a spatial-temporal analysis of wetland loss and ... · 1. develop improved wetland mapping and...
TRANSCRIPT
A Spatial-Temporal Analysis of Wetland Loss and
Vulnerability: Needs for Improved
Mapping and Monitoring of Wetland Dynamics for
Improved Resilience and Delivery of Ecosystem
Services in the Mid-Atlantic Region
In-Young Yeo a (with contribution from Quentin Stubbs a, e) M. Langb, C. Huanga, C. Jantzc, V. Thomasd, S. Princea, and M. Kearneya
a Department of Geographical Sciences, University of Maryland ; b USDA Forest Service Northern Research Station;
c Shippensburg University; d Virginia Tech; e USGS
Designing Sustainable Coastal Habitats
16-17 April 2013
Background: Wetlands in the CBW
The importance of
wetland systems are
recognized regionally,
nationally, and
internationally
(RAMSAR site).
The CBW has lost
over 60% of its
historic wetlands.
It is predicted that
existing wetlands are
at high risk for future
loss.
Risk Based on Historic Population/Construction
Data and 1980s NWI Wetland Density
Wetland Mapping Background Accurate, dynamic wetland maps can improve
society’s resilience to increasing
urbanization, population growth, and
climate change through
early detection
improved understanding of climate
change and LCLUC effects
enhanced management of wetlands to
target desired ecosystem services
Wetland hydroperiod (duration and
frequency of flooding and soil saturation at a
specified depth) is the most important abiotic
factor controlling wetland extent and function
Fall Spring
Hydroperiod
controls
Biota Vegetation
Water Quality
Soils
Wetland Mapping Background
Fundamental wetland data are lacking to conduct regional-scale
vulnerability assessment of wetlands and to identify hot spots, and key
drivers for wetland losses.
Available wetland maps are dated and do not represent wetland dynamics.
Water levels in wetlands can vary highly, in response to climate condition.
Rapid changes in land cover further confound wetland mapping in developing
areas.
Forested wetland maps are inaccurate, especially at moderate spatial
resolutions.
The vast majority of wetlands in the CBW are forested.
Forested wetlands continue to sustain high levels of loss.
Improved mapping of forested wetlands and wetland change is needed
to assess current impacts and vulnerabilities of wetlands to climate
and land cover change and to develop adaptation strategies.
1. Develop improved wetland mapping and change detection using remote-sensing data from multiple, complementary sensors at various temporal and spatial scales;
2. Study the socioeconomic, policy, and physical drivers of wetland change affecting wetland extent and function from regional to local scales, using existing data;
3. Assess the impacts of multiple environmental stressors (particularly the anthropogenic ones);
4. Quantify vulnerability of wetlands and wetland ecosystem services under multiple climate and land use change scenarios.
Research Objectives
The overall research design and
outcomes
(Part 1) Geospatial Analysis (obj 1) (Part 2) Impacts and Vulnerability
Assessment (obj 2&3)
(Part 3) Ecosystem Service Assessment (obj4)
Project benefits
Improved, consistent, recent wetland maps for
CBW – readily operational technique
Enhanced understanding of change drivers = greater
conservation of wetlands and ecosystem services
Near time tracking of wetland loss to enhance
regulatory abilities
Targeting for restoration and conservation
Improved parameterization of the Bay Model
Support for Ecosystem Markets
Greater understanding of climate change impacts
Progress and findings
Vulnerability Impact Assessment – regional
approach and case study
Innovation in mapping technique
Ecosystem service assessment – integrated
approach (literature review, numerical modeling,
and parameterization of existing model)
Vulnerability Impact Assessment: Wetland changes and their potential drivers – regional
assessment using existing dataset • What is the quantity of the wetland change over the last 30 years on
the Delmarva and CBW, measured by existing geospatial data sets?
• What physical and anthropocentric drivers of land use and land cover change are correlated with wetland loss?
• What information do the spatial and temporal distribution of wetland permits and wetland loss patterns provide regarding the influence of wetland change drivers and the impacts of the wetland permitting system?
• What wetlands, watersheds, and counties are most vulnerable to wetland loss due to physical, socioeconomic and policy drivers of wetland change?
(c)Wetland Loss (In Red)
between 1992-2006 (CCAP)
(d) Hot Spot and Point Density of Wetland Permits issued by the
State of MD, MDE
(b) Forest and Agricultural Conversion in the CBW between
1990 and 2000 (Jantz et al., 2005)
(a) Case study site
Regional assessment: land cover data
available on wetlands in the CBW The National Wetlands Inventory
Outlier Detection Analysis (Neilson et al.,
2008)
Land Cover Datasets
The National Land Cover Dataset
(NLCD)
The Coastal Change Analysis Program
(CCAP)
The Chesapeake Bay Watershed Land
Cover Data Series (CBLCD)
All of our Datasets cover the entire study
area, except for the CCAP
Area CCAP does cover accounts for
~98% of the wetlands in the study area.
