39
Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina Valerie Garcia Forestry Department, North Carolina State University Office of Research and Development, Environmental Protection Agency August 4, 2004

Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

  • Upload
    clove

  • View
    43

  • Download
    0

Embed Size (px)

DESCRIPTION

Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina. Valerie Garcia Forestry Department, North Carolina State University Office of Research and Development, Environmental Protection Agency August 4, 2004. Why is the research important? - PowerPoint PPT Presentation

Citation preview

Page 1: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North

Carolina

Valerie GarciaForestry Department, North Carolina State University

Office of Research and Development, Environmental Protection AgencyAugust 4, 2004

Page 2: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Why is the research important?

A large percentage of non-point source pollution is suspected to occur through headwaters streams

Current available USGS 1:24,000 Topographic maps lack accuracy in depicting the presence and location of headwaters streams

The lack of accurate maps for headwaters streams places an inordinate burden on both the regulatory agencies and the regulated communities in planning and implementing policy

Page 3: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Why is the research important?

Recent availability of LIDAR data for portions of North Carolina provide new opportunities for developing more accurate stream maps

This study focuses on the mapping of headwaters stream networks using Geographical Information System (GIS) approaches and LIDAR data

Study results limited to the Piedmont Ecoregion of North Carolina

Page 4: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Layout of Study

Phase I: Extensive literature search to investigate state-of-science Geographical Information System (GIS) approaches relevant for using LIDAR data to map headwaters streams

Phase II: Compare and evaluate approaches identified through the literature search by applying the techniques to a study site in the Piedmont Ecoregion of North Carolina

Page 5: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Phase I: Literature Search

Page 6: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Some Definitions…

Triangulated Irregular Network (TIN)

Digital Elevation Model (DEM)

Hydro-enforcement

Page 7: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Triangulated Irregular Network (TIN):

Formulation of non-overlapping triangles from irregularly spaced x, y, and z points (vector-based)

Page 8: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Digital Elevation Model (DEM):

As used in this study…uniformly spaced, 3-dimensional cartographic representation (x, y, z) in a grid or raster format

Page 9: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Both TINs and DEMs result in artifacts—artificial disruptions of the natural drainage of water

Incorporation of known stream center-lines (breaklines) into the modeled terrain ensure the downstream drainage of water

Hydro-Enforcement

Page 10: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Summary of Literature Search Findings fell into two major categories

Production of accurate topographic maps Extraction of stream networks

TINs produce more precise topographic maps maintains LIDAR elevation points as triangle vertices better retains linear structures (breaklines)

DEMs are better for automatically extracting headwaters streams can automatically correct drainage problems and

determine stream origin

Page 11: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Summary of Literature Search

The interpolation method and the resolution used to generate the topographic map can impact the accuracy of the map more complex interpolation methods (e.g., Spline,

Kriging) require more knowledge and are computationally demanding, but are expected to perform better in modeling terrain

resolution drives computation demands of the interpolation method and can impact the selection of which interpolation method can be used

Page 12: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Summary of Literature Search

Hydro-enforcement of the DEM enhances the accuracy of the extracted stream networks derived from orthophotos (can be expensive) typically not available for headwaters streams

Physical processes relevant at headwaters stream scales are different than watershed scales typically modeled stream origin and hillslope (diffuse) flow are critical at

headwaters stream scales

Page 13: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Phase II: Application of Stream Mapping Techniques to Study Site in

Falls Lake, North Carolina

Page 14: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Research Questions

What is the accuracy of the LIDAR surface elevation points?

When using densely spaced LIDAR data, does the interpolation method used to create a DEM make a difference in the accuracy of the DEM?

At what resolution do you begin to lose channel definition of headwaters streams, thereby affecting the production of headwaters stream maps?

Does using breakline data to hydro-enforce TINs and DEMs make a difference in the production of headwaters stream maps?

Do more complex stream flow algorithms and stream origin approaches produce better results than simpler methods in mapping headwaters streams? [not covered in this presentation]

Page 15: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Method Steps

Step 1: Evaluate the accuracy of LIDAR surface elevation points

Step 2: Compare the accuracy of topographic maps generated using various interpolation methods

Step 3: Evaluate the effect of scaling on topographic and stream map accuracy

Step 4: Assess the impact of using hydro-enforcement to extract stream networks

Page 16: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Study Site

Forested headwater stream catchment near Falls Lake in the Piedmont Ecoregion of North Carolina was selected as the study site Collected surveyed elevation points and mapping grade

Global Positioning System (GPS) along headwater stream

All modeling techniques were run for the study site and compared against survey data and available ancillary data

Page 17: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

GIS Applications Used in Study

ArcGIS used to generate topographic maps

Four interpolation methods compared: Natural Neighbor, Inverse Distance Weighting (IDW), Spline and Kriging

Four resolutions compared: 10 ft., 20 ft., 60 ft., 90 ft.

