be seen from Fig. 3 that Pilkington's curve fits the data very well. However, it is likely that the pro- posed lower-bound curve represents a better correlation of horizontal stress vs. depth for nonnally pressured for- mations than does Pilkington's curve. There are two reasons for this. I. The majority of data in Fig. 3 are from formation integrity tests. These tests frequently show some in- fluence of the hoop stress around the borehole and therefore lead to an overestimation of the minimum horizontal stress. In other words, the -small symbols in Fig. 3 generally represent values somewhat in excess of sHmin 2. The data set of Fig. 3 predominantly includes points from nonnally pressured or overpressured fonnations, known underpressures having been eliminated from the data set. Consequently. underpressured environments are underrepresented in this data set. U nderpressured fonnations generally have lower horizontal total stress than normally pressured or overpressured formations. To investigate further the effect of under- or over- pressures and the hoop stress effect. hydraulic fracturing data for S Hmin (lSIP's or fracture-propagation pressures) for which pore pressure data are available were col- lected. For the U.S. gulf coast, 77 data points were ob- tained, including those relating to underpressured forma- tions that have been omitted from Fig. 2. For these data points, Fig. 4 plots (s Hmin -SH,)/D (the measured horizontal stress gradient minus the trendline value of Fig. 2) vs. (Pc -Pcn)/D (actual minus nonnal pore pressure gradient, nonnal pore pressure for this area be- ing defined as 0.465 psi / ft). The straight line in Fig. 4 was obtained by linear regression, with a c0rrelation C'oeffiC'ient r=O. 79. The 2193 17500 6000 20000 (m) (It) Oopth zoo 400 600 7500 800 1000 "_(barl ZOO '0000 .Z5OO 15000 .7500 ~ Hydraulic frocturi"9 data o Formation-intqity t.,t doto interpretation of Pilki"9ton'. ov,rove Itr ... rotio lint (pli) Fig. 3-Comparison between proposed lowerbound trendline and average curve interpreted from Pilkington,6 on a plot showing U.S. gulf coast data. o . o , o ~ Fig. 4-Effect of abnormal pore pressures on horizontal total stress (U.S. hydraulic fracturing data) . o slope of the line is 0.46, which agrees well with the value of 0.50 obtained by Salz 12 for the Vicksburg for mation in south Texas . The intercept value of the line is almost zero. Since the hydraulic fracturing data of Fig. 4 do not contain a hoop stress effect, this nearzero in- tercept leads us to conclude that the lower-bound curve in Fig. 2 constitutes a good average trendline for the horizontal stress SHlllin in normally pressured formations. The fact that most data points in Fig. 2 fall above and to the right of the trendlinc results from the hoop stress ef feet (in formation integrity tests) and the existence of overpressures. The horizontal stress in abnormally pressured formations in the U.S. gulf coast region can be 2194 600 2 7500 10000 12500 "_(bor) 800 10000 (poi) 0 ~ ...J! 4000 o 0 Jl (m) (11) Ooplh Fig. 5-Minimum horizontal total stress obtained from hydraulic fracturing data from Venezuela. -..... o bar/," 0 .05 -0.05 'I 0.56 0.011 c ~ _ . _I", Fig. 6-Effect of pore pressure changes on horizontal total stress for data from Block 1, lake Maracaibo, Venezuela. estimated by combining the correlations of Figs. 2 and 4 into one single correlation: S Hillin =0.197D 1.145 + 0.46(P(, -PCI/) for D ~ 11 ,500 ft. and S Hillin = 1.167D-4,596+0.46(Pc -P ('1/) for D> 11,500 ft , where P (' /1 is a pressure corresponding to a gradient value of 0,465 psilft. Note. however, that for depths greater JOURNAL OF PETROLEUM TECHNOLOGY IHlftln (bar) o 200 400 600 800 __ __ 1000 2000 3000 (m) \ \ \ \ \ \ \ \ \ \ 5000 7500 10.000 (11) Oep.h 5000 10.000 (p,ol TREATMENT VOLUME \ ,,500 bbl, II TREATMENT VOLUME > bbl. Fig. 7-Hydraulic fracturing data from Venezuela converted to normal pore pressure. than 11,500 ft this correlation is not supported by hydraulic fracturing data. Venezuela Hydraulic fracturing data from Venezuela are plentiful, most information being available for Lake Maracaibo Block I. Nearly all these data were obtained from stimulation treatments carried out in the C4 and C5 sands of Eocene formation . They therefore cover only a relatively small depth range. . The lSIP values are plotted against depth in Fig. 5. Also indicated in this figure is the curve established previously for normally pressured formations in the U.S. gulf