A collaborative proposal funded by the Active Tectonics Program EAR Division National Science Foundation
by
Center for Earthquake Research and Information, The University of Memphis
and
Institute of Tectonics and Department of Earth Sciences, University of California, Santa Cruz
____________________________________________
Blind thrusts are important structures in a variety of tectonic settings and for a variety of reasons. In the short term, blind thrusts are a well known seismic hazard, particularly as they often underlie relatively flat and populated regions. One of the primary goals of the Working Group on California Earthquake Probabilities (WGCEP, 1995) was to better understand the blind faults that do not break the surface (p379-380) and concluded that blind thrusts remain one of the significant sources of uncertainty in the estimation of seismic hazard.
In the long term, blind thrusts are integral components of foreland-migrating fold-and-thrust belts, and their growth dictates both the gross topographic evolution of such belts (which is coupled to the evolving mechanical and thermal state of the crust) and the development of associated basinal deposits (relevant to issues of hydrocarbon migration). Such settings are also very fruitful places to investigate the interaction between climate and tectonics, since evidence for temporal variations in the relative roles of surface and tectonic processes may be preserved. Mugnier et al. (1997) have shown in analog models, for example, that the development of geological structures is strongly dependent on syntectonic erosion and sedimentation, Burbank et al. (1996b) have shown that the geomorphology of folds and piggy-back basins over blind thrusts is a sensitive indicator of fold and fault growth, and Molnar et al. (1994) have documented the competing effects of climate and tectonics in blind thrust settings in the northern Tien Shan mountains.
We propose to quantitatively analyze the landscape morphology over blind thrusts using the numerical landscape evolution model Zscape, which the PIs have developed over the last three years. Landscape morphology is a potentially quite useful recorder of a blind thrust's history, since the landscape is quite literally the interface between surface processes (climate) and tectonic processes. Landscapes respond to changes in these processes at a characteristic timescale that is on the order of 105 yr (Ellis et al., 1997); they thus preserve evidence of processes and responses that span the gap between geologic (> 105 yr) and seismologic or geodetic (100 to 102 yr) timescales.
We will modify Zscape in order to simulate the growth of topography and sedimentary deposits over a set of blind thrust faults. After calibrating our numerical results by comparison with topography and process rates from several well-constrained blind thrusts (the Ventura Avenue and Wheeler Ridge anticlines) in southern California, we will apply the model to nearby, less well understood thrusts, particularly those that underlie the Santa Susanna, and Santa Monica Mountains (Fig. 1).
Figure 1: Location map of southern California. Base map courtesy of the SCEC web page (http://www/scec.org). Shading indicates number of times per century that seismic shaking will exceed 0.2g. APF - Arroyo Parida fault , GF - Garlock fault, KH - Kettleman Hills, SAF - San Andreas fault, SaMF - Sierra Madre fault, VF - Ventura fault
The primary goal is to provide additional constraints on the long-term slip-rates, segmentation, and geometries of blind faults. A secondary, and independent, goal is to gain a quantitative understanding of the competing effects of tectonics and climate (proxied as surface processes) in, for example, the generation of drainage networks, drainage divides, and fluvial terraces. These goals may be cast in terms of two questions:
(1) How sensitive is the topography (including river and terrace profiles, slope and aspect distribution, channel networks, drainage areas, etc.) to variations in fault geometry and slip rate?
(2) Are we able to recognize and quantify the competing effects of tectonics and climate? Later in the proposal, we describe an interesting hypothesis that represents an example of this type of question.
We choose to focus here on blind (and related emergent) thrusts in southern California for several reasons, including (1) the relative wealth of existing, available geologic, geodetic, and seismologic data, (2) the logistical ease and economy of the region, and (3) the opportunity to contribute to the very real difficulties in constraining estimates of slip-rate and fault geometry. The choice of southern California makes this research particularly relevant to the seismic community, but at the same time allows us to address other, important geological and climate-change issues. The proposed research is part of a long-term research project by the PIs, the theme of which centers on the relation between climate, surface processes, and tectonics. Some of the longer term goals include the incorporation of glacial processes, a better coupling between fault slip and surface processes, the ability to generate sedimentary basin architecture, and importantly to include a fully 3d thermal evolution within the landscape model.
