Quantification Algorithm Updates:

 

Hello and welcome to our update page. This page serves as an active update link to the progress and development of a series of quanitifcation algorithms developed for morphology and rheology studies, as well as other non topographic applications. Below is a short summary of our work from the AGU 2006 fall meeting. We would also like to thank American Geophysical Union (AGU), Earth and Space Science Informatics Focus Group, for awarding us an oustanding student paper award, and for NSF MARGINS for an honorable mention award.

 

Note

We are interested in applying these algorithms to other fields and types of data, including non-topographic data such as atmospheric applications, hydrological contaminants, and gravity and magnetic anomalies. Please feel free to contact us about collaboration opportunities.

 

 

Quantitative Characterization of Topographical Features in Digital Elevation Models (DEM) and Bathymetry Data

T. Fedenczuk¹, P. Fryer¹, J. Escartin², A. Harris¹

¹ SOEST - HIGP, University of Hawaii, Honolulu, USA.

²Equipe de Géosciences Marines, Institut de Physique du Globe, Paris, France.

Contact

Tom Fedenczuk , 1680 East-West Rd POST504, Honolulu HI 96822, USA ( tmf@hawaii.edu )

 

Abstract #: IN11A-1148

We have developed an automated and quantitative approach to the interpretation of topographical features with an application to Em300 bathymetry data collected in 2003 over Big Blue Seamount, a serpentinite mud volcano in the Mariana forearc region (Western Pacific). The method employs a series of Matrix Laboratory (MATLAB) functions and scripts, and exports the results as data sets, figures, text files, and summary reports. These can be visualized and manipulated in MATLAB itself or in Geographic Information System (GIS) applications. The method can be used for any digital topographic/bathymetric data set in subaerial or submarine settings to characterize surface features such as flows, folds, distal edges of flows or mass-wasting deposits, slump scars, fault scarps, ridges, valleys, and plateaus. It provides a method for analyzing surface roughness as well as gradient magnitude and direction. It is automated to convert DEM and bathymetry into gradient data sets, render gradient maps, create profiles across morphologic features, determine the underlying slope of the features, calculate the slope-corrected heights (or thickness), and estimate volumes by assuming that thickness satisfies Laplace’s equation. This approach permits a more rigorous measure of surface morphology that can be applied to the interpretation of a wide variety of geologic settings, ranging from volcanic and mass-wasting processes to reef habitats.

Location of Sample Data

The Izu-Bonin/Mariana (IBM) convergent plate margin is host to a series of large serpentinite mud volcanoes along the forearc region (fig. 1) Big Blue Seamount is one of the largest active mud volcanoes and has been surveyed using high resolution bathymetric (EM300) and sidescan (DSL-120) data sets. Its summit lies at a depth of ~1240 m, with an overall height of more than 2 km above the surrounding ocean floor, and a maximum width of ~40 km. On the southwest flank at least four flows (ranging from simple to compound) are preserved in relatively undisturbed form (ABCD in fig. 2 and 4). We analyzed the three flows nearest the summit. We also used the fourth flow field (lower on the flank), as well as other surrounding features, to help complete the analysis for underlying slope.

Figure 1. Bathymetry map of Mariana Forearc, showing Big Blue Seamount relative to other seamounts, as well as the islands of Guam, Saipan, and Pagan.

Methods

The algorithm:

• imports XYZ bathymetry data into MATLAB

• processes bathymetry into a gradient (see box A) data set that includes latitude, longitude, slope in the x direction, slope in the y direction, and combined slope magnitude

• creates gradient maps using color to display the magnitude of the gradient (fig. 3 and 4)

• creates normalized vector arrows that can be overlain to indicate the direction of the gradient (fig. 4).

Figure 2. Bathymetry map of the SW flank of Big Blue Seamount with flows A through D outlined in white. See figure 4 for a gradient comparison of the same area. This figure has been cropped to retain data security.

Figure 3. Example of a gradient color map (area covered by red outlined box in fig 4), overlaid by normalized, reverse-gradient vector arrows

In GIS:

At this stage, the generated gradient map is imported into GIS and gradient defined features are identified and digitized in GIS as shapefile lines and exported back into MATLAB. In this application, we identified four main flows fields (A, B, C, D) (fig. 4).

Figure 4. Gradient map for Big Blue mud volcano, flow fields A through D. Color bar is optimized for relevant slope range. Dark blue represents a slope of 0° and red represent slopes > 20°, with green and yellow gradations in between. This figure has been cropped to retain data security.

Next the algorithm:

• automatically samples a gradient feature (such as a flow’s distal edge) at discrete, user-defined intervals (fig. 5)

• calculates the trend of the maximum gradient at each sample point to ensure profile direction is accurate

• creates profiles based on a user-defined, minimum-gradient threshold (fig. 6)

• plots profiles on the gradient map for error checking

• calculates horizontal and vertical length for each profile

• exports profiles into MATLAB workspace files (.mat), and ASCII files (for import into GIS)

• can be customized to measure other characteristics such as profile trend, slope, etc.