Temporal Extent
Each Dataset covers a range of different time periods, both yearly and seasonally. NWI – 1977 to 1998
NLCD – 1992 (unusable for wetlands), 2001, 2006
CCAP – 1992 (DE, MD, VA) 1996, 2001, 2006
CBLCD – 2001 baseline, 2006, 1992 and 1984 retroactively
ODA – 2000/2001 (baseline – NWI map)
Land Cover Mapping
Classification and Change Detection
All of the land cover products used Classification and Regression Tree (CART) approach
Change identification for updates
CCAP and CBLCD – MDA Federal’s Cross Correlation Analysis
NLCD – Multi-Index Integrated Change Analysis (MIICA)
Outlier Detection Method – major, minor, or no change in wetlands (using z score from Landsat spectra, similar to the cross correlation analysis) since NWI
Land Cover Mapping
All of these datasets incorporate ancillary data Not as specific about how data was used
DEM, NWI
NCLD uses extensive independent data to help map urban land cover Independently generated percent imperviousness layers,
NOAA city lights, tigerline roads
CCAP also uses an impervious surface layer and roads layer, but the roads are not as detailed in final map
NLCD uses a minimum mapping unit to preserve logic of land cover Urban land cover– 5 pixels
Agricultural land cover – 32 pixels
All other classes – 12 pixels
The CBLCD corrected for an overabundance of coastal emergent wetlands
Minimum mapping
unit and roads
example
Total Wetland Change by Land Cover
Product, 2001-2006 Area (Hectares) C-CAP NLCD CBLCD
Total Wetland Area
(2001)
785489.49 758100.60 779748.75
Total Wetland Area
(2006)
777587.58 750048.84 773297.19
Net Change -7901.91 -8051.76 -6451.56
Percent Change -1.00598 -1.06209 -0.82738
Wetland Loss
Development 1226.16 2367.27 1074.33
Agriculture 4188.15 2050.02 3564
Forest 669.69 3973.5 565.92
Grass/Scrub 769.23 2358.63 800.55
Bare 491.13 328.23 431.91
Water 2409.66 2334.24 2044.62
Total 9754.02 13411.89 8481.33
Wetland Gain
Development 28.44 0 2.34
Agriculture 68.67 1558.17 121.32
Forest 848.07 1950.66 872.91
Grass/Scrub 544.32 1032.93 486
Bare 34.83 25.83 34.2
Water 322.56 792.54 505.71
Total 1846.89 5360.13 2022.48
Net Change
Development -1197.72 -2367.27 -1071.99
Agriculture -4119.48 -491.85 -3442.68
Forest 178.38 -2022.84 306.99
Grass/Scrub -224.91 -1325.7 -314.55
Bare -456.3 -302.4 -397.71
Water -2087.1 -1541.7 -1538.91
Total -7907.13 -8051.76 -6458.85
Geographical distribution of
wetland changes
NWI 2007 Delaware Updates
The NWI has been updated for the year of 2007
The FWS released a status and trends report
covering 1992-2007, based on the NWI update
Comparison of the CCAP and CBLCD over 1992-
2006 with the FWS 1992-2007 report
FWS Report Comparison
FWS report (2007) 83,877 acres emergent wetland
72,839 acres of estuarine emergent wetland
10,938 acres of palustrine emergent wetland
225,546 acres of forested wetland
CCAP (2006) 83,453 acres emergent wetland
75,532 acres estuarine emergent
7,921 acres palustrine emergent
174,001 acres forested wetland
CBLCD (2006) 79,445 emergent wetlands
176,068 Forested wetlands
CCAP and FWS agreement on emergent wetlands indication that CBLCD overcorrected for emergent wetlands?
Spatial agreement between Land
Cover and NWI
• ~71% of NWI polygons agree with the
location wetlands in CCAP and CBLCD
• CCAP agrees with any NWI polygon: 83%
Emergent wetlands agree with NWI (86% for
estuarine), 65% Forested wetlands, 38%
Scrub/Shrub wetlands
• CBLCD agrees with any NWI polygon: 85%
emergent wetlands, 61% forested wetlands
FWS Report Comparison on
change (1992-2007)
Total Wetland Change: FWS Report – 3893.4 acres lost, 767.5 acres gained
CCAP – 8453.87 acres lost, 1821.64 acres gained
CBLCD – 7085.5 acres lost, 514.84 acres gained
FWS Change Estuarine wetland loss to – 84% open water, 11% intertidal
unconsolidated shore
Palustrine wetland loss to – 32% Agriculture, 32% Development, 18% Altered/Barren land
CCAP change Estuarine wetland loss to – 53% open water, 18% cultivated land, 8%
Development
Palustrine wetland loss to – 52% cultivated land, 16% scrub shrub, 14% forest, 5% development
Earlier land cover products (1992) are not reliable? (under-investigation- land cover vs. land use/cover)
NWI and Outlier Detection
Analysis The Outlier Detection Analysis was developed by
Nielsen et. al.(2008) as an update to the NWI, using
multiseasonal Landsat ETM+ images from 2000
Identify hotspots & estimate total amount of
wetland changes at different scales
Validation exercise resulted in an 89.9 percent
accuracy assessment with a Kappa score of 0.779.
