Вы находитесь на странице: 1из 30

Methods proposal for wetland and riparian area

mapping efforts across BLM-administered land


throughout the Western United States
Final Report
June 30th, 2017

Prepared for:

Bureau of Land Management


BLM-NOC West-wide Wetland and Riparian Mapping Assessment

Prepared by:

Hannah Hutchins
ON THE COVER:
Rio de Los Pinos in Northcentral New Mexico
Photo taken by Andy Robertson, SMUMN GSS
Table of Contents
Acronyms and Definitions .............................................................................................................. 5

Introduction ..................................................................................................................................... 8

Project Summary ..................................................................................................................... 8

Purpose of Methods Proposal .................................................................................................. 8

Wetland and Riparian Area Mapping Standards ............................................................................ 8

National Wetlands Inventory, Version 2 ................................................................................. 9

Object Based Image Analysis ......................................................................................................... 9

Wetland and Riparian Area Mapping Approaches ....................................................................... 10

Cross-Platform Methods ........................................................................................................ 11

Geodatabase Design .......................................................................................................... 11

Data Sources ..................................................................................................................... 11

Open Water Data............................................................................................................... 13

Fine Approach ....................................................................................................................... 14

Standards ........................................................................................................................... 15

Methodology ..................................................................................................................... 15

Advantages and Disadvantages......................................................................................... 17

Medium Approach ................................................................................................................. 18

Standards ........................................................................................................................... 19

Methodology ..................................................................................................................... 19

Advantages and Disadvantages......................................................................................... 20

Coarse Approach ................................................................................................................... 21

Standards ........................................................................................................................... 22

Methodology ..................................................................................................................... 22

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 3 of 30
Advantages and Disadvantages......................................................................................... 25

Methods Review Summary ........................................................................................................... 26

Literature Cited ............................................................................................................................. 29

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 4 of 30
Figures
Figure 1. Example of an OBIA output generated on CIR imagery by GeoSpatial Services that
has been further processed to highlight wetland features in Alaska. .........................................10

Figure 2. Example of color-infrared (CIR) imagery. On the left is the 2015 true color NAIP
image source (red, green, and blue band), and on the right is the same NAIP image source but
with a near infrared band included (red, green, blue, and near infrared band). CIR imagery
makes water appear black and vegetation appear red, allowing wetland features to be easily
distinguished from upland features (yellow circle). ..................................................................12

Figure 3. An example of smoothed open water data generated by GeoSpatial Services in


northwestern Minnesota. Once created, the photo interpreter can use it as input data to capture
open water, as well as, a collateral dataset indicating potential wetland features. ....................13

Figure 4. An example of the fine mapping approach near Carmen, Idaho (USGS Salmon Quad
45113-B8). The decision scale for this product was 1:12,000, while the mapping scale ranged
from 1:5,000 to 1:6,000, thus capturing small wetland and riparian area features. ...................14

Figure 5. An example of the medium mapping approach near Carmen, Idaho (USGS Salmon
Quad 45113-B8). The decision scale for this product was 1:24,000, while the mapping scale
ranged from 1:12,000 to 1:15,000. This approach has a larger MMU and “lumps” together
more wetland and riparian area features than the fine mapping approach. ................................18

Figure 6. An example of the coarse mapping approach near Carmen, Idaho (USGS Salmon
Quad 45113-B8). The decision scale for this product was 1:40,000, while the mapping scale
ranged from 1:24,000 to 1:30,000. This approach has the largest MMU and “lumps” together
more wetland and riparian area features than the medium mapping approach. Linear features
are not included in the coarse mapping approach. ....................................................................21

Figure 7. A comparison of all three mapping approaches. The fine approach provides the most
detail, while the coarse omits wetland features and lumps in upland. Linear features vary as
well between the fine and medium approach; what would be polygons in the fine approach,
become linears in the medium. .................................................................................................27

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 5 of 30
Tables
Table 1. Specifications of the Federal Geographic Data Committee (FGDC) Wetland Mapping
Standard for federally-funding wetland inventory mapping projects (FGDC 2009). ................. 8

Table 2. Wetland Cowardin et al. (1979) classification specifications of the FGDC Wetland
Mapping Standard for federally-funded wetland inventory mapping projects (FGDC 2009). ... 9

Table 3. The standards developed by GSS that are used with the fine mapping approach. These
standards comply with the FGDC Wetland Mapping Standard (FGDC 2009). .........................15

Table 4. Advantages and disadvantages of the fine mapping approach. Also included in this
table are the polygon and linear data generated for the Salmon USGS Quad in Idaho (Quad ID:
45113-B8). ...............................................................................................................................17

Table 5. The standards developed by GSS that are used with the medium mapping approach.19

Table 6. Advantages and disadvantages of the medium mapping approach. Also included in
this table are the polygon and linear data generated for the Salmon USGS Quad in Idaho (Quad
ID: 45113-B8). .........................................................................................................................20

Table 7. The standards developed by GSS that are used with the coarse mapping approach. ..22

Table 8. Advantages and disadvantages of the coarse mapping approach. Also included in this
table are the polygon data generated for the Salmon USGS Quad in Idaho (Quad ID: 45113-
B8). Linear features are not included in this approach ..............................................................25

