Reservoir rock classification, Arab-D reservoir, Ghawar field, Saudi Arabia

Fluid flow behavior and fluid saturation distributions in the subsurface are controlled by the pore network developed in reservoir rocks. Hence, the key to predicting the saturations and matrix flow behavior of a reservoir is to develop an understanding of the geologic controls producing this pore network. The first step in this understanding is to establish a direct relationship between rock types and pore networks. This petrophysical study reports the results of an effort to develop a reservoir rock classification that provides a better understanding of the relationships between rocks, pores and fluids within the Arab-D reservoir in Ghawar field (Figure 1).

The reservoir rock classification developed during this study provides a simple but practical method of dividing the complex carbonate rocks of the Arab-D into meaningful reservoir rock types. Each rock type has a distinct pore network as defined by porosity-permeability relationships and capillarity expressed as pore-size distributions and J-function curves. As a result of this work, a much better understanding of the reservoir characteristics of various rock types within the Arab-D is possible.

The objective of this study was to investigate the basic relationship between petrographic variables (such as a rock’s texture and pore types) and the pore network, as defined through its porosity, permeability and capillarity (J-function curves and pore-size distributions), of the Arab-D carbonates in Ghawar field. The purpose of this work was to provide additional insight into the reservoir rocks found in the Arab-D in order to better determine their fluid flow characteristics and saturation behavior.

Prior to this study, most petrographic work had been related to depositional or diagenetic facies employing either textural classifications (Lucia et al., 2001; Powers, 1962; Dunham, 1962) or lithofacies classifications (Meyer and Price, 1993; Mitchell et al., 1988). Rock descriptions using rock classifications based on texture and lithofacies provide the information required to develop the depositional and stratigraphic framework of the reservoir and its gross reservoir quality. However, they do not provide sufficient information to define pore network types within the reservoir. These textural and lithofacies classifications only indirectly relate to pore types and fail, in most instances, to subdivide the reservoir rocks into groupings that are significant relative to their reservoir performance.

The results described in this study are based on an analysis of 378 samples from 13 Arab-D wells (Table I, Figures 1 and 2). Approximately 53 percent of the samples were from Zone 2 and 43 percent from Zone 3, with the remainder from Zones 1 and 4. The large number of samples from Zones 2B and 3A in particular reflect their diverse rock types.

Detailed petrographic descriptions were conducted on all 378 samples, as were porosity and permeability measurements. In addition, mercury capillary pressure data were obtained for 165 samples using automated high pressure equipment. These data were then analyzed (using multivariate statistical procedures) to determine key petrographic variables controlling the reservoir properties. Scanning electron microscopy was also used on selected samples to examine the micropore system.

The selected samples contained a representative group of Arab-D rock textures and depositional lithofacies (Table II; Figures 2 and 3). In general, the upper Arab-D (Zones 1 and 2) is dominated by grainstones, mud-lean packstones and (locally) muddy packstones; the lower Arab-D (Zones 3 and 4) in contrast is characterized by more mud-dominated textures, especially mud-rich packstones and wackestones (Figure 3). Likewise, the proportions of the various lithofacies varies within the reservoir interval, with the upper Arab-D dominantly comprising skeletal-oolitic, Cladocoropsis and stromatoporoid-red algae-coral lithofacies, while the lower Arab-D contains bivalve-coated grain-intraclast, micritic and rare skeletal-oolitic lithofacies. Dolomite occurs throughout the Arab-D. While diagenesis - and especially grain leaching - is predominantly controlled by the composition of each lithofacies, significant differences in diagenesis do occur between the upper and lower Arab-D, as will be shown later. For a full discussion of the lithofacies nomenclature used in this study, refer to Mitchell et al. (1988).

Both the textural and lithofacies distributions used in this study are approximately equivalent to the general population of rock in the reservoir. By far the most abundant reservoir rock texture in the Arab-D and in this study is packstone; likewise, the most abundant lithofacies is skeletal-oolitic (SO). The other lithofacies, including the Cladocoropsis (CLADO), stromatoporoid-red algae-coral (SRAC), bivalve-coated grain-intraclast (BCGI), micritic (MIC) and dolomite (DOLO) lithofacies, are represented in approximate proportion to their abundance in the reservoir. Figure 4 shows the porosity-permeability distribution of the study samples and has the typical wide scatter found in Arab-D rocks. It also demonstrates the characteristic higher permeability of reservoir-quality dolomites relative to limestones, and the higher porosities typically found in Zones 1 and 2 rocks compared to those from Zones 3 and 4.

Work on these samples followed a multi-phase program. These phases included:

• (1) Petrographic description of both the plugs and thin sections were made using a standard binocular microscope, and observations of 36 lithologic parameters were recorded. Estimates of the abundance of various components were made using standard comparison charts (Terry and Chilinger, 1955), while size estimates (of modal grain and mold sizes) were made using grain size comparison charts overlain on the plug or thin section.

• (2) These descriptions were combined with Boyles Law measurements of porosity and permeability on 1.25 inch diameter horizontal core plugs, to produce an integrated computer data base of petrophysical data on these samples. This database was then analyzed using multivariate statistical analyses to determine significant petrographic variables and statistical groupings of the samples. Two types of multivariate statistics were used to analyze the data, principle component analysis (PCA) and cluster analysis; PCA was used to establish important variables and provide suitable components for use in cluster analysis, while cluster analysis was used to group the individual samples into similar petrophysical groups.

