ARTÍCULOS
Characterization of fluvial hydrocarbon reservoirs and aquifers: problems and solutions
Caracterización de reservorios de hidrocarburos y acuíferos fluviales: Problemas y Soluciones.
John S. Bridge
Department of Geological Sciences, Binghamton University, P.O. Box 6000, Binghamton, NY 13902-6000, USA.
Email: jbridge@binghamton.edu
Abstract
Fluvial deposits are important hydrocarbon reservoirs and aquifers in many parts of the world. In
order to assess, develop and manage these resources, it is necessary to determine the three-dimensional
geometry, orientation, spatial distribution and total volume of the reservoirs/aquifers. Such characterization
of reservoirs/aquifers normally involves (1) analysis of well logs, cuttings, cores and seismic data (2) correlation
of strata between wells, and (3) modeling of the three-dimensional volume between wells. These procedures
require a great deal of geological knowledge. Unfortunately, current approaches to reservoir/aquifer
characterization are problematical in many cases.
Specific fluvial depositional forms (e.g., channel bars, crevasse splays, lacustrine deltas) are commonly
interpreted from well logs and cores, because it is thought that deposits associated with particular depositional
forms will have particular geometries and reservoir quality. However, it is very difficult to unambiguously
interpret fluvial depositional forms from such one-dimensional data, and it is questionable that a particular
depositional form will have a distinctive stratal geometry. In particular, it is not possible to distinguish
braided river deposits from meandering river deposits based on their vertical facies sequences. Also, it is a
common myth that there is a relationship between channel-belt width/thickness, channel pattern, grain size
of sediment load, and bank stability. However, with very careful analysis, the lateral extent of subsurface
channel-belt deposits and overbank deposits can be estimated from their facies, thickness, and proportion
observed in well logs and cores.
Correlation of fluvial lithostratigraphic units (e.g., channel-belt sandstone bodies) between wells is commonly
based on untenable assumptions. The result is normally that stratigraphic units appear much more laterally
continuous than they really are. Before attempting correlation, it is desirable to estimate the likely lateral
extent and thickness variation of stratigraphic units, based on the interpretation of available well logs and
cores mentioned above. Then, it is wise to set up multiple working hypotheses for the correlation. Estimation
of the geometry of stratigraphic units is aided by use of modern analogs, ancient analogs, and depositional
models. Modern analogs are superior aids because it is possible to describe fluvial morphology and deposits
directly and unambiguously. Ancient analogs are less reliable, because their origin must be interpreted, and
because outcrops are commonly not large enough to allow definition of the facies geometry (e.g., channel-belt
width). Depositional models can be used to help estimate the lateral extent of stratigraphic units (e.g., zones
of high or low net-to-gross). However, such models may be only two-dimensional, not quantitative, or do not
represent depositional processes and products correctly. Sequence stratigraphic models are particularly
problematical in these regards.
In order to characterize the three-dimensional volume between wells, it is common to make use of ancient
analogs in combination with object-based stochastic models. The geometry and proportion of stratigraphic
objects such as channel-belt sandstone bodies are determined statistically from well data and ancient analogs,
and these objects are located in space using their known positions in wells and using quasi-random placement
between wells. Although this method honors well data, the shapes and locations of objects can be very
unrealistic. Process-based (forward) models are capable of producing much more realistic stratigraphy, and
have much greater predictive value, but it is very difficult to fit such models to well data exactly. For this
reason, process-based models are not routinely used for reservoir/aquifer modeling. However, it has been
shown recently that hybrid process-based and stochastic models can potentially be fitted exactly to well data
using a trial-and-error approach with new optimization procedures (genetic algorithms).
Keywords: Fluvial reservoirs and aquifers; Well logs and cores; Stratigraphic correlation; 3-D alluvial architecture models
Resumen expandido
Los depósitos fluviales
son importantes reservorios de hidrocarburos y acuíferos en
muchas regiones del mundo. A fin de valorar, desarrollar y
utilizar estos recursos es necesario determinar la geometría
tridimensional, la orientación, la distribución espacial y el
volumen total de los reservorios y acuíferos. Generalmente,
este tipo de caracterización de reservorios y acuíferos
involucra, (1) el análisis de perfiles de pozos, cuttings, testigos
corona y datos sísmicos, (2) la correlación de estratos
entre pozos, y (3) el modelado tridimensional de los depósitos
entre los pozos. Estos procedimientos requieren una elaboración
del conocimiento geológico. Desafortunadamente,
en muchos casos las propuestas actuales para la caracterización
de los reservorios y acuíferos son problemáticas.
Las formas fluviales de depositación (ejemplos: barras de
canal, crevasse splays, deltas lacustres) son interpretadas
comúnmente a partir de perfiles de pozo y testigos corona,
ya que se piensa que los depósitos asociados con formas
depositacionales determinadas tendrán geometrías particulares
y una calidad de reservorio determinada. Sin embargo,
es muy difícil interpretar inequívocamente las formas
depositacionales fluviales a partir de estos datos
unidimensionales y es cuestionable suponer que una forma
depositacional particular tendrá una geometría estratal distintiva.
En especial, no es posible distinguir depósitos de
ríos enlazados de depósitos de ríos meandrantes en base a
sus secuencias verticales de facies. Además, es un mito popular
que hay una relación entre ancho/profundidad de la
faja de canales, diseño del canal, tamaño de grano de la carga
de lecho y estabilidad de los márgenes. Sin embargo,
mediante un análisis muy detallado de las facies, el espesor
y la proporción observados en los perfiles de pozo y testigos
corona puede ser estimada la extensión en subsuelo de los
depósitos de las fajas de canales y de los depósitos de
albardón.
La correlación de unidades litoestratigráficas fluviales (por
ejemplo: cuerpos de areniscas de faja de canales) entre pozos
está basada comúnmente en suposiciones insostenibles.
Normalmente, el resultado de estas correlaciones es que las
unidades estratigráficas parecen mucho más continuas lateralmente
de lo que son en realidad. Antes de intentar la
correlación, es conveniente estimar la probable extensión
lateral y las variaciones de espesor de las unidades
estratigráficas, basándose en el análisis de los perfiles de
pozo y testigos corona mencionados anteriormente. A partir
de estas estimaciones es necesario establecer múltiples hipótesis
de trabajo para realizar la correlación. La estimación
de la geometría de las unidades estratigráficas es factible
mediante el uso de análogos actuales, análogos antiguos y
modelos depositacionales. Los estimadores más útiles son los obtenidos de análogos modernos porque en ellos es posible
describir sin ambigüedades la morfología y los depósitos
fluviales. Los análogos antiguos son menos confiables debido
a que su origen debe ser interpretado y porque los afloramientos
no son lo suficientemente extensos como para permitir
la definición de las geometrías de las facies (por
ejemplo: ancho de la faja de canales). Los modelos
depositacionales pueden ser utilizados para ayudar a estimar
la extensión lateral de las unidades estratigráficas (por
ejemplo: zonas de alta o baja relación porcentual entre espesor
útil y espesor total del reservorio, net-to-gross). Sin embargo,
tales modelos pueden ser análogos inapropiados, ya
que son solamente 2-D, no cuantitativos, o no representan
correctamente los procesos y productos depositacionales. En
estos análisis, los modelos de secuencia estratigráfica son
particularmente problemáticos.
A fin de caracterizar el volumen de los depósitos entre los
pozos, es común utilizar los análogos antiguos en combinación
con objetos basados en modelos estocásticos. La geometría
y la proporción de los objetos estratigráficos, tales como
cuerpos de areniscas de faja de canal, son determinados
estadísticamente a partir de datos de pozo y análogos antiguos,
estos objetos son ubicados en el espacio utilizando sus
posiciones conocidas en los pozos y un emplazamiento cuasi
aleatorio entre los pozos. Aunque este método señala la importancia
de los datos de pozos, las formas y ubicaciones de
los objetos son muy poco realistas. Los modelos basados en
procesos (ver más adelante) son capaces de producir
estratigrafías mucho más realistas y tienen un mayor valor
predictivo pero es muy difícil adecuar exactamente estos
modelos a los datos de pozos. Por esta razón, los modelos
basados en procesos no son usados rutinariamente en el
modelado de reservorios y acuíferos. Sin embargo, recientemente
ha sido demostrado que potencialmente los modelos
híbridos, entre estocásticos y basados en procesos, pueden
ser ajustados con exactitud a los datos de pozos utilizando
una aproximación por prueba y error con nuevos procedimientos
de optimización (algoritmos genéticos).
Los depósitos fluviales son importantes reservorios de hidrocarburos
y acuíferos en muchas regiones del mundo. A
fin de valorar, desarrollar y utilizar estos recursos es necesario
determinar la geometría tridimensional, la orientación,
la distribución espacial y el volumen total de los reservorios
y acuíferos. Generalmente, este tipo de caracterización de
reservorios y acuíferos involucra, (1) el análisis de perfiles
de pozos, cuttings, testigos corona y datos sísmicos, (2) la
correlación de estratos entre pozos, y (3) el modelado
tridimensional de los depósitos entre los pozos. Estos procedimientos
requieren una elaboración del conocimiento
geológico. Desafortunadamente, en muchos casos las propuestas
actuales para la caracterización de los reservorios y
acuíferos son problemáticas.
Las formas fluviales de depositación (ejemplos: barras de canal, crevasse splays, deltas lacustres) son interpretadas
comúnmente a partir de perfiles de pozo y testigos corona, ya
que se piensa que los depósitos asociados con formas
depositacionales determinadas tendrán geometrías particulares
y una calidad de reservorio determinada. Sin embargo,
es muy difícil interpretar inequívocamente las formas
depositacionales fluviales a partir de estos datos
unidimensionales y es cuestionable suponer que una forma
depositacional particular tendrá una geometría estratal distintiva.
En especial, no es posible distinguir depósitos de ríos
enlazados de depósitos de ríos meandrantes en base a sus
secuencias verticales de facies. Además, es un mito popular
que hay una relación entre ancho/profundidad de la faja de
canales, diseño del canal, tamaño de grano de la carga de
lecho y estabilidad de los márgenes. Sin embargo, mediante
un análisis muy detallado de las facies, el espesor y la proporción
observados en los perfiles de pozo y testigos corona
puede ser estimada la extensión en subsuelo de los depósitos
de las fajas de canales y de los depósitos de albardón.
La correlación de unidades litoestratigráficas fluviales (por
ejemplo: cuerpos de areniscas de faja de canales) entre pozos
está basada comúnmente en suposiciones insostenibles. Normalmente,
el resultado de estas correlaciones es que las unidades
estratigráficas parecen mucho más continuas lateralmente
de lo que son en realidad. Antes de intentar la correlación,
es conveniente estimar la probable extensión lateral
y las variaciones de espesor de las unidades estratigráficas,
basándose en el análisis de los perfiles de pozo y testigos
corona mencionados anteriormente. A partir de estas estimaciones
es necesario establecer múltiples hipótesis de trabajo
para realizar la correlación. La estimación de la geometría
de las unidades estratigráficas es factible mediante el
uso de análogos actuales, análogos antiguos y modelos
depositacionales. Los estimadores más útiles son los obtenidos
de análogos modernos porque en ellos es posible describir
sin ambigüedades la morfología y los depósitos fluviales. Los análogos antiguos son menos confiables debido a que su
origen debe ser interpretado y porque los afloramientos no
son lo suficientemente extensos como para permitir la definición
de las geometrías de las facies (por ejemplo: ancho
de la faja de canales). Los modelos depositacionales pueden
ser utilizados para ayudar a estimar la extensión lateral de
las unidades estratigráficas (por ejemplo: zonas de alta o
baja relación porcentual entre espesor útil y espesor total
del reservorio, net-to-gross). Sin embargo, tales modelos
pueden ser análogos inapropiados, ya que son solamente 2-
D, no cuantitativos, o no representan correctamente los procesos
y productos depositacionales. En estos análisis, los
modelos de secuencia estratigráfica son particularmente problemáticos.
