Mapping block-and-ash flow hazards based on Titan 2D simulations: a case study from Mt. Taranaki, NZ
Numerical models for simulation of mass flows are typically focussed upon accurately predicting the paths, travel times and inundation from a single flow or collapse event. When considering catchment-based hazards from a volcano, this is complicated by often being faced with several possible scenarios. Over the last 800 years at Mt. Taranaki/Egmont, a number of dome growth and collapse events have resulted in the genesis and emplacement of block-and-ash flows (BAFs). Each BAF was directed northwestward by a breach in the crater rim. The latest dome collapse events in the AD 1880s and AD 1755 inundated the northwestern flank and had run-out lengths 10 km from source. Future activity of this type could have a devastating effect on the Taranaki region’s communities, infrastructure and economy. Hazard planning has involved constructing volcanic hazard maps based upon the areas inundated by past volcanic flows, with little consideration of present-day topography. Here, a numerical geophysical mass flow modelling approach is used to forecast the hazards of future comparable BAF events on NW Mt. Taranaki. The Titan2D programme encompasses a “shallow water”, continuum solution-based, granular flow model. Flow mechanical properties needed for this approach include estimates of internal and basal friction as well as the physical dimensions of the initial collapse. Before this model can be applied to Taranaki BAFs, the input parameters must be calibrated by simulating a range of past collapse events. By using AD 1860 and AD 1755 scenarios, initial collapse volumes can be well constrained and internal and basal friction angles can be evaluated through an iterative approach from previous run-out lengths. A range of possible input parameters was, therefore, determined to produce a suite of potentially inundated areas under present-day terrain. A suite of 10 forecasts from a uniformly distributed range were combined to create a map of relative probabilities of inundation by future BAF events. These results were combined in a GIS package to produce hazard zones related to user-specified hazard thresholds. Using these input parameter constraints, future hazard forecasts for this scale and type of event can also take into account changing summit and topographic configurations following future eruptive or collapse events.
KeywordsBlock-and-ash flows Titan2D Mt. Taranaki (Mt. Egmont) Volcanic hazard map Lava dome
Repeated block-and-ash flows (BAFs) formed from the partial or total collapse of viscous lava domes are common and deadly hazards on many composite volcanoes with lava domes, as shown recently at Unzen, Japan (1991–1995; Nakada and Fujii 1993), Gunung Merapi, Indonesia (1984, 1994, 1997–1998, 2006; Schwarzkopf et al. 2005), Colima, Mexico (1991, 1998; Saucedo et al. 2004) and Soufrière Hills on Montserrat (1995-present-day; Sparks and Young 2002). Between eruptive events, current methods to evaluate BAF hazards rely primarily on reconstructions from historical and stratigraphic records to ultimately produce a hazard map. These hindcasting methods, particularly with respect to inundation maps, are based on paleo-topography and paleo-drainage, and it is not clear how these can be applied to current landscapes or the prediction of future events.
Methods to define volcanic hazard zones were outlined by Crandell et al. (1984) and Scott et al. (2001), who highlighted the use of detailed geological investigations, combined with return period information, to produce high-quality hazard assessments. By combining return periods of volcanic mass flows of differing inundation patterns, Crandell et al. (1984) displayed a progressively changing risk of the likelihood of impact by using a system of graduating colours. This process, however, was rarely transferred into practise. With the development of Geographical Information Systems (GIS) and increasing computer power, numerical simulations of volcanic events are becoming more common (Canuti et al. 2002; Iverson et al. 1998; Bonadonna et al. 2005, Magill et al. 2006). Typically, these methods are applied in combination with geological mapping and focus on individual scenarios, where impacts are on restricted points of interest.
For assessment of pyroclastic flow hazards, Malin and Sheridan (1982) used the Heim coefficient (∆H/L) to identify run-out length-based hazard zones in relation to the current topography. Sheridan et al. (2000) and Toyos et al. (2007) continued to apply computer models to identify hazard zones based on the run out and velocity of small-volume pyroclastic flows, yet these zones usually encompassed the entire cone and ignored confining topography. Similar BAF modelling was undertaken by Saucedo et al. (2004) and Itoh et al. (2000) on Colima and Merapi volcanoes, respectively, with the outputs from individual scenarios being the basis for hazard assessment. Related work in lahar hazards analysis (Canuti et al. 2002; Schilling 1998; Stevens et al. 2003) compared the effectiveness of the computer simulations between FLO-2D and LAHARZ.
The varying input parameters used in this broad range of models highlight a need for developing clearly definable constraints on the range of input parameters required for simulations. This would ensure accurate identification of hazard and provide a robust method of optimisation against real events. In addition, an effective display of the associated uncertainties or sensitivity of these scenarios also needs to be devised.