• (3) Mercury capillary pressure data for 165 samples were analyzed for characteristics of the pore network. These data were collected by two laboratories, Micromeritics Instruments Corp. and Core Laboratories. Micromeritics analyzed 132 samples using their highly automated AutoPore-II system to obtain 200 drainage capillary pressure data points from 1 to 30,000 psi, while Core Labs analyzed an additional 33 samples using routine mercury injection procedures that captured 30 equally-spaced pressure points from 2 to 2000 psi. For the purposes of this study, it was assumed that the higher resolution of the Micromeritics technique provided superior detail results over the Core Labs technique, so these samples were preferentially used to develop the pore throat-size relationships characteristic of each of the rock-types identified; the lower resolution Core Labs samples were only used for qualitative comparison and assessment of the 33 samples analyzed. Pore-size distribution curves were obtained by converting capillary pressure (Pc) to pore diameter (d) and then replotting d versus the change in wetting phase saturation according to the Wasburn equation (equation and technique are discussed in Cantrell and Hagerty, 1999).

• (4) J-function curves were calculated as a dimensionless function of the wetting phase saturation defined as:

  $J(s)=2.7PcσCOSθkø$

  where:

  Pc= capillary pressure, psi

  σ = interfacial tension, dynes/cm

  θ = contact angle, degrees

  k = permeability, millidarcys

  ø = porosity, percent

  The terms σ and θ are primarily functions of the fluids involved and when comparing rock properties in a given reservoir, they can be considered constants. Therefore, we can write:

  $J(s)(C)=PCkø$

  This relationship can be used to relate the pore-size distribution of rocks and initial fluid saturations. Theory and experience have shown that rocks with similar pore-size distributions will have similar J versus fluid saturation curves or J-curves (Leverett, 1941; Rose and Bruce, 1949). These curves can then be used to compare the effect of various rock types on initial fluid saturations.

• (5) Average pore-size distribution curves and average J-function curves were generated, and all data were combined into a final classification based on five easily quantified petrographic parameters.

• (6) The final phase of the study involved a limited scanning electron microscopy (SEM) study of the micropore structures of each rock type in Zones 2 and 3.

In general terms, the D Member of the Arab Formation was deposited during the Upper Jurassic in the shallow Tethys Sea on a broad, relatively stable shelf or platform. This broad shelf lay between 10-15 degrees south latitude, and was bounded to the north by the Basrah Basin and by the Rub’Al Khali Basin to the south. To the west lay the Arabian Shield, while the Qatar-Surmeh High was located to the east. The overall abundance of grain-rich, mud-poor sediments suggests that generally high water energy conditions prevailed across much of this shelf area. Overall, the paleoclimate was probably hot and arid, much like today’s climate on the Arabian Peninsula. The geology of this area during the Tithonian has been previously summarized (Al Husseini, 1997; Alsharhan and Kendall, 1986; Alsharhan and Magara, 1995; Ayres et al., 1982; Cecca et al., 1993; Fourcade et al., 1993; Grabowski and Norton, 1995; Handford et al., 2002 in press; Le Nindre et al., 1987; Mitchell et al., 1988; Murris, 1980; Powers, 1962; Wilson, 1975), and will not be reviewed in further detail here.

The Arab-D carbonates of Ghawar field have been grouped into a reservoir rock classification containing seven limestone and four dolomite rock types (Figure 5). The distinction between limestone and dolomite follows the standard Saudi Aramco practice of defining a dolomite as a rock with 75 percent or more dolomite mineral (Powers, 1962). Within each of these distinct lithofacies, different petrographic parameters are important in controlling the reservoir behavior of the rocks. The amount of matrix (lime mud) and the pore types are the primary controlling parameters for limestones, while dolomite texture is the controlling parameter for dolomites. The limestones also exhibit differences in the size of their micropores between Zones 1 and 2, and Zones 3 and 4.

The seven limestone reservoir rock types are defined based on the magnitude or character of five petrographic parameters. These are: (1) percentage of matrix (lime mud); (2) degree of grain sorting; (3) dominant pore type; (4) size of the largest molds; (5) percentage of cement.

Of these five parameters, matrix or lime mud is by far the most important (in this study, matrix is defined as all material smaller than 0.06 mm in diameter). Cement is not a major component of the Arab-D rocks studied but when it exceeds 10 percent of the rock, the pore size distribution is often significantly affected. Therefore, rocks with greater than 10 percent cement form a separate rock type (IV) regardless of their other attributes.

Excluding Type IV (cemented) rocks, we can then divide the Arab-D limestones into two families with each subdivided into three types depending on matrix content. Family “A” tends to be a coarse grained, poorly sorted rock with abundant and relatively large moldic porosity. Family “B” is generally a fine to medium grained, well sorted rock, dominated by interparticle and/or very small moldic porosity. These relationships are shown in Figures 6 and 7.