____________________________________________
The Working Group on California Earthquake Probabilities (WGCEP, 1995) makes the important observation that blind thrusts have demonstrated their importance in regional seismic hazard. However, there is considerable uncertainty about their extent, geometry at depth, and the nature of the postulated regional detachment into which they might root (p394). Virtually all historical and modern moderate earthquakes in the greater Los Angeles region have occurred on reverse faults (WGCEP, 1995), including the 1994 M 6.7 Northridge earthquake, which occurred on a blind section of the Oak Ridge fault (Yeats et al., 1995; the Pico fault of Davis and Namson, 1994). Dolan et al. (1995) identify six major fault systems in the Los Angeles region that contribute to the seismic hazard, all of which are thrust and many of which are blind. Relatively rapid strain rates across the region, coupled with a significant absence of seismicity, are consistent with a repeat time of ~140 years for the M 7.2-7.6 earthquakes that must occur on these thrust systems (Dolan et al., 1995).
The most successful approach to characterizing blind thrusts and their emergent segments has been through combined kinematic and seismic reflection modeling (e.g., Shaw and Suppe, 1996). In the absence of drill-hole data, well-defined seismicity or geodetic surveys, the characteristics of blind faults may be poorly known, if known at all. And these data may provide only limited information about the long-term slip rates and geometry of underlying thrusts. Interseismic activity, for example, may not necessarily "light up" faults that will later sustain the regions largest earthquakes, nor will it yield much information about aseismic parts of the fault. Interseismic geodetic data, although relatively straightforward to interpret over well-mapped and dominant strike-slip faults, are extremely difficult to interpret over distributed faults (e.g., Shen et al., 1996), all or only some of which may be accumulating strain. Geological cross-sections reveal perhaps the best estimate of blind fault geometry, but their resolution usually limits discussion to a few millions of years and, because rates can vary over time, slip rates averaged over millions of years do not necessarily provide adequate constraints on present behavior (WGCEP, 1995, p394-5).
An important and complementary source of data comes from the morphology of the landscape (including deformed fluvial terraces), which may provide information about fault activity over the time-scale of 103 to 106 years. With a few important exceptions (e.g, Anderson, 1994; Arrowsmith, 1995; Burbank et al., 1996b; Meigs, 1996; Brozovic et al., 1997), the potential information in the landscape morphology remains largely untapped. Zscape has the potential to extract a wide variety of information that depends on the availability of independent and complementary data. At one extreme, where there is no independent data (seismic lines, drill-holes, etc.), Zscape will be able to better constrain the geometry and long-term slip rate of a buried fault. At the other extreme, where slip rates and geometry are already well known, Zscape may be able to constrain estimates of local climate changes that must have occurred over the lifetime of the landscape over the growing fold. The latter approach (which importantly may be accomplished independently of the first) is similar to that taken by Koltermann and Gorelick (1992) in their reconstruction of the geologic evolution of a large alluvial fan sequence in central California. We consider each end of this spectrum to be an important contribution to the earth sciences.
____________________________________________
Zscape is a finite-difference landscape evolution model that incorporates the three-dimensional effects of both tectonic and surface processes. Tectonics are encoded by subjecting the region of interest to an appropriate three-dimensional deformation field. Far-field boundary conditions may be cast in terms of displacements, displacement gradients, strain, or stress. Alternatively, it is possible to specify displacements across individual faults within the elastic material. In either case, the material accommodates the imposed strain by displacement across those faults. Each infinitesimal increment of deformation is regarded as an earthquake or a suite of earthquakes and the deformation pattern (that may vary over time) becomes the template that is worked on by surface processes (e.g., Fig. 2) (Sorry - Figure 2 is too difficult and big too translate yet.)