Table 9. Three different mapping approaches (fine, medium, and coarse) developed by GSS
are outlined below with information on the level of detail that is produced along with
advantages and disadvantages. .................................................................................................28

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 6 of 30
Acronyms and Definitions
BLM = Bureau of Land Management

CIR = color infrared

FGDC = Federal Geographic Data Committee

GSS = GeoSpatial Services

HGM = hydrogeomorphic

ISO = International Organization for Standardization

LLWW = Landscape Position, Landform, Water Flow Path, Waterbody Type

MMU = minimum mapping unit

MSAVI = Modified Soil Adjusted Vegetation Index

NAIP = National Agriculture Imagery Program

NDVI = Normalized Difference Vegetation Index

NDMI = Normalized Difference Moisture Index

NED = National Elevation Dataset

NHD = National Hydrography Dataset

NWI = National Wetlands Inventory

OBIA = Object-Based Image Analysis

PCA = principal component analysis

PI = Photo Interpreter

QAQC = quality assurance and quality control

SSURGO = Soil Survey Geographic Database

USFWS = U.S. Fish and Wildlife Service

USGS = U.S. Geological Survey

WMS = Web Mapping Service

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 7 of 30
Introduction
Project Summary
The Bureau of Land Management (BLM) is interested in a comprehensive wetland and riparian
inventory that captures the extent and status of these ecosystems across BLM-administered lands in
the Western U.S. This inventory will help provide a basis for understanding land health and
management of BLM lands through four fundamental areas of functioning wetland and riparian
ecosystems. These four fundamental areas cover: properly functioning watersheds, compliance
with state water quality standards, habitats restored or maintained for federal threatened and
endangered species and others, and the maintenance of ecological processes that sustain healthy
biotic populations and communities. In addition, this effort will inform the investigation of wetland
status, condition, and trend across much of the U.S.

Purpose of Methods Proposal


This methods proposal report will describe three mapping approaches (i.e. fine, medium, and
coarse) that were developed by GeoSpatial Services (GSS). These approaches can be used to map
wetland and riparian areas on BLM-administered land in the Western United States. In addition to
descriptions of these mapping approaches, recommendations for efficient and accurate mapping
across the varied landscape will be provided.

Wetland and Riparian Area Mapping Standards


The U.S. Fish and Wildlife Service (USFWS) has developed wetland mapping standards that are,
as of 2009, required to be implemented with any federally-funded wetland mapping initiative
(Table 1 and Table 2). The purpose of these standards are to ensure consistency and accessibility
across the nation with wetland data, in addition to compatibility with other water data sources such
as the National Hydrography Dataset (NHD) (FGDC 2009). A number of specifications are in
place for different aspects of this mapping standards; everything from the resolution of imagery
used to the minimum mapping unit (MMU).

Table 1. Specifications of the Federal Geographic Data Committee (FGDC) Wetland Mapping Standard
for federally-funding wetland inventory mapping projects (FGDC 2009).
Conterminous U.S., Hawaii & Territories
Alaska
Palustrine, Riverine, Marine Estuarine, Lacustrine
Source (base) Imagery Resolution1 1m 3m 5m
Imagery Scale 1:12,000 1:24,000 1:63,360
Minimum Mapping Unit (MMU)2 0.5 acres 1.0 acres 5.0 acres
Feature Accuracy (wetland identification) 98% 98% 98%
Attribute Accuracy
85% 85% 85%
(FGDC wetland classification)
1 This standard refers to the resolution of the imagery that is used to detect wetland features.
2 When at the imagery scale, this standard refers to the smallest a wetland feature can be to be
included in the on-screen delineation

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 8 of 30
Table 2. Wetland Cowardin et al. (1979) classification specifications of the FGDC Wetland Mapping
Standard for federally-funded wetland inventory mapping projects (FGDC 2009).
Water Special
System Sub-system Class Subclass*
Regime Modifiers
Palustrine,
Conterminous
U.S., Hawaii &

Riverine, Yes Yes Yes Yes Yes Yes


Territories

Marine

Estuarine,
Yes Yes Yes Yes Yes No
Lacustrine

Alaska Yes Yes Yes Yes Yes Yes

* At most, users should use this level for forested and scrub-shrub classes

Riparian area mapping standards are not as comprehensive as wetland mapping standards, but the
USFWS has documented that MMU “will be established for specific projects based on funding,
scale and quality of aerial photographs, and agency needs” (Dall et al. 1997 p. 4) and that the
Federal Geographic Data Committee (FGDC) Wetland Mapping Standard can be applied to
riparian areas as well (Dick et al. 2009).

National Wetlands Inventory, Version 2


In 2016, the USFWS released a more comprehensive data model for National Wetlands Inventory
(NWI) data (NWI V2) to incorporate stream features and to connect segmented polygons. These
streams and segments are represented by linears and will provide a dataset that better represents all
surface water features on the landscape (USFWS 2016c).

