ABSTRACT Accurately predicting rates of coarse sediment transport in river channels is a central ... more ABSTRACT Accurately predicting rates of coarse sediment transport in river channels is a central goal of fluvial geomorphology and civil engineering. However, it is difficult to evaluate sediment transport and bedrock abrasion models in large rivers because quantitative measures of bedload transport are labor intensive and often dangerous to obtain in floods. Two recent studies show that the amplitude of seismic waves near rivers may record bedload flux, indicating that seismometers near rivers provide a potential means of monitoring bedload transport. In an effort to better interpret seismic waves generated by rivers, we seek a relationship between the variables governing bedload transport and seismic waves. Our approach relies on the fact that elastic waves are generated when momentum is transferred to the bed during a bedload particle impact. For an impacting particle of known mass and velocity, the momentum transfer can be computed from Hertzian impact theory. Here we combine analytic results based on Hertzian and elastic wave theories with empirical equations developed to describe the ballistics of bedload particles in terms of fluid shear stress and grain size. From this synthesis we arrive at a semi-analytic expression that predicts how the characteristic frequencies and amplitudes of seismic waves generated from saltating bedload particles should scale with fluid shear stress, grain size, and coarse sediment flux. Preliminary tests of our predictions using previously published and newly acquired laboratory data indicate that seismic signals near rivers can record information about the size, velocity and number of particles impacting the bed. Additionally, our analytical results help identify bedload transport events in seismic data collected along the Chijiawan River in Taiwan. Here the river is evolving rapidly in response to a dam removal - resulting in predictable changes in bedload transport efficiency in time and space that we can compare to local seismic data.
ABSTRACT The Chijiawan river, located in the Hsuehshan Range in central Taiwan, is a steep, coars... more ABSTRACT The Chijiawan river, located in the Hsuehshan Range in central Taiwan, is a steep, coarse-grained, mixed bedrock-alluvial channel subject to heavy monsoonal rainfall, occasional typhoons and frequent floods. In May 2011, a 15m-high check dam holding back around 200,000m3 of sediment was removed. To monitor subsequent channel response, an array of instruments were deployed prior to dam removal. Given the anticipated magnitude of sediment transport and excellent empirical constraints, the Chijiawan channel is an ideal location to study fluvial incision processes into bedrock. There is abundant bedrock exposed in the river bed downstream of the dam, and we are interested in how the pulse of sediment following dam removal will impact bedrock erosion rates. We monitor erosion using repeat high resolution laser scanning at 8 sites spread along a 3 km long reach, 2 sites upstream of the dam and 6 downstream. We scan two sites with a Konica Minolta Vivid 910 scanner (with an accuracy of 1.4mm), and the remaining sites with a Faro Photon 120 scanner (with an accuracy of 2mm). Scanning at ranges of ~1m (Vivid) to ~10m (Faro), we obtain resolutions of 1 to 3mm. Based on the system developed by Wilson and Hovius (2010), reference frames for repeat surveys consist of permanent stainless steel sockets installed in 14 cm deep holes drilled into the bedrock, combined with removable targets that screw into the sockets for scanning. Vivid data are processed and 3D models are constructed using Rapidform 2004 software. Faro data is processed using Faro's Scene software. In order to compare the high resolution scanning with more traditional low-tech monitoring methods, we have also installed two sets of concrete expansion bolts. We obtain profiles between bolts with both a contour gauge and the scanners. The high accuracy and resolution of these scanners, and the system of precise benchmarks enable us to detect extremely small amounts of erosion. By repeat scanning after individual flood events, we can directly relate magnitudes and patterns of erosion with flood characteristics and measures of sediment flux. Citation Wilson A, and Hovius N, 2010, Upstream facing convex surfaces and the importance of bedload in fluvial bedrock incision: observations from Taiwan, Geophysical Research Abstracts, Vol. 12, EGU2010-4978.
We have compiled a database of floods and landslides that occurred in Italy between AD 1279 and 2... more We have compiled a database of floods and landslides that occurred in Italy between AD 1279 and 2002 and caused deaths, missing persons, injuries, and homelessness. Analysis of the database indicates that more than 50,593 people died, went missing, or were injured in 2580 flood and landslide events. Harmful events were inventoried in 26.3% of the 8103 Italian municipalities. Fatal events were most frequent in the Alpine regions of northern Italy and were caused by both floods and landslides. In southern Italy, landslides were the principal agents of fatalities and were most numerous in the Campania region. Casualties were most frequent in the autumn. Fast-moving landslides, including rock falls, rockslides, rock avalanches, and debris flows, caused the largest number of deaths. In order to assess the overall risk posed by these processes, we merged the historical catalogs and identified 2682 “hydrogeomorphological” events that triggered single or multiple landslides and floods. We estimated individual risk through the calculation of mortality rates for both floods and landslides and compared these rates to the death rates for other natural, medical, and human-induced hazards in Italy. We used the frequency distribution of events with fatalities to ascertain the magnitude and frequency of the societal risks posed by floods and landslides. We quantified these risks in a Bayesian model that describes the probabilities of fatal flood and landslide events in Italy.