Recently developed depth-averaged 2D simulation codes offer a new potential to incorporate the effect of present (and changeable) terrain on the inundation and run out of mass flows. The SUNY-Buffalo developed “Titan2D” granular flow code has proven effective for modelling dry rock collapses, either in cold (Little Tahoma Peak: Sheridan et al. 2005) or hot conditions (Volcán Colima: Rupp et al. 2003, 2006). These provided validation information for individual scenarios. However, for long-term hazard forecasts, considerable uncertainty exists in factors controlling future events, including: volume of material collapsing, initiation points, triggering mechanisms and the conditions of collapsing and flowing materials. Hence, a single-modelled scenario may not lead to a reliable hazard map. In the ideal case, a map or dynamic hazard assessment should incorporate the probability of any area being inundated by a mass flow within a range of known (or geologically possible) events from the eruptive centre, or within a particular catchment studied.
2 Mt. Taranaki mass flow hazards
The andesitic stratovolcano Mt. Taranaki (2,518 m) is situated in the western North Island of New Zealand. Since inception at ≥130 ka, the cone has experienced a cyclic pattern of growth through accumulation of lava flows and pyroclastic deposits, alternating with destruction through debris avalanches (Alloway et al. 2005; Zernack et al. 2009). The latest constructional phase of the stratocone (<10,000 years) is focussed around two main vents; at the volcano’s summit and a parasitic cone (Fanthams Peak) located directly to the south (Fig. 1). The dominant Holocene style of volcanism has involved frequent dome emplacement and collapse events with associated tephra falls (Turner et al. 2008). The last 1,000 years of explosive and extrusive activity has produced at least ten major ash fallout-producing episodes (Platz et al. 2007; Turner et al. 2008). In addition, Neall (1979) identified up to 14 mass flow units (debris flow, pyroclastic flow deposits) associated with eruptive activity during the last 500 years concentrated on the NW flanks. An update of this work by Platz (2007) indicates that 10 separate eruptive episodes occurred, between 878 ± 39 years B.P. and c. AD 1850, almost all of which produced domes in the summit crater and one or several mass flows (BAFs or cold rock avalanches) on the NW flanks. The lack of a defined crater rim to the NW has resulted in mass flow hazards being directed to the NW flanks (Platz 2007).
The internal structures of the Mt. Taranaki BAFs are similar to those described from many historical examples (e.g., Boudon et al. 1993; Miyabuchi 1999; Cole et al. 2002). They commonly contain evidence of hot emplacement, including partially or fully charcoalised wood, gas escape (pipe) structures (typically above large charcoalised logs) and oxidised haematite-stained upper portions. Many units show poorly sorted, pinching, bedded ash ground surge units, containing soil rip-up clasts beneath the main body of deposit. Coarse-tail reverse grading is common (e.g., Palladino and Valentine 1995) along with weakly developed clast trains and localised clast-supported lenses.
Constraints on the best exposed BAF deposits preserved on the NW Flanks (from Platz 2007)
Mass flow type
Min. volume estimate (m3)
Run-out distance (km)
c. AD 1880
1–2 × 106
c. AD 1755
5 × 106
c. AD 1655
10–15 × 106
c. AD 1555
7–11 × 106
11–15 × 106
7–11 × 106
3 Titan2D geophysical mass flow simulation code
The Titan2D code was designed by the Geophysical Mass Flow Modelling Group at SUNY Buffalo (Pitman et al. 2003; Patra et al. 2005) to simulate a dry granular flow from an initial point of collapse over a natural terrain. These pre-conditions are highly suited to simulating landslides and BAFs that form from the collapse of large portions of lava domes. The code is based on a model for an incompressible Coulomb flow adapted from the work of Savage and Hutter (1989). It uses a “shallow-water”, depth-averaged approximation, simplifying the complex 3D phenomena (after Iverson and Denlinger 2001). This assumption is grounded on the fact that compared to the entire area over which a long-run-out mass flow travels and deposits, its thickness in comparison is small. Mass and momentum conservation equations are solved with a Coulomb friction term for the interface between the granular material and basal surface and for the internal friction of the flowing media (Pitman et al. 2003). Conservation of energy is neglected in the first order since it is assumed that a pure coarse-grained rock avalanche has insufficient heat to affect a propagating avalanche.
This hyperbolic system of equations is solved using a parallel adaptive mesh, Godunov solver (Patra et al. 2005). Adaptive gridding and the use of a Message Passing Interface (MPI) allows for calculations to be spread across multiple processes, increasing the computational power and decreasing the computing time. This combination can also produce a more accurate simulation (Patra et al. 2005), when the computational mesh is able to sample a grid of digital terrain data at a high resolution. The use of an adaptive grid refines the simulation according to the region more likely to be affected, or where rapid changes in mass distribution occur, i.e., at the flow front. The process also allows the mesh to be unrefined in tail areas. In general, this allows large (detailed) terrain grid files to be used without re-sampling, while also minimising memory requirements and computer power.