coast region. The Venezuelan data points can be seen to scatter widely around this line and appear to form a trend predominantly to the left of the gulf coast curve. This tendency to lower horizontal stress levels is ex- plained partly by the low pore pressure in these forma- tions, many of these wells having been stimulated after pressure depletion during production. The large scatter is caused partly by pore pressure differences, but two other factors also may playa role. First, more than one area is represented in Fig. 5, and there is evidence to suggest that the tectonic environment varies from one lake block to another and even within subregions of Block I. Sec- ond , the stimulation treatments quoted vary widely in volume from true mini fractures to treatments in excess of 4,000 bbl. This variation is bound to produce some scat- ter in the data, 15 and, in addition, large treatments can produce problems in correctly interpreting the value of the ISIP. 16 The influence of pore pressure variations can be taken into account by again plotting the difference between the measured and (U .S. gulf coast) trendline values of the horizontal stress gradient, S Hmin -s II ID. VS. actual r SEPTEMBER 1982 800 10000 (poll .. SIIoIo (ml (III Depth Fig. 8-Minimum horizontal total stress obtained from hydraulic fracturing data from Brunei. minus normal pore pressure gradient, (P c - P en )/ D, with the original Eocene pore pressure of 0.433 psi/ft be- ing taken as the normal pore pressure. Plotting all available data from Lake Maracaibo in this way reveals no correlation between the two variables . This is prob- ably because of location and fracturing treatment volume mentioned previously. Excluding data from treatments larger than 1,000 bbl, we attempted a correlation for data from the C4 and C5 Eocene sands of Block I. Fig. 6 shows a linear regres- sion based on 45 data points . The straight line has a cor- relation coefficient of 0.7, a gradient of 0.56, and an in- tercept of 0.05 psi/ft on the stress axis. This intercept in- dicates that normally pressured formations in this area are at a slightly higher stress level than U.S. gulf coast sediments (the intercept value corresponds to an extra 300 psi at a depth of 6,000 ft). Fig. 7 shows the original U.S. gulf coast trendline of horizontal total stress against depth (Curve 2) and the same curve repositioned for the above intercept (Curve I). It also plots the hydraulic fracturing data from Block I (same data as above), after adjustment of each point to normal pore pressure, with the straight-line relationship of Fig. 6 (see Appendix). The adjusted relationship ap- pears to hold reasonably well for the data presented; in- corporating the pore pressure adjustment for this area, we obtain the equation for Curve I (Fig. 7): S Hmin = 0.2 IOD 1.145 +0.56(Pe - P CII)' where S Hmin' Pc, and P en are in pounds per square inch and D is in feet. Normal pore pressure corresponds to 0.433 psi /ft. Note that the data do not support extrapolation of this trend beyond the Eocene formation of Block I . 2195 .t. SANOST(M.S SHAL.ES -0.0% ,.04"_ 0025 "-0.1t9} OOZ 0.04 00. 0.08 ~ c - .. 'N,/M) -.- Fig. 9-Effect of abnormal pore pressures on horizontal total stress for data from Brunei. Brunei An approach similar to that used for the U.S. gulf coast and Venezuela was adopted for data from Brunei. In this case data were far less plentiful, and a correlation had to be attempted from only IS hydraulic fracturing data points. On the positive side, however, these data were all from mini fracture tests carried out specifically to deter- mine stress levels, rather than from large-scale stimula- tion treatments. The small volumes of fracturing fluid in- volved increased the reliability of the interpretation of the ISIP's derived from each test. Fig. 8 shows the ISIP values plotted against depth, again with the U.S. gulf coast curve drawn in for com- parison. The rock type in which each test was carried out is indicated also. It is clear that the points from this region lie predominantly to the right of the U.S. gulf coast curve, but data are too few to discern any mean- ingful difference between rock types. If the data are plotted again on a graph of (SHmin-SH )/D vs. (Pc-Pcn)/D, the influence of ab- normal p o r ~ pressures can be analyzed. * The straight line in Fig. 9, which has a correlation coefficient of 0.9, has a gradient of 0.49 and an intercept on the vertical axis of 0.11 psi/ft. This intercept