A global database of 2,626 rainfall events that have resulted in shallow landslides and debris fl... more A global database of 2,626 rainfall events that have resulted in shallow landslides and debris flows was compiled through a thorough literature search. The rainfall and landslide information was used to update the dependency of the minimum level of rainfall duration and intensity likely to result in shallow landslides and debris flows established by Nel Caine in 1980. The rainfall intensity–duration (ID) values were plotted in logarithmic coordinates, and it was established that with increased rainfall duration, the minimum average intensity likely to trigger shallow slope failures decreases linearly, in the range of durations from 10 min to 35 days. The minimum ID for the possible initiation of shallow landslides and debris flows was determined. The threshold curve was obtained from the rainfall data using an objective statistical technique. To cope with differences in the intensity and duration of rainfall likely to result in shallow slope failures in different climatic regions, the rainfall information was normalized to the mean annual precipitation and the rainy-day normal. Climate information was obtained from the global climate dataset compiled by the Climate Research Unit of the East Anglia University. The obtained global ID thresholds are significantly lower than the threshold proposed by Caine (Geogr Ann A 62:23–27, 1980), and lower than other global thresholds proposed in the literature. The new global ID thresholds can be used in a worldwide operational landslide warning system based on global precipitation measurements where local and regional thresholds are not available..
Land-surface model schemes for catchment-scale hydrology ideally require a good parameterization ... more Land-surface model schemes for catchment-scale hydrology ideally require a good parameterization of the sub-grid scale geomorphology. In particular, the morphology of the landscape at the hillslope scale needs to be quantified so that any hydrologi- cal parameterization can be reliably calibrated. Schemes that rely on kilometre-scale digital elevation model (DEM) data require some form of topographic downscaling, because such data are far too coarse to describe hillslope-scale morphology. In this study, we address the task of topographic downscaling and look in detail at the statis- tical behaviour of surface properties such as slope, upstream area, and TOPMODEL topographic index as a function of DEM resolution. We demonstrate that downscaling of such properties is theoretically possible. We examine the viability of downscaling data such as Hydro1k using SRTM (Shuttle Radar Topographic Mission) and Star3i high-resolution DEMs. Our preliminary analysis shows that topographic downscaling schemes take a very general, widely applicable form - but that a key parameter, the hillslope-channel transition scale, requires calibration using very high resolution DEM data.
When a slope begins to fail, how big will the ensuing landslide be? In mountainous environments t... more When a slope begins to fail, how big will the ensuing landslide be? In mountainous environments the span of possible landslide sizes is very large, with potential failures ranging in scale from meters to kilometers, but we remain unable to predict such sizes or their frequency. This inability persists despite the fact that we now have rather good constraints on the size-frequency distribution of landslides at a regional scale. Empirical studies have amply demonstrated that such distributions exhibit power-law scaling across a broad range of landslide areas. What they have not established is why such scaling occurs - at least there is no consensus in this regard. I present a simple, stochastic calculus model for slope failure can explain in full the observed size-frequency distribution of landslides, including the mean landslide size and both the scaling (power-law) and non-scaling components of the distribution. This model distribution is remarkably consistent with the larger, more reliable landslide data sets where power-law scaling can be estimated with some confidence. The success of the model and its underlying assumptions pose some interesting questions about the process of landsliding and our ability to predict it. The model also presents a viable alternative to those that assume self-organized criticality.
The dynamical behavior of mountain landscapes plays a key role in the interaction between tectoni... more The dynamical behavior of mountain landscapes plays a key role in the interaction between tectonics and geomorphology, and understanding this behavior is a major challenge. At the heart of the dynamics lies the spatio-temporal response of a channel network to changes in boundary conditions, such as fluctuations in climate, changes in tectonic forcing, single earthquake events, and so on. The dynamics set the rate at which a landscape converges to a steady state, if at all, and determine the response of a steady-state landscape to perturbations in the driving forces. Without a good understanding of the channel network dynamics, no reliable interpretations of spatio-temporal changes in natural landscapes can be made. By and large, most theoretical treatments of channel network and landscape dynamics (typically as landscape evolution models) have avoided treatment of the dynamical properties of channel morphology. In this exploratory study, I address the issue directly. I present a very simple model for a single channel link that encapsulates the dynamic roles that variable channel width, depth, sinuosity, bed roughness and sediment storage play in modulating the rate of sediment throughput. I focus on how the evolution of channel morphology is intimately connected to the process of bedrock channel erosion, and on how these processes determine the spatio-temporal response of a channel network and the nonlinear response of the whole landscape. Model solutions show that a single channel link can exhibit oscillatory, coupled variations in channel morphology and slope. These oscillations may decay to a stable channel slope or may reach a limit cycle where oscillations persist indefinitely, albeit over a stable range. The behavior of a coupled, double channel link model is similar. Both models predict that real mountain channels may be autocyclic, with stable oscillations in morphology and slope over the long term. They suggest that the dynamical behavior of an evolving landscape can only be modeled if the dynamics of channel morphology are taken into account.