Titan2D operates in a LINUX environment via a Python scripted graphical user interface (GUI). Terrain data are entered into the simulation via the GRASS (Geographic Resources Analysis Support System; US Army Corps of Engineers’ CERL) GIS environment and format. Simulations on real terrain usually require a large amount of pre-processing and re-sampling of the original DEM to generate a new grid. This is avoided in Titan2D by integrating the model with GRASS GIS and adaptive gridding.
Defining dimensions of an initial “pile” of material; including shape, footprint, height, volume, position and initial velocity (if required);
A variable that nominally represents the angle of internal friction of the granular pile;
A variable that nominally represents the angle of basal friction between the granular pile and the substrate;
A grid by which differing substrates (based on roughness, vegetation, slope) can be defined, each with differing basal-friction angles (if needed);
Stopping criteria to halt the simulation (normally a limit on “simulated—real time” or the number of computational time steps);
Providing a 3D grid containing topographic information (x, y, z; i.e., a Digital Elevation Model (DEM) for the simulation area).
Titan2D has been applied to and evaluated against small-scale pyroclastic flows, BAFs and rock avalanches on Volcan de Colima, Mexico (Pitman et al. 2003; Rupp et al. 2006), El Misti Peru (Delaite et al. 2005) and Little Tahoma Peak (Sheridan et al. 2005). The outcomes of these studies have highlighted the uncertainties in objectively defining model input parameters for realistic simulation of local flow conditions. These have hence only taken a first-order approach to hazard evaluation. For the creation of a hazard map that encompasses all geologically reasonable possibilities, there is a need to; (1) create a more encompassing range of probable scenarios for each of these and (2) include a range of model controlling parameters.
4 Case study—BAF and rock avalanche scenarios at Mt. Taranaki
The last major sequence of dome building, AD 1700–1850, from Mt. Taranaki was followed by the westward collapse of around 5 × 106 m3 of rock from the summit area (Fig. 2), leaving a half-sectioned dome structure at the summit of approximately 2 × 106 m3 (Platz 2007). This collapse is characterised by at least two stages, one occurring in the late AD 1880s involved pre-cooled dome rock (<Curie Point temperatures of 350°C) and parts of the underlying hydrothermally altered crater rim. The unit was confined to within 5.3 km planimetric distance of the summit (Platz 2007). In contrast to this event, the penultimate eruption episode (Tahurangi episode) dated at AD 1755 was one of the largest eruptions of the last 800 years. Hot BAF deposits (Tahurangi Breccia, a and b) extend on the NW sector to between 8 km (b) and >10 km (a) from source (Fig. 2).
The low-energy characteristics of the recent Taranaki BAF and rock avalanche deposits (Table 1) mean that the Titan2D approach is well suited. In addition, at Mt. Taranaki, a well constrained fixed point for flow onsets can be assumed, with broad constraints on potential collapse volumes given from events in the recent past. The limited record of run out/volumes of past flows is used to determine a first-order estimate of initial volume.
4.1 Digital elevation model
The identification of the source area in the Titan2D computer simulation is one of the key input parameters, but developing an appropriate DEM provides the basis for any realistic flow modelling. As with all simulation studies that attempt to use existing depositional records to evaluate model outputs, the topography representation or DEM used is normally that of the present day, rather than the ideal of a pre-event terrain model. The only way to get around this in normal conditions is to create a detailed geological model of the stratigraphy of mass flow deposits and subtract these thicknesses from the current terrain (e.g., Daag 2003) This is made impractical by the fact that BAFs have highly variable deposit thicknesses within a network of deep, narrow channels and interfluve areas that may change considerably between events. Despite this, the present topography (as represented by a 20 m DEM) at Taranaki (over the last 500 years) retains the same overall broad morphology and channel system down the northwestern slopes into the Hangatahua River as well as the same type of channel/erosion landforms and vegetation (Lees and Neall 1993) that existed previously. In addition, using a DEM of the current landscape is more applicable to development of future hazard forecasting and hazard zonation.
The DEM used of Mt. Taranaki was produced from national topographic GIS datasets (NZMS 260 series, Land Information New Zealand), containing height data as contours and spot-heights. Using the ArcGIS TOPOGRID (ESRI 2005) command, a DEM was created with a cell dimension of 20 m. Major drainages and distinctive geomorphic features on the volcano are well defined on the DEM, although, in the surrounding ring plain area with low slopes and few dominant geomorpholoical features, morphological resolution is lost. The arcgrid DEM was then imported into GRASS GIS as an ascii file and converted to a GRASS DEM format for use in Titan2D.