Object Based Image Analysis


Since the late-1990s, Object-Based Image Analysis (OBIA) has become the method of choice with
automated mapping (Blaschke 2003); this includes both detection and delineation of wetlands to a
comprehensive analysis of within-wetland cover types which can provide insight into wetland
classification (Dronova 2015). Prior to OBIA, pixel-based image analysis was used which focused
just on the spectral properties of each pixel within an image. Due to this technique providing
limited analysis, OBIA was introduced due to its ability to take into account spatial or contextual
information along with the spectral properties of each pixel. The only required dataset to execute
an OBIA is satellite or aerial imagery, but the output can be further enhanced with input from
collateral datasets such as elevation, soils, and/or vegetation. OBIA uses a segmentation process
and an iterative learning algorithm to create homogenous geo-objects, also known as image objects
or segments (Blaschke 2003) (Figure 1), which ultimately feeds into the semi-automatic mapping
approach (Weih and Riggan 2010). Once these image objects are created, the data can be further
processed (i.e. incorporate collateral datasets for further analysis, smooth line work, etc.) to

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 9 of 30
ultimately end up with vector polygons and be incorporated into a geographic information system
(GIS) (O'Neil-Dunne et al. 2013).

Figure 1. Example of an OBIA output generated on CIR imagery by GeoSpatial Services that has been
further processed to highlight wetland features in Alaska.

Donnelly (2012), for example, has successfully completed segmentation work along parts of the
Rio Grande in semi-arid regions of Colorado and New Mexico, while Knight et al. (2010) has used
similar methods to complete data automation in much wetter areas such as Minnesota.

Wetland and Riparian Area Mapping Approaches


Due to many variables such as BLM-administered land covering a large area across various
regions, collateral dataset availability, and available funding, three mapping approaches were
developed by GSS to help streamline the wetland and riparian area mapping process. These
approaches can also be considered ‘scalable mapping’ due to their ability to fill in data gaps and
ultimately contribute to the nation-wide NWI database. The USFWS defines ‘scalable mapping’ in
the same manner but has a more generalized approach in creating the data (MNHP 2016). The
methods outlined below differ from the methods used by the USFWS.

To understand the mapping approaches below, GSS defines the “decision scale” to be the scale at
which wetland and riparian area features are detected on-screen, while the “mapping scale” is the
scale at which wetland and riparian area features are delineated (or digitized) and classified. The

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 10 of 30
FGDC Wetland Mapping Standards are implemented to the greatest extent for all three approaches.
Exceptions to the standard will be outlined where appropriate.

Cross-Platform Methods

Geodatabase Design
GeoSpatial Services will assemble an ArcGIS version 10.4 file geodatabase using a geographic
projection in decimal degrees and referenced to the North American Datum (NAD) 1983 High
Accuracy Reference Network (HARN) geodetic datum for all three approaches. The geodatabase
will utilize techniques such as parsing of classification codes to ensure attribution accuracy and to
simplify the application of attributes during mapping.

Data Sources
Base (Source) Imagery
The National Agriculture Imagery Program (NAIP) provides nationwide one-meter imagery to
government agencies and the public within a year of acquisition. Since 2009, NAIP imagery is
being collected every three years; prior to that, it was collected every five years. The majority of
NAIP imagery is “true color” or “natural color”, meaning it contains a three band spectral
resolution (i.e. red, green, and blue band). Starting in 2007, some areas of the nation have an
additional near infrared band (fourth band) included in the NAIP imagery (U.S. FSA 2016).
Through a process of shifting the imagery color bands, this four band image source, also known as
color-infrared (CIR) imagery (MnGeo 2016), can provide a richer product that becomes useful in
distinguishing wetland and riparian area features versus upland features. With the help of the near
infrared band, vegetation appears red, water appears black, and artificial structures like roads and
buildings appear white to light blue (Figure 2) (MnGeo 2016).

Unless other image sources are provided, NAIP will be utilized for decision support in all three
mapping approaches due to its nation-wide availability and high resolution. When applicable, CIR
NAIP will be used instead of the true color NAIP. Before mapping starts, multiple image years will
be analyzed to determine a source image year that best reflects the presence of wetlands (i.e., not
abnormally dry or high amounts of rain); this is not always the most recent image year. This source
image is used as the foundation for decision making, and then other collateral imagery is used for
support.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 11 of 30
Figure 2. Example of color-infrared (CIR) imagery. On the left is the 2015 true color NAIP image source
(red, green, and blue band), and on the right is the same NAIP image source but with a near infrared
band included (red, green, blue, and near infrared band). CIR imagery makes water appear black and
vegetation appear red, allowing wetland features to be easily distinguished from upland features (yellow
circle).

Collateral Datasets
In addition to NAIP imagery, collateral datasets will be downloaded or streamed through a Web
Mapping Service (WMS). Nation-wide datasets such as the U.S. Geological Survey (USGS)
topographic maps and National Elevation Dataset (NED), Soil Survey Geographic Database
(SSURGO), and the historic NWI data can be utilized across all BLM-administered land. In
addition, state, county, and BLM-specific data that will help with the decision making process will
be utilized.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 12 of 30
Open Water Data
Data representing open water will be generated through an OBIA process and incorporated into all
three mapping approaches (Figure 3). Due to variability in NAIP tone, texture, and emulsion across
the Western U.S., multiple spectral properties will be determined that best represent and define
open water features. Once these spectral properties are determined, image segmentation will be
performed to create the data, which will then be integrated into the NWI dataset. This automated
data will be helpful in expediting the wetland and riparian area mapping process.