A fin de caracterizar el volumen de los depósitos entre los
pozos, es común utilizar los análogos antiguos en combinación
con objetos basados en modelos estocásticos. La geometría
y la proporción de los objetos estratigráficos, tales como
cuerpos de areniscas de faja de canal, son determinados
estadísticamente a partir de datos de pozo y análogos antiguos,
estos objetos son ubicados en el espacio utilizando sus
posiciones conocidas en los pozos y un emplazamiento cuasi
aleatorio entre los pozos. Aunque este método señala la
importancia de los datos de pozos, las formas y ubicaciones
de los objetos son muy poco realistas. Los modelos basados
en procesos (ver más adelante) son capaces de producir
estratigrafías mucho más realistas y tienen un mayor valor
predictivo pero es muy difícil adecuar exactamente estos
modelos a los datos de pozos. Por esta razón, los modelos
basados en procesos no son usados rutinariamente en el
modelado de reservorios y acuíferos. Sin embargo, recientemente
ha sido demostrado que potencialmente los modelos
híbridos, entre estocásticos y basados en procesos, pueden
ser ajustados con exactitud a los datos de pozos utilizando
una aproximación por prueba y error con nuevos procedimientos
de optimización (algoritmos genéticos).
Palabras claves: Reservorios y acuíferos fluviales; Perfiles de pozo y testigos corona; Correlación estratigráfica; Modelos de arquitectura aluvial 3-D.
Introduction
Fluvial deposits are important hydrocarbon reservoirs in many parts of the world (e.g., Alaska, Argentina, southern USA and Gulf of Mexico, North Sea, China, Venezuela). Also, fluvial deposits are commonly important aquifers. In order to assess, develop, and manage these resources, it is necessary to determine: (1) the total volume of reservoir or aquifer in a particular volume of the sedimentary basin (the net-to-gross), and; (2) the three-dimensional geometry, orientation and spatial distribution of the sedimentary rocks that form the reservoirs or aquifers. In fluvial deposits, reservoir or aquifer rocks are mainly channel-belt sandstones and gravelstones (conglomerates), with subordinate sandstones attributable to crevasse splays and levees. Determination of the stratigraphic properties (characterization) of a reservoir or aquifer generally entails the following steps: (1) analysis of well logs, cuttings and cores in order to describe and interpret the nature and origin of the rocks; (2) stratigraphic correlation of well logs and cores in order to assess the continuity of distinctive rock types (facies) between wells; (3) use of seismic data to assess the orientation and structural continuity of relatively thick (> 100 m thick) sequences of strata, and to recognize distinctive seismic patterns that can be related to distinctive sedimentary facies; (4) modeling of the geometry and distribution of lithofacies in the three-dimensional space between wells, and; (5) distribution of rock properties such as porosity and permeability as a function of lithofacies or using stochastic models. Successful completion of these steps involves a great deal of geological knowledge, the use of a range of stratigraphic models, and many assumptions that are difficult to test. Some methods of fluvial reservoir/aquifer characterization that are in common use are ill-conceived and unjustifiable. In this review, I will explain and justify this criticism, and attempt to offer constructive alternatives to reservoir/aquifer characterization. The comments in this review also have general implications for describing and interpreting fluvial deposits.
Analysis of well logs and cores
Description and interpretation of distinctive rock types (facies)
Well logs and cores are used to describe or interpret the spatial distribution of sedimentary rock composition, texture, sedimentary structures, porosity and permeability, the nature of the fluids in the void spaces, and the thickness of the different rock types (e.g., sandstone, shale, coal). It is common practice to interpret the specific fluvial depositional environment (e.g., channel bar, crevasse splay, levee, floodbasin) of the various lithofacies or petrofacies using vertical trends in grain size or some other sedimentary or petrophysical property. The objective is to use this interpretation to help estimate the three-dimensional geometry and connectivity of the reservoir rocks and permeability barriers. The assumptions in this practice are: (1) it is possible to interpret specific fluvial depositional environments from vertical logs, and; (2) there is an unambiguous relationship between the geometry and sedimentary processes of a specific depositional environment and the threedimensional geometry of the deposits formed in that environment. Both of these assumptions are suspect, as explained below.
Difficulties interpreting fluvial depositional environments from well logs and cores
A simple view of fluvial deposits, as seen in well logs and cores, is that the thickest (say, meters thick) sandstone-gravelstone bodies are deposits of channel belts, and very thick bodies (say, tens of meters thick) may be superimposed (amalgamated) channel belts. Thinner sandstones bodies (dm to m thick) interbedded with shales might be taken as overbank deposits such as crevasse splays, levees, lacustrine deltas, or fills of floodplain-drainage channels. Thick sequences of shale with minor sandstones would be floodbasin deposits (Fig. 1). However, it is difficult to distinguish: (1) relatively thin parts of isolated channel-belt sandstone bodies (e.g., parts of channel fills, or shallow parts of the channel) from the thicker overbank sandstone bodies; (2) floodbasin muds from muddy channel fills, and; (3) the different types of overbank sandstone bodies (Fig.1).
Figure 1. Example of interpreted gamma ray logs from the Travis Peak Formation, East Texas Basin (from Tye, 1991). Interpretations of
depositional environment were aided by cores from well S.F.E. No.2. The sandstone bodies vary in thickness, and the thicker ones are
in most cases interpreted as channel-belts. The thinner, overbank sandstones were interpreted by Tye (1991) as either crevasse splays or
lacustrine deltas. The sandstone bodies may fine upward, coarsen upward or appear "blocky", irrespective of their thickness and
interpretation. It is possible that some of the overbank sandstones are channel-belt sandstones, and vice versa. Also, it is possible that
some of the mudstone could be deposited in channel fills or as upper-bar deposits within channel belts (Bridge and Tye, 2000). Notice
also that, in correlating the stratigraphic units between wells, the tops and bases of sandstone bodies were not assumed to be horizontal.
Figura 1. Ejemplo de perfiles de rayos gama interpretados para la Formación Travis Peak, Cuenca East Texas (de Tye, 1991). Las
interpretaciones del ambiente depositacional fueron apoyadas por testigos corona del pozo S.F.E. Nº2. Los cuerpos de areniscas varian de
espesor y en la mayoría de los casos los más espesos se interpretaron como fajas de canales. Los cuerpos más delgados, las areniscas de
albardón fueron interpretadas por Tye (1991) como crevasse splays o deltas lacustres. Los cuerpos de arenisca pueden ser granodecreciente,
granocreciente o no presentar cambios (blocky) independientemente de su espesor e interpretación. Es posible que algunas de las
areniscas de albardón sean areniscas de faja de canal y viceversa. Es posible además que algunos de los estratos de pelitas podrían estar
depositados en los rellenos de canal o como depósitos de tope de barra dentro de las fajas de canales (Bridge and Tye, 2000). Note que
también, en las correlaciones de las unidades estratigráficas entre pozos, las bases y techos de los cuerpos de areniscas no fueron tomados
como horizontales.
The folly of trying to interpret paleochannel pattern from well logs and cores
Sandstone bodies interpreted from well logs are commonly classified as fining upward, coarsening upward, or "blocky". Fining-upward sandstone bodies might be interpreted as point-bar deposits or channel-fill deposits of a meandering river, whereas blocky sandstone bodies (no clear vertical trends in grain size) are commonly assigned to braided river bars (e.g., Galloway and Hobday, 1983, 1996; Flores et al., 1985). Coarsening-upward sandstone bodies might be interpreted as deposits of distributary-mouth bars, lacustrine deltas or progradational crevasse splays/levees. The reality is that all of these vertical sequence patterns can form in all of these depositional environments. Figures 2 and 3 show how the vertical variation of grain size, sedimentary structure, and thickness of single channel-belt deposits varies depending on the topographic features of the bars and channels, the position within a channel bar or channel fill, and the nature of channel migration (determining what parts of channel bars are preserved). Fining-upward, coarsening-upward, and "blocky" sequences can occur in any type of channel deposit. Therefore, it is not possible to distinguish the deposits of the different river patterns (braided, meandering) from single vertical logs.
Figure 2. Typical channel deposits, with gamma ray logs, from different parts of sandy channel bars and channel fills (from Bridge and
Tye, 2000).
Figura 2. Depósitos típicos de canal, con perfiles de rayos gama, de distintos sectores de las barras arenosas y de los rellenos de canal
(tomado de Bridge and Tye, 2000).
Figure 3. Topographic features of braided and meandering rivers: curved channels, unit bars, compound bars, cross-bar channels
(modified from Bridge, 1993a).
Figura 3. Rasgos topográficos de ríos enlazados y meandrantes: canales curvados, barras unidad, barras compuestas, canales transversales
en el tope de las barras (modificado de Bridge, 1993a).
Why is it so important to interpret ancient channel patterns such as braided, meandering, and anastomosing anyway? This preoccupation with determining ancient channel pattern is based on the misguided notion that different channel patterns are associated with distinctive sedimentary facies and geometry (e.g., Allen, 1965; Collinson, 1996; Galloway and Hobday, 1996; Miall, 1996; Selley, 1996). For example, meandering river deposits are commonly thought of as relatively fine-grained sandstones with ribbon-like geometries (low width/ thickness) set in voluminous muddy floodplain deposits. In contrast, braided river deposits are commonly thought of as relatively coarse-grained sandstones and/or gravelstones with sheet-like geometry (large width/thickness) with very little associated floodplain deposits. Anastomosing river deposits have been thought of as a connected network of channel sandstone bodies with low width/thickness set in a matrix of floodplain mud. There are some very serious misconceptions in all of this.
Attempts to distinguish meandering and braided river deposits based on their grainsize and geometry
This practice owes its origin to Schumm's (1963, 1971, 1972, 1977, 1981, 1985) classification of channel patterns. Schumm (followed by many sedimentologists such as Galloway and Hobday, 1983, 1996; Miall, 1996) postulated that rivers that transport large amounts of bed load relative to suspended load tend to have relatively low sinuosity and high degree of braiding. Such bed-load streams have been associated with relatively easily eroded banks of sand and gravel, large channel slope and large stream power, such that they are laterally unstable. In contrast, rivers with relatively large suspended loads were postulated to be characteristic of undivided rivers of higher sinuosity. Such suspended-load streams were associated with cohesive muddy banks, low stream gradient and power, and lateral stability. Schumm defined suspended- load channels as those carrying more than 97% suspended-sediment load (presumably at flood stage), whereas bed-load channels are defined as those carrying less than 89% suspended sediment. Such a definition of bed-load channels is misleading to say the least! The correlation between channel pattern, type of sediment load and bank stability is not generally supported by data. This mythical correlation probably arose because early studies of braided rivers were in mountainous areas of sandygravelly outwash and those of single-channel sinuous streams were from temperate lowlands (e.g., the U.S. Great Plains). In fact, many braided rivers are sandy and silty (e.g., Brahmaputra in Bangladesh, Yellow in China, Platte in Nebraska), and many single- channel, sinuous rivers are sandy and gravelly (Madison in Montana, South Esk in Scotland, Yukon in Alaska) (Rust, 1978; Jackson, 1978; Bridge, 1985). The key controls on whether or not a river is braided or meandering are the amounts of water and sediment supplied during seasonal floods (Fig. 4).
Figure 4. Qualitative variation of equilibrium channel patterns
with channel-forming water discharge, valley slope, and sediment
size. Valley slope can be thought of as a measure of sediment
transport rate, because sediment transport rate increases with
increasing slope and water discharge and with decreasing sediment
size.
Figura 4. Variación cualitativa de los diseños de equilibrio de los
canales con respecto al caudal de agua del canal formador, la pendiente
del valle y el tamaño de sedimento. La pendiente del valle
puede ser analizada como una medida de la tasa de transporte de
sedimento debido a que la tasa de transporte de sedimento se
incrementa con el incremento de la pendiente, el caudal de agua y
con el decrecimiento del tamaño del sedimento.