4.2 Source area and initial pile
The bulk of the present (and pre-collapse) dome mainly comprises a single endogenous lobe (Platz 2007), within a 420 m radius crater (Fig. 3b). The present remnant dome is ~100 m high, a height-limit apparently constrained by magma viscosity at Mt. Taranaki (Platz 2007). The crater is essentially flat, but has a distinctive slope to the northwest of about 32º. This apparently generated the ellipsoid shape of the last dome.
Titan2D requires an initial starting pile. To effectively simulate a BAF or dome collapse the initial volume and shape of the collapsing pile must be defined within the context of the present crater configuration, including an accurate representation of the sub-dome surface. An estimation of the pre-dome surface was needed, along with an idealised complete dome reconstruction. To carry this out, a combination of GPS and aerial ortho-photography was used in GIS, ArcMap/ArcScene (ESRI 2005), to extrapolate a full parabolic dome shape (Fig. 3). Using the volume and shape of the remnant dome (1.6 × 106 m3), the maximum height of the original dome was 105 m and a (parabolic) basal outline of 450 and 320 m radii in the x and y directions, respectively. This yields a parabolic dome shape stretched over the sloping surface with a full volume of 6 × 106 m3 (Fig. 3). The centre-point of the re-constructed dome and its dimensions were applied to the Titan2D simulations. Most dome collapses involve only small proportions of the overall structure (e.g., Ui et al. 1999), although at Soufriere Hills large proportions of domes were incorporated in dome collapse at times. For the purposes of these simulations, a typical volume of 1 × 106 m3 was used to replicate partial dome collapse events or pulses on Mt. Taranaki based.
4.3 Titan 2D parameters—basal/internal friction angle
The two “friction angle” parameters used in Titan2D have not been directly related to physical parameters that can be measured in the laboratory or field. These are normally calibrated by matching simulation travel times and inundation areas to known values (e.g., Schilling 1998; Oramas Dorta et al. 2007). Attempts to determine these parameters experimentally (e.g., Iverson et al. 1998; Iverson and Denlinger 2001; Bursik et al. 2005) have focused on the measurement of collapsing piles on tilt tables and on flume experiments. Results are highly variable and it is arguable whether the laboratory conditions can be scaled to real-world situations. The most commonly used comparator between diverse types of mass flows is the H/L ratio (Heim friction coefficient; Heim 1932, revised by Hayashi and Self 1992), which can be used to estimate resistance to flow of a sliding avalanche by interaction with its underlying surface. Sheridan et al. (2005) used this method to determine the optimal basal friction parameter for simulating the Tahoma Peak Avalanche with Titan2D. Sensitivity analysis showed a dramatic effect on increasing run-out lengths with reductions in the basal friction angle parameter, whereas changes in the internal friction angle have shown less dramatic or no noticeable impacts in middle-value ranges (Sheridan et al. 2005).
For the Mt. Taranaki flows, the only comparative measures available are inundation area, estimated volume, run-out length and minimum H/L estimates. Rather than choosing specific values based on this imperfect dataset, a batch of 80 simulations were run for a controlled initial volume and location with every possible combination of internal and basal friction angle parameters at intervals of 5°.
4.4 Simulation results
Most simulated BAF deposition surfaces compared well to mapped deposits, without anomalies, such as overtopping known major topographic barriers or crater walls and without entering alternative catchments. All simulations exhibited a partial collapse from the initial pile with flow to the northwest of the current amphitheatre. Within the first 2.5 km from source, simulated flows were confined to a 500 m-wide amphitheatre-like structure. The flows either stopped on this slope (with high basal friction angles) or continued into box-shaped gullies of major stream channels (Maero and Pyramid Streams). After 3.5 km, these tributary flows combined in the Hangahatua (Stony) River valley for a further 9.5 km of travel.
The inundation areas, H/L ratio and run-out planar distance from mapped BAF deposits and those of simulated BAF’s were most similar with basal friction angles between 15–25° (Fig. 4).
4.5 Deriving hazard representations from the modelled flows
Converting Titan2D simulations of BAFs at Mt. Taranaki to an overall hazard representation required a method that combined all valid outputs. The approach developed takes into account variation resulting from the possible range in model input parameters that were used to infer the variation within realistic geological (or rheological) possibilities for hot and cold flows and source regions as described earlier.
