Detection of high-silica lava �ows and lava morphology at the Alarcon Rise, Gulf of California, Mexico using automated classi�cation of the morphological-compositional relationship in AUV multibeam bathymetry and sonar backscatter
Christina A. Maschmeyer ([email protected])1, Scott M. White1, Brian M. Dreyer2, David A. Clague3
University of South Carolina, Columbia, SC1 University of California, Santa Cruz, CA2 Monterey Bay Aquarium Research Institute, Moss Landing, CA3
Introduction and Objectives�e Alarcon Rise is the southernmost spreading ridge in the Gulf of California, Mexico. �e Alarcon Rise is an intermediate-rate spreading center with a full spreading rate of 50 cm/yr. Eruption processes of lava e�usion rate (which corresponds to spreading rate), underlying topography, and lava viscosity control the formation of submarine lava morphologies for mid-oceanic ridge basalt (MORB) [Gregg and Fink (1995)]. Typical MORB exists at the Alarcon Rise, but researchers from the Monterey Bay Aquarium Research Institute (MBARI) and Scripps Oceanographic Institute collected volcanic samples with elevated silica content [Castillo et al. (2002)].
Because submarine lava morphology depends on eruption processes, we propose to use morphologic signatures to identify evolved lavas. Our three primary questions for this research are:1. Can lava composition be determined by lava �ow morphology? 2. What morphologic signatures are critical to identi�cation of lavas with elevated silica? 3. How much high-silica lava exists at the Alarcon Rise?
We built an arti�cial intelligence classi�er to begin to answer these questions at the Alarcon Rise.
Expert Classi�cation- Construction of an adaptive neuro-fuzzy inference system (ANFIS) enabled the classi�cation of lava composition- Inputs were local slope and texture datasets from high-resolution bathymetry and sonar backscatter collected by AUV D. Allen B. in 2012.- Training cycle data included of 134 ground-referenced igneous samples collected by MBARI in 2012 and 2015 plus 48 points on MORB domes generated in ArcGIS along ROV track lines- Testing data consisted of 47 expert-picked MORB dome samples and 200 ground-referenced igneous samples collected in 2012 and 2015- We applied the classi�cation to two regions along the Alarcon Rise: a high-silica region and a MORB region
AUV BathymetryAUV Bathymetry Sidescan Sonar Backscatter
Sidescan Sonar Backscatter
GLCM TexturesGLCM TexturesLocal SlopeLocal Slope GLCM TexturesGLCM Textures
ANFISANFIS Classi�cationClassi�cation
Accuracy AssessmentAccuracy
Assessment
Ground Referenced Igneous Rock Samples
Ground Referenced Igneous Rock Samples
Training DataTraining Data Testing DataTesting Data
Submarine Lava MorphologyTypical oceanic ridge lava exhibits three distinct morphologies: sheet �ows, lobate �ows and small pillow mounds (~150 m diameter). Lava e�usion rate, viscosity, and underlying topography determine the morphologic di�erences between �ows [Gregg and Fink (1995)]. Lava e�usion rate is the most in�uential variable at spreading ridges with compositionally uniform lava [White and Klein (2014)].
Spreading ridges with silica-rich lavas do not exhibit submarine lava �ow morphologies. Large pillows with extreme radial jointing and corrugated surfaces and asymmetrical lava blocks build large domes. Angular, blocky talus is also common at silica-rich domes [Wanless et al. (2010)]. �e high-silica region of interest includes a large (~500 m diameter) dome.
MORB Morphology
Shee
tLo
bate
Pillo
w
High-Silica Lava Morphology
Rad
ial J
oint
ing
Lava
Blo
cks
Cor
ruga
ted
Pillo
w
High-Silica Dome
Distance (m)550500450400350300250200150100500
Dep
th (m
)
-2,340
-2,360
-2,380
-2,400
-2,420
-2,440
-2,460
MORB Dome
Depth (m)550500450400350300250200150100500
Dep
th (m
)
-2,330
-2,340
-2,350
-2,360
-2,370
-2,380
Sea�oor Mapping and Sampling- February-April 2012 and 2015, MBARI mapped the Alarcon Rise at 1 m horizontal resolution using the AUV (Autonomous Underwater Vehicle) D. Allen B. We created inputs for the expert classi�er from bathymetry and sonar backscatter collected during 2012 - MBARI collected 410 rock samples over 19 dives by the ROV (remotely operated vehicle) Doc Ricketts -364 igneous samples collected along the spreading ridge were used during classi�cation
Completed Dives Trackline Coverage
Basalt Basaltic Andesite Andesite Dacite Rhyolite Total
9 36.3 km 135 7 6 7 35 190
10 38.6 km 97 14 37 16 10 174
Igneous Rock Samples
References- Castillo, P.R., Hawkins, J.W., Lonsdale, P.F., Hilton, D.R., and Shaw, A.M., 2002, Petrology of the Alarcon Rise lavas, Gulf of California: Nascent intracontinental ocean crust: Journal of Geophysical Research, v. 107, no.B10, p. 1-15.- Gregg, P.M., Fink, J.H., 1995, Quanti�cation of submarine lava-�ow morphology through analog experiments: Geology, v. 23, n.1, p. 73-76.- Wanless, V.D., Per�t, M.R., Ridley, W.L., and Klein, E., 2010, Dacite petrogenesis on mid-ocean ridges: Evidence for oceanic crustal melting and assimilitaion: Journal of Petrology, v. 51, n. 12, p. 2377-2410.-White, W.M., and Klein, E.M., 2014, Composition of the Oceanic Crust, in Saunders, J.A.,Hofstra, A.H., Goldfarb, R.J., and Reed, M.H., eds., Treatise on Geochemistry: Elsevier, v. 4, p. 457-496.