value indicates that normally pressured formations offshore Brunei are (on the evidence of this small data set) subject to higher com- pressive stresses than in the U.S. gulf coast region. Ad- justing the U.S. gulf coast trendline (Curve 2 in Fig. 10) to take this intercept into account, we obtain Curve I in Fig. 10. The same figure also shows the IS hydraulic fracturing data and some formation-integrity test data (from other Brunei wells), both after adjustment to nor- mal pore pressure by use of the slope 0.49 of the straight- line relationship of Fig. 9 (see Appendix). If we again incorporate the pore pressure adjustment, the full rela- tionship for Brunei (Curve I, Fig. 10) becomes S Hmin =0.227D 1.145 +0.49 (Pc -P en). The North Sea More than 200 formation integrity (leakoff) test values are available for the North Sea, if data from offshore The Netherlands, the U.K. , and Norway are combined (Fig. 11); known or suspected overpressures among these data are very few. P en is the normal pore pressure corresponding to a gradient value of 0.433 psi/ft . 2196 200 400 "_Ibar) 600 900 5000 iOOOO {poi} iOOO 2000 7!lOO ~ 10.000 (m) (II) Fig. 10-Data from Brunei converted to normal pore pressure. From this large data set, a best-fit leakoff curve (Curve I in Fig. II) emerges that is very similar to the U.S. gulf coast curve (Curve 2) but lies between 21 % (at 2,000 ft) and 8 % (at 10,000 ft) to the right of it. Of course, it should be expected that best-fit curves derived from for- mation integrity (Ieakoff) test data from tectonically relaxed environments would fall some distance to the right of the U.S. gulf coast curve, which was derived from hydraulic fracturing data. This results mainly from the hoop stress effect in leakoff tests. While no direct comparison can be made between results from these two test types, similarities of trend with a migration toward higher stresses for the leakoff test values are indicative of similar stress environments. A difference in apparent stress level because of rock type is shown in Fig. 12, where the best-fit line for shales is seen to fall some 0 to 5% (depending on depth) to the right of the best-fit line for sandstones. At 8,000 ft this difference amounts to 280 psi. However, it is doubt- ful whether this difference of up to 5 % is statistically significant. The Netherlands, Onshore More than 100 leakoff test data are available for The Netherlands onshore fields. In addition, six hydraulic fracturing treatments have been carried out. Because of this small number of hydraulic fracturing data, a separate correlation has not been attempted, and the data are shown together with the leakoff test data in Fig. 13. The scatter in the data in Fig. 13 is somewhat larger than that seen previously. This can be attributed partly to pore pressure effects, particularly in the marine Lower Cretaceous, which was partly depleted by production before the drilling of the majority of wells from which data were taken for this study. Abnormal pore pressures also are known to occur in some of the floating carbonate JOURNAL OF PETROLEUM TECHNOLOGY 'Hrnfn (bar) -NORWAY Fig. 11-Leakoff test data from the North Sea. blocks within the Zechstein fonnation, and some 18 data points for carbonates refer to tests carried out in these blocks. As seen in Fig. 13, data from the Carboniferous give very low apparent stress levels. These cannot be ex- plained in tenns of pore pressure effects and indicate that an extensional regime exists in this fom1ation. All data from the Carboniferous are from wells that penetrate this fonnation immediately below the Zechstein, and it would appear that the low leakoff test values seen in these tests could result from stress reductions induced by salt movements in the overlying Zechstein. To some ex- tcnt, the same comment applies to data from the Siochteren fonnation, which also directly underlies the Zechstein salt. The deviation from the general Netherlands sandstone trend is, however, far less pro- nounced for the Slochteren than for the Carboniferous. The difference in trend for the different rock types is indicated by Curves I through 5 in Fig. 13. The U.S. gulf coast curve is shown for comparison (Curve 6). As for the North Sea, the Netherlands leakoff curves should be expected to lie somewhat to the right of the U.S. gulf coast curve. Three curves are shown for sandstone data: Curve 5 is a best fit to the Carboniferous points only, while Curves 3 and 4 exclude the Carboniferous