In the 1990s, a flurry of research into river network self-organization culminated in the theory ... more In the 1990s, a flurry of research into river network self-organization culminated in the theory of Optimal Channel Networks and the idea that fluvial landscapes arise through a global minimization of stream power. Despite its elegance, OCN theory has not achieved widespread traction among geomorphologists, in part because its variational principle is not well understood, and in part because it solves only for the limit, stationary network configuration and not for the dynamics, i.e., it doesn't describe how the landscape evolves to the limit state. Here I present an alternate model in which self-organization is driven by asymmetric rates of landsliding at drainage divides. This new approach is termed Topographically Consistent Network theory, because topographic continuity between adjacent streams is achieved directly by relief minimization driven by hillslope erosion, in contrast to OCN theory where it arises as by-product of energy minimization. TCNs reproduce the area distributions and width functions seen in OCNs and in nature, with the advantage that the minimization principle is physically based (slope instability). The question then is: can TCN theory yield a functional that describes the dynamics of landscape evolution? Example TCN river network derived on a Voronoi lattice. Drainage divide motion and stream capture driven by landsliding reduces relief between adjacent sub-catchments and leads to self-organization.
The process of meander formation is commonly thought to be restricted to alluvial rivers and cons... more The process of meander formation is commonly thought to be restricted to alluvial rivers and constructive submarine channels, and when meanders are observed in bedrock rivers they are typically interpreted as an antecedent feature. In Taiwan we have found widespread evidence that this is not always the case and that instead bedrock rivers may actively meander. By active we mean that meanders can naturally initiate within a straight bedrock channel and grow into long-lived, highly sinuous reaches. The evidence for active bedrock meandering comes from across Taiwan and from two Taiwanese catchments in particular: the Jukou (western Central Range) and the Hsiukuluan (Coast Range). Our key field observations are the following: concave-up strath terraces merging into the active bedrock channel in established meander loops; flights of such straths topping gooseneck ridges and recording the growth of meander loops; lateral erosion by landsliding concentrated on the outer bends of meanders; breaching of goosenecks; cutoff meander loops. These features indicate a dynamic process of meandering; by contrast, erosional meanders inherited from an alluvial river would have a static geometrical form and would not systematically display any of these features. Moreover, it is almost inconceivable that bedrock meandering could be inherited on the scale observed in the Taiwanese uplands. We conclude that active river meandering can be a purely erosional process. This conclusion poses challenging questions about the basic process of meander formation.
It is notoriously difficult to extract from landscapes any quantitatively meaningful signatures o... more It is notoriously difficult to extract from landscapes any quantitatively meaningful signatures of climate or tectonics. While some topographic studies have been able to infer basic tectonic boundary conditions, they have shown little success so far in inferring the values of key climatic variables that shape the relief. We set ourselves the challenge of doing just that, and we chose as our focus the sinuosity of mountain rivers as measured using the SRTM digital elevation model. Bedrock rivers adopt a sinuous, sometimes meandering form, just as do alluvial rivers, and this sinuosity is achieved through both lateral and vertical erosion. Measurement of DEM-derived mountain river sinuosity, together with analysis of discharge statistics, could therefore provide constraints on channel incision processes. Our analysis of mountain catchments across the western North Pacific cyclone basin has revealed that sinuosity is not controlled by mean flow conditions. Rather, for geologically similar terrain, sinuosity appears to relate directly to the shape of the discharge distribution. Rivers that have more relative variance in discharge tend to have greater lateral mobility. Discharge variability is a function of rainfall variability, which is dominated in our study area by the climatology of tropical cyclones. We conclude that mountain river sinuosity in the western North Pacific is controlled by typhoon strike frequency.
Empirical constraints on the dynamics of very large landslides are hard to obtain. In principle, ... more Empirical constraints on the dynamics of very large landslides are hard to obtain. In principle, such constraints are to be found in the long-period seismic waves radiated by a landslide mass sliding down and variably loading the substrate beneath. Here we present inversions of long-period seismograms for the time-varying forces driving catastrophic landslides at a number of locations around the world. Once calibrated using remote-sensing imagery and differential topographic data (where possible), these ``Landslide Force History'' (LFH) inversions indicate masses of 0.4-7× 1011kg accelerated at 1-3 m/s2. Integration of the LFH inversions indicates the mean landslide masses reached speeds of 25-75 m/s and traveled distances of 800m to 6000m in 50-160s along trajectories strongly controlled by relief. Analysis of the rates of change of potential and kinetic energy indicate mean effective friction coefficients of 0.2--0.3. Fundamental properties of catastrophic landsliding can therefore be estimated by remote observation, and global assessment of major landslides is feasible in near-real-time.
The Calabrian forearc of southern Italy is characterized by rapid post-Miocene uplift (~1 mm/y) w... more The Calabrian forearc of southern Italy is characterized by rapid post-Miocene uplift (~1 mm/y) whose tectonic origin remains a matter of dispute. Evidence for this uplift is recorded in the Calabrian peninsula in at least two ways. First, the landscape is cut by paleo-cliffs and marine abrasion surfaces that form staircases of abandoned shorelines. Second, valley-fringing sequences of coarse clastics exhibit magnificent Gilbert-type fan deltas of marine origin. These two phenomena have generally been studied separately, even though both provide excellent and likely coeval constraints on relative sea-level during the late Pleistocene. Here we present a preliminary synthesis of the Crati Valley deltas and the strandlines of the Sibari Plain. Through a combination of field observation and analysis of air photographs and digital elevation (SRTM DEM) data, we examine the relative timing of fan delta deposition, marine terrace formation, and penecontemporaneous faulting. This expanded correlation of coeval surfaces may help to establish whether the uplift is uniform or time transgressive. We evaluate our results in light of an oft-quoted model for Calabrian uplift, in which the rapid rise of the peninsula is thought to have begun at around 700~ka. Our results highlight the urgent need for better age control of marine deposits and surfaces tentatively assigned (through correlation and extrapolation) to the late Pleistocene.