Rock Types I, II and III are determined by the amount of matrix present. Natural splits of the data suggest that 15 and 40 percent matrix are significant breaks in matrix content. Type I rocks have 15 percent or less matrix. Rocks with as much as 15 percent matrix have similar reservoir properties to those with no matrix. This fact is significant in respect to the Dunham textural classification because Type I rocks include all grainstones, all mud-lean packstones (packstones containing 10% or less matrix) and some packstones. Rocks with more than 40 percent matrix (Type III) have pore systems that are dominated by matrix and the grain to matrix ratio ceases to be important. Type III rocks include some packstones and all wackestones and mudstones of the Dunham classification. Type II rocks are those with matrix contents from 16 to 40 percent. These rocks tend to have a wide range of reservoir characteristics and are all considered to be packstones in the Dunham classification.

Type I, or the low matrix rocks, are subdivided into IA and IB based on the sorting of the carbonate grains.

$Sorting=largest grainsmallest grain(KrumbeinandPettijohn,1938)$

A sorting coefficient of 0.5 is used to divide those with poor sorting (IA) from those with good sorting (IB). A sorting coefficient of 0.5 occurs when the largest grains are 10 times bigger than the smallest grains. The poorer the sorting, the larger the sorting coefficient becomes. In the Arab-D, sorting also is an approximate indicator of grain size; poorly sorted rocks tend to be coarse grained while well sorted rocks tend to be fine grained.

Sorting ceases to be a dominant factor in controlling reservoir quality when the matrix content increases in the Type II and Type III rocks. For these rocks, pore type and size of the leached molds become significant variables. Type II rocks (16-40 percent matrix) are subdivided into Types IIA and IIB based first on the dominant pore type, and then on the size of the largest leached molds. Type IIA rocks have moldic (MO) and intraparticle (WP) pores in excess of the interparticle (BP) pores and have leached molds larger then 0.3 mm in diameter. Type IIB rocks are dominated by interparticle porosity or have molds smaller than 0.3 mm. The Type III rocks are subdivided only on the basis of the leached mold size since there is essentially no inter-particle porosity in this rock type. Type IIIA has molds larger than 0.3 mm and Type IIIB smaller.

The final element of the limestone classification relates to the stratigraphic position of the rock within the reservoir. As will be discussed later, the size of the smaller or micropore pore system of the rocks is larger in Zones 1 and 2 than it is in Zones 3 and 4 for each of the six major rock types. This size difference is believed to be significant in regards to the reservoir properties of the rocks, but it is not a petrographic element which can be easily identified. Only by knowing the stratigraphic position of the rock or by obtaining its pore-size distribution can the micropore size of Arab-D limestones be determined.

Each rock type has its unique combination of rock attributes. At the same time, there is a complex interplay between the attributes which results in a blending of the types so that sharp boundaries between them do not exist. Direct relationships can be seen in the histograms in Figure 8. These include the correlations between porosity and permeability; poor sorting and grain size (large sorting values correspond to poor sorting); and large molds and grain size. There are also many inverse relationships such as: matrix and permeability; interparticle porosity and matrix; interparticle and moldic porosity; and cement and matrix. More complex relationships are also apparent such as the fact that the very poorly sorted Type IA rock has permeabilities as high as the well sorted Type IB rock. This relationship is contrary to the usual situation in clastic rocks. In the Arab-D, however, the effect of sorting is offset by the coarse nature of the Type IA rocks relative to IB rocks and the presence of large moldic pores in the former.

The histograms in Figure 8 also highlight the differences between the “A” and “B” families for such variables as interparticle porosity, moldic porosity, grain size, sorting and especially the largest mold size. The histogram of porosity also demonstrates its relative uniformity between many rock types.

Most carbonate reservoirs exhibit a poor relationship between porosity and permeability and the Arab-D is no exception (Figure 4). This scatter is a result of three factors: (1) variations in rock type, (2) non-uniformities in the samples used for analysis, and (3) analytical errors. In this study, much of the scatter attributed to the first two factors can be accounted for.

Figure 9 shows that rather than one porosity-permeability relationship for all Arab-D limestones, there is a series of parallel or sub-parallel porosity-permeability groups formed by the six primary reservoir rock types. (Type IV samples are scattered among these primary groups and have not been plotted.) Rock types from the two families (A and B) tend to overlap; however, their capillary properties, which affect saturations and relative permeabilities, are different as will be discussed later. These capillary properties are important classifying elements in addition to porosity and permeability. Overlap of the groups also results because of their gradational nature and the previously mentioned causes of data scatter.

Also indicated on Figure 9 are those samples with visible core plug non-uniformities or textural characteristics that altered the fundamental porosity-permeability relationships resulting from the basic rock framework. While they are a normal aspect of most carbonate reservoirs, such non-uniformities serve to complicate the task of understanding the fundamental controls on reservoir quality, so for this study many non-uniform plugs were rejected during sample selection because of their obviously heterogeneous nature. However, the majority of Arab-D plugs have some degree of non-uniformity in them; at times this badly distorts the porosity-permeability relationship while at other times it does not. In this study, the most obvious causes for permeabilities higher than expected for a given porosity were either microfractures, burrows, or textural variations. In all of these cases, flow channels exist within the plug allowing rather low porosity plugs to have significant permeability. On occasions, burrows or textural variations can also decrease the permeability. Textural variations probably also result from intense bioturbation that does not leave distinct burrows.