and also the marine Lower Cretaceous. Curve 3 also excludes data from the Siochteren fonnation . Finally, Curve 2 is for the carbonates. and Curve I is for shales. It is seen that. at a depth of 8,000 ft, a difference of 400 to 1.000 psi exists between apparent stress levels in shales and sandstones. This difference is significant and may repre- sent a tme difference in horizontal total stress between nonnally pressured sandstone and shales. This may have impl ications for the problem of containment of hydraul ic . '" tractures .. , SEPTEMBER 1982 z 5000 7500 10000 (m) (II) Depth _(liar) 200 400 600 800 5000 10000 (pli) SA/()STONES CARBONATES @ SHALES a CLAYSTONES Fig. 12-Leakoff test data from the North Sea indicating rock type. Relationship Between Leakoff Test and ISIP Values The data from Bmnei provide the only opportunity to compare leakoff with ISIP values obtained during the same test. On the average , the leakoff values exceed the ISIP's by II % (with a standard deviation of 9%) for the 14 wells for which both values are known. Of course, it is somewhat doubtful whether a relation- ship based on so few points from one location may be ex- trapolated to other areas in the world. However, it is noteworthy that Pilkington's curve,6 which is an average trendline compiled from the work of several authors and derived mainly from leakoff tests, also falls between II and 15% above the previously mentioned U.S. gulf coast curve based on ISIP's in the same depth range. If the relationship between ISIP and leakoff values detennined from the data for Brunei is assumed to apply generally. one might expect to find a best-fit leakoff curve that falls about II % to the right of the U,S. gulf coast (ISIP) curve-i .e., leakoff pressure =0.219D 1.145 for D 11,500 ft. In Fig. 14 this relation (Curve 2) is compared with Pilk- ington's correlation (Curve 3) and with the best-fit curves for leakoff test data from The Netherlands for all fonnations except the Carboniferous (Curve 4) and from the North Sea (Curve 5): leakoff pressure=0.332D 1.104 (Netherlands). leakoff pressure = 0. 353D 1.091 (North Sea). The agreement between Curves 2 through 5 in Fig. 14 is 2197 2 3 lQooo (m) (tI) Deplh 400 Fig. 13-Leakoff test data from onshore Netherlands indicating rock type. sufficiently close that we can be fairly confident in using the U. S. gulf coast curve (Curve 1) to describe the trend of horizontal stress with depth in tectonically relaxed areas such as the North Sea and The Netherlands (ex- cluding the Carboniferous) for normally pressured formations. Conclusions Hydraulic fracturing data from the U.S. gulf coast, Venezuela, and Brunei have been analyzed to give rela- tionships between horizontal total stress and depth for these regions. If we take into account the effect of abnor- mal pore pressures, these relationships become as follows. u. S. Gulf Coast. s Hmin =0.197D I . )45 +0.46(P c -P en) for D< 11,500 ft. S Hmin = 1.167 -4,596+0.46(Pc - Pen) for D> 11,500 ft. Venezuela. s Hmin =0.2IOD 1.145 +0.56(P c -Pc,,) for 5,900 ft<D<9,200 ft. Brunei. sHmin =O.227DI 145 +0.49(pc -P cn ) for D< 10,000 ft. It should be emphasized that, while the U.S. gulf coast curve is probably fairly reliable over the depth range of 0 2198 2 3 tOOOO (ml (fl) Depth 200 5000 400 SHmin(bor) BOO 10000 (psi) I Gulf Coast ('SIp) Curve 2 Gulf Coast curve + 11 % 3 PilkIngton's correlation 4 Netherlands onshore 5 North Sea Fig. 14-Comparison of trend lines for U.S. gulf coast, the North Sea, and onshore Netherlands. to 11,500 ft and the Brunei curve is reliable for 0 to 10,000 ft. the Venezuelan data cover only sandstones within the Eocene formation of Lake Maracaibo Block I . Extrapolation of the above relationship beyond this for- mation cannot be supported by the available data. Extrapolating a relationship between leakoff values and ISIP's determined from data for Brunei, we modified the U.S. gulf coast ((SIP) curve to enable comparison with formation-integrity (Ieakoff) test data from the U.S. gulf coast (Pilkington's curve), the North Sea, and The Netherlands. Good agreement between the four eurves was achieved, indicating a close similarity between the horizontal total stress trends with depth for these three areas. This implies that the U.S. gulf coast curve of Fig. 2 can be used with a fair degree of confidence to describe the horizontal total stress as a function of depth in nor- mally pressured formations in tectonically relaxed areas such as the North Sea