Repeated templates generate a topographic envelope that is simultaneously acted on by a set of surface processes. In particular, we simulate the conversion of bedrock to erodible regolith with a modified gamma function, in which the regolith production rate is highest beneath a thin but non-zero cover of existing regolith (Ahnert, 1972; Heimsath et al., 1996). Downslope regolith transport is modeled as linear, weathering-limited diffusion, which serves to smooth out short-wavelength topographic roughness (e.g., Hanks et al., 1984; Anderson and Humphrey, 1989). For steep slopes, most of the material leaving the hillslopes does so in the form of large, bedrock landslides. In the model, these landslides are considered to be a stochastic process; the instantaneous probability of landsliding at any point is equal to the ratio of the actual hillslope height to the maximum stable hillslope height (as defined by the Culmann criterion; Schmidt and Montgomery, 1995), which is in turn a function of the landscape-averaged rock strength. Finally, a fluvial channel network is allowed to self-define on the basis of spatial variations in stream power (Bagnold, 1977), proportional to the product of the drainage area and local slope. Fluvial sediment transport, and bedrock incision in the absence of sediment, is directly proportional to the available stream power.
Our initial set of experiments with Zscape support the idea that large bedrock landslides are the dominant geomorphic agent in many actively deforming mountainous environments (e.g., Schmidt and Montgomery, 1995; Burbank et al., 1996a; Hovius et al., 1997). In addition, we find that simple tectonic geometries (such as those arising from single or paired, horst-style normal faults) give rise to relief-limited mountain ranges that attain a quasi-steady state form in 105 to 106 yrs (Densmore et al., in press; Ellis and others, 1997 and in prep.). This is an important result, since it tells us that, at typical Basin and Range process rates, initial (and usually unknown) conditions are forgotten beyond about half a million years; thereafter, the landscape contains information only about relatively recent and current conditions.
Bedrock channel incision is the rate-limiting process in these landscapes; relief is dictated by the nearly constant spacing between major catchments (Hovius, 1996), which is established early in the evolution of the drainage network (Anderson, 1994; Burbank et al., 1996b), and the rock strength. We further observe that bedrock landslides allow hillslopes to respond rapidly to changes in local and regional baselevel, and that our experimental landscapes exhibit many of the same features and landforms as normal fault-bounded ranges in the Basin and Range province of the western US (Densmore et al., 1997; Ellis and others, 1997 and in prep.)
Zscape, in its current form, is the only landscape evolution model that incorporates detailed three-dimensional tectonic displacements, and that accounts explicitly for the stochastic nature of bedrock landslides. As such, the model is ideally suited to model three-dimensional variations in erosion and deposition over timescales of 104 to 106 years. Two primary additions to Zscape are necessary for the proposed investigation - the inclusion of isostasy, and the accommodation of variable bedrock strength and erodibility. These additions will be accomplished in the first of three phases of the proposed research.
____________________________________________
The first phase of our proposed research will involve extending Zscape to allow for spatial variability in the erodibility of the bedrock. Such variability is primarily due to changes in lithology, but may also reflect geologically induced differences in material strength. Spatial changes in erodibility, or in the related concept of rock mass strength, have been frequently cited as fundamental controls on long-term landscape development (e.g., Selby, 1980; Augustinus, 1992; Schmidt and Montgomery, 1995). We will incorporate this into Zscape by allowing for spatial variability of the following, existing parameters:
We will also incorporate the isostatic response to surface mass redistribution by tectonic and geomorphic processes. This has shown to be an important process when displacement across the active fault is more than several km (Buck, 1988; King et al., 1988; Wernicke and Axen, 1988; King and Ellis, 1990). This will be accomplished by assuming that the model space behaves as a linear elastic solid, with a reduced effective elastic thickness (thereby simulating a ductile lower crust). Depositional loading and erosional unloading, as well as advection of topography by tectonic deformation, may then be represented as a sum of point loads, and the resulting vertical displacements calculated by solving the flexure equation (e.g., Weissel and Karner, 1989). The effective elastic thickness can be constrained at our calibration sites.
