Mapping seafloor surficial geological habitat in Massachusetts state waters

1
Mapping seafloor surficial geological habitat in
Massachusetts state waters
Daniel W. Sampson, GIS/Data Manager Anthony R.
Wilbur, Marine ecologist Seth D. Ackerman,
Geologist
2
Presentation objectives

• What are we trying to map?

3
Presentation objectives

• What are we trying to map?
• What was our approach?

4
Presentation objectives

• What are we trying to map?
• What was our approach?
• What were the results?

5
Study area

• Approx. area 134 km2
• Depth 4 48 m
• A variety of bottom types ranging from soft mud
  to bedrock

6
What are we trying to map?

• Potential habitat (term from Greene et al., in
  press)
• Specific habitat associations of a species or
  population are not often known during compilation
  and interpretation of seafloor data.  Therefore,
  it is not appropriate to describe interpretive
  maps of the seafloor as habitat maps.

7
Greene et al., cont.

• To avoid misconception of habitat the term
  potential habitatis used and is applied to
  describe a set of distinct seafloor conditions
  that may be found in the future to qualify as
  habitat.

8
What was our approach?

• Create potential habitat polygons based on
• Sediment type
• Rugosity
• Classify the polygons according to Greene et al.
  (in press) habitat classification schema
• Scale
• Physiography
• Induration
• Geomorphology

9
Sediment type rugosity

• Fundamental information to predict potential
  biological associations
• Relatively stable features

10
Data sources

• High-Resolution Geologic Mapping
• of the Inner Continental Shelf
• Nahant to Gloucester, Massachusetts
• U.S. Geological Survey Open File Report
  2005-1293
• Walter A. Barnhardt et al., 2006
• usSEABED Atlantic Coast Offshore
• Surficial Sediment Data Release,
• version 1.0
• U.S. Geological Survey Data Series 118
• Jamey M. Reid et al., 2005

11
Sediment type classification

• Classify backscatter data into discrete
  sediment classes
• not an easy task!
• a myriad of classification approaches
  classification schemas

12
Sediment classification
Prepare backscatter data
Create final sediment map
Create signatures apply to backscatter data
Create training data
13
Sediment classification
Prepare backscatter data
Create final sediment map
Create signatures apply to backscatter data
Create training data
14
Sidescan Backscatter
DN range 0-255 (8-bit grey scale data )
15
Grey Level Co-occurrence Matrix (GLCM)

• Quantifies relations between groups of
  backscatter pixels ? Texture
• Grey level differences (contrast)
• Defined size of area where change occurs (moving
  window)
• Directionality, or lack thereof

16
GLCM Mean
A measure of spatial dependency
17
GLCM Entropy
A measure of orderliness
18
GLCM Homogeneity
A measure of contrast
19
Sediment classification
Prepare backscatter data
Create final sediment map
Create signatures apply to backscatter data
Create training data
20
usSEABED training data
134 total points 27 Gravel (G) 13 gravelly
Sand (gS) 33 Hard (H) 30 Mud (M) 9 muddy Sand
(mS) 22 Sand (S)
21
Folk triangle
Added H _____________ Hard bottom to represent
rocky bottom types (as done in usSEABED data).
22
Sediment classification
Prepare backscatter data
Create final sediment map
Create signatures apply to backscatter data
Create training data
23
Classification process

• Buffered usSEABED training point data to create
  10m diameter circular polyons
• Created signatures on GLCM data using ArcGIS
  Create Signatures Tool
• Applied a Maximum Likelihood classifier (also in
  ArcGIS)

24
Initial classification results
Insert Jackson Pollack classification here
No. 9 Jackson Pollack
25
Sediment classification
Prepare backscatter data
Create final sediment map
Create signatures apply to backscatter data
Create training data
26
Six sediment classes quickly became three Hard
(H) ? Rocky Zone Gravel gravelly Sand (gS) ?
Coarse Sediments Mud (M) muddy Sand (mS)
Sand (S) ? Fine Sediments
(and H)
27
Reclassified sediment types
Brown Fine sediments Blue Coarse
sediments Red Rocky zone
28
Accuracy assessment
29
Bathymetry
Red shallow waters Blue deep waters
30
Rugosity

• Used Sappington et al. (2005) Vector
  Ruggedness Measure
• Directly measures variability in aspect and
  gradient
• Not strongly correlated with slope
• Better measure of habitat complexity than other
  common rugosity measures

31
Vector Ruggedness Measure
Blue low rugosity Green moderate rugosity Red
high rugosity
32
Potential habitat polys

• Upgridded both the sediment and rugosity grids to
  25m x 25m pixels (625m2 of seafloor)
• Best compromise (?) ? typically contains one
  sediment class yet is large enough to show
  backscatter texture
• Combined the two data sets with a logical
  statement in ArcGIS

33
Potential habitat polygons 4,828 polys total
Average size 0.026 km2 Std. Dev. 0.38 km2
34
Greene et al. classification

• Habitat codes
• Megahabitat (physiography approx. depth)
• Bottom induration (substrate hardness and/or
  consolidation)
• Meso-/Macrohabitat (geomorphology, structure,
  sedimentary features)
• Mega-/Macrohabitat modifier (sediment texture,
  bedform or rock type)
• Small-scale slope
• Small-scale rugosity

• Scale
• Megahabitat (gt1100K)
• Mesohabitat (lt1250K)
• Macrohabitat (lt150K)
• Microhabitat (ltlt150K)

35
Bottom Induration
36
Meso- / Macrohabitat
37
Mega- / Macrohabitat Modifier
Mega
38
Rugosity
39
Slope
40
What were the results?

• 84 unique Greene potential habitat code
  combinations
• Caveats
• Did not carefully take scale into account when
  adding attributes
• Im not a geologist so few geologic attributes
  were used

41
Assign attributes to polys
i

• All per Greene et al.
• Upgridded both the sediment and rugosity grids to
  25m2
• Combined the data sets with a logical statement
  in ArcGIS

Ss(s/m)f_u1A
Soft bottom continental shelf zone
composed of unconsolidated sand mud flat with
very low rugosity slope
42
Next steps

• Better define minimum mapping unit (MMU)
• Refine sediment classification
• Use additional training data
• Try different classifiers, e.g. neural network
• Reprocess backscatter data if possible
• Create an overall accuracy assessment or error
  budget
• Thematic accuracy
• Spatial accuracy

43
Questions?daniel.sampson_at_state.ma.us