Figure 3. An example of smoothed open water data generated by GeoSpatial Services in northwestern
Minnesota. Once created, the photo interpreter can use it as input data to capture open water, as well as,
a collateral dataset indicating potential wetland features.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 13 of 30
Fine Approach
GeoSpatial Services has developed a fine scale mapping approach that is based on a combination
of data automation techniques and traditional interpretation methods (i.e. on-screen detection and
digitizing of wetlands along with manual classification). In compliance with the FGDC Wetland
Mapping Standard (FGDC 2009), this mapping approach provides the most detailed data (Figure
4).

Figure 4. An example of the fine mapping approach near Carmen, Idaho (USGS Salmon Quad 45113-
B8). The decision scale for this product was 1:5,000, while the mapping scale ranged from 1:2,500 to
1:5,000, thus capturing small wetland and riparian area features.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 14 of 30
Standards
The fine mapping approach captures all wetland and riparian area polygon features that are at least
one-tenth to one-quarter of an acre (approximately 404 to 1,102 sq m); this is the MMU for this
approach. The interpreter scans the landscape for wetland and riparian area features at a scale of
1:5,000 (decision scale), then zooms in to 1:2,500 or 1:5,000 and delineates (digitizes) the feature
(mapping scale); this provides FGDC Wetland Mapping Standard compliant data. All six levels
(Table 2) of the Cowardin et al. (1979) classification are used. Depending on wetland and riparian
area density, complexity of the landscape, and base imagery and collateral dataset availability, a
USGS 1:24,000 quad can be completed in approximately ten to 20 hours by a GSS Photo
Interpreter (PI) (Table 3).

Table 3. The standards developed by GSS that are used with the fine mapping approach. These
standards comply with the FGDC Wetland Mapping Standard (FGDC 2009).
Approach Decision Scale Mapping Scale MMU Hours per Quad (USGS 24k)
Fine 1:5,000 1:2,500 – 1:5,000 0.1 - .25 acre 10 – 20

Methodology
GeoSpatial Services will implement the following procedures for the fine mapping approach:

1. Photo Interpreter scans the assigned project area [comprised of U.S. Geological Survey
(USGS) quadrangles (quads)] and performs traditional interpretation methods to capture
wetland and riparian area features (both polygon and linear features) according to the fine
scale approach standards outlined above. Due to interpretation being a subjective process, it
is important for the PI to have a basic understanding of climatology, geomorphology,
geology, ecology, and hydrology. Deductive reasoning, based on the physical
characteristics of the imagery being assessed and the scientific characteristics of the terrain
elements that are represented on the image, allows a PI to map wetland data from remotely
sensed imagery. The line work generated from the open water data will be “cleaned-up”
where necessary to ensure data accuracy and quality line work.
a. NAIP (CIR when applicable) imagery is used as the source for decision making; the
year of imagery used will be determined based on availability in the project area.
Other collateral datasets will be utilized to inform the decision making process.
b. Once a wetland and riparian area feature is created, the Cowardin et al. (1979)
classification system and the Western Riparian Classification system (Dick et al.
2009) will be applied, respectively

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 15 of 30
2. Undertake quality assurance and quality control (QAQC) reviews as wetland and riparian
area data are completed. These reviews will be done by the PI, known as “self-QAQCs”,
and then primarily be conducted by a QAQC Specialist.
a. Reviews consists of a complete visual inspection by the QAQC Specialist to assess
delineation and classification accuracy, quality of digital line work, check for errors
of omission and commission, and identify revisions that are necessary to meet the
FGDC Wetland Mapping Standard (FGDC 2009); multiple iterations between the
QAQC Specialist and the PI can occur if necessary.
b. Internally developed error-checking scripts that search for topology errors (i.e. gaps
between adjacent polygons) and attribute errors (i.e. adjacent polygons that contain
the same attribute) will be executed as well.
3. Once the data has gone through the QAQC process ensuring accurate line work and NWI
classification, the Landscape Position, Landform, Water Flow Path, and Waterbody Type
(LLWW) classification (Tiner 2014) will be applied at a scale of approximately 1:24,000.
The LLWW classification used for this project will be designed to fit the needs of the arid
west landscape. In addition, a hydrogeomorphic (HGM) subclass will be assigned based on
wetland characteristics derived from both the Cowardin et al. (1979) and the arid-specific
LLWW. Key variables that contribute to assignment of these subclasses will be geomorphic
setting (e.g. isolated depression, flowing, fringe, etc.), water source (e.g. precipitation,
groundwater, lateral flow, etc.), and hydrodynamic properties (e.g. vertical fluctuation,
dynamic flow, etc.).
4. As quads get completed, edge matching and topological structuring of adjacent quads will
be performed to create a seamless wetland and riparian area database. Topology validation
scripts and other verification tools will be run on the final merged dataset to ensure data
integrity.
5. Develop full project metadata for the final wetland and riparian area database. This
metadata will be formatted to meet the FGDC Metadata Guidelines, which work in
conjunction with the International Organization for Standardization (ISO) metadata
standards (FGDC 2016).

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 16 of 30
Advantages and Disadvantages
Along with a visual representation of the fine mapping approach (Figure 4), the advantages and
disadvantages of this approach is described below. In addition, data on the number of polygons and
linears created, the acres and miles mapped, and the number of unique NWI codes generated are
included in Table 4.