The method accurately identified areas of
wetland change 89.9 percent of the time;
There is substantial agreement between model
predictions of wetland change and on-the-
ground evidence for change.
Reliably discerned “no change” in wetland pixels
from “minor change,” and “minor” from “major”
change.
Change types in
ODA analysis
Wetland Change in the Chesapeake
Nielsen, E. M., S. D. Prince & G. T. Koeln (2008)
Upper
Potomac
Blackwater
NWR
Norfolk
Upper Potomac—few wetlands
> majority are palustrine (99.7%)
> 8% show major change
> 12% show minor change
> 80% show no change
Blackwater—many wetlands
> majority are palustrine (55%)
> 45% estuarine
> 4% show major change
> 35% show minor change
> 61% show no change
Norfolk—moderate wetlands, high change
> majority are palustrine (65%)
> 35% estuarine
> 4% show major change
> 14% show minor change
> 82% show no change
NWI Aggregation
Upper Potomac, shows wetland
change
> 20 HUC-11 watersheds
> 9.53 km^2 of wetlands
> 2 show major change
> 7 show minor change
> 11 show no change
Blackwater, minimal change
> 7 HUC-11 watersheds
> 219 km^2 of wetlands
> none show major change
> 1 shows minor change
> 6 show no change
Norfolk, minimal change
> 4 HUC-11 watersheds
> 74 km^2 of wetlands
> none show major change
> 1 shows minor change
No Change
Change
Minor Change
HUC-11 Watersheds
HUC-11
Aggregations
Upper Potomac
> Loudoun, VA and
Mongtomery, MD
> 45% of pixels show change
> 10% major, 35% minor
Blackwater
> Dorchester County, MD
> 14% of pixels show change
> All minor
Norfolk
> Cities of Norfolk, Portsmouth,
portions of Suffolk
> very few (<1%) wetland pixels
show change
No Change Minor Change Major Change
County
Aggregations
NWI Categories
and Wetland Area
HUC-11
Watersheds and
Counties
Physiographic
Provinces
Preliminary Results
Net Wetland Loss between 1992-2006 (Acres) Watershed (HUC-8) NLCD CCAP CBCLCD Pocomoke-W Lower Delmarva 204.19 -8.42 -6.55 Nanticoke 39.02 -5.48 -6.13 Tangier 46.71 -5.18 -5.48 Chincoteague 53.38 -5.07 -4.41 E Lower Delmarva 19.50 -2.79 -2.57 Broadkill-Smyrna 16.94 -2.31 -1.80 Choptank 33.40 -1.82 -1.59 Chester-Sassafras 48.34 -1.14 -0.90 Brandywine Christina 10.65 -1.03 -0.99
Wetland Loss between
1992-2006 (CCAP)
Wetland Loss (In Red)
between 1992-2006 (CCAP)
Decennial Census Total Housing Units
per Acre at HUC 8 Watershed Scale
Wetland Loss (In Red)
between 1992-2006 (CCAP)
Hot Spot and Point Density of Wetland Permits issued by the State of MD, MDE
2. Developing Technologies & Techniques
Landsat historic record (1984-Present)
Advanced geometric and radiometric correction
Knowledge from North American Forest Dynamics
Project: Vegetation Change Tracker (Co-I Huang)
0
2
4
6
8
10
12
85 87 89 91 93 95 97 99
Year (19xx)
IFZ
Masking and
normalization Masking and
normalization Masking and
normalization Masking and
normalization Masking and
normalization
LTSS
Masks
Time series
analysis
Disturbance
products
Indices
Yearly Information on Vegetation Change from 1984- Present
2.Developing Technologies &
Techniques
Wetness Change Tracker!
Monitoring hydroperiod in
response to weather and
land cover change
DEM based topographic
wetness indices
complement optical data
Implications for climate
change
Monitoring the Connection between Weather and Hydroperiod
with Landsat Based Wetness Trends through Time
Note that different trends in wetland gains are probably due to natural
versus human based drivers and differences amongst human based
restoration implementation practices.