To appreciate the fact that channel pattern may not have a major influence on the geometry of a single channel-belt sandstone body, consider the Mississippi and Brahmaputra Rivers, classic examples of meandering and braided rivers, respectively. Both of these rivers flow through extensive floodplains. The width of the Mississippi River meander belt in the lower Mississippi Valley (Yazoo Basin) is 15 to 25 km, and the maximum thickness of the channel-belt deposits ranges from 20 to 40 m, giving an average channel-belt width/ thickness of about 650 (Bridge, 1999). In the Atchafalaya Basin and delta plain, the Mississippi channel-belt width is 10 to 15 km and the maximum thickness is about 50m, giving a maximum width/ thickness of about 300. The Brahmaputra channel belt is about 10 km wide, and 40 m in maximum thickness, giving a minimum width/thickness of about 250 (Bristow, 1987). It appears that the width/ thickness ratio of the meandering Mississippi channel belt is greater than that of the braided Brahmaputra, or at least comparable. The reason for the similar width/thickness ratios is not hard to understand. A braided river occupies several channels across the width of its channel belt simultaneously. A meandering river may only occupy one channel at a time, but the channel still wanders through a channel belt that may be comparable in width to a braided channel belt of the same thickness.
The significance of planar cross strata
It is commonly stated in the literature (e.g., Collinson, 1996; Miall, 1977, 1996) that planar cross strata produced by transverse unit bars is characteristic of braided rivers, but not meandering rivers. This is quite wrong (review in Bridge, 2002; Fig. 3). Transverse unit bars occur in all alluvial river types. Furthermore, it appears from recent studies that the deposits of most unit bars are composed of medium-scale trough cross strata formed by the dunes that migrate over the unit bars during high flow stages. Planar cross strata associated with migration of a unit bar with an avalanche face apparently only occurs at the margins of unit bars. However, this is a moot point when dealing with well logs or cores, because it is not possible to distinguish planar cross strata from trough cross strata when the sets are dm to m thick (i.e., medium-scale cross strata).
Relationship between discharge variability, channel pattern and deposits
Another common myth (perpetuated by Miall,
1977,1996, and many others) is that discharge
variability is greater for braided rivers than for
meandering rivers. This myth probably originated
from the early studies of pro-glacial braided rivers
in mountainous regions of North America, where
discharge varied tremendously during snowmelt. In
contrast, many single-channel rivers were studied
in temperate lowland regions where discharge
variations were moderated by groundwater supply.
In fact, discharge variability does not have a major
influence on the existence of the different channel
patterns, because all patterns can be formed in laboratory
channels at constant discharge. Moreover,
many rivers with a given discharge regime show
along-stream variations in channel pattern.
However, variability of water and sediment supply during floods does have an influence on the nature
of flood sedimentation units, and these units might
be recognizable in well logs and cores (Fig. 2).
Related to this issue of discharge variability is
the claim that extremes of discharge will have a
strong influence on the nature of the river deposits.
Specifically, the deposits of ephemeral rivers are
considered to have certain distinctive features, such
as a predominance of planar laminated and lowangle
cross-stratified sandstone, and either a lack
of well-defined channels or channels with high
width/depth ratios (e.g., North and Taylor, 1996).
These features are not restricted to ephemeral rivers.
The key features that distinguish deposits of ephemeral
rivers from those of perennial rivers are those
that indicate that the channel flow ceased (i.e., plant
roots and desiccation cracks in the bottom of channels).
In reality, the main characteristics of channel
deposits are controlled by flood conditions, and
floods occur in ephemeral and perennial rivers alike.
Are anastomosing river deposits distinctive?
A distinction has been made between rivers where channels split around bars (braided) and those where channels split around floodplain areas (anastomosing or anabranching) (Lane, 1957; Brice, 1964, 1984; Chitale, 1970; Smith, 1976; Schumm, 1977, 1985; Knighton and Nanson, 1993; Nanson and Knighton, 1996; Makaske, 2001). The characteristic and definitive features of anastomosing (anabranching) channel segments are that they are longer than a curved channel segment around a single braid or point bar, and the patterns of flow and sediment transport in adjacent anastomosing segments are essentially independent of each other. This means that each anastomosing segment contains bars appropriate to the imposed discharge and sediment load, enabling assignment of its own channel pattern, based on degree of channel splitting around bars and sinuosity. This means also that the terms anastomosing and braiding are not mutually exclusive, as implied in the channel-pattern classifications of Rust (1978) and Miall (1992, 1996). In fact, many rivers are braided and/or meandering as well as anastomosing (Fig. 5). The term anastomosing belongs in classifications that describe how channel belts split and join on floodplains, such as distributive and tributive. The number of coexisting anastomosing or distributive channels is controlled by the nature of channel-belt deposition and avulsion rather than the water and sediment discharge in channels (which controls whether channels are meandering or braided). The nature of channel-belt deposition and avulsion has a major control on the superposition of channel belts, the preservation of overbank deposits, and the overall net-to-gross, as discussed below. Whether or not the ancient channel belt was braided or meandering is largely irrelevant when analyzing net-to-gross in the subsurface.
Figure 5. Brahmaputra River immediately north of its confluence
with the Ganges River, showing braided channels and undivided
sinuous channels that are also anastomosed. Scale bar is 20 km.
Photo courtesy of C.S.Bristow, also published in Bridge (1993a).
Figura 5. Río Brahmaputra inmediatamente al norte de su confluencia
con el río Ganges, se observan canales enlazados y
monocanales sinuosos que también son anastomosados. La barra
de escala es 20 km. La fotografía es cortesía de C.S. Bristow, también
publicada en Bridge (1993a).
Superimposed channel bars and channel belts
Channel bars and channel fills can be superimposed within a single channel belt and by superposition of different channel belts (Fig. 6). Recognition of such superposition is not a simple task. Within a single channel belt, simple and compound bars can be superimposed in complicated ways that cannot be interpreted easily from well logs. It is very important to recognize superimposed channel belts in well logs and cores, because superimposed channel belts generally produce sandstone-gravelstone bodies with much larger width/thickness than single channel belts. It is not possible to rationally correlate channel-belt sandstone bodies between wells without assessing their degree of superposition, as discussed below.
Figure 6. Vertical sequences of lithofacies and gamma-ray logs for superimposed channels bars and channel belts (from Bridge and Tye,
2000).
Figura 6. Secuencias verticales de litofacies y perfiles de rayos gama correspondientes a barras de canal superpuestas y fajas de canales
(tomado de Bridge and Tye, 2000).
Spatial variations in proportion of channelbelt deposits (net-to-gross)
The proportion and grain size of channel-belt deposits commonly vary vertically in wells, and vary between wells at a given stratigraphic level. For example, an increase in channel-deposit proportion may be associated with a zone of superimposed channel bars or channel belts. Variations in net-to-gross can occur over different spatial scales. For example, vertical variations in channel-deposit proportion over scales of tens of meters, hundreds of meters and kilometers have been documented (e.g., Willis, 1993a,b; Khan et al., 1997; Zaleha, 1997a,b). Spatial variations in alluvial architecture have been explained by a variety of different kinds of stratigraphic models (e.g., Leeder, 1978; Allen, 1978; Bridge and Leeder, 1979; Paola et al, 1992; Bridge and Mackey, 1993a; Shanley and McCabe, 1993, 1994; Wright and Marriott, 1993; Mackey and Bridge, 1995). There may be many different explanations for a given alluvial architecture, and great care must be taken to keep an open mind about the possible explanatory hypotheses. This is an important point, because choice of any particular interpretation of alluvial architecture has far-reaching implications for lateral correlation of stratigraphic units.
Proposed procedure for interpreting channeldeposits in well logs, image logs, and cores
When interpreting channel deposits in vertical
logs, it is important to estimate maximum and mean
bankfull-channel depth with some confidence,
because these parameters are commonly used to
estimate the width of subsurface channels and
channel belts (Bridge and Tye, 2000). Quantitative
estimation of channel depth from subsurface data
requires three preliminary steps. First, major
channel-belt sandstones and gravelstones must be
distinguished from floodplain sandstones. This is
not always an easy task, because of the gradation
in thickness and facies between these types of
sediment bodies. Second, it is essential to try to
recognize the different scales of channel deposits:
cross sets, flood sedimentation units, unit-bar
deposits, compound-bar deposits, individual
channel belts, and compound channel belts. This
must be done by inspection of spatial variations in
grain size, sedimentary structures, paleocurrents and
degree of disruption. It is very difficult to do this
using data from one well. As shown above, it is easy to confuse flood-sedimentation units with bar
deposits, and to confuse individual channel belts with superimposed channel belts. Third, the
thickness of as many untruncated channel bars or
fills (from the tops of channel belts) as possible must
be measured to get an idea of the range of maximum
channel depths.
Maximum bankfull-channel depth is commonly
estimated from the (decompacted) thickness of
untruncated channel-bar and channel-fill deposits.
It is not always easy to correctly identify untruncated
channel-bar and channel-fill deposits in logs or
cores. Furthermore, the thickness of such channel
sandstones is generally less than the bank-full
channel depth (Bridge and Mackey, 1993b). The
presence of sandy-muddy upper-bar deposits, and
the uncertainty in distinguishing these from proximal
overbank deposits makes it difficult to identify
paleo-bankfull level. Thick channel-fill mudstones
above thin channel-bar sandstones can look very
similar to overbank deposits. Furthermore,
maximum channel depth and bar thickness vary in
space quite markedly (by at least a factor of two)
in channel belts, such that limited data from a single
well may not be representative. An independent
means of estimating bankfull flow depth would be
beneficial. This is possible using an understanding
of the relationship between dune height and
thickness of associated cross sets, and of the known
relationship between dune height and water depth.
The method for calculating the distribution of
dune height from the distribution of cross-set
thickness is described by Bridge (1997), Leclair et
al. (1997) and Leclair and Bridge (2001). The
method is based on the justifiable assumption that
the distribution of cross-set thickness is due
primarily to variability in dune height, and that
variation in deposition rate plays a minor role. This
method is also limited to homogeneous cosets of
cross strata, meaning that there are no obvious
spatial changes in the type of strata or mean grain
size. The assumption is, therefore, that such cross
sets were formed by migration of dunes whose mean
geometry did not vary appreciably in time and
space. To use this method, the thickness, s, of as
many cross sets as possible should be measured,
such that the mean set thickness, sm, can be
calculated. Mean dune height, Hm , is approximately
equal to 3 sm . To avoid confusing cross strata of
dune origin with solitary sets formed by unit bars,
abnormally thick, isolated cross sets should be
avoided. As dune height is expected to vary with
position on channel bars, cross-set thickness
measured in different positions in the vertical profile
should be grouped into subsets. Mean dune height
generally increases with formative flow depth, d. It
appears that, for all types of river dunes (including
those not in equilibrium with the flow), d/Hm averages between 6 and 10. Although estimation
of flow depth from dune height is imprecise, such
an estimate is still a useful complement to flow
depth calculated from channel-bar thickness.
Re-interpretation of the geometry and spatial distribution of subsurface fluvial sandstone bodies
There are many examples in the literature of interpretation of river-channel deposits using data from cores and well logs (e.g., various papers in the volumes edited by Barwis et al., 1990; Ethridge and Flores, 1981; Ethridge et al., 1987; Galloway and Hobday, 1996; Lomando and Harris, 1988; Miall and Tyler, 1991; Selley, 1996). In most examples, a range of possible interpretations of well logs is not considered, and the cores are not described in sufficient detail to allow use of the methods proposed here. In particular, the thickness of medium-scale cross sets is rarely presented. However, Bridge and Tye (2000) demonstrated how application of the new techniques discussed here had an impact upon previous interpretations of paleochannel depths, channel-belt widths, and fluvial sandstone-body dimensions and connectedness.