Projected time to completion: ~6 months after start date
Deliverable: Modified, freely available version of Zscape
As described above, there are a relatively small number of parameters in Zscape that control the rate of surface processes. In order to model a specific landscape, we need to find values for each parameter that are appropriate for the present climate, fault slip rate, etc. Again, this stage of the work is similar to the calibration stage of Koltermann and Gorelick (1992), in which they used the first 150 kyr of the Alameda fan sequence to calibrate their (far more numerous) model parameters. We will start by modeling two regions (the Ventura Avenue and Wheeler Ridge anticlines) where there exists a variety of well constrained cross sections, slip rates, subsurface information, etc. These should allow us to constrain the more significant parameter values that control fluvial incision and bedrock landsliding.
The Ventura Avenue anticline, near the city of Ventura, has been active over approximately the past 200 to 400 kyr (Rockwell et al., 1988). The fold has developed in response to north-south directed shortening of approximately 9 mm/yr, and has formed above a decollement zone in Miocene rocks at a depth of ~5 km. A flight of at least five fluvial terraces of the Ventura River have been deformed by the growth of the fold, and thus form relatively well-dated marker horizons in the landscape (Rockwell et al., 1988). Rockwell (1988) estimates that the rates of uplift for progressively younger time periods defined by terrace ages (200 kyr to 30 kyr) range from 20 mm/y to 5 mm/y, and that rates of tilting show a similar decrease from 5.8 mrad/y to 1.2 mrad/y. The anticline is truncated to the south by the Pitas-Ventura fault (Yerkes et al., 1987). It would not be possible to calibrate Zscape without the extensive work over the past twenty years on the Ventura Avenue anticline and the Ventura basin by R. Yerkes, E. Keller and K. Lajoie of UC Santa Barbara, R. Yeats of Oregon State University, and T. Rockwell of UC San Diego, which has yielded one of the best constrained data sets on a fault-controlled fold and its topography. As each of these investigators have stated, the topography and terraces have been formed by the interaction of base-level change (sea-level), climate change, and tectonic forcing (see also Stein and Yeats, 1989). The process-based Zscape now allows us to quantify these relations, which includes the placing of uncertainties on estimates of both fault slip, geometry, and signatures of climate change.
Wheeler Ridge fold is an eastward plunging anticline that lies in a classic foreland setting, at the northern edge of San Emigdio Mountains, and developed above a wedge thrust that merges into a southward dipping detachment (Medwedeff, 1992). The fold is particularly well known because it is a productive oil source (e.g., Namson and Davis, 1988). The eastern half of the fold intiated at ~90-120 kyr (Keller et al., 1989; Zepeda et al., 1990) and the nose has propagated eastwards at ~25 m/kyr (Medwedeff, 1992). The average rate of hinge uplift has been ~3.5 m/kyr and mean sedimentation rate in the adjacent plains has been ~ 2 m/kyr during the past 100 kyr (Medwedeff, 1992). The Wheeler Ridge structure has the additional advantage that the cumulative amount of fault slip appears to vary monotonically along strike, meaning that we can employ the ergodic hypothesis to examine anticline development and topographic evolution through time. Burbank et al (1996) recognized that with this much information, there exists the potential to extract quantitative relations among the competing effects of surface processes and structural development. In lieu of a tool to do that, however, these authors have provided a conceptual model of the growth of Wheeler Ridge, which provides us with a rigorous site to first calibrate Zscape and subsequently quantitatively explore the climate-tectonic interactions.
As the tectonic and geomorphic parameters are refined, we will generate experimental landscapes that will be directly compared with USGS 3 arcsec and 30 m DEMs of the target areas (e.g., Meigs, 1996). We will use both visual comparison (i.e., does the landscape `look' right?) and more rigorous statistical tests, such as drainage density, overall and intracatchment relief, and the probability distributions of elevation, relief, and slope (Densmore et al., in press). Such simple statistical measures have proven to be robust and useful in characterizing landscapes (Meigs, 1996, Brozovic et al., 1997; Densmore et al., in press). We generate uncertainties by running a series of forward models that bound the parameter set that yield acceptable results. Because part of the model is stochastic (bedrock landslides), we cannot simply "least-squares difference" model output and real topography; there are simply too many local complexities. But we have been successful in using the types of measures referred to above in constraining estimates of parameter values in normal-faulted terrains (Densmore, 1997; Ellis et al, in prep.). Again, we point to the similarity in our approach to that of Koltermann and Gorelick (1992), with the advantage that our computation times are significantly less and that we have far fewer model parameters to constrain.