2. Innovative Mapping Technique Mapping forested wetlands and their change dynamics in response to
variable weather conditions (Landsat based approach)
Spectral
Index Formula Reference
NDVI (Band 4 – Band 3) / (Band
4 + Band 3) Rouse et al., 1974
NDWI-
1
(Band 4 – Band 5) / (Band
4 + Band 5) Gao, 1996
NDWI-
2
(Band 3 – Band 5) / (Band
3 + Band 5)
Rogers &
Kearney, 2004
TCWG
D TCW - TCG In this study
TCA Arctan (TCG / TCB ) Powell et al., 2010
IVR Band 5 / Band 2 Ozesmi & Bauer,
2002
IR (Band 5 – Band 7) / (Band
5 + Band 7) Ruan et al., 2007
Data pre-processing
Mapping inundation
dynamics
Landsat corrected
by LEDAPS
Landsat surface
reflectance
LiDAR 1 m
inundation
map
LiDAR
intensity
data
2007 Landsat reflectance
& spectral indices
2007 LiDAR 30 m
inundation
percentage map
Landsat-based
inundation model
2005, 2007, 2009 &
2010 Landsat Images
Predicted inundation percentage maps in
2005, 2007, 2009 & 2010
Spatial
aggregation
Model calibration,
validation & post-
adjustment
Model development
(Stepwise Regression Tree)
Fall Spring
Highly variable wetland hydrology
Spectral mixture with vegetation and soil
2. Innovative Mapping Technique Mapping forested wetlands and their change dynamics in response to
variable weather conditions (Landsat based approach)
Feb. 2005 b. Mar. 2007 c.
Mar. 2007 d. Mar., 2009
(1) Results from reference image (2007, normal year)
(2) Results from other target years (2005, 2009, 2010)
Subpixel Inundation Percent (2009)
a. b.
c. SPOT imagery on Apr., 2005 d. SPOT imagery on Apr., 2010
2. Innovative Mapping Technique Mapping forested wetlands and their change dynamics in response to
variable weather conditions (Landsat based approach)
a. Feb. 2005 b. Mar. 2007 c. Mar. 2009 d. Mar., 2010
(A) Inundation percentage maps for the catchment
(B) Total inundated area vs. the Palmer
seasonal drought index
(C) Total inundated area vs. Stream
discharge data
2. Innovative Mapping Technique Topographic metrics for improved mapping of forested wetlands
(LiDAR-based approach)
(A) Study site
(B) The overall design of the methodology
(C)Topographic index vs. products and forested
wetlands (a) the local terrain normalized relief; (b) enhanced topographic
wetness index (ETWI); (c)the relief enhance topographic wetness
index; (d) LiDAR intensity during a dry spring (e) LiDAR intensity
during average spring ; (f) false color near infrared aerial photograph
(collected coincident to e)
3. Ecosystem Service Assessment
Wetland function: Integrated approach 1. Regional Assessment for the CBW: use of regional models and existing
dataset and literature to assess the ability of wetlands to improve water quality
- Currently, regional water quality models (SPARROW, Chesapeake Bay
Watershed Model) for the CBW do not simulate natural wetland processes.
- Wetland function is based on the removal efficiency and it only applies to a
specific type of wetlands (e.g., constructed wetlands).
- We are in the process of compiling literature and field data to evaluate the
range of nutrient removal efficiency of wetlands, key functional drivers
affecting the removal efficiencies, and their applicability for the CBW
2. Simulation based approach at the local watershed: Choptank Watershed
- We are evaluating the effectiveness and accuracy of watershed process
model (Soil and Water Assessment Tool, SWAT) to simulate natural wetland
processes and its water quality improvement benefits at the watershed
scale.
(A) Study site (B) Calibration and validation of SWAT model
Thank you!
For additional information please contact In-Young at [email protected]
Physical Indicators
Variables
• Climate Change Data (Acquired)
– Palmer Drought Severity Index
– Temperature
– Precipitation
– Source: NOAA
• Sea Level Rise (Acquired)
– Mean Eustatic Sea level
– Inundation frequency
– Sources: SLAMM & PSMSL
• Hydrogeomorphology (HGM)
– Elevation and soil type (Acquired)
– Sources: USGS and USDA RCS
Indices (To be calculated)
HGM/LLWW
Slope
Proximity (Distance)
Wetland Loss
Land Cover Type
Areas: Urban and Protected
Density
Edges
Anthropogenic Indicators
Variables Demographics
Population Growth Housing Density Sources: US Census, Brown
University, and CBP (Acquired)
Wetland Permitting Records Total quantity of Permits Density of permits (by acre)
National Regional Letter of Permission Specific
Sources (Acquired) Federal:
US Army Corps of Engineers
State: MD MDE, VIMS, DE DNREC
Indices (To be calculated)
• Proximity (Distance) – Wetland permits
• land cover types
• pixels and patches of wetland loss
• urban areas
• shoreline
• protected areas
• Proportions – Wetland permits by
wetland type
– By population
– By housing density