Stratigraphic correlation of well logs and cores: use of seismic data
Method of correlation
Correlation of lithofacies between wells is normally accomplished within a framework of regional stratigraphic markers established using seismic sections, well logs and cores. In order to facilitate correlation, it is common practice to plot the stratigraphic markers as horizontal surfaces. Lithostratigraphic correlation must entail certain assumptions about, or models of, the geometry and lateral continuity of lithofacies such as channel-belt sandstone bodies or floodplain shales (Fig. 1, 7 and 8). For example, must the tops and bases of correlated sandstone bodies be horizontal and parallel? If they are not horizontal and parallel, how inclined can the boundaries be relative to the datum, and how variable can the thickness of a sandstone body be? What is a reasonable lateral extent in any given direction? These are very difficult questions to answer. In many cases, the assumptions made in correlation are either not stated explicitly, or are based on various kinds of stratigraphic models. If correlating stratigraphic sequences that are less than the order of 100 m thick, seismic data will be of limited use in resolving the lateral continuity of strata and the presence of small faults. Stratigraphic resolution using seismic data improves as interval thickness and impedance contrast increase. Therefore, an infinite number of correlations is possible, but only one will be used in practice.
Figure 7. Different ways of correlating rock units between two
wells, depending on assumptions about the physical continuity
and orientation of unit boundaries.
Figura 7. Diferentes formas de correlacionar unidades de roca
entre dos pozos, dependiendo de las suposiciones de la continuidad
física y de la orientación de los límites de la unidad.
Figure 8. Correlation of two wells within the Ness Formation of
the Brent Field, UK North Sea (from Bryant and Flint, 1993). GR =
Gamma Ray; FDC = Density Log; CNL = Compensated Neutron
Log. Coals have been used as correlation markers. There has been
some imaginative definitions of the shapes and lateral extents of
channel sands and crevasse splay sands.
Figura 8. Correlación de dos pozos en la Formación Ness del Distrito
Brent, Mar del Norte-Inglaterra (de Bryant and Flint, 1993). GR
= Rayos Gama; FDC = Perfil de Densidad; CNL = Perfil Compensado
de Neutrones. Las capas de carbón han sido utilizadas como
niveles de correlación. Algunas de las definiciones de las formas y
extensiones laterales de las arenas de canal y de las arenas de crevasse splay han sido imaginativas.
Using well-to-well correlation to estimate lateral extent of reservoirs or aquifers
It is ironic, but not surprising, that the most
common method for estimating channel-belt widths
and orientations is by correlating specific channelbelt
sandstone bodies between well logs (e.g., Nanz,
1954; Berg, 1968; Cornish, 1984; Tye, 1991). With
this technique, there is generally no attempt to
predict reasonable geometries and lateral extents
of different lithofacies from analysis of well logs.
The minimum possible width of sediment bodies
that can be resolved by this method is the well
spacing. This approach to estimating lateral extent lithofacies is commonly compromised by
simplistic or erroneous assumptions, such as: (1)
basal erosion surfaces and tops of channel-belt
sandstone bodies are flat; (2) sandstone bodies
positioned at the same stratigraphic level must be
connected between adjacent wells; (3) sandstone
body width/thickness ratios are closely related to
paleochannel pattern, and; (4) vertical sequences through channel deposits indicate the paleochannel
pattern and hence the geometry of channel-belt
sandstone bodies.
Concerning assumption (1), the depositional
models discussed above clearly show that the basal
erosion surfaces and tops of channel belts are not
generally flat. Concerning assumption (2), sandstone
bodies at the same stratigraphic level in
adjacent wells are not necessarily connected, and
some assessment of the probability of connection
is required, perhaps using empirical data on channelbelt
width/maximum channel depth. However, if
two sandstone bodies are indeed continuous
between wells, they are not necessarily from a single
channel belt. This is of particular concern if
sandstone-body proportion exceeds 0.4 (Bridge and
Mackey, 1993b). Assumptions (3) and (4) were
shown to be wrong above.
In order to rationalize the correlation process
as much as possible, it is desirable to use
interpretations of depositional environments and
stratigraphic models to constrain the correlation
assumptions. In the case of isolated channel belts,
it is unlikely that the tops and bases will be horizontal
and parallel. Data from modern and ancient analogs can be used to assess these characteristics.
Models and real-world examples of channel belts
indicate that their thickness may vary laterally by
a factor of at least two, and that there are systematic
lateral variations in thickness. A range of likely
channel-belt widths can be obtained from the
thickness of single channel belts, as indicated
below. It may turn out that sandstone bodies at the
same stratigraphic level should not be correlated
over the distance between two wells if this distance
greatly exceeds the expected lateral extent of a
channel belt. There needs to be much more work
done on the geometry of different lithofacies in
modern fluvial environments, if modern analogs are
to be used to predict the geometry of subsurface
fluvial sediment bodies.
Geometry of modern channel-belt deposits
Several attempts have been made to predict
channel-belt width in the subsurface using empirical
equations derived from modern rivers that relate
maximum channel depth, channel width, and
channel-belt width (Collinson, 1978; Lorenz et al.,
1985, 1991; Fielding and Crane, 1987; Davies et
al., 1993; discussed by Bridge and Mackey, 1993b).
This approach requires reliable estimates of
maximum bankfull-channel depth from one-dimensional subsurface data. As discussed previously,
correct estimation of maximum bankfull-channel
depth is not straightforward, because complete
channel-bar or channel-fill sequences may be
difficult to identify, and the thickness of the
sandstone-gravelstone parts of these coarse
members is not always as great as the bankfullchannel
depth. Furthermore, within a single channel
belt, maximum channel depth and bar thickness can
vary spatially by a factor of at least 2 (Bridge, 1993a;
Salter, 1993). Therefore, limited data from a single
well may not be representative.
Empirical regression equations relating
maximum channel depth, channel width, and
channel-belt width have large standard errors, and
are dependent on channel-pattern parameters such
as channel-bend wavelength and sinuosity (Bridge
and Mackey, 1993b). Most of the empirical
equations used to date (Collinson, 1978; Lorenz et
al., 1985, 1991; Fielding and Crane, 1987; Davies
et al., 1993) do not include such dependencies.
Channel-bend wavelength and sinuosity are actually
very difficult to reconstruct from outcrops and are
impossible to determine from well data. Lorenz et
al. (1985) used empirical equations derived from
rivers with sinuosity greater than 1.7, and assumed
that sandstone bodies in both outcrops and
subsurface strata were deposited in highly sinuous
rivers. Equations presented in Bridge and Mackey
(1993b: their Table 2) are based on larger data sets
than previous equations or on theoretical principles.
Channel-belt widths predicted by these equations
agreed with those observed in outcrops by Bridge
et al. (2000).
Use of outcrop analogs to aid well-to-wellcorrelation
The geometry and lithofacies of channel and channel-belt deposits determined in outcrops are commonly used as analogs for subsurface strata (Collinson, 1978; Walderhaug and Mjos, 1991; Lowry and Raheim, 1991; Cuevas Gozalo and Martinius, 1993; Dreyer, 1993; Dreyer et al., 1993; Robinson and McCabe, 1997; Bridge et al., 2000). Despite the popularity of this approach, it has many pitfalls. First, it must be established that the depositional setting interpreted for the subsurface strata is indeed analogous to that interpreted for the outcrops. This requires two interpretation steps, the reliability of which depends on the quality of the outcrops, of the subsurface data, and of the depositional models used for interpretation. It is difficult to make detailed interpretations of depositional environments using typical subsurface data. Because our understanding of modern depositional environments is incomplete, most depositional models are qualitative, lacking in detail, and not fully three-dimensional (Bridge, 1985, 1993b). The deficiency of most depositional models severely limits their use in detailed interpretation of ancient deposits. Rarely are outcrops extensive enough to allow unambiguous determination of the three-dimensional geometry and orientation of channels and channel belts. Channel-belt width is particularly difficult to determine, even in large outcrops (Geehan and Underwood, 1993). Moreover, channel-belt width and thickness are known to vary spatially within one channel belt and between different channel belts. In general, limited data from a few large exposures are unlikely to be generally representative of fluvial-deltaic channel belts. This is why it is desirable to use analog data from Holocene depositional environments, where channel-belt dimensions can be determined easily, and the relationship between the nature of the deposits and the geometry, flow and sedimentary processes of the environment can be established unambiguously.
Amplitude analysis of 3-D seismic horizon slices
Amplitude analysis of 3-D seismic horizon slices (Weber, 1993; Hardage et al., 1994, 1996; Burnett, 1996) is the only method capable of yielding directly the width of channel belts, and imaging the channel pattern (sinuosity, channel splitting) of subsurface sandstone bodies (Fig. 9). This is also the only method that can be used to predict the spatial distribution of channel-belt thickness and lithofacies. However, high quality 3-D seismic data are rarely available, the most sandstone bodies of interest are too thin and buried too deeply to be completely resolved using seismic data.
Figure 9. Amplitude analysis of a 3-D seismic horizon slice showing the width of a channel belt and channel pattern of subsurface channel sandstone bodies. The cross section shows correlated logs and the position of the horizon slice. Log 3 cuts through the variablewidth, straight channel belt trending north-south. Logs 4 to 6 cut through a point bar and channel fill of a slightly older channel belt.
Figura 9. Análisis de amplitud de una sección horizontal de sísmica 3-D mostrando el ancho de la faja de canales y el diseño de los canales de los cuerpos de areniscas de canal en subsuelo. Las secciones transversales muestran los perfiles correlacionados y la posición de la sección horizontal. El perfil 3 atraviesa longitudinalmente la faja de canal de ancho variable en dirección norte-sur. Los perfiles 4 y 6 atraviesan una barra de punta y el relleno de canal de una faja de canal un poco más antigua.
Estimating width of superimposed cannel belts from subsurface data
When correlating channel-belt sandstone bodies between wells, it is critical to be able to distinguish connected channel belts. Two connected channel belts may be up to twice the width of a single channel belt. Bridge and Mackey (1993b) used their revised two-dimensional model of alluvial architecture (Bridge and Mackey, 1993a) to study the width and thickness of sandstone (or gravelstone) bodies comprising single or connected channel belts. The width and thickness of sandstone bodies depend critically on the proportion of channel-belt deposits in the stratigraphic interval, as this proportion controls the degree of connectedness of individual channel belts. For low values of channel-deposit proportion (less than about 0.4), channel belts are unconnected, sandstone-body width equals channel-belt width, and sandstone-body thickness equals aggraded channel-belt thickness (Fig. 10). As channel-deposit proportion increases, some channel belts become connected, the mean and standard deviation of sandstone-body width and thickness increase, and their frequency distributions become polymodal with the largest modes at the lowest values of width and thickness. As channel-deposit proportion continues to increase, the distributions are still polymodal but the larger modes are at higher values of width and thickness (Fig. 10). If channel-deposit proportion exceeds about 0.75, all channel belts are connected, and the single sandstone body has a width equal to floodplain width and a thickness equivalent to the whole section.
Figure 10. Relationships between channel-belt deposit proportion and connectedness, and the width and thickness of channel-belt sandstone bodies (based on Bridge and Mackey, 1993b).
Figura 10. Relaciones entre la proporción de los depósitos de la faja de canales, la interconexión, el ancho y el espesor de los cuerpos de areniscas de la faja de canales (basado en Bridge and Mackey, 1993b).
Channel-deposit proportion, width and
thickness increase as bank-full channel depth and
channel-belt width increase, and as floodplain
width, aggradation rate and avulsion period
decrease. Channel-deposit proportion is also
influenced by depth of burial (degree of
compaction) and tectonic tilting of the floodplain.
For example, channel-deposit proportion and
connectedness of channel belts increase on the
down-tilted sides of floodplains but are reduced on
the up-tilted sides.
The three-dimensional model of alluvial
architecture of Mackey and Bridge (1995) gives
results that are generally similar to those described
above for two-dimensional models. However, the
three-dimensional model predicts that channeldeposit
proportion and connectedness, and the
dimensions of channel sandstone bodies, vary with
distance from points of channel-belt splitting (due
to avulsion). Upstream of avulsion points,
sandstone bodies have lower than average width/
thickness because of aggradation in a fixed channel
belt. Immediately down valley from avulsion points,
channel belts are connected, resulting in sandstone
bodies with higher than average width/thickness.