Basal friction angles used in the Titan2D simulations for the hazard zone model for BAFs at Mt. Taranki
Gausian distribution (%)
These simulations provide a range of scenario layers that can be further analysed in a GIS to display estimated flow depths, velocity and discharge. Using ArcGIS (ESRI 2005), the outputs of each simulation were processed to produce a point layer containing the maximum pile height at each processed point. The point dataset was interpolated to create a raster grid of pile-heights across the region. This produces a laterally and longitudinally gradational continuous layer. Raster layers from each of the simulations were combined into a single height/thickness layer using the ArcGIS raster calculator tool. This combined “thickness” layer was chosen to be directly represented as a gradational hazard zone (Fig. 5). A similar process can also be paralleled for velocity or momentum outputs depending on user engineering or human hazard needs.
This combined layer highlights the Stony River and catchments as areas of greatest hazard as expected. It can be queried at user-defined points and displayed using a variety of underlying images, maps or infrastructure networks, as well as displayed in 3D with a shaded-relief model or ortho-image draped over a DEM (Fig. 5).
5.1 Combined simulated results
The results and a 1:300 year (based on Neall and Alloway 1996, return period for hazard zone A) hazard zone from BAFs are defined by the area (darker red central hazard zone on Fig. 5) associated with the drainage patterns of the northwestern sector of the volcano. These areas have been inundated by all BAFs in the past c. 800 years (Platz 2007). The gradational zone, however, shows a lessening decrease in hazard on the interfluves between the main drainages, indicating the lower possibility of flows inundating these (Fig. 5). Of particular interest is the length of the hazard zone highlighting that flows would only rarely reach lower altitudes (700 m, >10 km run-out), which is consistent with field evidence (Platz 1997).
5.2 Mass flow hazard simulation
The majority of volcanic flow hazard maps are generalisations based solely on extrapolating the past inundation areas of each type of flow (e.g., pyroclastic flow, lahar, debris avalanche) onto present topography (Scott et al. 1995; Waitt et al. 1995; Hoblitt et al. 1998; Wolfe and Pierson 1995). These become redundant when either new mapping identifies additional constraints on event frequency or inundation or when eruptions/collapses substantially change the topography. In addition, hazard maps for land-use planning, workplace and recreational area management require more precision to reliably indicate differing degrees of relative hazard. In addition, other information, such as flow velocity, depth and mass flux is essential for assessing the stability of any structures in the flow path and planning any new infrastructure. However, hazard maps showing increasingly more complexity, including those with some probabilistic component (e.g., Nevado del Ruiz; Parra and Cepeda 1990) are typically less well understood by users.
For current tephra fall hazards, Magill et al. (2006) used the numerical model ASHFALL (Hurst 1994) to forecast the probability of volcanic ashfalls from all major sources to affect the city of Auckland. From these data, they determined and displayed tephra inundation maps of a similar nature to those presented here. Using an alternative approach, Bebbington and Lai (1996) and Turner et al. (2008) developed a time-variable, probabilisitic, ashfall forecast for Taranaki. These statistical approaches may be divorced by varying degrees from the physical hazard generation processes, particularly those that combine various physical models without considering their inherent uncertainties. Other limiting factors with these methods are the ability to produce spatial representations (i.e., hazard maps) of the predictions that can be easily understood and applied by local authorities, land-use planners and hazard/emergency managers.
A similar approach to the Taranaki example presented here was undertaken for mass flows landslide and rock/debris flow hazards in the District of North Vancouver (Jacob 2005; KWL Ltd. 2003) using Flo-2D software (FLO-2D Software Inc.) to recreate the discharge/inundation of flows of a particular recurrence period. Inundation maps of each simulated flow in the various catchments studied were combined in a GIS to give a representation of the hazard to downstream areas. A gradation of increasing relative hazard was defined on the basis of the likelihood of areas being affected. The range of volumes modelled also provides a good method for encompassing and compensating for the lack of uncertainty around the volume of future events. This method does not, however, take into account the variability in the physical properties of the flow and whether these can be accounted for by the model. The recognition of the variability that may exist due to changing or contrasting flow sediment concentrations and differing flow paths was entirely assumed to be a factor of volume. The methods presented for the Taranaki example differ by attempting to identify a range within the input (physical) parameters that influence the flow rheology (largely unknown here) and resulting inundation area. This is then used as a proxy to encompass all the realistic possible outcomes within the constraints of recent past events. As identified by Platz et al. (2007), we can identify or infer from deposits that flows and their behaviour (linked to emplacement processes) can vary greatly and this can have an effect on run-out and inundation. It is assumed that in the Titan2D model friction input values are the most influential parameters in the flow model that control the modelled flows mobility. For the Mt. Taranaki BAFs, the rheology of the flow (rather than source conditions) is one of the more dominant factors in determining the run-out and inundation.