Mercury capillary pressure measurements allowed the average pore throat-size distributions to be calculated by zones for each of the reservoir rock types. Figure 10 is a summary of these data and it shows significant differences between the various rock types and between reservoir zones. Cumulative curves are also given in Figure 11. In the distribution curves shown in these Figures, the “pore diameter” is the pore throat diameter which connects the pore volume indicated on the plots. For example in Figure 11, for Type IA rocks in Zones 1 and 2, approximately 25 percent of the pore space is connected by pore throats of one micron or less and approximately 48 percent of the pore space is connected by pore throats of less than 10 microns.

These distribution curves also dramatically illustrate the bimodal (dual porosity) nature of the porosity in Arab-D limestones and that the size of the micropores is larger in Zones 1 and 2 than in Zones 3 and 4. The bimodal character is best developed in Zones 3 and 4 where a pore-throat diameter of about one micron appears to be the separation point between the two modes of “micro” and “macro” porosity. Also, as the amount of matrix increases from Type I to III, the modal size of the macro-size pores becomes smaller, while that of the micro-size pores remains at about 0.1 micron. The result is that the two modes move together. Also, as the amount of matrix increases, the amount of micro-size pores increases to as much as 90 percent of the entire pore system (Figure 10).

The data for Zones 1 and 2 still show an obvious dual porosity system for the Type I samples but the two modes tend to merge in the Type II and IIIA rocks and have only one mode in the IIIB rocks. This difference between zones is the result of a systematic increase in pore throat size in the upper Arab-D in other words, typical pore throat sizes are larger in the upper Arab-D than in the lower. The micro-size mode for Zones 1 and 2 is about one micron, or an order of magnitude larger than the mode for Zones 3 and 4. The macro-size mode, however, is almost identical for all zones for Type I and II rocks. In the Type III rocks, where most of the pores are micro-sized, the entire pore-size distribution is shifted to larger sizes in Zones 1 and 2.

It is believed that the differences between the pore systems of the upper and lower Arab-D are related to increased leaching in the upper Arab-D. Such leaching has previously also been suggested as a dominant cause of the microporosity in the Arab-D (Cantrell and Hagerty, 1999). Repeated flushing by hypersaline brines in the Upper Arab-D is also postulated by Cantrell et al. (2001) and Swart et al. (in press) in their dolomitization model. Additional scanning electron microscopy (SEM) evidence for leaching as the explanation for micropore size variation will be presented in a later section.

The impact of moldic porosity is also displayed in Figures 10 and 11. Large moldic pores are an important attribute in the “A” family but not in the “B” family of rock types. This characteristic is seen in the pore throat-size distribution curves as a consistent larger amount of “large” pore throats in family “A” than in family “B” for the same major rock type. Likewise, the median pore throat size is always larger for the “A” family than the “B” family for comparable rock types.

J-functions are one means of normalizing capillary pressure data to account for variations in porosity and permeability. J-functions based on lithology and permeability ranges have historically been used to initialize simulation model saturations in many carbonate reservoirs (Amyx et al., 1960).

The average J-curves by zone for the six primary reservoir rock types developed in this study are shown in Figure 12. Several observations are apparent. The general groupings of the curves are similar for the two zonal sets, but they do differ in detail and probably should not be combined. Each rock type has a different curve and there is a complex relationship between them. Also, different rock properties control the curves depending on the saturation range involved. In situations of high wetting phase saturations (mercury vapor in the cases illustrated but water in the reservoir), the major control on the curves is the size and amount of moldic porosity. This control is evidenced by grouping of A and B family curves. However, when wetting saturations are low, the curves are strongly controlled by the amount of matrix and the two families tend to overlap, especially in Zones 3 and 4. At intermediate saturations the curves are similar. The major implication of these relationships is that if imbibition relationships follow these drainage relationships then initially, the amount of matrix in the rock greatly affects the reservoir saturations but as the reservoir is depleted and water encroachment occurs, the pore types become an important component that affect fluid saturations and displacement behavior.

The preceding account has concentrated on the limestone reservoir rock classification and relationships between its various types. This section contains a brief discussion of the significant characteristics of each limestone rock type, and focuses on how these rock types relate to other types of geologic information such as depositional texture and lithofacies. Please note, however, that significant variability occurs within each rock type and that clear, sharp boundaries do not exist between rock types. Very small changes in the value of a variable can move a sample from one rock type to another.

Type IA reservoir rocks are basically grainstones, mud-lean packstones and packstones of the BCGI, SRAC and CLADO lithofacies. Some SO facies also occur in this type. Type IA rocks are characterized by their low matrix content, coarse grain size, poor sorting and large moldic pores. They typically have permeabilities in the range of 100 to 1,000 mD over a wide porosity range. Its modal pore throat size is on the order of 20 to 50 microns, which is the largest of all rock types. This rock type constitutes the best reservoir rock in the lower part of Zone 3A and all of Zones 3B and 4. It is also a dominant rock type in much of the middle and lower portions of Zone 2B.