Projected time to completion: 18 to 24 months after start date
Deliverable: Full-length mss to JGR and/or GSA plus abstracts at national mtgs
At this stage, we divide the research into two parts that may be accomplished independently of each other.
We will apply Zscape next to immediately adjacent areas in the region, such as the Santa Susana mountains north of Northridge (site of the 1971 San Fernando M 6.9 earthquake), and the Santa Monica Mountains along the Malibu coast.
The Santa Susana fault extends 28 km from San Fernando County westwards into Ventura County, and farther east it may merge with the Sierra Madre fault at the southern edge of the San Gabriel Mountains. The fault dips north, gently near the surface and more steeply at depth, and places deposits of middle Miocene to Pliocene sediments of the Ventura basin against Late Cretaceous to Late Tertiary age (Yeats,1987). This juxtaposition is overlain by the Pleistocene Saugus Formation, the offset of which across the Santa Susana fault yields a rate of slip over the past 500 ky of 8mm/y, which has since been revised for the northern (Newhall-Protrero) and southern (Placerita) segments to 2.0 to 2.3 mmy and 5.7 to 6.7 mm/y, respectively (Huftile, 1992). This system is particularly difficult to get at with paleoseismologic techniques because the fault is not well expressed at the surface, although it is not considered a blind fault. We would concentrate our efforts on post-Saugus Formation (~500 ka) faulting and pay careful attention to the eastern section of the fault, which is not well constrained by well data.
The Santa Monica Mountains represent an actively growing, southward verging anticline above south-dipping blind thrusts, including most notably the Malibu Coastal thrust system. Meigs (1996) has provided a thorough analysis of the geomorphology of the range from a 30m DEM that, again, provides a rigorous set of observations that can be replicated and explained quantitatively by Zscape. Meigs has, for example, described how the drainage divide is not coincident with the maximum structural relief of the underlying fold, and that there exists evidence of probable drainage capture. Meigs also subdivides the range according to geomorphic characteristics (relief, slopes, drainage geometry, etc.), all of which is amenable to a process-based numerical landscape model that is linked to fault slip rate.
There are a number of other sites, of course, that would be amenable to this phase of the research. As we gain more experience in the relatively simpler structures in the region, we will broaden our scope and examine more complex structures. We emphasize also that the modified Zscape will be available to others toward the end of this research
.
Once we have established specific model parameters for the southern California region, we can begin to address the interaction between climate, tectonics, and topography. In particular, we will investigate the effects of two particular climatic variables - mean annual precipitation (MAP) and the degree of `storminess' - on topographic development. For example, blind thrust fault propagation is generally modelled as the upward migration of a fault tip, which carries with it axial surfaces and which produces an anticline that grows and steepens in the direction of fault tip migration (Suppe and Medwedeff, 1990). For a fold exposed at the surface as a positive topographic feature, this forward growth will alter the boundary conditions (i.e., drainage divide position, slope distribution, and spatial pattern of incision rate necessary to keep pace with uplift) of the catchments draining the fold (Burbank et al., 1996b; Jackson et al., 1996). If process rates are such that tectonic processes are much more efficient at creating topography than surface processes are at removing it, then the catchments will always be out of equilibrium with the boundary conditions imposed by fault propagation (Fig. 3).
Figure 3. If channel incision processes are efficient enough, the drainage divide will be "frozen in" to the landscape; further propagation of the fault tip beyond this time will result in a disconnection of the structural axis and the drainage divide, and fluvial terraces will reflect the deformation pattern. In more resistant lithologies,such early drainages may be defeated, resulting in barbed tributary junctions in the headwaters of the east-flowing stream.
This scenario might occur, for example, in an arid climate with low MAP, and result in a drainage pattern and drainage divide position which is constantly changing in an effort to `keep up' with the propagation of the fold. By contrast, very efficient geomorphic processes (and, by extension, higher MAP) might maintain existing drainage patterns and divides in the face of changes in the tectonic displacement field. That is, the topography might `freeze in' the initial spatial pattern of tectonic uplift, and be insensitive to subsequent fault and fold propagation (Fig. 3).