Because of this, relationships between channeldeposit
proportion and connectedness and
sandstone-body dimensions derived from 2-D models are strictly applicable only to parts of the
floodplain located some distance down valley from
avulsion points.
Correlation using sequence-stratigraphicmodels
Correlation of lithofacies associations between
wells may be aided by other kinds of stratigraphic
models, such as sequence-stratigraphic models
(Fig.11). Sequence-stratigraphic models (e.g.,
Jervey, 1988; Posamentier and Vail, 1988;
Posamentier et al., 1988; Ross, 1990; Shanley and
McCabe, 1993, 1994; Wright and Marriott, 1993)
predict the effects of relative sea-level change on
alluvial deposition rate and alluvial architecture, and
are therefore potentially applicable to near-coastal
fluvial successions. There are substantial differences
between these models, and all of them can be
criticized. Most of these models are severely limited
because they are qualitative, essentially two-dimensional,
and do not account for all of the factors that
influence alluvial architecture. Actually, eustatic
sea-level change may not have the influence on
alluvial systems that has been assumed in the past.
Factors such as climate, vegetation and tectonism
may also have an important influence on rivers and
floodplains during eustatic sea-level change
(Schumm, 1993; Wescott, 1993; Blum, 1994; Koss
et al., 1994; Leeder and Stewart, 1996; Miall, 1996;
Blum and Price, 1998; Ethridge et al., 1998; Blum
and Tornqvist, 2000). Miall (1986, 1991, 1996) has
criticized some of the earlier models (e.g., Jervey,
1988; Posamentier et al., 1988) for the effects of
sea-level change on near-coastal alluvial deposition.
His main point is that a relative fall in sea level is
not normally associated with alluvial aggradation,
except for the newly exposed part of the sea bed,
and even then only under special circumstances. In
fact, whether or not a river valley is incised or
aggraded during sea-level fall depends, among other
things, on the slope of the exposed shelf relative to
that of the river valley (Pitman, 1986; Pitman and
Golovchenko, 1988; Schumm, 1993; Leeder and
Stewart, 1996). In general, the effects of sea-level
change are expected to decrease up-valley (Saucier,
1981, 1994; Autin et al., 1991; Schumm, 1993;
Shanley and McCabe, 1994). Only three-dimensional
models can describe these effects (Bridge, 1999:
Karssenberg et al, 2001b).
Most alluvial sequence-stratigraphic models
have an erosional base to the sequence (Shanley
and McCabe, 1993, 1994; Wright and Marriott,
1993; Gibling and Bird, 1994; Miall, 1996; Fig. 11).
This is ascribed to river incision arising from relative
fall in base level and exposure of a slope that is
steeper than the river slope upstream. Although
incision may be associated with base-level fall near
the coast, incision further inland may not be
associated with falling base level (Blum, 1994; Blum
and Price, 1998: Blum and Tornqvist, 2000).
Therefore, the basal erosion surface of the sequence
may not be coeval with, or coincident with, erosion
surfaces inland. The erosional base of the sequence
in these models is overlain by amalgamated fluvial
channel deposits (the so-called lowstand systems
tract: LST) that are ascribed to deposition under
conditions of low deposition rate and restricted
floodplain width (due to valley incision). Many
workers assume that zones of high channel-deposit
proportion (high net-to-gross) in alluvial deposits
represent these basal parts of sequences, and that
the basal erosion surface of the lowest sandstone
body represents an incised valley (e.g., Aitken and
Flint, 1995). In many cases, evidence for an incised
valley is lacking. Criteria for incised valleys include
(Dalrymple et al., 1994): (1) erosional relief that is
greater than the thickness of a single channel fill;
(2) multiple, vertically stacked channel bars within
the valley; (3) evidence for extended periods of non
deposition (mature paleosols) on interfluves; (4)
alluvial channel deposits resting erosively upon
shallow marine sands and muds. Commonly, a large
amount of erosional relief on the base of a single
channel deposit can be misinterpreted as an incised
valley margin (Salter, 1993; Best and Ashworth,
1997).
Falling sea level (marine regression) does not
necessarily result in valley incision near the coast.
The land exposed by marine regression may
experience erosion or deposition depending on the
slope of the exposed surfaces (Schumm, 1993;
Wescott, 1993; Wood et al., 1993). Erosion will
occur on relatively steep slopes given enough time
and flow competence. However, erosion of channels
and floodplains is a long-term process, and the shortterm
response may be to increase sinuosity of rivers.
Also, upstream avulsion may cause channel
abandonment before incision is complete (Leeder
and Stewart, 1996). Incision of channels and
floodplains is associated with terrace formation,
reduction of valley width, and up-valley migration
of knick-points. Avulsion frequency is expected to
be low in areas of erosion. Channel-deposit
proportion and connectedness may increase during
sea-level lowstand because of reduced deposition rate (or erosion) and reduced floodplain width.
However, channel-deposit proportion may decrease
as a result of reduced avulsion frequency, or if the
channel-belt width and thickness were reduced as
a result of a decrease in water discharge. Sequence
stratigraphic models must account for changes in
all of these factors that influence alluvial
architecture.
Figure 11. Examples of alluvial sequence-stratigraphic models (upper model from Shanley and McCabe, 1993; lower model from Wright and Marriott, 1993).
Figura 11. Ejemplos de modelos de secuencia estratigráfica aluvial (modelo superior de Shanley and McCabe, 1993; modelo inferior de Wright and Marriot, 1993).
Thick and laterally extensive amalgamated
channel deposits are commonly important oil and
gas reservoirs. Therefore, it is important to interpret
such deposits accurately when predicting their
thickness, lateral extent, and bounding facies. If
zones of high channel-deposit proportion are
incorrectly interpreted as incised valley fills, their
extent normal to the valley direction will be
underestimated, and their extent parallel to the
valley will be overestimated. There are many
examples in the literature where zones of high
channel-deposit proportion are not associated with
filling of incised valleys. For example, they may be
associated with high deposition rates and high
avulsion frequencies of large channels on megafans
(Willis, 1993a,b; Khan et al, 1997; Zaleha, 1997a,b).
In this case, their lateral extent may be considerable.
According to the sequence-stratigraphic
models, the deposits above the lowstand systems
tract were deposited under conditions of relatively
high deposition rate on a broad alluvial plain,
associated with rising relative base level and
drowning of incised valleys (the so-called
transgressive systems tract: TST). The channeldeposit
proportion and connectedness are taken to
be relatively low as a result. According to Gibling
and Bird (1994), coal is likely to occur at the top of
the TST, immediately below the so-called maximumflooding
surface. Paleosols are likely to reflect high
groundwater table. Deposits associated with the
maximum-flooding surface may contain evidence
of marine influence.
During relative sea-level rise, slopes of rivers
and floodplains are reduced near shore due to
backwater effects, and the width and depth of
valleys increase due to drowning. These changes
result in reduced grain size of transported sediment,
deposition, change in channel pattern, and increased
frequency of avulsion. Tornqvist (1993, 1994)
associated high avulsion frequency and
anastomosing river patterns with periods of rapid
base-level rise and deposition. Increases in avulsion
frequency increase channel-deposit proportion and
connectedness, but increases in deposition rate and
width of floodplains decrease them. It is also
necessary to know how channel-belt geometry
changes during sea-level change in order to
adequately predict channel-deposit proportion. In
the case of the Mississippi coastal plain, the channeldeposit
proportion in the Holocene deposits is low
mainly because of high floodplain width relative to
channel-belt width. However, in the coeval deposits
of the lower Mississippi Valley, the channel-deposit
proportion is very high because of low floodplain
width relative to channel-belt width (Bridge, 1999).
The highstand systems tract (HST) is also
associated with relatively low channel-deposit
proportion according to Shanley and McCabe
(1993). However, Wright and Marriott (1993)
predict an increase in channel-deposit proportion
in the HST, and an increase in soil maturity, both
related to reduced deposition rate.
Based on the comments above, it is unlikely
that extant alluvial sequence-stratigraphic models
are generally applicable. Karssenberg et al (2001b)
have attempted to improve upon this situation by
producing a 3-D model that takes account of most
of the factors that control alluvial architecture
during changing base level.
Modeling of the three-dimensional Geometry and distribution of Facies between wells
Information derived from seismic profiles, cores, well logs, and well tests is rarely sufficient to provide comprehensive three-dimensional description and understanding of subsurface reservoir rocks. As a result, recourse is commonly made to outcrop analogs and quantitative depositional models. A common approach is to use interpreted outcrop analogs (discussed above) to provide supplementary data on sedimentary architecture, and to use stochastic, structure-imitating models conditioned by subsurface data to distribute the sediment types in 3-D space (reviews by Bryant and Flint, 1993; Koltermann and Gorelick, 1996; North, 1996; Anderson, 1997: Fig.12). Structure-imitating stochastic models do not simulate processes of deposition: they directly simulate the stratigraphy. Methods used are indicator geostatistics (e.g., Journel, 1983; Bierkens and Weerts, 1994), simulated annealing (e.g., Deutsch and Cockerham, 1994), Markov chains (e.g., Doveton, 1994, Carle et al., 1998), and Boolean object models that use probabilistic rules to define the geometry and location of stratigraphic units (e.g., Budding et al., 1992; Deutsch and Wang, 1996; Hirst et al., 1993; Holden et al., 1998). An advantage of these models is that they match the available data exactly. However, input parameters are very difficult to obtain, and the alluvial stratigraphy simulated is commonly very unrealistic (e.g., Tyler at al., 1994; Deutsch and Wang 1996; Holden et al., 1998). In the case of object-based models, the lack of realism is associated with the choices available for object shapes, and the essentially random placement of objects where well data are not available. For example, channel belts are modeled as one or more sinuous ribbons with channel-shaped bases (Fig. 13).
Figure 12. An example of use of a two-dimensional, object-based stochastic model (from Srivastava, 1994). The width and thickness of
channel-belt sandstone bodies are normally obtained from Monte Carlo sampling from empirical distributions derived from outcrop
analogs. Notice that the "sand bodies" have been given a channel-shaped form. Channel-belt sandstone bodies do not normally have this
form.
Figura 12. Un ejemplo del uso de un modelo bidimensional estocástico basado en objetos (de Srivastava, 1994). El ancho y el espesor
de los cuerpos de areniscas de faja de canales son obtenidos del muestreo Monte Carlo a partir de las distribuciones empíricas derivadas
de análogos de afloramiento. Note que los «cuerpos de arena» han sido adjudicados a una forma de canal. Los cuerpos de areniscas de faja
de canales normalmente no tienen esta forma.
Figure 13. (A) Example from the MOHERES software of a single channel belt consisting of four "channel beds" (from Tyler et al., 1994).
In this software, internal elements such as channel bars and channel fills can be placed randomly within these "channel beds". This is
not a realistic representation of the geometry of deposits in channel belts. Notice that the "channel beds" are not even in contact with
eachother in places. (B) Example of a channel belt from Patagonia showing a single active channel and many abandoned channel bars and
channel fills. The channel-bar deposits comprise most of the volume of this channel belt.
Figura 13. (A) Ejemplo del programa MOHERES consistente en una faja de canales simple con cuatro "lechos de canales" (de Tyler et al.,
1994). En este programa, los elementos internos tales como barras de canal y relleno de canales pueden estar ubicados al azar dentro de
los "lechos de canales". Esta no es una representación relista de la geometría de los depósitos en las fajas de canales. Note que los "lechos
de canales" no están en contacto entre sí en otros sectores (figura inferior). (B) Ejemplo de una faja de canales de Patagonia mostrando
un canal activo simple y muchas barras de canal abandonadas y rellenos de canal. Los depósitos de barras de canal conforman la mayor
parte del volumen de esta faja de canales.