Type IB reservoir rocks are generally quite different petrographically from Type IA, but have similar porosity and permeability values. Petrographically they are grainstones, mud-lean packstones and packstones of the SO lithofacies. Some well-sorted BCGI and SRAC samples also occur in this group. Type IB rocks are characterized by low matrix content, good sorting, generally fine to medium grain size and almost complete lack of moldic porosity. The porosity and permeability data cluster around 25 percent porosity and 400 mD permeability, with no firmly established trend. The lack of large moldic pores is evident on the pore-size distribution curve (Figure 10) when compared to Type IA. There is a distinct shift of the modal size towards smaller pore sizes and the larger pores in the 80 to 100-micron range are nearly absent. The effect of these differences in pore-size distribution is also seen by comparing the J-curves of the two types (Figure 12). The curves of Type IA and IB are quite similar except at high wetting phase saturations. Type IB rocks are most common in Zone 2A and the upper part of Zone 3A where the SO lithofacies is most abundant (Mitchell et al., 1988).

Type IB rocks illustrate the effect that only a small amount of matrix has on the permeability of a rock. Figure 13 shows that much of the scatter in permeability for this rock type is a result of as little as 3 percent matrix. Rocks with 3 percent or less matrix consistently have permeabilities nearly an order of magnitude higher than those with 3 to 15 percent matrix. While the presence of matrix does not explain all of the scatter, it appears to be the major control.

Type IIA reservoir rocks are the muddy equivalents of Type IA rocks. The porosities of the two types are quite similar (Figure 9) but because of their increased matrix content, Type IIA rocks have distinctly lower permeabilities and a quite different pore-size distribution (Figures 10 and 11) and J-curve (Figure 12) than Type IA. Petrographically, Type IIA is always a packstone and normally is the BCGI or SRAC lithofacies although a few SO and CLADO samples do occur in this rock type. Perhaps the most distinctive petrographic property of the Type IIA samples is their large molds and the dominance of moldic over interparticle porosity. This rock attribute could result in poor oil recovery from this rock because large pores are connected by small pore throats as a result of high matrix content in combination with moldic porosity. Type IIA rocks are intimately associated with Type IA in the reservoir and on the basis of porosity alone, the two types cannot be distinguished.

Type IIB reservoir rocks are the muddy version of Type IB rocks but unlike its family “A” counterpart, all of its reservoir attributes differ from Type IB (Figures 9 through 12). Rock Type IIB is always a packstone and is usually found in the SO or BCGI lithofacies although it occurs in all lithofacies. The major difference between Type IIB and Type IIA is the absence of large (>0.3 mm) moldic pores in the Type IIB. The lack of large molds and the increase in muddy matrix in Type IIB results in a decrease in porosity as well as permeability compared to Type IB and IIA rocks. During deposition, mud filled the interparticle pore space in both Type IIA and IIB rocks but in the case of Type IIA, later leaching removed many large unstable grains and thus increased porosity. In the case of Type IIB rocks, however, few large unstable grains were present so later leaching had little effect on the total porosity. Type IIB rocks are found throughout the reservoir.

Type IIIA reservoir rocks are dominated by muddy matrix and are packstones and wackestones of all lithofacies with molds larger than 0.3 mm. The most common lithofacies, however, are the BCGI and SRAC. The porosity and permeability relationships of this rock type are similar to those of the Type IIB rocks (Figure 9); however, their pore systems are quite different. In the case of Type IIB, interparticle porosity is the dominant macroporosity while moldic porosity dominates Type IIIB (Figure 8). The fact that the moldic pores are larger than the interparticle pores appears to compensate for the fact that Type IIIA rocks average about 30 percent more matrix than the Type IIB rocks (Figure 8) but have similar permeabilities. The pore-size distribution of Type IIIA rock is almost log normal for Zones 3 and 4 (Figure 10) as the micro and macro modes combine. However there is a shift in the total distribution for Zones 1 and 2 reflecting an apparent increase in leaching. In fact, for Zones 1 and 2, there is little difference in the pore-size distribution between Type IIA and Type IIIA rocks (Figures 11 and 12). Type IIIA rocks are found throughout Zones 2B and 3 and occasionally 2A. It should be noted that in Zone 2, this rock type cannot be identified based on porosity alone, but in Zone 3 it generally occurs in lower porosity intervals.

Type IIIB is the least permeable of all the rock types and often is of non-reservoir quality (Figure 9). It probably represents the most abundant rock type in Zones 3 and 4. Type IIIB rocks are mostly wackestones and mudstones of the MIC facies. However, some packstones and samples from other facies also occur in this rock type. Type IIIB rocks are the most dolomitic of all the rock types although some dolomitic samples occur in all reservoir types. All visible porosity (which is often minimal) occurs either as very small moldic pores or as intragranular porosity within scattered foraminifer tests (Figure 6). The pore-size distribution of this rock type in Zones 3-4 is unique in that it is all connected by micropores but has a pronounced bimodal distribution. This bimodality is thought to be related to variable degrees of leaching of the matrix. Some samples have only one mode between 0.01 and 0.1 microns (minor leaching), others have only one mode between 0.1 and 1.0 microns (leached), and others have both modes. Type IIIB rocks in Zones 1-2 are often very porous (Figure 9) and have a unimodal pore-size distribution larger than the Zone 3-4 samples. Leaching is again believed to be the cause for this improvement in porosity size and amount. Type IIIB rocks occur throughout Zone 3 and while much of it is non-reservoir (Ø<5%, k<0.1 mD), it still constitutes a significant amount of Zone 3 reservoir rock. This rock type also occurs to a limited degree in Zone 2 where it is often highly porous but with low permeability.