To expand this by including the variability amongst all known or possible BAFs on Taranaki events (such as variations in initial volume and source) would require either having a complete event history with accurately known volumes and inundation areas, or developing a probabilistic model to describe the known limits of variation in these that could then be coupled to Titan2D.
Given the relative dominance of BAF and related mass flows over the last 800 years, the simulations made are more likely to encompass the main initial hazards of the next unrest at Mt. Taranaki. The present configuration means that collapses down the NW slope will be highly likely if a new dome were to grow. The maximum volume that could be accommodated in the present crater is around ~10 × 106 m3, similar to the total volume of the largest sets of flow-packages mapped to date. It is common that only small fractions of dome volumes collapse at any one time (Unzen Volcano (eruption c. 1991), Nakada and Fujii 1993; Gunung Merapi Volcano (eruption c. 1993), Schwarzkopf et al. 2005), which further reduces the probable volumetric range of flows expected.
5.3 Taranaki BAF hazard mapping
Grant-Taylor (1964) and Neall (1972) considered that potential risk from future pyroclastic flow (BAF) activity at Mt. Taranaki would be restricted to within a 9 km radius of the summit. This was followed by a hazard map for volcanic flows (Neall and Alloway 1996) that defined six zones of risk, from various types of mass flows. This was based on mapped historical and prehistoric deposits of pyroclastic flows, lahars and debris avalanches along with their estimated return periods. The two zones relating to pyroclastic flows are defined by a radius of 15 km from the summit, with the highest hazard zone having a 1 in 300 year return period of impact. Recent studies (Turner et al. 2008; Platz 2007) have revealed a much higher frequency of small-volume dome collapse and pyroclastic-flow producing eruptions, necessitating a revision of the hazard zones. In addition, questions remain as to how accurate the map remains in the context of present-day topography, particularly in relation to a future “typical” sequence of dome growth and collapse with a crater area open and sloping towards the NW.
The 15 km-circular hazard zone of Neall and Alloway (1996) incorporates the extent of all possible pyroclastic flows that may originate from the volcano. This may be a useful overall “exclusion zone” for emergency management during a volcanic crisis due to the unpredictable nature and path of pyroclastic flows, associated surges, ballistics, tephra and possible blast events. However, this approach does not allow distinction of varying degrees of mass flow risk in the area for the managers of recreational area (and workplace) between eruptions and even during prolonged eruptions. Nor does it allow identification of any key infrastructure elements that may be affected with greatest likelihood by mass flows, such as critical bridges, pipe lines and power cables. Given that BAFs and other mass flows since ca. 800 years have almost exclusively affected the NW sector, this is the area of highest apparent risk under present topographical conditions yet this is not represented in hazard zones due to larger events (with less probability of occurrence) overshadowing this area of higher risk. However, it must be recalled that the newly created hazard zone is scenario-based rather than encompassing all of these events possibly originating from the summit and considered by Neall and Alloway (1996) to have a return period of 1:300 years based on the last 1,000 years of events.
There are few examples of applying a numerical model or computer simulation to hazard analysis and determination of future high risk zones other than that of Saucedo et al. (2005). The majority of computer model applications are related to the comparison of real and modelled simulations or the optimisation of parameters within a model. The method presented here explores this premise in relation to a scenario in which BAFs from the same part of the volcano or sequence of dome collapses may vary greatly. It is apparent that for the Titan2D model, friction input values are the most influential parameters, which appears also to be the case for the Mt. Taranaki examples, where source differences matter less.
By varying these factors we can account for expected variance in future Mt. Taranaki BAFs, during the course of a dome growth and collapse sequence. By modelling a range around an optimally compared flow, we can account for the variation that might exist by combining and displaying all results. The Gaussian distribution was used to encompass the variability observed in inundation areas. These give rise to GIS layers that can be displayed to show the gradational nature of hazard assessment, as implied by the early workers quoted earlier in this article.
Haynes et al. (2007) provided important observations about the effectiveness of traditional hazard maps in relation to public interpretations of them. They found that respondents could better orientate themselves and interpret hazard zone information in 3D format or data that was superimposed onto photographs or aerial photographs. Titan2D outputs and use of a GIS easily enable overlays of hazard zones onto aerial/satellite imagery or shaded-relief models to facilitate this knowledge transfer (Fig. 5).