Landslide-driven erosion is controlled by the scale and frequency of slope failures and by the co... more Landslide-driven erosion is controlled by the scale and frequency of slope failures and by the consequent fluxes of debris off the hillslopes. Here I focus on the magnitude-frequency part of the process and develop a theory of initial slope failure and debris mobilization that reproduces the heavy-tailed distributions (PDFs) observed for landslide source areas and volumes. Landslide rupture propagation is treated as a quasi-static, non-inertial process of simplified elastoplastic deformation with strain weakening; debris runout is not considered. The model tracks the stochastically evolving imbalance of frictional, cohesive, and body forces across a failing slope, and uses safety-factor concepts to convert the evolving imbalance into a series of incremental rupture growth or arrest probabilities. A single rupture is simulated with a sequence of weighted ``coin tosses'' with weights set by the growth probabilities. Slope failure treated in this stochastic way is a survival process that generates asymptotically power-law-tail PDFs of area and volume for rock and debris slides; predicted scaling exponents are consistent with analyses of landslide inventories. The primary control on the shape of the model PDFs is the relative importance of cohesion over friction in setting slope stability: the scaling of smaller, shallower failures, and the size of the most common landslide volumes, are the result of the low cohesion of soil and regolith, whereas the negative power-law tail scaling for larger failures is tied to the greater cohesion of bedrock. The debris budget may be dominated by small or large landslides depending on the scaling of both the PDF and of the depth-length relation. I will present new model results that confirm the hypothesis that depth-length scaling is linear. Model PDF of landslide volumes.
We present a preliminary tectonic and geomorphologic analysis of SRTM data from western Turkey. O... more We present a preliminary tectonic and geomorphologic analysis of SRTM data from western Turkey. Our method of analysis is based on a delineation of surface flow patterns using a fast, multipixel outflow algorithm. The algorithm is designed to map, on a regional scale, surface drainage patterns that may include diffuse flow (e.g., alluvial fans) and braided channels. The algorithm is a signficant improvement over off-the-shelf (e.g., ArcGIS) schemes that impose single pixel outflow and everywhere-dendritic drainage, and over specialist multipixel outflow schemes that cannot handle >10^7 pixel DEMs. Images of model surface flow provide an excellent means of using SRTM data to study interactions between faulting and drainage patterns. As an example, we discuss some observations of drainage perturbation induced by the propagation of the North Anatolian Fault.
Ancient coastlines, in the form of abrasion platforms or coastal deposits, provide markers in the... more Ancient coastlines, in the form of abrasion platforms or coastal deposits, provide markers in the landscape that can be correlated and dated to determine patterns and rates of uplift. We present new data on an excellently exposed flight of Pleistocene coastlines to constrain patterns of uplift associated with rollback subduction and arc-continent collision in the Calabrian arc and the southern Apennines. While many aspects of this type orogen of Mediterranean tectonics are coming into focus, major issues are still unresolved. A more complete and detailed resolution of vertical tectonics is expected to discern between competing hypotheses: for instance, whether or not subduction is still ongoing. We present results from a preliminary database of marine terraces mapped on a regional scale with an SRTM-3arcsec DEM and field observations. Supplementary to the geometric correlation of terraces, we are currently processing samples for cosmogenic isotope analysis to determine age relationships. Four prominent, seaward dipping terrace levels were identified on the west coast at heights of 100m, 170m, 400m and 700m, which we call T1, T2, T3 and T4 respectively. The upper terrace levels (T2, T3 and T4) comprise both bedrock-carved platforms and fossiliferous coastal deposits, with rounded paleocliff edges that form clear scarps in the DEM. The lower terrace level (T1) consists of shallow marine deposits with fossiliferous horizons containing in-situ coral. The geographic distribution of the T4 terrace level is particularly extensive, forming a continuous ridge along the backbone of Calabria, in the footwall of a major N-S trending extensional fault bounding the Mésima Valley. Similar terraces with well-defined sea-cliffs and thick accumulations of shallow marine deposits were identified on the eastern flank of Calabria. We believe this is the first documentation of terraces at this level on the east coast, and is a particularly interesting find, as it suggests that the earliest uplift affected both sides of Calabria. In the Mésima Valley, fossiliferous shallow marine sediments, with abundant oysters, dip at 10° E, which we interpret as hangingwall tilt toward the eastern-bounding fault of the Mésima Valley. These results demonstrate that ancient marine shorelines provide a reliable datum for calibration of (1) regional vertical motion since the formation of the earliest-formed terrace, and (2) local tectonic deformation following terrace formation.