Type IV reservoir rocks are those with greater than 10 percent cement (Figure 14). Only 12 samples of the 378 described were found to contain this much cement. Generally, significant amounts of cement are found only in relatively mud-free samples, especially those with moldic pores. The limited number of Type IV samples makes it impractical to generalize on its pore network. Based on the available samples it appears that depending on the type and amount of cement the pore network is variously affected, the samples do not fit with their noncemented counterparts and meaningful average reservoir properties are impractical to calculate.

Four dolomite reservoir rock types have been identified within the Arab-D based on the dolomite texture (Figure 5). These textures are: fabric preserving, sucrosic, intermediate and mosaic, and they correspond to reservoir rock types Vfp, Vs, Vi and Vm respectively. Unlike the limestones, the dolomite reservoir rock types can be determined by their stratigraphic position or their porosity value. Vfp dolomite only occurs in Zone 1 and most dolomite in Zone 1 is Vfp. As will be shown below, Vs, Vi and Vm dolomites have characteristic porosity ranges which allows them to be identified based on their porosity.

Dolomite constitutes only about 14 percent of the Arab-D in Ghawar (Cantrell et al., 2001) and much of it is of nonreservoir quality. Therefore, dolomite was not emphasized in this study. A more exhaustive study with more samples might produce additional dolomite rock types.

The fabric preserving texture (Vfp) has previously been recognized in the Arab-D (Mitchell et al., 1988; Cantrell et al., 2001). This type of dolomite is composed of 10- to 30-micron sized crystals which preserves the original limestone texture (Figure 14). The pore network of Vfp dolomite remains essentially the same as that of the precursor limestone. This type of replacement is in contrast to the more typical non-fabric preserving textures (Figure 14) found in the Arab-D, and most other dolomites forming petroleum reservoirs. Non-fabric preserving dolomite has two textural end members, sucrosic (Vs) or sugar-like and mosaic (Vm). Vs dolomite is composed almost completely of dolomite rombs with point contacts. Vm dolomite is composed completely of anhydral dolomite crystals with face-to-face contacts. Between these two end members occur the bulk of the Arab-D dolomites which have textures intermediate between sucrosic and mosaic; the Vi dolomites.

A common variation on the microscopic intermediate texture is the macroscopic mixture of sucrosic and mosaic textures in the same plug. This macroscopic mixture results in “patchy” porosity within the samples and produces heterogeneities which cause wide scatter of data for all types of core analysis procedures. Patchy porosity development is very common in Arab-D dolomites and has been previously ascribed to changes in the original texture of the precursor limestone (Mitchell et al., 1988), probably reflecting bioturbation shortly after deposition.

Two other aspects of Arab-D dolomites, crystal size and moldic pores, were examined as to their likely influence on reservoir rock types. Based on the limited number of samples, no consistent relationships between these variables and porosity, permeability or capillarity were found. Both of these petrographic variables should influence the reservoir characteristics of the rock but it is believed that the heterogeneities from the patchy porosity and irregular development of dolomite textures within small plugs masks any effect of crystal size or moldic pores in the samples studied. In addition, the small sample population of these rock types may also account for the lack of a consistent relationship between crystal size and moldic pores and reservoir quality.

Arab-D dolomites exhibit a strong relationship between crystal texture and porosity and permeability (Figure 15). Samples with sucrosic texture (Vs rocks) have high porosity and permeability while samples with mosaic textures (Vm rocks) have low porosity and permeability and are usually non-reservoir rock. Between these two end members occur the bulk of the samples with intermediate textures (Vi rocks). Porosity ranges can be assigned to these three textural types; less than 5 percent for Vm, 5 to 12 percent for Vi and greater than 12 percent for Vs. The patchy porosity development in several of the Vi samples results in outlier points on the porosity-permeability cross plots.

Also shown in Figure 15 are five data points for the Vfp dolomite found in Zone 1. In these particular samples, the original rocks were grainstones of the skeletal-oolitic facies or Type IB reservoir rocks. Three of the samples have oomoldic porosity in which the original grains (mostly ooids) have been leached and the original interparticle porosity has been largely infilled with cement. This fabric results in significantly lower permeability for these samples than would be expected from their high porosity. Petrophysical relationships of fabric-preserving dolomite textures are more closely related to their original limestone equivalents than to the other dolomites. However, since this Vfp dolomite is rare and restricted to the overall very thin Zone 1, it is of minor importance within an overall Arab-D framework.

Dolomites tend to have rather uniform pore-size distributions compared to limestones. Figure 16 shows the distribution of the Vs, Vi and Vm dolomites based on 15 mercury capillary pressure curves. (No average for the Vfp type dolomite is presented because of limited data.) These curves demonstrate the uniform nature of the pore network in the Vs and Vm dolomites and also show the influence of both end members in the Vi type.