Table 4. Advantages and disadvantages of the fine mapping approach. Also included in this table are the
polygon and linear data generated for the Salmon USGS Quad in Idaho (Quad ID: 45113-B8).
Advantages Disadvantages Salmon Quad, Idaho (45113-B8)
 Captures small wetland and  Takes the most time to create Polygon Features
riparian area features  Generates a large dataset;  593 polygon wetland and riparian
 Includes wetland linears takes up storage space area features created
 Provides a detailed  Requires a high resolution  1,164 acres mapped
representation of the wetland image source  43 unique NWI codes
and riparian area landscape  Requires fieldwork Linear Features
 Does not lump together
wetland features  395 linear wetland features
created
 Provides most support for
decision making  160 miles mapped
 Can be integrated into a rapid  29 unique NWI codes
assessment methodology

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 17 of 30
Medium Approach
GeoSpatial Services has also developed a medium scale mapping approach that is based on a
combination of data automation techniques and traditional interpretation methods (i.e. on-screen
detection and digitizing of wetlands along with manual classification). This approach provides a
moderate amount of detail in terms of feature inclusion and level of classification (Figure 5).

Figure 5. An example of the medium mapping approach near Carmen, Idaho (USGS Salmon Quad
45113-B8). The decision scale for this product was 1:24,000, while the mapping scale ranged from
1:12,000 to 1:15,000. This approach has a larger MMU and “lumps” together more wetland and riparian
area features than the fine mapping approach.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 18 of 30
Standards
The medium mapping approach captures all wetland and riparian area features that are at least two
to five acres (approximately 8,094 to 20,234 sq m); this is the MMU for this approach. The PI
scans the landscape for wetland and riparian area features at a scale of 1:24,000 (decision scale),
then zooms in to 1:12,000 or 1:15,000 and delineates (digitizes) the feature (mapping scale). All
six levels of the Cowardin et al. (1979) classification are used when applicable; subclasses and
special modifiers could be excluded if the interpreter does not have appropriate collateral data to
make a decision. Depending on wetland and riparian area density, complexity of the landscape, a nd
base imagery and collateral dataset availability, a USGS 1:24,000 quad can be completed in
approximately eight to 12 hours by a GSS PI (Table 5).

Table 5. The standards developed by GSS that are used with the medium mapping approach.
Approach Decision Scale Mapping Scale MMU Hours per Quad (USGS 24k)
Medium 1:24,000 1:12,000 – 1:15,000 2 – 5 acre 8 – 12

Methodology
GeoSpatial Services will implement the following procedures for the medium mapping approach:

1. Photo Interpreter scans assigned project area (comprised of USGS quads) and performs
traditional interpretation methods to capture wetland and riparian area features (both
polygon and linear features) according to the medium scale approach standards outlined
above. Due to interpretation being a subjective process, it is important for the PI to have a
basic understanding of climatology, geomorphology, geology, ecology, and hydrology.
Deductive reasoning, based on the physical characteristics of the imagery being assessed
and the scientific characteristics of the terrain elements that are represented on the image,
allows a PI to map wetland data from remotely sensed imagery. The line work generated
from the open water data will be “cleaned-up” where necessary to ensure data accuracy and
quality line work.
a. NAIP (CIR when applicable) imagery is used as the source for decision making; the
year of imagery used will be determined based on availability in the project area.
Other collateral datasets will be utilized to inform the decision making process.
b. Once a wetland and riparian area feature is created, the Cowardin et al. (1979)
classification system and the Western Riparian Classification system (Dick et al.
2009) will be applied, respectively.
2. Undertake QAQC reviews as wetland and riparian area data are completed. These reviews
will be done by the Photo Interpreter, known as “self-QAQCs”, and then will primarily be
conducted by a QAQC Specialist.
a. Reviews consists of a complete visual inspection by the QAQC Specialist to assess
delineation and classification accuracy, quality of digital line work, check for errors
of omission and commission, and identify revisions that are necessary to meet the

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 19 of 30
FGDC Wetland Mapping Standard (FGDC 2009); multiple iterations between the
QAQC Specialist and the PI can occur if necessary.
b. Internally developed error-checking scripts that search for topology errors (i.e. gaps
between adjacent polygons) and attribute errors (i.e. adjacent polygons that contain
the same attribute) will be executed as well.
3. Once the data has gone through the QAQC process ensuring accurate line work and NWI
classification, the LLWW classification (Tiner 2014) will be applied at a scale of
approximately 1:24,000. The LLWW classification used for this project will be designed to
fit the needs of the arid west landscape. In addition, an HGM subclass will be assigned
based on wetland characteristics derived from both the Cowardin et al. (1979) and the arid-
specific LLWW. Key variables that contribute to assignment of these subclasses will be
geomorphic setting (e.g. isolated depression, flowing, fringe, etc.), water source (e.g.
precipitation, groundwater, lateral flow, etc.), and hydrodynamic properties (e.g. vertical
fluctuation, dynamic flow, etc.).
4. As USGS quads get completed, edge matching and topological structuring of adjacent
quads will be performed to create a seamless wetland and riparian area database. Topology
validation scripts and other verification tools will be run on the final merged dataset to
ensure data integrity.
5. Develop full project metadata for the final wetland and riparian area database. This
metadata will be formatted to meet the FGDC Metadata Guidelines, which work in
conjunction with the ISO metadata standards (FGDC 2016).