In addition to considering the effects of precipitation, we will explore the role of changes in the magnitude-frequency distribution of storms that deliver that precipitation. This can be parameterized by the `storminess', or the proportion of the precipitation that falls in large, infrequent events. All else being equal, we expect that large storms will move larger amounts of sediment than small storms, and that the sediment load does not track linearly with storm size. This is, in part, because small storms may lead to little runoff depending on antecedent moisture conditions, whereas large storms overwhelm the infiltration capacity of the landscape and lead to considerably greater runoff (e.g., Wilson and Wieczorek, 1995). Simple 1d numerical experiments on the development of graded river profiles have shown us that storms are the critical ingredient in generating profiles that minimize the total energy dissipation rather than distribute it uniformly (Ellis et al., 1997), and which therefore governs the shape of the river profile. We anticipate therefore that incorporating storms (as spatially and temporally nonuniform precipitation) has a profound effect on the landscape.
To address these issues, we will first generate generic, endmember experimental landscapes in which the tectonic processes are held fixed and the precipitation or storminess is stepped through a range of plausible values. We will also follow the lead of Koltermann and Gorelick (1992) who simulated stormy hydrographs and discharge by the combined use of global marine data, the oxygen isotope record, and an ergodic use of climate up and down California. These initial, insight-generating results will then be used to guide us in analyzing the climatic history of a real blind thrust fault. We ideally require a blind thrust for which the fault geometry and slip history are relatively simple and particularly well-constrained, and which has been active for enough time to have accrued significant structural displacements and experienced a range of climatic environments. The Kettleman Hills/Coalinga anticline system in the Central Valley is an outstanding example of these requirements. The three domes of the Kettleman Hills are the result of slip along west-dipping thrust faults. Bloch et al. (1993) published a kinematic model of the growth of Kettleman Hills South Dome, in which folding over a propagating blind thrust in Pliocene time was superseded by duplex development and a subsequent shift in the locus of maximum uplift from Pliocene to Holocene time. The anticlines expose strata that range in age from the Miocene Etchegoin Formation in their cores to the Pleistocene Tulare and Paso Robles Formations at their edges. This lateral variation in surface lithology leads to a corresponding variation in drainage density and catchment relief between the different rock types.
By applying Zscape, with parameters calibrated on the nearby Wheeler Ridge anticline, to the Kettleman Hills, we will examine the topography for indications that past climatic conditions are recorded in the landscape. In the absence of reliable paleoclimate data for the region, we will construct endmember cases using `extreme' values of the climatic parameters, as well as use the method outlined in Koltermann and Gorelick (1992), and will compare those experimental results to the observed topography of the Kettleman Hills using the visual and statistical techniques outlined above. In particular, we will look for evidence that areas of the landscape (for example, the catchment headwaters, near the fold axis) do not resemble topography formed under the current climatic conditions, but instead are more visually and statistically similar to topography formed under considerably more humid or arid conditions. We expect that the headwaters may retain this information the longest, as they are the farthest removed from local baselevel (in this case, the elevation of the Kettleman Plain to the west and the Central Valley to the east) and thus are the last to `feel' tectonically-induced changes in that baselevel (e.g., Densmore et al., 1997).
Projected time to completion: 18 to 36 months after start date
Deliverable: Full-length mss to JGR or GSA plus abstracts at national mtgs
____________________________________________
Over the three-year course of the proposed research, we will develop a unique and potentially very powerful tool for the study of blind thrusts and related folds, both in southern California and elsewhere. Once calibrated, Zscape may be used in one of two complementary ways - in the absence of fault geometry and slip data, it may be used to predict the orientation and distribution of faults that give the best match to the observed topography. Alternatively, if detailed fault information is available, it may be used to back out the likely climatic conditions present during the growth of the fold. We emphasize that no other existing landscape evolution model has the power to do this, and that this fills a significant gap now occupied by qualitative, conceptual models of topographic and structral development.
____________________________________________