The average mercury J-curve for the dolomites differs significantly from those of the limestone reservoir rock types. Figure 17 compares these curves and supports the accepted practice of treating limestones and dolomites as separate rock types. Figure 18 shows the average J-curves for the three individual non-fabric preserved dolomite types. Note how the curve for the Vm dolomites has a similar shape, but is displaced from, the curve for the Vs dolomites. This difference in J-curve profile is in agreement with other research which suggests that pores and pore throats for the Vm dolomites would tend to be sheet-like while those for Vs dolomite would be more or less polyhedral or tetrahedral (Wardlow, 1976). The Vi dolomites are interpreted to contain a mixture of both sheet-like and polyhedral pores, and as a result the J-curve for this dolomite type has a profile that is intermediate between that seen in the Vm and Vs dolomites.

In summary, the carbonate reservoir rock classification developed in this study is a function of: grain to matrix relationships; the size relationships of the grains; the nature of the pores in terms of both their type and size; and in the case of dolomites, the crystal textures. These same rock attributes are those considered in various ways by other investigators. Dunham (1962) stressed the importance of matrix (mud) and grain abundance in terms of their rock support function but maintains the basic idea of the important distinction between these two end members. Powers (1962) recognized the importance of grain to matrix relationships and grain size in his classification of Arab-D rocks. Neither of these workers, however, incorporated pore types or sizes in their classifications. Lucia (1983) and Lucia et al. (2001) consider “grain” size (including matrix as well as grains) and recognizes the importance of interparticle versus moldic (vuggy) porosity. Other workers (Archie, 1952; Aschenbrenner and Chilingar, 1960; Stout, 1964; and Jodry, 1966) have stressed porosity types and sizes to the exclusion of grain to matrix considerations in developing petrophysical relationships. Choquette and Pray (1970) developed the now almost universally accepted carbonate porosity classification and stressed the importance of carbonate pore types in controlling reservoir characteristics. They did not, however, develop a reservoir rock classification.

Historically, the rocks in the Arab-D have been described using the Dunham textural classification and a variety of depositional facies classifications, most commonly the classification presented by Mitchell et al. (1988). Figure 19 shows a comparison between classifying the limestone samples used in this study in terms of reservoir rock types and by these two classifications. In a general way, the Dunham and Mitchell et al. classifications provide a guide for reservoir descriptions. Grainstones (GRN) and mud-lean packstones (MLP) are Type I rocks but may be either Type IA or IB. Packstones (PCK) may be any of the seven reservoir rock types but are most commonly Type II rocks or Type IIIA. Wackestones (WCK) and mudstones (MUD) are usually Type IIIB rocks but wackestones can also be Type IIIA rocks.

The Mitchell et al. depositional facies tend to divide the rocks into the “A” and “B” families. The bivalve-coated grain-intraclast (BCGI) and the stromatoporoid-red algae-coral (SRAC) facies are predominately “A” type rocks, although a significant number of the BCGI samples are Type IIB rocks (these tend to be samples without the bivalve molds). The“B” type rocks tend to be mostly skeletal-oolitic (SO) and micritic (MIC) facies. The Cladocoropsis (CLADO) facies tends to be associated with the “A” family but the limited samples from this facies makes the assignment inconclusive.

Therefore, if one knows the texture and depositional facies of a sample, a first approximation of its reservoir rock type can be made. For example, a SO grainstone nearly always is a IB rock, a MIC wackestone is probably a IIIB rock and a BCGI or SRAC mud-lean packstone is very likely a IA rock. As mentioned before, packstones present problems but making them IIA or IIB rocks depending on their facies is a good estimate. Because of the overlap of Type II and III samples in their reservoir behavior, it is more important to place a packstone sample in the correct family than in the correct II or III type (Figures 9 through 12). Any descriptions of rocks which highlight large amounts of cement are probably Type IV rocks.

Applying previous dolomite descriptions to the reservoir rock classification should be straightforward. Zone 1 dolomites can be assumed to be Vfp unless they have been specifically described as sucrosic or mosaic. Other dolomites can be assigned either on the basis of the textural description and/or their porosity ranges.

A significant finding of this study is that the micropore system within the Arab-D differs within individual samples, between rock types and stratigraphically. These differences are believed to result from variations in the original carbonate mineralogy and to the degree of leaching that the rock has undergone since deposition. These relationships were investigated by analyzing the porosity, permeability and pore size distributions of the samples and by scanning electron microscopy (SEM).

The differences between rock types are largely a function of the lithology (limestone versus dolomite), the amount of matrix present in the limestones or the dolomite texture (Figures 10, 11 and 16). Compared to limestones, dolomites have very little microporosity that, when present, occurs at the intercrystal boundaries between anhedral crystal development in Vi and Vm dolomites (Cantrell and Hagerty, 1999). These micropores are sheet-like in structure as opposed to the thread-like structure of the micropores developed in limestone matrix and grains. The differences between the amount and nature of the micropores in limestones is primarily controlled by the amount of matrix present within the sample (Figure 10).