ArcHydro used to extract stream networks

Networks extracted with and without hydro-enforcement

Networks extracted at four resolutions: 10 ft., 20 ft., 60 ft., 90 ft.

Page 18: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Data Used in Study

Existing data used in modeling

LIDAR mass elevation points (NC State Floodplain Mapping Program, 2003)

Breakline data (stream centerlines and shorelines) (NC State Floodplain Mapping Program, 2003)

Data used for comparisons

Field-collected data (survey, mapping-grade GPS) Medium and high resolution National Hydrography Dataset (NHD) (2000) USGS 1:24,000 Topographic Digital Raster Graph (DRG) (1994) Wake County Hydrography Lines (2000) derived from 1:12,000 aerial

photography 1999 Wake County Color Digital Orthophotography

Page 19: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 1: Evaluation of LIDAR Surface Elevation Points

Collection of high-accuracy ground truth data

Control Benchmarks established along ridge of study catchments (vertical +/- 2cm)

Transect surveyed across study catchment (vertical +/- 8cm)

GPS measurements taken along headwater stream (horizontal 1-2m)

Compared LIDAR data to field-collected survey points

TIN generated from LIDAR data (without breaklines) and used for comparison with survey data

Page 20: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 1: Evaluation of LIDAR Surface Elevation Points

RMSE was 1.32’ or 40.1 cm (no points removed) Recalculated using the “95 percentile” methodology with adjusted RMSE of 28.7 cm Published accuracy of LIDAR data is 25 cm (95 percentile) -- but study did not limit

survey to uniform slope and used only one landcover type

300.0

310.0

320.0

330.0

340.0

350.0

360.0

104.4 9.7 23.4 36.6 28.0 28.8 28.0 24.8 30.3 27.4 35.8 26.5 15.3 4.8 4.0 51.3 2.0 1.9 2.1 33.9 41.7 34.4 25.0 21.7

C ontr ol P oi nts Sur vey P oints T IN-no br eaks

Page 21: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 2: Comparison of Topographic Maps

Generated DEMs from LIDAR data using four different interpolation techniques and compared to ground-truth data 20 ft. resolution used as LIDAR data averaged at least

one point every 20 sq. ft.

Four interpolation methods compared (Natural Neighbor, IDW, Spline and Kriging)

Results compared to survey data

Page 22: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

300.0

310.0

320.0

330.0

340.0

350.0

360.0

104.4 9.7 23.4 36.6 28.0 28.8 28.0 24.8 30.3 27.4 35.8 26.5 15.3 4.8 4.0 51.3 2.0 1.9 2.1 33.9 41.7 34.4 25.0 21.7

C ontr ol P oi nts

Sur vey P oints IDW Kr ig

Step 2: Comparison of Topographic Maps

Method vs. Survey Points

RMSE

LIDAR 1.3'

Spline 1.4'

Natural Neighbor 1.6'

IDW 2.2'

Kriging 2.3'

IDW(brick), Kriging (majenta) vs. Survey Points (blue)

Page 23: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

300.0

310.0

320.0

330.0

340.0

350.0

360.0

104.4 9.7 23.4 36.6 28.0 28.8 28.0 24.8 30.3 27.4 35.8 26.5 15.3 4.8 4.0 51.3 2.0 1.9 2.1 33.9 41.7 34.4 25.0 21.7

C ontr ol P oi nts

Sur vey P oints Spl ine Natur al Neighbor

Method vs. Survey Points

RMSE

LIDAR 1.3'

Spline 1.4'

Natural Neighbor 1.6'

IDW 2.2'

Kriging 2.3'

Spline (red), Natural Neighbor (green) vs. Survey Points (blue)

Step 2: Comparison of Topographic Maps

Page 24: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

300.0

310.0

320.0

330.0

340.0

350.0

360.0

104.4 9.7 23.4 36.6 28.0 28.8 28.0 24.8 30.3 27.4 35.8 26.5 15.3 4.8 4.0 51.3 2.0 1.9 2.1 33.9 41.7 34.4 25.0 21.7

C ontr ol P oi nts

Sur vey P oints IDW Spl ine Natur al Neighbor Kr ig

Method vs. Survey Points

RMSE

LIDAR 1.3'

Spline 1.4'

Natural Neighbor 1.6'

IDW 2.2'

Kriging 2.3'

Step 2: Comparison of Topographic Maps

Page 25: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 3: Evaluate Impact of Scaling to Various Resolutions

Evaluate impact of scaling on accuracy of topographic map

Natural Neighbor, IDW, and Regularized Spline used to generate 10 ft., 20 ft., 60 ft. and 90 ft. resolution DEMs

Results compared to survey data

Page 26: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 3: Evaluate Impact of Scaling to Various Resolutions

Evaluate impact of scaling on stream extraction

ArcHydro used to extract stream networks at various resolutions (10 ft., 20 ft., 60 ft. 90 ft.)