Unlike the structure-imitating models, process based models (sometimes referred to as processimitating
models) simulate the sedimentary
processes acting to produce a deposit (Koltermann
and Gorelick, 1996; Anderson, 1997). Process-based
models can be deterministic and/or stochastic, and
empirical and/or theoretical. Examples of such
models include random-walk sedimentation models
of braided rivers (Webb, 1994), models based on
the fundamental equations of fluid flow and
sediment transport (e.g., Bridge, 1977, 1992; Tetzlaff
and Harbaugh, 1989; Stam, 1996), and avulsionbased
alluvial stratigraphy models (e.g., Bridge and
Leeder, 1979; Mackey and Bridge, 1995; Heller and
Paola, 1996). Process-based models are forward
models in the sense that they predict the nature of
deposits given a set of initial starting parameters. It
is not known a priori what the deposits will look
like. Advantages of process-based models are that they utilize fundamental physical principles to
generate stratigraphy. Therefore, process-based
models can help provide genetic interpretations of
deposits, and can predict more realistic stratigraphy
than structure-imitating (stochastic) models.
However, a perceived disadvantage of process-based
models is that it is difficult or impossible to make
the simulated deposits fit observational data in
sufficient detail (Clemetsen et al., 1990; North,
1996; Koltermann and Gorelick, 1996; Anderson,
1997). Therefore, process-based models have had
limited application in quantitative simulation of
reservoir stratigraphy.
Recent work shows, however, that fitting of
process-based models to well data is possible in
principle, using a trial-and-error approach with some
optimization (Karssenberg et al., 2001a).
Karssenburg et al. (2001a) fitted a simplified version
of the Mackey-Bridge (1995) three-dimensional
model of alluvial stratigraphy to five hypothetical
wells (Fig. 14). The approach is to run the model
many times, each time with different input defined
using Monte Carlo simulation techniques.
Successful runs are those where output fits well data
within certain tolerance limits. Figure 15 shows the
temporal evolution of the floodplain for one run of
the process-based model with the simulated
deposits fitting the five wells. On the basis of all of
the realizations of the model that fit the wells, the
probability of occurrence of channel-belt deposits
within the area can be calculated (Fig. 16).
Figure 14. (A) Hypothetical model area with location of five wells
(B) Hypothetical well data used for model fitting (from Karssenberg et al., 2001a).
Figura 14. (A) Área hipotética del modelo con la ubicación de
cinco pozos, (B) Datos hipotéticos del pozo usados para ajustar el
modelo (de Karssenberg et al., 2001a).
Figure 15. Model run that fits the well data. Each map shows position of the old and new channel belt, and the surface elevation (m) at the time of an avulsion (from Karssenberg et al., 2001a).
Figura 15. Modelo simulado que se ajusta a los datos de los pozos. Cada mapa muestra la posición de la faja de canales antigua y nueva, y la elevación de la superficie (m) en el momento de una avulsión (de Karssenberg et al., 2001a).
Figure 16. 3-D image of probability of occurrence of channel belt deposits based on model runs that fit the well data. Top figure shows
volume with probability > 0.8. Bottom diagram shows transects through the 3-D block, with grayscale representing probability of
occurrence of channel-belt deposits (from Karssenberg et al., 2001a).
Figura 16. Imagen 3-D de probabilidad de ocurrencia de los depósitos de faja de canales basada en los modelos simulados que se ajustan
a los datos de los pozos. La figura en la parte superior muestra el volumen con probabilidad > 0,8. El diagrama de la base muestra las
transectas a través del bloque 3-D, la escala de grises representa la probabilidad de ocurrencia de los depósitos de la faja de canales (de
Karssenberg et al., 2001a).
It was concluded that using this trial-and-error approach in practice will require: (1) refinement of the process-based model; (2) improved algorithms for efficiently fitting the model to the data, and; (3) overcoming the anticipated large amounts of computing time. Recent progress on these three problems has facilitated the implementation of this inverse modeling approach and its application to real-world data. Moreover, the effectiveness of process-based models in simulating alluvial stratigraphy can now be compared with stochastic, structure-imitating models. The 3-D process-based model can be fitted to well data exactly by inverse modeling (Cross and Lessenger, 1999; Bornholdt et al, 1999; Karssenberg et al., 2001a). With inverse modeling, fitting of the model to well data is achieved by repeatedly running the model, comparing its output with well data, and adjusting the model input parameters until the model output fits the well data within specified tolerance limits. For each model run, a fitness function (called objective function or goal function by others) expresses the difference between model output and well data. The objective is to minimize the difference between the model output and well data, or to maximize the fitness function. Since the fitting procedure will put a high demand on computational resources, the choice of the fitting method and the software implementation of this method and the process-based model is critical in order to minimize run times.
Acknowledgements
Sincere thanks to my friend and colleague Bo Tye for many discussions of the issues raised in this review. Bo also generously reviewed the manuscript. Sincere thanks also to another friend and colleague, Sergio Georgieff, who invited me to write this paper, and who was responsible for the Spanish translations.
References
1. Aitken, J.F. and S. S. Flint, 1995. The application of highresolution sequence stratigraphy to fluvial systems: a case study from the Upper Carboniferous Breathitt Group, eastern Kentucky, USA. Sedimentology 42: 3-30.
2. Allen, J.R.L., 1965. A review of the origin and characteristics of recent alluvial sediments. Sedimentology 5: 89-191.
3. Allen, J.R.L., 1978. Studies in fluviatile sedimentation: an exploratory quantitative model for architecture of avulsion-controlled alluvial suites. Sedimentary Geology 21: 129-147.
4. Anderson, M.P., 1997, Characterization of geological heterogeneity. In: G. Dagan, and S.P. Neuman (Eds.), Subsurface Flow and Transport: A Stochastic Approach. Cambridge University Press, Cambridge, 23-43.
5. Autin, W.J., S.F. Burns, R.T. Miller, R. T. Saucier and J. I. Snead, 1991. Quaternary Geology of the lower Mississippi valley. In: R. B. Morrison (Ed.), Quaternary nonglacial geology: conterminous U.S. The Geology of North America. Geological Society of America K-2: 547-582.
6. Barwis, J. H., J. G. Mcpherson and R. J. Studlick, 1990. Sandstone Petroleum Reservoirs. Springer-Verlag, New York, 583 pp.
7. Berg, R.R., 1968. Point-bar origin of Fall River Sandstone reservoirs. American Association of Petroleum Geologists Bulletin 52: 2116-2122.
8. Best, J.L. and P. J. Ashworth, 1997. Scour in large braided rivers and the recognition of sequence stratigraphic boundaries. Nature 387: 275-277.
9. Bierkens, M.F.P. and H.J.T Weerts, 1994. Application of indicator simulation to modelling the lithological properties of a complex confining layer. Geoderma 62: 265-284.
10. Blum, M.D. and D.M. Price, 1998. Quaternary alluvial plain construction in response to glacio-eustatic and climatic controls, Texas Gulf Coastal Plain. In: Shanley, K.J. and P. J. McCabe (Eds.), Relative Role of Eustasy, Climate and Tectonism on Continental Rocks. Society of Economic Paleontologists and Mineralogists Special Publication 59: 31-48.
11. Blum, M.D., 1994. Genesis and architecture of incised valley fill sequences: a late Quaternary example from the Colorado River, Gulf Coastal Plain of Texas. In: P. Weimer and H.W. Posamentier (Eds.), Siliciclastic sequence stratigraphy. American Association of Petroleum Geologists Memoir 58: 259-283.
12. Blum, M.D. and T.E. Törnqvist, 2000. Fluvial responses to climate and sea-level change: a review and look forward. Sedimentology 47 (Suppl.1): 2-48.
13. Bornholdt, S., U. Nordlund and H. Westphal, 1999. Inverse stratigraphic modeling using genetic algorithms. In: Harbaugh, J.W. et al. (Eds.), Numerical Experiments in Stratigraphy: Recent Advances in Stratigraphic and Sedimentologic Computer Simulations. Society of Economic Paleontologists and Mineralogists Special Publication 62: 85-90.
14. Brice, J.C., 1964. Channel patterns and terraces of the Loup Rivers in Nebraska. U.S. Geological Survey ProfessionalPaper 422D: 1-41.
15. Brice, J.C., 1984. Planform properties of meandering rivers. In: C.M. Elliot (Ed.), River Meandering. American Society of Civil Engineers, 1-15.
16. Bridge, J.S., 1977. Flow, bed topography, grain size and sedimentary structure in open channel bends: a threedimensional model. Earth Surface Processes 2: 401-416.
17. Bridge, J.S., 1985. Paleochannel patterns inferred from alluvial deposits: A critical evaluation. Journal of Sedimentary Petrology 55: 579-589.
18. Bridge, J.S., 1992. A revised model for water flow, sediment transport, bed topography and grain size sorting in natural river bends. Water Resources Research 28: 999-1023.
19. Bridge, J.S., 1993a. The interaction between channel geometry, water flow, sediment transport and deposition in braided rivers. In: J.L. Best and C.S. Bristow (Eds.), Braided Rivers. Geological Society of London, Special Publication 75: 13-72.
20. Bridge, J.S., 1993b. Description and interpretation of fluvial deposits: a critical perspective. Sedimentology 40: 801- 810.
21. Bridge, J.S., 1997. Thickness of sets of cross strata and planar strata as a function of formative bed-wave geometry and migration, and aggradation rate. Geology 25: 971-974.
22. Bridge, J. S., 1999, Alluvial architecture of the Mississippi Valley: predictions using a 3D simulation model. In: S.B. Marriott and J. Alexander (Eds.), Floodplains: Interdisciplinary Approaches. Geological Society of London, Special Publication 163: 269-278.
23. Bridge, J.S., 2002. Alluvial Rivers and Floodplains. Blackwells, Oxford
24. Bridge, J.S., G. A. Jalfin and S. M. Georgieff, 2000. Geometry, lithofacies and spatial distribution of Cretaceous fluvial sandstone bodies, San Jorge Basin, Argentina: outcrop analog for the hydrocarbon-bearing Chubut Group. Journal of Sedimentary Research 70: 341-359.
25. Bridge, J.S. and M. R. Leeder, 1979. A simulation model of alluvial stratigraphy. Sedimentology 26: 617-644.
26. Bridge, J.S. and S. D. Mackey, 1993a. A revised alluvial stratigraphy model. In: Marzo, M. and C. Puidefabregas (Eds.), Alluvial sedimentation. International Association of Sedimentologists Special Publication 17: 319-337.
27. Bridge, J.S. and S, D. Mackey, 1993b. A theoretical study of fluvial sandstone body dimensions. In: S.S. Flint and I.D. Bryant (Eds.), Geological modeling of hydrocarbon reservoirs. International Association of Sedimentologists Special Publication 15: 213-236.
28. Bridge, J.S. and R. S. Tye, 2000. Interpreting the dimensions of ancient fluvial channel bars, channels, and channel belts from wireline-logs and cores. American Association of Petroleum Geologists Bulletin 84: 1205-1228.
29. Bristow, C.S., 1987. Brahmaputra River: Channel migration and deposition. In: Ethridge, F.G., Flores, R. M., and M. D. Harvey (Eds.), Recent developments in fluvial sedimentology. Society of Economic Paleontologists and Mineralogists Special Publication 39: 63-74.
30. Bryant, I.D., and S. Flint, 1993. Quantitative clastic reservoir geological modeling: problems and perspectives. In: Flint, S. and I. D. Bryant (Eds.), Geological modeling ofhydrocarbon reservoirs. International Association of Sedimentologists Special Publication 15: 3-20.
31. Budding, M.C., A. H. M. Paardekam and S. J. Van Rossem, 1992. 3D connectivity and architecture in sandstone reservoirs. SPE Technical Paper 22342, presented at SPE International Meeting on Petroleum Engineering, Beijing, China, 24-27 March 1992.
32. Burnett, M., 1996. 3-D seismic expression of a shallow fluvial system in west central Texas. In: P. Weimer and T.L. Davis (Eds.), Applications of 3-D Seismic Data to Exploration and Production. AAPG Studies in Geology No. 42 and SEG Geophysical Developments Series No. 5, AAPG/SEG, Tulsa: 45-56.