Advantages and Disadvantages


Along with a visual representation of the medium mapping approach (Figure 5), the advantages
and disadvantages of this approach is described below. In addition, data on the number of polygons
and linears created, the acres and miles mapped, and the number of unique NWI codes generated
are included in Table 6.

Table 6. Advantages and disadvantages of the medium mapping approach. Also included in this table are
the polygon and linear data generated for the Salmon USGS Quad in Idaho (Quad ID: 45113-B8).
Advantages Disadvantages Salmon Quad, Idaho (45113-B8)
 Captures a moderate amount  Lumps together a moderate Polygon Features
of wetland and riparian area amount of wetland features  173 polygon wetland and riparian
features  Uses a generalized NWI area features created
 Does not take as much time as classification system; less  862 acres mapped
the fine scale approach information for decision  33 unique NWI codes
 Includes wetland linears support
Linear Features
 Moderate representation of the
wetland and riparian area  361 linear wetland features
landscape created
 Does not require fieldwork  160 miles mapped
 30 unique NWI codes

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 20 of 30
Coarse Approach
Finally, GSS has developed a coarse scale mapping approach that is primarily based on data
automation techniques and then incorporates traditional interpretation methods to QAQC the
automated data and make minor adjustments where necessary. This approach provides the least
amount of detail in terms of feature inclusion and level of classification; a large amount of feature
lumping and generalized attribution occurs (Figure 6).

Figure 6. An example of the coarse mapping approach near Carmen, Idaho (USGS Salmon Quad 45113-
B8). The decision scale for this product was 1:40,000, while the mapping scale ranged from 1:24,000 to
1:30,000. This approach has the largest MMU and “lumps” together more wetland and riparian area
features than the medium mapping approach. Linear features are not included in the coarse mapping
approach.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 21 of 30
Standards
The coarse mapping approach captures all wetland and riparian area features that are at least eight
to 11 acres (approximately 32,375 to 44,515 sq m); this is the MMU for this approach. Linear
features are not included. All six levels (Table 2) of the Cowardin et al. (1979) classification are
used to the best of the PI’s ability. Depending on wetland and riparian area density, complexity of
the landscape, and base imagery and collateral dataset availability, a USGS 1:24,000 quad can be
completed in approximately ten to 20 hours after data automation techniques has been performed
(Table 7).

Table 7. The standards developed by GSS that are used with the coarse mapping approach.
Approach Decision Scale Mapping Scale MMU Hours per Quad (USGS 24k)
Coarse 1:40,000 1:24,000 – 1:30,000 8 – 11 acre 6 to 10

Methodology
The coarse mapping approach is broken down into two processes: image segmentation using NAIP
imagery and feature classification using Landsat Surface Reflectance Higher Level Data Products.
GeoSpatial Services will implement the following procedures for this mapping approach:

Data Assembly and Preprocessing


1. Import individual NAIP images into a mosaic dataset for mosaicking and color balancing.
a. Generate a CIR composite image for the study area using the extract bands function
inside the mosaic dataset.
b. Perform 3-band segmentation on the CIR images using the Mean Shift
segmentation algorithm.
c. Convert raster segments to smoothed vector polygon segments.

Due to the presence of water being highly variable in the arid west, it is important to utilize
multiple years of Landsat data to capture changes in vegetation phenology and hydrologic cycles.
The principal component analysis (PCA) is used throughout this process. This analysis removes
data redundancy of the segmented raster data and emphasizes variation and strong patterns in the
data.

2. Perform a PCA on Landsat CIR spectral bands for multiple image years.
a. Extract the first three components of the spectral band PCA due to these
components containing the strongest variance in spectral properties.
b. Perform segmentation of these first three spectral band components for input into
the automated classifier process.
3. Perform a PCA on Landsat spectral indices datasets for multiple image years.
a. Indices will include the Normalized Difference Vegetation Index (NDVI) or the
Modified Soil Adjusted Vegetation Index (MSAVI), along with the Normalized

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 22 of 30
Difference Moisture Index (NDMI). The use of NDVI or MSAVI will depend on
the sparsity of vegetation and the potential effect of soil background reflectance.
b. Extract the first three components of the spectral indices bands PCA due to these
components containing the strongest variance in spectral properties.
i. These components are inputs into the automated classifier process.
4. Assemble best available elevation data (LiDAR or NED).
a. Generate a Compound Topographic Index (CTI) using the TauDEM and the Dinf
flow model (Knight et al. 2009).