Extracted stream networks compared to each other, GPS points and Wake County hydrography lines

Page 27: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 3: Evaluate Impact of Scaling to Various Resolutions - DEM

Comparison of DEMs at different resolutions Very little difference exists between the 10 ft. and 20 ft. resolution DEMs

300.0

310.0

320.0

330.0

340.0

350.0

360.0

104.4 9.7 23.4 36.6 28.0 28.8 28.0 24.8 30.3 27.4 35.8 26.5 15.3 4.8 4.0 51.3 2.0 1.9 2.1 33.9 41.7 34.4 25.0 21.7

Sur vey P oi nts

Sur vey P oints Spl ine - 10' Spl ine - 20'

Page 28: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 3: Evaluate Impact of Scaling to Various Resolutions – Stream Extraction

Comparison of DEMs at different resolutions At 60 ft. resolution, 2 ft. headwater stream channel becomes a 120 ft. depression At 90 ft. resolution, the entire drainage is lost

300.0

310.0

320.0

330.0

340.0

350.0

360.0

104.4 9.7 23.4 36.6 28.0 28.8 28.0 24.8 30.3 27.4 35.8 26.5 15.3 4.8 4.0 51.3 2.0 1.9 2.1 33.9 41.7 34.4 25.0 21.7

Sur vey P oi nts

Sur vey P oints Spl ine - 60' Spl ine - 90'

Page 29: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Impact of Scaling to Various Resolutions

Comparison of drainages extracted at different resolutions

Similar to the topographic map results, very little difference exists between the extracted stream drainages generated from the 10 ft. and 20 ft. resolution DEMs

Page 30: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Impact of Scaling to Various Resolutions

Comparison of drainages extracted at different resolutions

At 60 ft. resolution, drainage lines become much more unnaturally linear and have more occurrences of drainage interruptions

At 90 ft. resolution, this problem is more extreme

Page 31: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 4: Assess the effect of Hydro-enforcement on the Extraction of Stream Networks

Accuracy of source breakline data Overlain on NHD, Wake County hydrography lines, Wake County

digital orthophotography and USGS 1:24,000 Topographic DRG Evaluate the impact of hydro-enforcement on extracting headwaters

stream networks ArcHydro used to extract stream networks for each DEM, with and

without hydro-enforcement Extracted networks compared to stream GPS points, NHD, Wake

County hydrography lines, Wake County digital orthophotography and USGS 1:24,000 Topographic DRG

Page 32: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

NC Floodplain Program breakline data aligns well with the 1999 Wake County digital orthophotography

Step 4: Impact of Hydro-enforcement – Accuracy of Breakline Data

Page 33: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 4: Impact of Hydro-enforcement – Accuracy of Breakline Data

NHD (high resolution (blue) and medium resolution (red)) aligns well with USGS 1:24,000 Topographic DRG

Page 34: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 4: Impact of Hydro-enforcement – Accuracy of Breakline Data

NHD (high resolution) does NOT align with 1999 orthophotography (beige polygons emphasize alignment problems)

Page 35: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 4: Impact of Hydro-Enforcement

Most of the differences were in the higher-order streams

The headwaters stream network was substantially the same regardless of whether the grid was hydro-enforced or not

Hydro-enforced Natural Neighbor DEM produced the cleanest drainage lines in the breakline areas

Drainage created from hydro-enforced TIN (converted to grid) produces a confused drainage in the lake area

Page 36: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Step 4: Impact of Hydro-Enforcement

“Best” drainage generated from Natural Neighbor DEM with hydro-enforcement (blue) overlain on Wake County hydrological lines

(green)

Page 37: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Conclusions

Simpler interpolation methods (e.g., Nearest Neighbor, IDW) did as well or better than the more complex interpolation methods (e.g., Kriging) for generating the base DEMs that are used for extracting headwaters stream networks

Hydro-enforcement did not improve the results in extracting the headwaters stream networks

Hydro-enforcement did generate more direct drainages in the lake area, indicating that flatter areas or any area prone to flooding, will be aided by breakline data

Page 38: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

Conclusions

Breakline data available through the NC State Floodplain Program are better aligned than currently available NHD or USGS 1:24,000 Topographic Map

Because misalignments are carried throughout the stream network, only the highest quality breakline data should be used for hydro-enforcement; alternatively, no hydro-enforcement should be used

Page 39: Automated Techniques to Map Headwaters Stream Networks in the Piedmont Ecoregion of North Carolina

“Zooming in” shows that the model does better than the Wake County hydrography lines as compared to the GPS stream points

Bottomline…

For the study catchment—LIDAR data and GIS modeling approaches did a better job than the best stream data currently available for the area