33. Carle, S.F., E. M. Labolle, G. S. Weissmann, D. Van Brocklin and G. E. Fogg, 1998. Conditional simulation of hydrofacies architecture: a transition probability/Markov approach. In: Fraser, G.S. and J.M. Davis (Eds.), Hydrogeologic models of sedimentary aquifers. Concepts in Hydrogeology and Environmental Geology No.1, Society of Economic Paleontologists and Mineralogists: 147-170.
34. Chitale, S.V., 1970. River channel patterns. Journal of the Hydraulics Division, American Society of Civil Engineers 96: 201-221.
35. Clemetsen, R., A. R. Hurst, R. Knarud and H. Omre, 1990. A computer program for evaluation of fluvial reservoirs. In: Buller, A.T., E. Berg, O. Hjemeland, J. Kleppe, O. Torsaeter and L. O. Aasen (Eds.), North Sea oil and gas reservoirs-II. Graham and Trotman, London: 373-385.
36. Collinson, J.D., 1978. Vertical sequence and sand body shape in alluvial sequences. In: A.D. Miall (Ed.), Fluvial Sedimentology. Canadian Society of Petroleum Geologists Memoir 5: 577-586.
37. Collinson, J. D., 1996. Alluvial sediments. In: H. G. Reading (Ed.), Sedimentary Environments and Facies, 3rd edition, Elsevier, New York: 37-82.
38. Cornish, F.G., 1984. Fluvial environments and palaeohydrology of the Upper Morrow "A" (Pennsylvanian) meander belt sandstone, Beaver County, Oklahoma. Shale Shaker, 34: 70-80.
39. Cross, T.A. and M. A. Lessenger, 1999. Construction and application of a stratigraphic inverse model. In: Harbaugh, J.W. et al. (Eds.), Numerical Experiments in Stratigraphy: Recent Advances in Stratigraphic and Sedimentologic Computer Simulations, Society of Economic Paleontologists and Mineralogists Special Publication 62: 69-83.
40. Cuevas Gozalo, M.C. and A. W. Martinius, 1993. Outcrop data-base for the geological characterization of fluvial reservoirs: an example from distal fluvial fan deposits in the Loranca Basin, Spain. In C.P. North and D.J. Prosser (Eds.), Characterization of Fluvial and Aeolian Reservoirs. Geological Society of London Special Publication 73: 79-94.
41. Dalrymple, R. W., R. Boyd and B. A. Zaitlin (Eds.), 1994. Incised-valley systems: origin and sedimentary sequences. Society of Economic Paleontologists and Mineralogists Special Publication 51.
42. Davies, D. K., B. P. J. Williams and R. K. Vessell, 1993. Dimensions and quality of reservoirs originating in low and high sinuosity channel systems, Lower Cretaceous Travis Peak Formation, East Texas, USA. In: C. P. North and D. J. Prosser (Eds.), Characterization of Fluvial and Aeolian Reservoirs. Geological Society of London Special Publication 73: 95-121.
43. Deutsch, C. and P. Cockerham, 1994. Practical considerations in the application of simulated annealing to stochastic simulation. Mathematical Geology 26: 67-82.
44. Deutsch, C.V. and L. Wang, 1996. Hierarchical Object-Based Stochastic Modeling of Fluvial Reservoirs. Mathematical Geology 28: 857-880.
45. Doveton, J.H., 1994, Theory and applications of vertical variability measures from Markov Chain analysis. In: Yarus, J.M. and R. L. Chambers (Eds.), Stochastic Modeling and Geostatistics. American Association of Petroleum Geologists Computer Applications in Geology 3: 55-64.
46. Dreyer, T., 1993. Quantified fluvial architecture in ephemeral stream deposits of the Esplugafreda Formation (Palaeocene), Tremp-Graus Basin, northern Spain. In: M. Marzo and C. Puigdefábregas (Eds.), Alluvial Sedimentation. International Association of Sedimentologists Special Publication 17: 337-362.
47. Dreyer, T., L. M. Falt, T. Hoy, R. Knarud, R. Steel and J. L. Cuevas, 1993. Sedimentary architecture of field analogues for reservoir information (SAFARI): a case study of the fluvial Escanilla Formation, Spanish Pyrenees. In: Flint, S. and I. D. Bryant (Eds.), The geological modeling of hydrocarbon reservoirs and outcrop analogues. International Association of Sedimentologists Special Publication 15: 57-80.
48. Ethridge, F. G. and R. M. Flores (Eds.), 1981. Recent and ancient nonmarine depositional environments: models for exploration. Society of Economic Paleontologists and Mineralogists Special Publication 31.
49. Ethridge, F. G., R. M. Flores and M. D. Harvey (Eds.), 1987. Recent developments in fluvial sedimentology. Society of Economic Paleontologists and Mineralogists Special Publication 39, Tulsa, 389 pp.
50. Ethridge, F.G., L. J. Wood and S. A. Schumm, 1998. Cyclic variables controlling fluvial sequence development: problems and perspectives. In: Shanley, K.J. and P. J. McCabe (Eds.), Relative Role of Eustasy, Climate and Tectonism on Continental Rocks. Society of Economic Paleontologists and Mineralogists Special Publication 59: 17-29.
51. Fielding, C.R. and R. C. Crane, 1987. An application of statistical modelling to the prediction of hydrocarbon recovery factors in fluvial reservoir sequences. In: Ethridge, F.G., R. M. Flores and M. D. Harvey (Eds.), Recent developments in fluvial sedimentology. Society of Economic Paleontologists and Mineralogists Special Publication 39: 321-327.
52. Flores, R. M., F. G. Ethridge, A. D. Miall, W. E. Galloway and T. D. Fouch, 1985. Recognition of fluvial depositional systems and their resource potential. Society of Economic Paleontologists and Mineralogists Short Course 19, 290 pp.
53. Galloway, W. E. and D. K. Hobday, 1983. Terrigeneous Clastic Depositional Systems: Applications to Petroleum, Coal, and Uranium Exploration, 1st ed., Springer-Verlag, New York,
54. Galloway, W. E. and D. K. Hobday, 1996. Terrigeneous Clastic Depositional Systems: Applications to Petroleum, Coal, and Uranium Exploration, 2nd ed., Springer-Verlag, New York, 489 p.
55. Geehan , G. and J. Underwood, 1993. The use of shale length distributions in geological modeling. In: Flint, S. and Bryant, I.D. (Eds.), The geological modeling of hydrocarbon reservoirs and outcrop analogues. International Association of Sedimentologists Special Publication 15: 205-212.
56. Gibling, M.R. and D. J. Bird, 1994. Late Carboniferous cyclothems and alluvial paleovalleys in the Sydney Basin, Nova Scotia. Geological Society of America Bulletin 106: 105-117.
57. Hardage, B.A., R. A. Levey, V. Pendleton, J. Simmons and R. Edson, 1994. A 3-D seismic case history evaluating fluvially deposited thin-bedded reservoirs in a gasproducing property. Geophysics 59: 1650-1665.
58. Hardage, B.A., R. A. Levey, V. Pendleton, J. Simmons and R. Edson, 1996. 3-D seismic imaging and interpretation of fluvially deposited thin-bed reservoirs. In: P. Weimer and T.L. Davis (Eds.), Applications of 3-D Seismic Data to Exploration and Production. AAPG Studies in Geology No. 42 and SEG Geophysical Developments Series No. 5, AAPG/SEG, Tulsa: 27-34.
59. Heller, P.L. and C. Paola, 1996. Downstream changes in alluvial architecture: an exploration of controls on channel-stacking patterns. Journal of Sedimentary Research 66: 297-306.
60. Hirst, J. P. P., C. R. Blackstock and S. Tyson, 1993. Stochastic modeling of fluvial sandstone bodies. In: S. S. Flint and I. D. Bryant (Eds.), The Geological Modeling of Hydrocarbon Reservoirs and Outcrop Analogues. International Association of Sedimentologists Special Publication 15: 237-251.
61. Holden, L., R. Hauge, Ø. Skare and A. Skorstad, 1998. Modeling of Fluvial Reservoirs with Object Models. Mathematical Geology 30: 473-496.
62. Jackson, R.G., 1978. Preliminary evaluation of lithofacies models for meandering alluvial streams. In: Miall, A.D. (Ed.), Fluvial Sedimentology. Canadian Society of Petroleum Geologists Memoir 5: 543-576.
63. Jervey, M.T., 1988. Quantitative geological modeling of siliciclastic rock sequences and their seismic expression. In: C.K. Wilgus, B. Hastings, C. Kendall, H.W. Posamentier, C.A. Ross and J.C. Van Wagoner (Eds.), Sea-level changes: an integrated approach. Society of Economic Paleontologists and Mineralogists Special Publication 42: 47-69.
64. Journel, A.G., 1983. Nonparametric estimation of spatial distributions. Mathematical Geology 15: 445-468.
65. Karssenberg, D., T. E. Tornqvist and J. S. Bridge, 2001a. Conditioning a process-based model of sedimentary architecture to well data. Journal of Sedimentary Research 71: 868-879.
66. Karssenberg, G, D., J. S. Bridge and T. E. Tornqvist , 2001b. Modeling cycles of fluvial aggradation and degradation due to base-level change using a revised process-based, 3-D alluvial stratigraphy model. VII International Conference on Fluvial Sedimentology, Lincoln, Nebraska.
67. Khan, I.A., J. S. Bridge, J. Kappelman and R. Wilson, 1997. Evolution of Miocene fluvial environments, eastern Potwar plateau, northern Pakistan. Sedimentology 44: 221-251.
68. Koltermann, C.E. and S. M. Gorelick, 1996. Heterogeneity in sedimentary deposits: a review of structure-imitating, process-imitating and descriptive approaches. Water Resources Research 32: 2617-2658.
69. Knighton, A.D. and G. C. Nanson, 1993. Anastomosis and the continuum of channel pattern. Earth Surface Processes and Landforms 18: 613-625.
70. Koss, J.E., F. G. Ethridge and S. A. Schumm, 1994. An experimental study of the effects of base-level change on fluvial, coastal plain and shelf systems. Journal of Sedimentary Research B64: 90-98.
71. Lane, E.W., 1957. A study of the shape of channels formed by natural streams flowing in erodible material. Missouri River Division Sediment Series No. 9, U.S. Army Engineer Division, Missouri River, Corps of Engineers, Omaha, Nebraska.
72. Leclair, S.F. and J. S. Bridge, 2001. Quantitative interpretation of sedimentary structures formed by river dunes. Journal of Sedimentary Research.71: 713-716.
73. Leclair, S. F., J. S. Bridge and F. Wang, 1997. Preservation of cross-strata due to migration of subaqueous dunes over aggrading and non-aggrading beds: comparison of experimental data with theory. Geoscience Canada 24: 55- 66.
74. Leeder, M.R., 1978. A quantitative stratigraphic model for alluvium with special reference to channel deposit density and interconnectedness. In: Miall, A.D. (Ed.), Fluvial sedimentology. Canadian Society of Petroleum Geologists Memoir 5: 587-596.
75. Leeder, M.R. and M. Stewart, 1996. Fluvial incision and sequence stratigraphy: alluvial responses to relative base level fall and their detection in the geological record. In: Hesselbo, S.P. and D. N. Parkinson (Eds.), Sequence stratigraphy in British Geology. Geological Society of London Special Publication 103: 47-61.
76. Lomando, A. J. and P. M. Harris, 1988. Giant oil and gas fields: a core workshop. Society of Economic Paleontologists and Mineralogists Core Workshop 12, 853 pp.
77. Lorenz, J.C., D. M. Heinze, J. A. Clark and C. A. Searls, 1985. Determination of widths of meander-belt sandstone reservoirs from vertical downhole data, Mesaverde Group, Piceance Creek Basin, Colorado. American Association of Petroleum Geologists Bulletin 69: 710-721.