Data Processing
1. Input the extracted Landsat PCA spectral components (CIR and indices) and CTI data into
an unsupervised automated classifier such as the ISO Cluster classification algorithm.
2. For unsupervised classification, the resulting clusters will be assigned to wetland, upland,
and open water classes. A cluster busting process can be used in additional unsupervised
classification iterations to resolve confused clusters and to then assign them to appropriate
classes.
3. Assign the majority class from the unsupervised classification results to overlapping
smoothed vector polygon segments. Results of this process is the final coarse mapping
result of Landsat data which contains the classified water and wetland polygon segments.
4. Compute weighted sum overlays with classified Landsat wetland segments over the
following wetland indicators: rasterized hydric soils, CTI. Classify results into high,
intermediate, and low probability of wetland occurrence based on overlap of these wetland
indicators. High probability of wetland occurrence includes all three wetland indicators,
medium probability includes two of the three, while low probability contains only one of
the three indicators.
5. Extract Landsat polygon segments that intersect and have greater than 50 percent of area
overlapped by wetland indicators. Results of this process is a dataset of potential wetland
segments that have a generalized measure of probability of wetland occurrence based on
the available indicators.

Post Processing
1. The Cowardin et al. (1979) classification system and the Western Riparian Classification
system (Dick et al. 2009) will be applied to the data.
2. Undertake QAQC reviews as wetland and riparian area data are completed. These reviews
will be done by the Photo Interpreter, known as “self-QAQCs”, and then will primarily be
conducted by a QAQC Specialist.
a. Reviews consists of a complete visual inspection by the QAQC Specialist to assess
delineation and classification accuracy, quality of digital line work, check for errors
of omission and commission, and identify revisions that are necessary to meet the

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 23 of 30
FGDC Wetland Mapping Standard (FGDC 2009); multiple iterations between the
QAQC Specialist and the PI can occur if necessary.
b. Internally developed error-checking scripts that search for topology errors (i.e. gaps
between adjacent polygons) and attribute errors (i.e. adjacent polygons that contain
the same attribute) will be executed as well.
3. Once the data has gone through the QAQC process ensuring accurate line work and NWI
classification, the LLWW classification (Tiner 2014) will be applied at a scale of
approximately 1:24,000. The LLWW classification used for this project will be designed to
fit the needs of the arid west landscape. In addition, an HGM subclass will be assigned
based on wetland characteristics derived from both the Cowardin et al. (1979) and the arid-
specific LLWW. Key variables that contribute to assignment of these subclasses will be
geomorphic setting (e.g. isolated depression, flowing, fringe, etc.), water source (e.g.
precipitation, groundwater, lateral flow, etc.), and hydrodynamic properties (e.g. vertical
fluctuation, dynamic flow, etc.).
4. As USGS quads get completed, edge matching and topological structuring of adjacent
quads will be performed to create a seamless wetland and riparian area database. Topology
validation scripts and other verification tools will be run on the final merged dataset to
ensure data integrity.
5. Develop full project metadata for the final wetland and riparian area database. This
metadata will be formatted to meet the FGDC Metadata Guidelines, which work in
conjunction with the ISO metadata standards (FGDC 2016).

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 24 of 30
Advantages and Disadvantages
Along with a visual representation of the coarse mapping approach (Figure 6), the advantages and
disadvantages of this approach is described below. In addition, data on the number of polygons
created, the acres mapped, and the number of unique NWI codes generated are included in Table 8.
Linear wetland features are not included in this approach.

Table 8. Advantages and disadvantages of the coarse mapping approach. Also included in this table are
the polygon data generated for the Salmon USGS Quad in Idaho (Quad ID: 45113-B8). Linear features
are not included in this approach
Advantages Disadvantages Salmon Quad, Idaho (45113-B8)
 May rely on data automation  A fair amount of wetland Polygon Features
techniques feature lumping occurs  90 polygon wetland and riparian
 Lowest cost approach  Does not include wetland area features created
linears  537 acres mapped
 Uses a generalized NWI  20 unique NWI codes
classification system; less
information for decision
support
 Data only useful for regional-
level planning; lack of detail

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 25 of 30
Methods Review Summary
Due to many variables such as BLM-administered land covering a large area across various
regions, collateral dataset availability, and available funding, three mapping approaches (i.e. fine,
medium, and coarse) were developed by GSS to help streamline the wetland and riparian area
mapping process. Each approach provides varying degrees of the same NWI dataset (Figure 7).
The fine approach, which complies with the FGDC Wetland Mapping Standards (FGDC 2009),
provides the most detailed data out of all three approaches. This means small wetland features are
included (the MMU is one-tenth of an acre or 404 square meters) and upland is completely
excluded from the data. Data generated using the fine approach can be integrated into a rapid
assessment methodology and provides the most support for decision making. The coarse approach
provides the least amount of detail. Small wetland features are excluded from the data and a fair
amount of lumping occurs with features. For example, if there is a large sloped wetland (Cowardin
code: PEM1B) with little basins or depressions where water gets held (Cowardin code: PEM1C)
within, with the coarse approach the depressions will more than likely be lumped in with the
sloped wetland. The coarse approach, despite having the lowest cost, provides a lack in data
comprehension and can only be used for region-based planning. Table 9 provides a summary of
each.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 26 of 30
Figure 7. A comparison of all three mapping approaches. The fine approach provides the most detail, while the coarse omits wetland
features and lumps in upland. Linear features vary as well between the fine and medium approach; what would be polygons in the fine
approach, become linears in the medium.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 27 of 30
Table 9. Three different mapping approaches (fine, medium, and coarse) developed by GSS are outlined below with information on the level
of detail that is produced along with advantages and disadvantages.
Decision Minimum Average Hours per
Approach Mapping scale2 Pro of Approach Con of Approach
scale1 mapping unit USGS quad (24k)
 Captures small  Large time investment
wetlands  Requires high resolution
 Upland completely image source
0.1 to 0.25
Fine3 1:5,000 1:2,500 to 1:5,000
acres
10 to 20 hours excluded  Requires fieldwork
 Provides support for
decision making
 Includes wetland linears
 Moderate amount of  Moderate amount of
defined wetland types wetland lumping
 Includes wetland  Upland features can be
linears included
Medium 1:24,000 1:12,000 to 1:15,000 2 to 5 acres 8 to 12 hours
 Does not require  Generalized
fieldwork classification
 Less information for
decision support
 Lowest cost approach  Large amount of wetland
 May rely on data lumping
automation  Upland features will be
included
Coarse 1:40,000 1:24,000 to 1:30,000 8 to 11 acres 6 to 10 hours  Generalized
classification
 Data only useful for
regional- based planning