ABSTRACT Accurately predicting rates of coarse sediment transport in river channels is a central ... more ABSTRACT Accurately predicting rates of coarse sediment transport in river channels is a central goal of fluvial geomorphology and civil engineering. However, it is difficult to evaluate sediment transport and bedrock abrasion models in large rivers because quantitative measures of bedload transport are labor intensive and often dangerous to obtain in floods. Two recent studies show that the amplitude of seismic waves near rivers may record bedload flux, indicating that seismometers near rivers provide a potential means of monitoring bedload transport. In an effort to better interpret seismic waves generated by rivers, we seek a relationship between the variables governing bedload transport and seismic waves. Our approach relies on the fact that elastic waves are generated when momentum is transferred to the bed during a bedload particle impact. For an impacting particle of known mass and velocity, the momentum transfer can be computed from Hertzian impact theory. Here we combine analytic results based on Hertzian and elastic wave theories with empirical equations developed to describe the ballistics of bedload particles in terms of fluid shear stress and grain size. From this synthesis we arrive at a semi-analytic expression that predicts how the characteristic frequencies and amplitudes of seismic waves generated from saltating bedload particles should scale with fluid shear stress, grain size, and coarse sediment flux. Preliminary tests of our predictions using previously published and newly acquired laboratory data indicate that seismic signals near rivers can record information about the size, velocity and number of particles impacting the bed. Additionally, our analytical results help identify bedload transport events in seismic data collected along the Chijiawan River in Taiwan. Here the river is evolving rapidly in response to a dam removal - resulting in predictable changes in bedload transport efficiency in time and space that we can compare to local seismic data.
ABSTRACT The Chijiawan river, located in the Hsuehshan Range in central Taiwan, is a steep, coars... more ABSTRACT The Chijiawan river, located in the Hsuehshan Range in central Taiwan, is a steep, coarse-grained, mixed bedrock-alluvial channel subject to heavy monsoonal rainfall, occasional typhoons and frequent floods. In May 2011, a 15m-high check dam holding back around 200,000m3 of sediment was removed. To monitor subsequent channel response, an array of instruments were deployed prior to dam removal. Given the anticipated magnitude of sediment transport and excellent empirical constraints, the Chijiawan channel is an ideal location to study fluvial incision processes into bedrock. There is abundant bedrock exposed in the river bed downstream of the dam, and we are interested in how the pulse of sediment following dam removal will impact bedrock erosion rates. We monitor erosion using repeat high resolution laser scanning at 8 sites spread along a 3 km long reach, 2 sites upstream of the dam and 6 downstream. We scan two sites with a Konica Minolta Vivid 910 scanner (with an accuracy of 1.4mm), and the remaining sites with a Faro Photon 120 scanner (with an accuracy of 2mm). Scanning at ranges of ~1m (Vivid) to ~10m (Faro), we obtain resolutions of 1 to 3mm. Based on the system developed by Wilson and Hovius (2010), reference frames for repeat surveys consist of permanent stainless steel sockets installed in 14 cm deep holes drilled into the bedrock, combined with removable targets that screw into the sockets for scanning. Vivid data are processed and 3D models are constructed using Rapidform 2004 software. Faro data is processed using Faro's Scene software. In order to compare the high resolution scanning with more traditional low-tech monitoring methods, we have also installed two sets of concrete expansion bolts. We obtain profiles between bolts with both a contour gauge and the scanners. The high accuracy and resolution of these scanners, and the system of precise benchmarks enable us to detect extremely small amounts of erosion. By repeat scanning after individual flood events, we can directly relate magnitudes and patterns of erosion with flood characteristics and measures of sediment flux. Citation Wilson A, and Hovius N, 2010, Upstream facing convex surfaces and the importance of bedload in fluvial bedrock incision: observations from Taiwan, Geophysical Research Abstracts, Vol. 12, EGU2010-4978.
We have compiled a database of floods and landslides that occurred in Italy between AD 1279 and 2... more We have compiled a database of floods and landslides that occurred in Italy between AD 1279 and 2002 and caused deaths, missing persons, injuries, and homelessness. Analysis of the database indicates that more than 50,593 people died, went missing, or were injured in 2580 flood and landslide events. Harmful events were inventoried in 26.3% of the 8103 Italian municipalities. Fatal events were most frequent in the Alpine regions of northern Italy and were caused by both floods and landslides. In southern Italy, landslides were the principal agents of fatalities and were most numerous in the Campania region. Casualties were most frequent in the autumn. Fast-moving landslides, including rock falls, rockslides, rock avalanches, and debris flows, caused the largest number of deaths. In order to assess the overall risk posed by these processes, we merged the historical catalogs and identified 2682 “hydrogeomorphological” events that triggered single or multiple landslides and floods. We estimated individual risk through the calculation of mortality rates for both floods and landslides and compared these rates to the death rates for other natural, medical, and human-induced hazards in Italy. We used the frequency distribution of events with fatalities to ascertain the magnitude and frequency of the societal risks posed by floods and landslides. We quantified these risks in a Bayesian model that describes the probabilities of fatal flood and landslide events in Italy.