78. Lorenz, J. C., N. R. Warpinski and P. T. Brannagan, 1991. Subsurface characterization of Mesaverde reservoirs in Colorado: geophysical and reservoir-engineering checks on predictive sedimentology. In: A. D. Miall and N. Tyler (Eds.), The three-dimensional facies architecture of terrigenous clastic sediments and its implications for hydrocarbon discovery and recovery. Society of Economic Paleontologists and Mineralogists Concepts in Sedimentology and Paleontology 3: 57-79.
79. Lowry, P. and A. Raheim, 1991. Characterisation of delta front sandstones from a fluvial-dominated delta system. In: L. W. Lake, H. B. Carroll Jr., and T. C. Wesson (Eds.), Reservoir characterisation - II. Academic Press, London: 665-676.
80. Mackey, S.D. and J. S. Bridge, 1995. Three dimensional model of alluvial stratigraphy: theory and application. Journal of Sedimentary Research B65: 7-31.
81. Makaske, B., 2001. Anastomosing rivers: a review fo their classification, origin and sedimentary products. Earth-Science Reviews 53: 149-196.
82. Miall, A.D., 1977. A review of the braided river depositional environment. Earth Science Reviews 13: 1-62.
83. Miall, A.D., 1986. Eustatic sea level changes interpreted from seismic stratigraphy: a critique of the methodology with particular reference to the North Sea Jurassic record. American Association of Petroleum Geologists Bulletin 70: 131-137.
84. Miall, A.D., 1991. Stratigraphic sequences and their chronostratigraphic correlation. Journal of Sedimentary Petrology 61: 497-505.
85. Miall, A.D., 1992. Alluvial deposits. In: Walker, R.G. and N. P. James (Eds.), Facies models: response to sea level changes. Geological Association of Canada, St Johns, Newfoundland: 119-142.
86. Miall, A.D., 1996. The geology of fluvial deposits. Springer-Verlag, New York, 582 p.
87. Miall, A.D. and N. Tyler (Eds.), 1991. The Three-Dimensional Facies Architecture of Terrigeneous Clastic Sediments and Its Implications for Hydrocarbon Discovery and Recovery. Society of Economic Paleontologists and Mineralogists Concepts in Sedimentology and Paleontology 3, 309 pp.
88. Nanson, G.C. and A. D. Knighton, 1996. Anabranching rivers: their cause, character and classification. Earth Surface Processes and Landforms 21: 217-239.
89. Nanz, R.H. JR., 1954. Genesis of Oligocene sandstone reservoir, Seeligson field, Jim Wells and Kleberg Counties, Texas. American Association of Petroleum Geologists Bulletin 38: 96-117.
90. North, C.P., 1996. The prediction and modelling of subsurface fluvial stratigraphy. In: P.A. Carling and M.R. Dawson (Eds.), Advances in Fluvial Dynamics and Stratigraphy. John Wiley, Chichester: 395-508.
91. North, C. P. and K. S. Taylor, 1996. Ephemeral-fluvial deposits: integrated outcrop and simulation studies reveal complexity. American Association of Petroleum Geologists Bulletin 80: 811-830.
92. Paola, C., P. L. Heller and C. L. Angevine, 1992. The large-scale dynamics of grain-size variation in alluvial basins. 1 - Theory. Basin Research 4: 73-90.
93. Pitman, W.C. III, 1986. Effects of sea level change on basin stratigraphy. American Association of Petroleum Geologists Bulletin 70: 1-62.
94. Pitman, W.C. III and X. Golovchenko, 1988. Sea-level changes and their effect on the stratigraphy of Atlantictype margins. In: Sheridan, R.E. and J. A. Grow (Eds.), The geology of North America, vol I-2, The Atlantic continental margin, United States. Geological Society of America: 429-436.
95. Posamentier, H.W., M. T. Jervey and P. R. Vail, 1988. Eustatic controls on clastic deposition I - Conceptual framework. In: C.K. Wilgus, B.S. Hastings, C.G.St.C. Kendall, H.W. Posamentier, C.A. Ross and J.C. Van Wagoner (Eds.), Sea-level changes: an integrated approach. Society of Economic Paleontologists and Mineralogists Special Publication 42: 109-124.
96. Posamentier, H.W. and P.R. Vail, 1988. Eustatic controls on clastic deposition II - Sequence and systems tract models. In: C.K. Wilgus, B.S. Hastings, C.G.St.C. Kendall, H.W. Posamentier, C.A. Ross and J.C. Van Wagoner (Eds.), Sea-level changes: an integrated approach. Society of Economic Paleontologists and Mineralogists Special Publication 42: 125-154.
97. Robinson, J. W. and P. J. McCabe, 1997, Sandstone-body and shale-body dimensions in a braided fluvial system: Salt Wash Sandstone Member (Morrison Formation), Garfield County, Utah. American Association of Petroleum Geologists Bulletin 81: 1267-1291.
98. Ross, W.C., 1990. Modeling base-level dynamics as a control on basin-fill geometries and facies distribution: a conceptual framework. In: Cross, T.A. (Ed.), Quantitative Dynamic Stratigraphy, Prentice-Hall, Englewood Cliffs, New Jersey: 387-399.
99. Rust, B.R., 1978. A classification of alluvial channel systems. In: Miall, A.D. (Ed.), Fluvial Sedimentology Canadian Society of Petroleum Geologists Memoir 5: 187-198.
100. Salter, T., 1993. Fluvial scour and incision: models for their influence on the development of realistic reservoir geometries. In: C. P. North and D. J. Prosser (Eds.), Characterization of Fluvial and Aeolian Reservoirs. Geological Society of London Special Publication 73: 33- 51.
101. Saucier, R.T., 1981. Current thinking on riverine processes and geologic history as related to human settlement in the southeast. Geoscience and Man 22: 7-18.
102. Saucier, R.T., 1994. Geomorphology and Quaternary Geologic History of the Lower Mississippi Valley. Mississippi River Commission, Vicksburg, 364 pp.
103. Schumm, S.A., 1963. Sinuosity of alluvial channels on the Great Plains. Geological Society of America Bulletin 74: 1089-1100.
104. Schumm, S.A., 1971. Fluvial geomorphology. In: Shen, H.W. (Ed.), River Mechanics I. Water Resources Publications, Colorado, Chapter 5: 1-22.
105. Schumm, S.A., 1972. Fluvial paleochannels. In: J.K. Rigby and W.K. Hamblin (Eds.), Recognition of ancient sedimentary environments. Society of Economic Paleontologists and Mineralogists Special Publication 16: 98-107.
106. Schumm, S.A., 1977. The Fluvial System. John Wiley and Sons, New York, 338 p.
107. Schumm, S.A., 1981. Evolution and response of the fluvial system, sedimentologic implications. In Ethridge, F.G. and R. M. Flores (Eds.), Recent and ancient nonmarine depositional environments: models for exploration. Society of Economic Paleontologists and Mineralogists Special Publication 31: 19-29.
108. Schumm, S.A., 1985. Patterns of alluvial rivers. Annual Reviews Earth and Planetary Science 13: 5-27.
109. Schumm, S.A., 1993. River response to base level change: implications for sequence stratigraphy. Journal of Geology 101: 279-294.
110. Selley, R.C., 1996. Ancient Sedimentary Environments and their Subsurface Diagnosis, 4th ed., Chapman and Hall, London, 300 pp.
111. Shanley, K. W. and P. J. McCabe, 1993. Alluvial architecture in a sequence stratigraphic framework: a case history from the Upper Cretaceous of southern Utah, USA. In: Flint, S. and Bryant, I.D. (Eds.), The geological modeling of hydrocarbon reservoirs and outcrop analogues. International Association of Sedimentologists Special Publication 15: 21-56.
112. Shanley, K. W. and P. J. McCabe, 1994. Perspectives on the sequence stratigraphy of continental strata. American Association of Petroleum Geologists Bulletin 78: 544-568.
113. Smith, D.G., 1976. Effect of vegetation on lateral migration of anastomosed channels of a glacier meltwater river. Geological Society of America Bulletin 87: 857-860.
114. Srivastava, R.M., 1994. An overview of stochastic methods for reservoir characterization. In: Yarus J.M. and R. L. Chambers (Eds.), Stochastic modeling and Geostatistics. American Association of Petroleum Geologists Computer Applications in Geology 3: 3-16.
115. Stam, J.M.T., 1996. Migration and Growth of Aeolian Bedforms. Mathematical Geology 28: 519-536.
116. Tetzlaff, D.M. and J. W. Harbaugh, 1989. Simulating Clastic Sedimentation. New York, Van Nostrand Reinhold.
117. Törnqvist, T.E., 1993. Holocene alternation of meandering and anastomosing fluvial systems in the Rhine-Meuse Delta (central Netherlands) controlled by sea-level rise and subsoil erodibility. Journal of Sedimentary Petrology 63: 683-693.
118. Törnqvist, T.E., 1994. Middle and late Holocene avulsion history of the River Rhine (Rhine-Meuse Delta, Netherlands). Geology 22: 711-714.
119. Tye, R. S., 1991. Fluvial-sandstone reservoirs of the Travis Peak Formation, East Texas Basin. In: A. D. Miall and N. Tyler (Eds.), The Three-Dimensional Facies Architecture of Terrigeneous Clastic Sediments and Its Implications for Hydrocarbon Discovery and Recovery. Society of Economic Paleontologists and Mineralogists Concepts in Sedimentology and Paleontology 3: 172-188.
120. Tyler, K., A. Henriquez and T. Svanes, 1994. Modeling heterogeneities in fluvial domains: a review of the influence on production profiles. In Yarus J.M. and R. L. Chambers (Eds.), Stochastic modeling and Geostatistics. American Association of Petroleum Geologists Computer Applications in Geology 3: 77-89.
121. Walderhaug, O. and R. Mjos, 1991. Sandbody geometry in fluvial systems - an example from the Jurassic of Yorkshire, England. In: L. W. Lake, H. B. Carroll Jr. and T. C. Wesson (Eds.), Reservoir characterisation - II. Academic Press, London: 701-703.
122. Webb, E.K., 1994. Simulating the three-dimensional distribution of sediment units in braided stream deposits. Journal of Sedimentary Research B64: 219-231.
123. Weber, K. J., 1993. The use of 3-D seismic in reservoir geological modelling. In: S.S. Flint and I.D. Bryant (Eds.), The Geological Modelling of Hydrocarbon Reservoirs and Outcrop Analogues. International Association of Sedimentologists Special Publication 15: 181-188.
124. Wescott, W.A., 1993. Geomorphic thresholds and complex response of fluvial systems - some implications for sequence stratigraphy. American Association of Petroleum Geologists Bulletin 77: 1208-1218.
125. Willis, B.J., 1993a. Ancient river systems in the Himalayan foredeep, Chinji village area, northern Pakistan. Sedimentary Geology 88: 1-76.
126. Willis, B.J., 1993b. Evolution of Miocene fluvial systems in the Himalayan foredeep through a two kilometer-thick succession in northern Pakistan. Sedimentary Geology 88: 77-121.
127. Wood, L.J., F. G. Ethridge and S. A. Schumm, 1994. An expermental study of the influence of subaqueous shelf angles on coastal plain and shelf deposits. In: Weimer, P. and H. W. Posamentier (Eds.), Siliciclastic sequence stratigraphy. American Association of Petroleum Geologists Memoir 58: 381-391.
128. Wright, V.P. and S. B. Marriott, 1993. The sequence stratigraphy of fluvial depositional systems: the role of floodplain sediment storage. Sedimentary Geology 86: 203-210.
129. Zaleha, M.J., 1997a. Fluvial and lacustrine palaeoenvironments of the Miocene Siwalik Group, Khaur area, northern Pakistan. Sedimentology 44: 349- 368.
130. Zaleha, M.J., 1997b. Intra- and extrabasinal controls on fluvial deposition in the Miocene Indo-Gangetic basin, northern Pakistan. Sedimentology 44: 369-390.