1 The scale at which wetland and riparian area features are detected on-screen.
2 The scale at which wetland and riparian area features are delineated (or digitized) and classified.
3 Federal Geographic Data Committee (FGDC) National Wetland Mapping Standard for the conterminous U.S.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 28 of 30
Literature Cited
Blaschke, T. 2003. Object-based contextual image classification built on image segmentation. Pages
113-119 in IEEE Workshop on Advances in Techniques for Analysis of Remotely Sensed Data,
Washington, D.C.

Cowardin, L. M., V. Carter, F. C. Golet, and E. T. LaRoe. 1979. Classification of wetlands and
deepwater habitats of the United States. Version 04DEC98. U.S. Fish and Wildlife Service,
Washington, D.C.

Dall, D., C. Elliot, and D. Peters. 1997. A system for mapping riparian areas in the Western United
States. U.S. Fish and Wildlife Services, Arlington, Virginia.

Dick, J., E. Blok, K. Bon, B. Kirchner, T. E. Dahl, M. T. Bergeson, and J. Miner. 2009. A system for
mapping riparian areas in the Western United States. U.S. Fish and Wildlife Services, Arlington,
Virginia.

Donnelly, P. 2012. Image segmentation for wetlands inventory: Data considerations and concepts.
Webinar at Association of State Wetland Managers 2012 Wetland Mapping Consortium, 16 May,
2012.

Dronova, I. 2015. Object-based image analysis in wetland research: A review. Remote Sensing
7:6380-6413.

Federal Geographic Data Committee (FGDC). 2009. Wetlands Mapping Standard. FGDC-STD-015-
2009. Federal Geographic Data Committee (FGDC), Reston, Virginia.
https://www.fws.gov/wetlands/Documents/FGDC-Wetlands-Mapping-Standard.pdf (accessed 22
November 2016).

Federal Geographic Data Committee (FGDC). 2016. ISO Geospatial Metadata Standards.
https://www.fgdc.gov/metadata/iso-standards. (accessed 30 March 2017).

Knight, J. F., J. Corcoran, L. Rampi, B. Tolcser, and M. Voth. 2009. Wetland mapping methods for
the Twin Cities Metropolitan area. University of Minnesota, Department of Forest Resources,
Remote Sensing and GeoSpatial Analysis Laboratory, St. Paul, Minnesota.

Knight, J. F., J. Corcoran, L. Rampi, B. Tolcser, and M. Voth. 2010. Wetland mapping methods for
the Arrowhead Region of Minnesota. University of Minnesota, Department of Forest Resources,
Remote Sensing and GeoSpatial Analysis Laboratory, St. Paul, Minnesota.

Minnesota Geospatial Information Office (MnGeo). 2016. Color-Infrared (CIR) Imagery.


http://www.mngeo.state.mn.us/chouse/airphoto/cir.html. (accessed 05 April 2017).

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 29 of 30
Montana Natural Heritage Program (MTNHP). 2016. Wetlands Information.
http://mtnhp.org/wetlands/. (accessed 20 February 2017).

O'Neil-Dunne, J. P. M., S. W. MacFaden, A. R. Royar, and K. C. Pelletier. 2013. An object-based


system for LiDAR data fusion and feature extraction. Geocarto International 28(3):227-242.

Tiner, R. W. 2014. Dichotomous keys and mapping codes for wetland landscape position, landform,
water flow path, and waterbody type descriptors. Version 3.0. U.S. Fish and Wildlife Service,
Hadley, Massachusetts.

U.S. Farm Service Agency (FSA). 2016. NAIP Imagery. https://www.fsa.usda.gov/programs-and-


services/aerial-photography/imagery-programs/naip-imagery/. (accessed 05 April 2017).

Weih, R. C., Jr. and N. D. Riggan, Jr. 2010. Object-based classification vs. pixel-based
classifications: Comparitive importance of multi-resolution imagery. Pages 6-11 in GEOBIA
2010: Geographic Object-Based Image Analysis, Ghent, Belgium.

GEOSPATIAL SERVICES, Saint Mary’s University Of Minnesota


700 Terrace Heights ∙ Winona, Minnesota ∙ 55987 ∙ 507-457-8746 ∙ www.geospatialservices.org Page 30 of 30