A global database of 2,626 rainfall events that have resulted in shallow landslides and debris fl... more A global database of 2,626 rainfall events that have resulted in shallow landslides and debris flows was compiled through a thorough literature search. The rainfall and landslide information was used to update the dependency of the minimum level of rainfall duration and intensity likely to result in shallow landslides and debris flows established by Nel Caine in 1980. The rainfall intensity–duration (ID) values were plotted in logarithmic coordinates, and it was established that with increased rainfall duration, the minimum average intensity likely to trigger shallow slope failures decreases linearly, in the range of durations from 10 min to 35 days. The minimum ID for the possible initiation of shallow landslides and debris flows was determined. The threshold curve was obtained from the rainfall data using an objective statistical technique. To cope with differences in the intensity and duration of rainfall likely to result in shallow slope failures in different climatic regions, the rainfall information was normalized to the mean annual precipitation and the rainy-day normal. Climate information was obtained from the global climate dataset compiled by the Climate Research Unit of the East Anglia University. The obtained global ID thresholds are significantly lower than the threshold proposed by Caine (Geogr Ann A 62:23–27, 1980), and lower than other global thresholds proposed in the literature. The new global ID thresholds can be used in a worldwide operational landslide warning system based on global precipitation measurements where local and regional thresholds are not available..
Land-surface model schemes for catchment-scale hydrology ideally require a good parameterization ... more Land-surface model schemes for catchment-scale hydrology ideally require a good parameterization of the sub-grid scale geomorphology. In particular, the morphology of the landscape at the hillslope scale needs to be quantified so that any hydrologi- cal parameterization can be reliably calibrated. Schemes that rely on kilometre-scale digital elevation model (DEM) data require some form of topographic downscaling, because such data are far too coarse to describe hillslope-scale morphology. In this study, we address the task of topographic downscaling and look in detail at the statis- tical behaviour of surface properties such as slope, upstream area, and TOPMODEL topographic index as a function of DEM resolution. We demonstrate that downscaling of such properties is theoretically possible. We examine the viability of downscaling data such as Hydro1k using SRTM (Shuttle Radar Topographic Mission) and Star3i high-resolution DEMs. Our preliminary analysis shows that topographic downscaling schemes take a very general, widely applicable form - but that a key parameter, the hillslope-channel transition scale, requires calibration using very high resolution DEM data.
When a slope begins to fail, how big will the ensuing landslide be? In mountainous environments t... more When a slope begins to fail, how big will the ensuing landslide be? In mountainous environments the span of possible landslide sizes is very large, with potential failures ranging in scale from meters to kilometers, but we remain unable to predict such sizes or their frequency. This inability persists despite the fact that we now have rather good constraints on the size-frequency distribution of landslides at a regional scale. Empirical studies have amply demonstrated that such distributions exhibit power-law scaling across a broad range of landslide areas. What they have not established is why such scaling occurs - at least there is no consensus in this regard. I present a simple, stochastic calculus model for slope failure can explain in full the observed size-frequency distribution of landslides, including the mean landslide size and both the scaling (power-law) and non-scaling components of the distribution. This model distribution is remarkably consistent with the larger, more reliable landslide data sets where power-law scaling can be estimated with some confidence. The success of the model and its underlying assumptions pose some interesting questions about the process of landsliding and our ability to predict it. The model also presents a viable alternative to those that assume self-organized criticality.
The dynamical behavior of mountain landscapes plays a key role in the interaction between tectoni... more The dynamical behavior of mountain landscapes plays a key role in the interaction between tectonics and geomorphology, and understanding this behavior is a major challenge. At the heart of the dynamics lies the spatio-temporal response of a channel network to changes in boundary conditions, such as fluctuations in climate, changes in tectonic forcing, single earthquake events, and so on. The dynamics set the rate at which a landscape converges to a steady state, if at all, and determine the response of a steady-state landscape to perturbations in the driving forces. Without a good understanding of the channel network dynamics, no reliable interpretations of spatio-temporal changes in natural landscapes can be made. By and large, most theoretical treatments of channel network and landscape dynamics (typically as landscape evolution models) have avoided treatment of the dynamical properties of channel morphology. In this exploratory study, I address the issue directly. I present a very simple model for a single channel link that encapsulates the dynamic roles that variable channel width, depth, sinuosity, bed roughness and sediment storage play in modulating the rate of sediment throughput. I focus on how the evolution of channel morphology is intimately connected to the process of bedrock channel erosion, and on how these processes determine the spatio-temporal response of a channel network and the nonlinear response of the whole landscape. Model solutions show that a single channel link can exhibit oscillatory, coupled variations in channel morphology and slope. These oscillations may decay to a stable channel slope or may reach a limit cycle where oscillations persist indefinitely, albeit over a stable range. The behavior of a coupled, double channel link model is similar. Both models predict that real mountain channels may be autocyclic, with stable oscillations in morphology and slope over the long term. They suggest that the dynamical behavior of an evolving landscape can only be modeled if the dynamics of channel morphology are taken into account.