Titan2D provides a tool that can simulate granular flows and provide outputs that are testable within geological mapping constraints. Multiple runs can be undertaken rapidly on a desktop computer and allow for timely hazard assessments to be made on present-day terrain, or on terrain that is rapidly changing. This case study shows that by undertaking an initial series of Titan2D runs that incorporate all reasonable combinations of basal and internal friction angles for BAFs while keeping the source conditions constant, best-fit simulations can be determined by comparison to the spatial distribution of known BAF events (Platz 2007). Here, reasonable Titan2D input parameters were 15–20º for basal friction and 30º for internal friction. Using this, a Gaussian distribution of new input parameters could be derived to produce a series of new titan simulations. The combination of the inundation areas from these new simulations allowed for the creation of a graduated zone of inundation likelihood. This method provides a means to account for uncertainty related to differences in flow behaviour of BAFs or similar granular flows. The digital outputs from Titan2D can be displayed in GIS and presented with aerial, satellite imagery or in 3D to allow the user a greater understanding of the relationship between the possible hazard and the landscape.
Iverson (2005) remarks that ample scepticism should be used when scientific interpretations are made from models and all physical, mathematical and computational aspects need to be understood. This is applicable for hazard analysis also, particularly when volcanic flow hazard analysis methods and the display of hazard zones have developed little since Crandell et al. (1984). Flow modelling is becoming a more common practise in volcanic hazard analysis, but mostly in relation to single events or scenarios. Despite the rapid use and uptake of the results produced from computer simulations of flows (particularly by local government/authority in land use planning), there has been very little effort to transfer that information into publicly available hazard maps. This comes, however, with the danger that appropriate representation of inherent uncertainties may be overlooked. Even in traditional hazard analysis and mapping based on geological field identification, uncertainty exists in both the inundation area and volume of a particular deposit as well as in the stratigraphic interpretation of the return period. However, these uncertainties are rarely transferred onto the cartographic display of the hazard. The identification and quantification of uncertainty and display of model simulations related to hazard analysis is an area that requires more attention to provide both the geologist and the user with a simple display of the potential risk.
This work is supported by a NZ FRST TPMF PhD fellowship (JP), and forms part of the FRST PGST programme contract MAUX0401 “Learning to live with volcanic risk” (SJC). We also thank the George Mason Trust and Freemasons for scholarship support and Dr. R.B. Stewart for comments on an earlier version. We also thank Prof. J-C. Thouret and Prof. C. Siebe for their thoughtful review comments and assistance.
- Alloway B, McComb P, Neall V, Vucetich C, Gibb J, Sherburn S, Stirling M (2005) Stratigraphy, age, and correlation of voluminous debris avalanche events from an ancestral Egmont Volcano: implications for coastal plain construction and regional hazard assessment. J R Soc NZ 35:229–267Google Scholar
- Cole PD, Calder ES, Sparks RJS, Clarke A, Druitt TH, Young SR, Herd RA, Harford CL, Norton GE (2002) Deposits from dome collapse and fountain collapse pyroclastic flows at Soufrière Hills Volcano, Montserrat. In: Druitt TH, Kokelaar BP (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Memoirs, vol 21. Geological Society, London, pp 231–262Google Scholar
- Crandell DR, Booth B, Kazumadinata K, Shimozuru D, Walker GPL, Westercamp D (1984) Source book for volcanic hazards zonation. UNESCO, ParisGoogle Scholar
- Cronin SJ, Stewart RB, Neall VE, Platz T, Gaylord D (2003) The AD1040 to present Maero Eruptive Period of Egmont Volcano, Taranaki, New Zealand. Geol Soc NZ Misc Publ 116A:43Google Scholar
- Delaite G, Thouret J-C, Sheridan M, Labazuy P, Stinton A, Souriot T, Van Westen C (2005) Assessment of volcanic hazards of El Misti and in the city of Arequipa, Peru, based on GIS and simulations, with emphasis on lahars. Zeitschrift für Geomorphol NF 140:209–231Google Scholar
- ESRI ArcGIS (2005) Environmental systems research Inc. USAGoogle Scholar
- Fisher RV, Schmincke H-U (1984) Pyroclastic rocks. Springer, Berlin, p 472Google Scholar