200Meters
KEY
Ground-Referenced Samples
Classi�cationMORB
High-Silica
Expert De�ned MORB Dome
MORB
Transitional
High-Silica
¯
High-Silica Region
MORB Region
High-Silica Region
MORB Region
¯
Regional Classi�cation- Classi�cation result data distribution determined three classes (red, yellow, blue) majority �ltered with a 25 m radius gridded at 2 m2- Blue represents areas of MORB classi�ed ANFIS with values <1.2- Yellow represents mixed classi�cation ranging in ANFIS values between 1.2 - 1.7- Red represents exclusively high-silica lavas with ANFIS values >1.7
Alarcon Rise Bathymetry- AUV bathymetry gridded at 1 m2 collected by AUV D. Allen B. in 2012 overlaying ESRI World-Ocean basemap obtained from multiple contributors-ROV Doc Ricketts dive tracks during 2012 and 2015
-Training and testing data in the high-silica region and MORB region were classi�ed using the 25 m majority �ltered ANFIS result-Classi�cation values at high-silica region range between 0.9 and 1.5 for MORB and between 1.2 and 2.0 for high-silica lavas-Classi�cation of MORB region data range from 0.9 to 1.7
Num
ber o
f Sam
ples
MORB
Expert De�nedMORB Domes
High-Silica LavaExpert De�nedMORB Domes
MORB
MORB RegionHigh-Silica Region
Ground-Referenced Data Classi�cation
-ANFIS classi�cation for the 100 m2 sampling square on the high-silica dome resulted in a value of 1.3-Two sampling zones of 100 m2 were constructed on a MORB dome and �at MORB area-Classi�cation of the MORB dome ranged from 1.3-2.0-Classi�cation of the �at MORB area ranged from 0.9-1.5
Num
ber o
f Sam
ples
Focus-Area Classi�cation
- Of the 0.98 km2 high-silica region, ANFIS classi�ed 20.6% as MORB, 64.9% as transitional, and 14.5% as high-silica lava-Of the 4.45 km2 MORB region, ANFIS classi�ed 69.8% as MORB, 21.0% as transitional, and 9.2% as high-silica lava
Class ANFIS Total Producer's Accuracy User's Accuracy
MORB 167.0 80.2% 73.2%
Evolved 80.0 38.8% 48.8%
Rela
tive A
bund
ance
of
Lava
Com
posit
ion
(%) High-Silica Lava
Transitional
MORB
Conclusions- Because spreading rate remains constant at the Alarcon Rise, di�erences in lava �ow morphology result from di�erences in lava viscosity and can be used to distinguish high-silica lava �ows from MORB �ows - Local slope, BPI texture, and sonar backscatter texture do not sign�cantly distinguish between high-silica lava and MORB domes- �e amount of high-silica lava present at the Alarcon Rise cannot be accurately determined by a classi�er with 66.8% overall accuracy
Future Work- Dome classi�cation may improve classi�cation of compositional di�erences between high-silica lavas and MORB because high-silica lava �ows only form coorugated pillows and blocky domes- �e current classi�er cannot distinguish between large silica-rich domes and small MORB domes there-fore training and testing data should only be picked at domes rather than on MORB sheet and lobate �ows- A shape-recognition parameter may aid in classi�cation because MORB domes form circular outlines and high-silica domes form rectangular outlines
Acknowledgements- Science team, AUV team, ROV pilots, and crews of the 2012 and 2015 Alarcon Rise cruises on the R/V Zephyr and Western Flyer- �e Monterey Bay Aquarium Reserach Institute- �e Department of Earth and Ocean Sciences at the University of South Carolina, Columbia
ANFIS Results
