NETVOLC: An algorithm for automatic delimitation of volcano edifice boundaries using DEMs
Introduction
Landforms are discrete features of the continuous land surface having characteristic, recognizable shapes (Evans, 1972, Neuendorf et al., 2005). Quantitative landform mapping and analysis (specific geomorphometry) is an important task of geomorphology. To measure geometric characteristics of individual landforms, complete delimitation by a closed boundary is necessary. Currently, most landform delimitation is performed manually by visual identification and digitization on maps, airphotos, satellite images or digital elevation models (DEMs) (Evans, 2012). Many landforms are bounded by changes in slope gradient (breaks in slope). The general approach for delimitation of these landforms is manual tracing of slope breaks using DEM-based techniques (Smith and Clark, 2005, Evans, 2012). However, manual delimitation is time-consuming and subject to user subjectivity. Algorithms for automatic boundary delimitation of landforms are thus desirable, but remain a research frontier (Evans, 2012). In recent years, several automatic and semi-automatic approaches have been developed for general landform classification (Bue and Stepinski, 2006, Drăguţ and Blaschke, 2006, Iwahashi and Pike, 2007, Schneider and Klein, 2010) and for extraction of specific landforms, such as seamounts (Hillier, 2008) or drumlins (Saha et al., 2011).
Volcanism produces a wide range of landforms. They can be excavational (calderas, maars) or, more commonly, constructional (monogenetic cones and domes; polygenetic or composite volcanoes). The ideal constructional volcano is a cone with radial symmetry and slopes steeper than its surroundings; it is thus bounded by concave breaks in slope at its base. However, volcanoes tend to exhibit various degrees of complexity, up to very irregular volcanic chains or massifs. In addition, volcanoes can contain several landform elements such as summit craters, small parasitic cones, collapse scars, erosional channels or valleys, ridges, etc. A further complication is that volcanoes tend to merge gradually with the surrounding landscape and some products can be deposited at great distances from the central source. All this makes volcanic landform delimitation for geomorphometric analysis a complicated endeavor. Consequently, most volcano morphometric studies have focused on monogenetic scoria cones because of their simpler morphologies with clearer boundaries (Favalli et al., 2009, Kervyn et al., 2012 and references therein). Morphometric studies of composite volcanoes are relatively few and have used different manual delimitation methods based on product extent (geology), elevation contours and/or breaks in slope (Pike, 1978, Plescia, 2004, Grosse et al., 2009, Völker et al., 2011, Karátson et al., 2012). Recently, Grosse et al. (2012) developed a manual delimitation method using DEM-derived curvature and slope data; they emphasized that delimitation by DEMs should be restricted to volcano edifices and not include their aprons which do not have a clear morphometric signature and tend to merge with the surroundings.
Automatic closed-contouring algorithms, based on searching for the lowest elevation contour with a quasi-elliptical shape that completely encloses a topographic high, have been used to detect and delineate seamounts (Behn et al., 2004, Bohnenstiehl et al., 2008). This method has a major limitation in that closed-contours lie at constant topographic elevation, leading to erroneous solutions when the landform lies on a sloping landscape. Bohnenstiehl et al. (2012) overcome this limitation by implementing a modification to the closed-contouring approach (MBOA) that adjusts base elevation by evaluating the area/perimeter ratio along radial profiles. Application of the MBOA algorithm to a cinder cone field shows that it performs much better than the standard closed-contour method (Howell et al., 2012). However, it depends on several user-defined parameters and thresholds, introducing some degree of subjectivity.
Here we present the algorithm NETVOLC for automatic delimitation of volcano edifice boundaries using DEMs and applying minimum cost flow networks. The algorithm, based on the premise that volcano edifices are bounded by concave breaks in slope, calculates the best possible outline by solving a minimum cost flow problem. Thus, it delimits only the edifice as a specific landform and excludes the surrounding apron and other far reaching products. Application of the algorithm is intended for any volcano with recognizable positive topography, independently of its size. The paper is organized as follows. Section 2 describes the proposed algorithm. Evaluation of the algorithm by real data is presented in Section 3. Conclusions are presented in Section 4.