and The Netherlands (excluding Carboniferous). Nomenclature D = depth, ft (m) k = parameter in Eq. P b = formation breakdown (fracture initiation) pressure, psi (MPa) Pc = pore pressure, psi (MPa) Pen = normal pore pressure, psi (MPa) PI = fracturing pressure, psi (MPa) PIp = fracture-propagation pressure, psi (MPa) s Hmin = minimum horizontal total stress, psi (MPa) s Hmax = maximum horizontal total stress, psi (MPa) s H, = trend line value of minimum horizontal total stress , psi (MPa) s v vertical total stress (overburden stress), psi (MPa) JOURNAL OF PETROLEUM TECHNOLOGY (J H minimum horizontal effective stress, psi (MPa) (J v vertical effective stress, psi (MPa) 1> porosity, % References I. Pennebaker, E.S.: "An Engineering Interpretation of Seismic Data," paper SPE 2165 presented at the SPE 1968 Annual Meeting, Houston, Sept. 29-0ct. 2. 2. Matthews, W.R. and Kelly. J.: "How to Predict Fonnation Pressure and Fracture Gradient," Oil and Gas 1. (Feb. 20, 1967) 92- 106. 3. Costley. R.D.: "Hazards and Costs Cut by Planned Drilling Pro- grams." World Oil (Oct. 1967) 93-96. 4. Eaton. B.A.: "Fracture Gradient Prediction and Its Application in Oilfield Operations," 1. Pet. Tech. (Oct. 1969) 1353-60; Trans .. AIME. 246. 5. Christman. S.A.: "Offshore Fracture Gradients." 1. Pet . Tech (Aug. 1973) 910-14. 6. Pilkington. P.E.: "Fracture Gradient Estimates in Tertiary Basins," Pet. Eng. Iml. (May 1978) 138-48. 7. Taylor, D.B. and Smith, T.K.: "Improved Fracture Gradient Estimates in Offshore Drilling Operations , " Proc., API. Houston, TX (March 1970) paper 926-15-L. 8. Matthews, W.R. and Cesmirosky. LJ .: "Programmed Casing Seats Can Lower Well Costs." Oil and Gas 1. (Jan. 24. (972) 60- 64. 9. MacPherson. L.A. and Berry. L.N.: "Prediction of Fracture Gra- dients From Log Derived Elastic Moduli," Log Analyst (Sept.lOct. 1972) 12-19. 10. Anderson. R.A., Ingram. D.S .. and Zanier, A.M.: "Fracture Pressure Gradients From Well Logs ," 1. Pel. Tech. (Nov. 1973) 1259- 68. II. Althaus, V.E.: "A New Model for Fracture Gradient." 1. Cdn. Pel. Tech. (April/June 1977) 98-108. 12. Salz. L.B.: "Relationship Between Fracture Propagation Pressure and Pore Pressure, " paper SPE 6870 presented at the SPE 1977 Annual Technical Conference and Exhibition, Denver, Oct. 9-12. 13. Kehle. R.O.: "The Detcnnination of Tectonic Stresses Through Analysis of Hydraulic Well Fracturing." 1. Geophys. Res. (1964) 69. No.2. 259-73. 14. Felsenthal. M. and Ferrell, H.H.: "Fracturing Gradients in Waterlloods of Low-Permeability. Partially Depleted Zones," 1. Pel. Tech. (June 1971) 727-30. 15. Fast, C.R.: "Injection and Instantaneous Shut-in Pressure Changes During Hydraulic Fracturing." paper presented at the 30th Annual ASME Petroleum Div. Petroleum Engineering Con- ference, Tulsa, Sept. 1975. 16. Nolte. K.G.: "Detennination of Fracture Parameters From Frac- turing Pressure Decline, " paper SPE 8341 presented at the SPE 1979 Annual Technical Conference and Exhibition, Las Vegas. Sept. 23-26. SEPTEMBER 1982 17. Van Eekelen. H.A. M.: " Hydraulic Fracture Geometry: Fracture Containment in Layered Fonnations, " Soc. Pet. Ellg. 1. (June 1982) 341-49. APPENDIX Example of Application of Trendlines To Assess Horizontal Stress An example of the application of the total horizontal stress relationship with depth and pore pressure is given below. Knowing the pore pressure at the depth of in- terest, we can select the equation to calculate horizontal total stress. If we assume that an imaginary well in the U.S. gulf coast region encounters a pore pressure of 4,000 psi at a depth of 7,500 ft, the horizontal total stress can be ap- proximated by SHmin =0.197D 1.145 +0.46(P c -P cn ), with D in feet and S Hmin' Pc, and Pen in pounds per square inch. The normal pore pressure gradient for the U.S. gulf coast is 0.465 psi/ft, so we have S Hmin =0. 197(7500) 1.145 +0.46[4000-(0.465 x75OO)] =5388+236=5,624 psi . The slight overpressure results in a horizontal total stress 236 psi higher than if the formation had been at normal pore pressure. SI Metric Conversion Factors bar x 1.0* E+05 Pa bbl x 1.589 873 E-OI m 3 ft x 3.048* E-Ol m pSI X 6.894757 E-03 MPa Conversion factor is exact. JPT Original manuscript received in Society of Petroleum Engineers office July 27, 1981. Paper accepted for publicalion March 2, 1982. Revised manuscript received June 1 e, 1982. Paper (SPE 10336) first presented at the 1981 SPE Annual Technical Con ference and Exhibition held in San Antonio Oct . 5-7. 2199