• determines underlying slope of each flow using two customized methods for two separate sets of calculations (volume and rheology)

• calculates the slope corrected distal edge thickness for each point for use in yield strength calculations (Box B)

• calculates the minimum, maximum, mean, and median of distal edge thickness for each feature

• creates a report with a summary of results

• calculates volume using boundary element method and Jacobi iteration (box C)

• compares volume calculation to the simpler volume estimation of area * average thickness

Figure 5. Close up of profile lines and associated components: Sample Point (P), Sample Interval (Int), Profile Line, and Shapefile Line.

a) τ = 11° b) τ = 15° c) τ = 12.5°

Figures 6. Profile line results of three minimum threshold (τ) values used to define the upslope and downslope boundaries of the same section of a distal edge. At low τ (a), the profile lines are over extended and include parts of other features. At high τ (b), the profiles lines are too short and under represent the distal edge. A τ of 12.5 (c) was the optimal threshold for flows A, B, and C, on Big Blue Seamount.

Benefits of the Technique:

• thresholds and criteria are very precise (eg slope of 12 degrees a sample interval of 10m) and can be established consistently throughout all of the samples

• the number of samples can be increased by a magnitude of 10 or more, and helps verify the measurements and give them greater resolution

• the entire extent of the feature is sampled evenly at the regular intervals, and omits the human predisposition for rendering prominent land forms that are easier to digitize but may not be proportionally representative of the feature as a whole

• much faster than the manual method.

• thresholds and criteria can be easily changed and the boundaries redrawn, thus allowing for quick comparison and optimization

• reproducible results with an open and editable algorithm

• fast error correction

Figure 7. 3-D perspective of a distal edge of flow B with profile lines extended to a threshold of 12.5°.

A. Gradient Equations

The gradient (Λz) at point (x₀,y₀) is defined by:

Λz(x₀,y₀) = (z_x(x₀,y₀), z_y(x₀,y₀))

where z is the bathymetry value (or height for DEMs); z_x and z_y and denotes differentiation with respect to x (latitude) and y (longitude), respectively. The magnitude (M) of the gradient is then calculated at each point as:

B. Yield Strength Equations

Assuming that the slope corrected thickness, t, at which the flow came to rest represents the critical thickness (t_c), the following equation can be solved to calculate yield strength using density (ρ), gravity (g) and underlying slope (θ):

Sy = t_cρg sinθ

Mean bulk density (mbρ) data of similar volcanoes in the area is available from two ocean drilling programs, ODP Leg 195 (Salisbury and Shinohara et al., 2002) and ODP 125 (Fryer, Pearce and Stokking et al., 1992), and can be used to constrain the upper and lower density limits in yield strength calculations. The combined data sets have 111 samples, with an average density of 1.82 g/cm³, ranging from 1.35 g/cm³ to 2.02 g/cm³ (Table 2).

C. Volume Equations

To determine the volume of individual flows, as defined by its perimeter thickness, we assume that thickness within the boundaries of the flow satisfies Laplace’s equation:

We used Jacobi iteration to solve Laplace’s equation numerically for thickness inside the flow, given values along the boundary. We create a mesh grid of an irregular boundary area, and assign the boundary values to the closest grid points, and assign all of the points inside the boundary with an initial value of zero. Then, for each mesh-grid point inside of the boundary, the function successively solves:

Where t_j,k is thickness at longitude (j) and latitude (k), and n is the iteration number (Strauss 1992).

Results

At a sample interval of 10m, we measured 303 profiles for flow A, 982 for flow B, and 2931 profiles for flow C (table 1).

In order to retain security of the data, we have temporarily removed most of the results section.

Conclusion and Acknowledgments

The automated morphology-quantification method presented in this paper has enabled us to calculate the thickness of the distal edges of three mudflows under three different underlying slope conditions using more than 5000 sample points each. All calculations are repeatable and use the same threshold criteria. The most important advantage to this quantification method is that we can conduct the same calculations using different parameters rapidly. We can also edit or append the MATLAB functions and scripts for other applications.

The application of this research has been funded by the NSF-Margins program, grant number OCE-0002584, and by the Maui High Performance Computing Center.

For more information on this algorith, please email Tom Fedenczuk. For more information on our work with serpentinite mud volcanoes please follow the links or email Patricia Fryer.

References

Fryer, P., Pearce, J.A., Stokking, L.B., et al., 1992, Proc. ODP, Sci. Results, 125:593-614, College Station, TX, .

Rowland, S. K., 1996, Slopes, lava flow volumes, and vent distributions on volcan Fernandina, Galapagos islands, J. Geophys. Res., 101, 27,657-27,672.

Salisbury, M.H., Shinohara, M., Richter, C., et al., 2002. Proc. ODP, Init. Repts., 195, College Station TX.

Stevens, N.F., Wadge, G. & Murray, J.B., 1999, Lava flow volume and morphology from digitised contour maps. Geomorphology, 28, 251-261.

Strauss, W. A. , 1992, Partial Differential Equations, John Wiley & Sons, New York. pp. 208-210.