Section snippets
Rationale
The general problem is to compute a closed path around a volcanic center that represents the edifice boundary. Considering a DEM of an idealized volcanic cone emerging over an arbitrary landscape, we can define a closed path by intersection of the cone surface with the pre-cone landscape surface. This outline represents the edifice boundary (Fig. 1) and is precisely located by concave breaks in slope: points with maximum negative profile curvature. Thus, the straightforward solution for a DEM
Algorithm evaluation
In this section we apply NETVOLC to the Mauna Kea pyroclastic cone field (Hawaii, USA) in order to evaluate performance of the algorithm. The Mauna Kea cone field contains hundreds of cones with a wide range of morphologies and is covered by the 10 m National Elevation Dataset (NED) DEM (Gesch et al., 2002, Maune, 2007). Furthermore, Kervyn et al. (2012) recently carried out a morphometric analysis of this cone field using the NED DEM and considering slope breaks as the key criteria for manual
Conclusions
NETVOLC is an innovative algorithm for automatically computing the boundary of volcanic edifices. It has the advantage of being more objective and much faster than manual delineation, and requires only a DEM of the volcano as input. In most cases, the main cost function produces accurate outlines, but can yield erroneous results for complex edifices. The alternative cost functions give the flexibility to re-process these cases and obtain good results, although this implies that results have to
Acknowledgments
Matthieu Kervyn is thanked for kindly providing his manual outlines and resulting data of the Mauna Kea cone field. Reviewers Richard J. Pike and Ian S. Evans are acknowledged for their very insightful comments and corrections to the original manuscript. Pablo Grosse acknowledges CONICET and Fundación Miguel Lillo for their support.
References (30)
- et al.
Effect of the Galápagos hotspot on seafloor volcanism along the Galápagos spreading center (90.9–97.6°W)
Earth and Planetary Science Letters
(2004) - et al.
A modified basal outlining algorithm for identifying topographic highs from gridded elevation data, Part 1: Motivation and methods
Computers & Geosciences
(2012) - et al.
Automated classification of landforms on Mars
Computers & Geosciences
(2006) - et al.
Automated classification of landform elements using object-based image analysis
Geomorphology
(2006) Geomorphometry and landform mapping: what is a landform?
Geomorphology
(2012)- et al.
Morphometry of scoria cones located on a volcano flank: a case study from Mt. Etna (Italy), based on high-resolution LiDAR data
Journal of Volcanology and Geothermal Research
(2009) - et al.
Systematic morphometric characterization of volcanic edifices using digital elevation models
Geomorphology
(2012) - et al.
A modified basal outlining algorithm for identifying topographic highs in gridded elevation data, part 2: Application to Springerville Volcanic Field
Computers & Geosciences
(2012) - et al.
Automated classifications of topography from DEMs by an unsupervised nested-means algorithm and a three-part geometric signature
Geomorphology
(2007) - et al.
Erosion rates and erosion patterns of neogene to quaternary stratovolcanoes in the Western Cordillera of the Central Andes: an SRTM DEM based analysis
Geomorphology
(2012)
Geomorphometric variability of “monogenetic” volcanic cones: evidence from Mauna Kea, Lanzarote and experimental cones
Geomorphology
An object-oriented approach to automated landform mapping: a case study of drumlins
Computers & Geosciences
Comparative mass balance of volcanic edifices at the southern volcanic zone of the Andes between 33°S and 46°S
Journal of Volcanology and Geothermal Research
Morphometric evolution of cinder cones
Journal of Volcanology and Geothermal Research
Network Flows: Theory, Algorithms, and Applications
Cited by (38)
Monogenetic scoria cone and associated lava flow volume estimates and their controlling factors
2023, Journal of Volcanology and Geothermal ResearchGrowth and evolution of long-lived, large volcanic clusters in the Central Andes: The Chachani Volcano Cluster, southern Peru
2022, Journal of Volcanology and Geothermal ResearchCitation Excerpt :With the same principle but using (2) interpolation of Triangular Irregular Networks (TIN) obtained from 30-m DEM data, and reference inclined contact surface between bedrock and C-LVC derived from exposed outcrops, we computed a volume of 346 km3 on ArcMap® software. In addition, (3) considering the volcanoclastic deposits in the western wall of the Río Chili canyon (base of Estribo edifice) and inclined reference basal surface, we obtained a volume as large as 390 km3 with the NETVOLC (Euillades et al., 2013) and MORVOLC algorithms (Grosse et al., 2012). Thus, we consider the two first estimated values, which are similar within ~60 km3 as more accurate than the results using NETVOLC and MORVOLC algorithms which deviate significantly from each other (190–390 km3).
Evolution of the Azufre volcano (northern Chile): Implications for the Cerro Pabellón Geothermal Field as inferred from long lasting eruptive activity
2022, Journal of Volcanology and Geothermal Research