- Grant-Taylor TL (1964) Geology of Egmont National Park. In: Scanlan AB (ed) Egmont National Park. Egmont National Park Board, New Plymouth, pp 13–26Google Scholar
- Heim A (1932) Der Bergstruz von Elm. Geol Gesell Zeitschr 34:74–115Google Scholar
- Hoblitt RP, Walder JS, Driedger CL, Scott KM, Pringle PT, Vallance JW (1998) Volcano hazards from Mount Rainier, Washington, Revised 1998: US Geo Surv Open-File Report, 98–428Google Scholar
- Hurst AW (1994) ASHFALL—a computer program for estimating volcanic ash fallout. Report and users guide. Institute of Geological & Nuclear Sciences Science Report 94/23. 22Google Scholar
- KWL Ltd (2003) Debris flow study and risk mitigation alternatives for Percy Creek and Vapour Creek. (Final Report, December). District of North VancouverGoogle Scholar
- Lees CM, Neall VE (1993) Vegetation response to volcanic eruptions on Egmont volcano, New Zealand, during the last 1500 years. J R Soc NZ 23:91–127Google Scholar
- Neall VE (1972) Tephrochronology and tephrostratigraphy of western Taranaki, New Zealand. NZ J Geol Geophys 15:507–557Google Scholar
- Neall VE (1979) Sheets P19, P20 and P2 l New Plymouth, Egmont and Manaia, Geological Map of New Zealand. New Zealand Department of Science and Industrial Research, Wellington, scale 1:50 000, 3 sheets, 36ppGoogle Scholar
- Neall VE, Alloway BE (1996) Volcanic hazard map of Western Taranaki. Massey Uni Dep Soil Sci Occ Report 12Google Scholar
- Patra AK, Bauer AC, Nichita CC, EPitman EB, Sheridan MF, Bursik M, Rupp B, Webber A, Stinton AJ, Namikawa LM, Renschler CS (2005) Parallel adaptive numerical simulation of dry avalanches over natural terrain. J Volcanol Geotherm Res 139(1–2):89–102Google Scholar
- Platz T (2007) Aspects of dome-forming eruptions from Andesitic Volcanoes exemplified through the Maero Eruptive Period (1000 yrs B.P. to Present) activity at Mt. Taranaki, New Zealand. Unpublished PhD thesis, Institute of Natural Resources, Massey University, Palmerston North, New ZealandGoogle Scholar
- Rupp B, Bursik M, Patra A, Pitman B, Bauer A, Nichita C, Saucedo R, Macias J (2003) Simulation of pyroclastic flows of Colima volcano, Mexico, using the TITAN2D program, AGU/EGS/EUG Spg. Meet. Geophys Res Abstracts 5:12857Google Scholar
- Rupp B, Bursik M, Namikawa L, Webb A, Patra AK, Saucedo R, Marcias JL, Renschler C (2006) Computational modelling of the 1991 block-and-ash flows at Colima Volcano, Mexico. In: Seibe C, Marcias JL, Aguirre-Diaz GJ (eds) Neogene-Quaternary Continental Margin Volcanism: a perspective from Mexico. Geol Soc Am Spec Paper, 402:237–250Google Scholar
- Savage SB (1987) Interparticle percolation and segregation in granular materials: a review. In: Selvadurai APS (ed) Developments in engineering mechanics. Elsevier, New York, pp 347–363Google Scholar
- Schilling SP (1998) LAHARZ: GIS programs for automated mapping of Lahar-inundation hazard zones. US Geo Surv Open-File Report 98–63Google Scholar
- Scott WE, Iverson RM, Vallance JW, Hildreth W (1995) Volcano hazards in the Mount Adams Region, Washington: US Geo Surv Open-File Report 95–492, 11ppGoogle Scholar
- Scott KM, Macías JL, Naranjo J, Rodriguez S, McGeehin JP (2001) Catastrophic debris flows transformed from landslides in volcanic terrains: mobility, hazard assessment and mitigation strategies. US Geol Surv Prof Pap 1630:59Google Scholar
- Sheridan MF, Hubbard B, Carrasco-Nuñez G, Siebe C (2000) GIS model for volcanic hazard assessment: pyroclastic flows at Volcán Citlaltépetl, México. In: Parks BO, Clarke KM, Crane MP (eds) Proceedings of the 4th international conference on integrating geographic information systems and environmental modeling: problems, prospects, and needs for research; 2000 September 2–8; Boulder, CO. Boulder: University of Colorado, Cooperative Institute for Research in Environmental ScienceGoogle Scholar
- Sparks RSJ, Young SR (2002) The eruption of Soufrière Hills Volcano, Montserrat (1995–1999). In: Druitt TH, Kokelaar BP (eds) The Eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999, Geol Soc London Memoirs 21:45–69Google Scholar
- Waitt RB, Mastin LG, Begét JE (1995) Volcanic-hazard zonation for Glacier Peak Volcano, Washington. US Geo Surv Open-File Report 9:5–499Google Scholar
- Wolfe EW, Pierson TC (1995) Volcanic-Hazard Zonation for Mount St. Helens, Washington, 1995. US Geo Surv Open-File Report 9:5–497Google Scholar
- Zernack A, Procter J, Cronin SJ (2009) Sedimentary signatures of cyclic growth and destruction of stratovolcanoes: a case study from Mt. Taranaki, New Zealand. In: Németh K, Manville V, Kano K (eds) Source to sink: from volcanic eruptions to volcaniclastic deposits on the Pacific Rim. IUGS, Special Volume Sedimentary Geology (in press)Google Scholar