In the 1990s, a flurry of research into river network self-organization culminated in the theory ... more In the 1990s, a flurry of research into river network self-organization culminated in the theory of Optimal Channel Networks and the idea that fluvial landscapes arise through a global minimization of stream power. Despite its elegance, OCN theory has not achieved widespread traction among geomorphologists, in part because its variational principle is not well understood, and in part because it solves only for the limit, stationary network configuration and not for the dynamics, i.e., it doesn't describe how the landscape evolves to the limit state. Here I present an alternate model in which self-organization is driven by asymmetric rates of landsliding at drainage divides. This new approach is termed Topographically Consistent Network theory, because topographic continuity between adjacent streams is achieved directly by relief minimization driven by hillslope erosion, in contrast to OCN theory where it arises as by-product of energy minimization. TCNs reproduce the area distributions and width functions seen in OCNs and in nature, with the advantage that the minimization principle is physically based (slope instability). The question then is: can TCN theory yield a functional that describes the dynamics of landscape evolution? Example TCN river network derived on a Voronoi lattice. Drainage divide motion and stream capture driven by landsliding reduces relief between adjacent sub-catchments and leads to self-organization.
The process of meander formation is commonly thought to be restricted to alluvial rivers and cons... more The process of meander formation is commonly thought to be restricted to alluvial rivers and constructive submarine channels, and when meanders are observed in bedrock rivers they are typically interpreted as an antecedent feature. In Taiwan we have found widespread evidence that this is not always the case and that instead bedrock rivers may actively meander. By active we mean that meanders can naturally initiate within a straight bedrock channel and grow into long-lived, highly sinuous reaches. The evidence for active bedrock meandering comes from across Taiwan and from two Taiwanese catchments in particular: the Jukou (western Central Range) and the Hsiukuluan (Coast Range). Our key field observations are the following: concave-up strath terraces merging into the active bedrock channel in established meander loops; flights of such straths topping gooseneck ridges and recording the growth of meander loops; lateral erosion by landsliding concentrated on the outer bends of meanders; breaching of goosenecks; cutoff meander loops. These features indicate a dynamic process of meandering; by contrast, erosional meanders inherited from an alluvial river would have a static geometrical form and would not systematically display any of these features. Moreover, it is almost inconceivable that bedrock meandering could be inherited on the scale observed in the Taiwanese uplands. We conclude that active river meandering can be a purely erosional process. This conclusion poses challenging questions about the basic process of meander formation.
It is notoriously difficult to extract from landscapes any quantitatively meaningful signatures o... more It is notoriously difficult to extract from landscapes any quantitatively meaningful signatures of climate or tectonics. While some topographic studies have been able to infer basic tectonic boundary conditions, they have shown little success so far in inferring the values of key climatic variables that shape the relief. We set ourselves the challenge of doing just that, and we chose as our focus the sinuosity of mountain rivers as measured using the SRTM digital elevation model. Bedrock rivers adopt a sinuous, sometimes meandering form, just as do alluvial rivers, and this sinuosity is achieved through both lateral and vertical erosion. Measurement of DEM-derived mountain river sinuosity, together with analysis of discharge statistics, could therefore provide constraints on channel incision processes. Our analysis of mountain catchments across the western North Pacific cyclone basin has revealed that sinuosity is not controlled by mean flow conditions. Rather, for geologically similar terrain, sinuosity appears to relate directly to the shape of the discharge distribution. Rivers that have more relative variance in discharge tend to have greater lateral mobility. Discharge variability is a function of rainfall variability, which is dominated in our study area by the climatology of tropical cyclones. We conclude that mountain river sinuosity in the western North Pacific is controlled by typhoon strike frequency.
Empirical constraints on the dynamics of very large landslides are hard to obtain. In principle, ... more Empirical constraints on the dynamics of very large landslides are hard to obtain. In principle, such constraints are to be found in the long-period seismic waves radiated by a landslide mass sliding down and variably loading the substrate beneath. Here we present inversions of long-period seismograms for the time-varying forces driving catastrophic landslides at a number of locations around the world. Once calibrated using remote-sensing imagery and differential topographic data (where possible), these ``Landslide Force History'' (LFH) inversions indicate masses of 0.4-7× 1011kg accelerated at 1-3 m/s2. Integration of the LFH inversions indicates the mean landslide masses reached speeds of 25-75 m/s and traveled distances of 800m to 6000m in 50-160s along trajectories strongly controlled by relief. Analysis of the rates of change of potential and kinetic energy indicate mean effective friction coefficients of 0.2--0.3. Fundamental properties of catastrophic landsliding can therefore be estimated by remote observation, and global assessment of major landslides is feasible in near-real-time.
The Calabrian forearc of southern Italy is characterized by rapid post-Miocene uplift (~1 mm/y) w... more The Calabrian forearc of southern Italy is characterized by rapid post-Miocene uplift (~1 mm/y) whose tectonic origin remains a matter of dispute. Evidence for this uplift is recorded in the Calabrian peninsula in at least two ways. First, the landscape is cut by paleo-cliffs and marine abrasion surfaces that form staircases of abandoned shorelines. Second, valley-fringing sequences of coarse clastics exhibit magnificent Gilbert-type fan deltas of marine origin. These two phenomena have generally been studied separately, even though both provide excellent and likely coeval constraints on relative sea-level during the late Pleistocene. Here we present a preliminary synthesis of the Crati Valley deltas and the strandlines of the Sibari Plain. Through a combination of field observation and analysis of air photographs and digital elevation (SRTM DEM) data, we examine the relative timing of fan delta deposition, marine terrace formation, and penecontemporaneous faulting. This expanded correlation of coeval surfaces may help to establish whether the uplift is uniform or time transgressive. We evaluate our results in light of an oft-quoted model for Calabrian uplift, in which the rapid rise of the peninsula is thought to have begun at around 700~ka. Our results highlight the urgent need for better age control of marine deposits and surfaces tentatively assigned (through correlation and extrapolation) to the late Pleistocene.