In the following sections, the relationship of micropore size to total porosity and permeability and how this relates to stratigraphic position are examined. SEM work is also presented to illustrate differences in micropore morphology between reservoir zones and within individual samples.

The variation in limestone micropore sizes between zones has been discussed and illustrated in Figures 10 and 11. However, only by implication has the general size of the micropores (as measured by the mode) been related to the absolute porosity of the sample. Figure 20 shows this relationship. The correlation coefficient for all samples is 0.55, which is significant but fairly low. Also shown in Figure 20 is the cross plot of permeability versus micropore mode. This plot shows a total lack of correlation as indicated by a correlation coefficient of 0.10.

If these relationships are examined by reservoir rock type, the correlations improve (Table III and Figures 21 and 22). The porosity data support the previous contention that microporosity size is related to the overall magnitude of the porosity for a given rock type. The permeability-micropore mode correlation is not as good as that for porosity but shows improvement as the amount of matrix increases in both the “A” and “B” rock type families (Type I rock has the least matrix and Type III the most). This permeability relationship is as one would expect. In Type I rocks, permeability is controlled by the large moldic and interparticle pore system. The micropores contribute nothing to the permeability. But as the amount of matrix increases, the micropores within the matrix begin to control more of the permeability. Finally in a Type III rock, all permeability is controlled by the size of the pore throats in the matrix. Therefore, in the Type III rocks the micropore size is closely related to the sample’s permeability.

It can also be seen in Figures 21 and 22 that there is a close relationship between large micropores and reservoir zone. When a porosity value was selected for each reservoir rock type that would best divide the samples into groups of high and low porosity as well as small and large micropores (Figure 21), it was found that these porosity splits effectively divided the Arab-D reservoir into two groups at the Zone 2-3 boundary. Rocks in the upper Arab-D (Zones 1-2) have a micropore mode at about 1.0μm while those in the lower Arab-D (Zones 3-4) have a micropore mode at about 0.1μm.

SEM examination of pore casts of 18 samples provided a possible explanation for these variations seen in the pore-size distributions and micropore modes determined from the mercury capillary pressure work. Comparison of samples from Zone 2 and Zone 3 reveal that Zone 2 samples often have an “eroded” morphology while those from Zone 3 display more angular and “sharp” micropore edges for the same reservoir rock type (Figure 23). This observation suggests that leaching has enlarged and smoothed the pores and pore-throats in the Zone 2 samples.

The reservoir rock classification developed in this study of the Arab-D in Ghawar is important because it emphasizes the petrographic characteristics that are significant for reservoir performance. The classification itself is of secondary importance and the limits assigned to define each rock type may need to be modified with additional data and use of the classification. The value of considering both texture and pore types in assigning rock types and the recognition of the size differences in the micropore system between the upper and lower Arab-D are major contributions to the knowledge and understanding of Ghawar field.

A key benefit of this classification is that is provides a superior mechanism for assigning reservoir-significant rock fabric properties in 3-D across the reservoir, and so can used to build more accurate reservoir models. The standard description of a rock by a Dunham texture name and a depositional lithofacies typically does not provide the necessary reservoir characterization of the rock for an understanding of its influence on fluid flow and saturations within the reservoir. This classification provides a way to incorporate the full range of reservoir quality variations into reservoir mapping and modeling efforts in Ghawar field. As such, while this specific classification may or may not be useful in other Arab-D or even other carbonate reservoirs, the principles developed for the Arab-D in Ghawar should have widespread application.

Probably the greatest weakness suffered by this classification - and most other rock type classifications - is its reliance on petrographic data for identification of reservoir rock type. Future development efforts in this area should focus on extending these reservoir rock types using wireline log data, especially NMR data. Application of this dataset could provide valuable information about pore volume amounts and pore size distributions that would be independent of and could substitute for the petrographic data utilized by this classification.

Appreciation is given to the Saudi Arabian Oil Company for permission to publish this article. We also extend our thanks to the two anonymous reviewers for GeoArabia, who reviewed an earlier version of this manuscript. We also greatfully acknowledge GeoArabia and Gulf PetroLink for the design and drafting of the final figures.

David (Dave) Cantrell has over 20 years of worldwide exploration and development experience in the oil industry. He graduated from the University of Tennessee with an MSc in Geology in 1982. Dave began his industry career with Exxon where he conducted numerous reservoir characterization and geological modeling studies on reservoirs in the Middle East; the Permian, Powder River, Williston, and Gulf of Mexico Basins of the USA; and the Maracaibo and Barinas Basins of Venezuela; among others. He has been responsible for several studies on large carbonate reservoirs since joining Saudi Aramco in 1997. Dave is presently Chief Geologist with Saudi Aramco’s Geological Technical Services Division.

dave.cantrell@aramco.com

Royal M. Hagerty made numerous contributions to Exxon Production Research Company and to the oil and gas industry over many years before retiring from Exxon in 1996. After an initial 2-year stint at Jersey Production Research Company, Royal left Exxon to attend Texas Tech to study oceanography. After graduating with a PhD in Oceanography, he worked as Chief Scientist for a deep-sea mining company operating in the Pacific. Royal rejoined Exxon in 1981 and, during the next 15 years, conducted numerous field and modeling studies for Exxon around the world.