Landslide-driven erosion is controlled by the scale and frequency of slope failures and by the co... more Landslide-driven erosion is controlled by the scale and frequency of slope failures and by the consequent fluxes of debris off the hillslopes. Here I focus on the magnitude-frequency part of the process and develop a theory of initial slope failure and debris mobilization that reproduces the heavy-tailed distributions (PDFs) observed for landslide source areas and volumes. Landslide rupture propagation is treated as a quasi-static, non-inertial process of simplified elastoplastic deformation with strain weakening; debris runout is not considered. The model tracks the stochastically evolving imbalance of frictional, cohesive, and body forces across a failing slope, and uses safety-factor concepts to convert the evolving imbalance into a series of incremental rupture growth or arrest probabilities. A single rupture is simulated with a sequence of weighted ``coin tosses'' with weights set by the growth probabilities. Slope failure treated in this stochastic way is a survival process that generates asymptotically power-law-tail PDFs of area and volume for rock and debris slides; predicted scaling exponents are consistent with analyses of landslide inventories. The primary control on the shape of the model PDFs is the relative importance of cohesion over friction in setting slope stability: the scaling of smaller, shallower failures, and the size of the most common landslide volumes, are the result of the low cohesion of soil and regolith, whereas the negative power-law tail scaling for larger failures is tied to the greater cohesion of bedrock. The debris budget may be dominated by small or large landslides depending on the scaling of both the PDF and of the depth-length relation. I will present new model results that confirm the hypothesis that depth-length scaling is linear. Model PDF of landslide volumes.
We present a preliminary tectonic and geomorphologic analysis of SRTM data from western Turkey. O... more We present a preliminary tectonic and geomorphologic analysis of SRTM data from western Turkey. Our method of analysis is based on a delineation of surface flow patterns using a fast, multipixel outflow algorithm. The algorithm is designed to map, on a regional scale, surface drainage patterns that may include diffuse flow (e.g., alluvial fans) and braided channels. The algorithm is a signficant improvement over off-the-shelf (e.g., ArcGIS) schemes that impose single pixel outflow and everywhere-dendritic drainage, and over specialist multipixel outflow schemes that cannot handle >10^7 pixel DEMs. Images of model surface flow provide an excellent means of using SRTM data to study interactions between faulting and drainage patterns. As an example, we discuss some observations of drainage perturbation induced by the propagation of the North Anatolian Fault.
Ancient coastlines, in the form of abrasion platforms or coastal deposits, provide markers in the... more Ancient coastlines, in the form of abrasion platforms or coastal deposits, provide markers in the landscape that can be correlated and dated to determine patterns and rates of uplift. We present new data on an excellently exposed flight of Pleistocene coastlines to constrain patterns of uplift associated with rollback subduction and arc-continent collision in the Calabrian arc and the southern Apennines. While many aspects of this type orogen of Mediterranean tectonics are coming into focus, major issues are still unresolved. A more complete and detailed resolution of vertical tectonics is expected to discern between competing hypotheses: for instance, whether or not subduction is still ongoing. We present results from a preliminary database of marine terraces mapped on a regional scale with an SRTM-3arcsec DEM and field observations. Supplementary to the geometric correlation of terraces, we are currently processing samples for cosmogenic isotope analysis to determine age relationships. Four prominent, seaward dipping terrace levels were identified on the west coast at heights of 100m, 170m, 400m and 700m, which we call T1, T2, T3 and T4 respectively. The upper terrace levels (T2, T3 and T4) comprise both bedrock-carved platforms and fossiliferous coastal deposits, with rounded paleocliff edges that form clear scarps in the DEM. The lower terrace level (T1) consists of shallow marine deposits with fossiliferous horizons containing in-situ coral. The geographic distribution of the T4 terrace level is particularly extensive, forming a continuous ridge along the backbone of Calabria, in the footwall of a major N-S trending extensional fault bounding the Mésima Valley. Similar terraces with well-defined sea-cliffs and thick accumulations of shallow marine deposits were identified on the eastern flank of Calabria. We believe this is the first documentation of terraces at this level on the east coast, and is a particularly interesting find, as it suggests that the earliest uplift affected both sides of Calabria. In the Mésima Valley, fossiliferous shallow marine sediments, with abundant oysters, dip at 10° E, which we interpret as hangingwall tilt toward the eastern-bounding fault of the Mésima Valley. These results demonstrate that ancient marine shorelines provide a reliable datum for calibration of (1) regional vertical motion since the formation of the earliest-formed terrace, and (2) local tectonic deformation following terrace formation.
Uploads
Papers by Colin Stark