The Oregon LiDAR Hydrography Pilot Project Evaluation of Existing GIS Hydrological Toolsets for Modeling Stream Networks with LiDAR and Updating the National Hydrography Dataset (NHD) Bill Kaiser (USFS), Craig Ducey (BLM), and Dan Wickwire (BLM) August 9, 2010
27
Embed
Evaluation of Existing GIS Hydrological Toolsets for ...€¦ · The Oregon LiDAR Hydrography Pilot Project Evaluation of Existing GIS Hydrological Toolsets for Modeling Stream Networks
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The Oregon LiDAR Hydrography Pilot Project
Evaluation of Existing GIS Hydrological Toolsets for Modeling Stream
Networks with LiDAR and Updating the National Hydrography Dataset (NHD)
Bill Kaiser (USFS), Craig Ducey (BLM), and Dan Wickwire (BLM)
Evaluation and Issues .................................................................................................................................. 18
LiDAR-based Stewardship Process ........................................................................................................ 21
OVERALL CONCLUSIONS AND RECOMMENDATIONS ........................................................................ 23
LITERATURE CITED .......................................................................................................................................... 25
Discontinuities in the modeled stream networks occurred wherever sinks were not
filled (Figure 5). Inconsistencies in sink locations between the rescaled surfaces led to high
variability in flow accumulation estimates and, consequently, contrasting representations
of the hydrological stream network. If addressed with high-quality ancillary road
information, not filling all sinks may be useful for identifying potential culvert locations
within the study area. Known culvert locations could be serve as break-points to either
modify the existing surface models or inform conversion of the original LiDAR point cloud
to new DEMs.
14
Summary and Future Recommendations
No single combination of cell size, sink depth, or flow accumulation threshold
produced completely satisfactory results. Spurious stream segments branching off main
channels occurring at each cell size suggested flow accumulation thresholds were not
constant throughout the study area. Stream channel predictions appeared qualitatively
better or worse in different regions of the study area depending on which combination of
parameters was used. Attempts to rectify issues with flow lines generated after all sinks
were filled were frustrated by the large number of streams captured by road ditches and
complex channel definitions in areas experiencing active soil movement. Increasing cell
size alleviated many of these situations, but resulted in only slight improvements in stream
geometry predictions over the existing NHD and well below the potential accuracy and
precision of the original LiDAR dataset.
Stream features modeled using the ESRI ArcGIS D8 hydrology tools do not account
for the numerous biophysical and human processes influencing the hydrology at Panther
Creek. Consequently, the model results are accurate if the landscape is viewed as an
impervious surface devoid of vegetation and without human modification. Future
modeling efforts including ancillary information predictive of the spatial patterns resulting
from these types of processes combined with spatial metrics descriptive of topographic
conditions are likely to extend beyond the sophistication of the ESRI toolset. Trials
conducted during this pilot exercise suggest adjusting the measurement scale while
maintaining the cell size using alternative image processing techniques such as Gaussian
convolution have the potential to more effectively mitigate localized variance in the LiDAR
dataset. Future research should also consider more realistic models of flow direction such
as the D-Infinity approach and the influence of topographic conditions beyond a focal cell’s
eight immediate neighbors (Tarboton, 2009)
Panther Creek represents a unique opportunity to explore the use of LiDAR for
stream channel delineation. Presently, five LiDAR acquisitions spanning the past four years
are available for conducting comparative analysis within the study area. A multi-partner
LiDAR research consortium with participants from the BLM, EPA, NRCS, non-government
15
groups, and others are investigating the value of LiDAR for a wide-range of environmental
science applications. These efforts are being informed by extensive field-based information
including weather monitoring, vegetation and stream surveys, soil classifications, etc.
Incorporating these data in future work identifying stream channels with LiDAR will allow
more detailed evaluations of the accuracy and precision of results, the effects of alternative
image pre-processing techniques, differences between multiple LiDAR acquisitions, as well
as help reduce subjectivity during parameterization (e.g., stream initiation flow
accumulation thresholds).
16
INTEGRATION OF LiDAR DERIVED STREAM NETWORKS INTO NHD
Introduction
PNWHF partners are currently updating the NHD using a variety of data sources.
The partners have implemented agreed upon stewardship processes that accommodate
these various data sources. The partnership anticipates that the percentage of NHD
updates derived from LiDAR elevation data will increase through time. This project
evaluated methodologies for updating the NHD with LiDAR derived stream networks
within the context of established stewardship agreements and procedures.
Generating a stream network from the LiDAR DEM is only the first step in
integrating these data in to the NHD. Processes that incorporate existing NHD edit tools
and stewardship relationships needed to be developed to support these LiDAR based edits.
Secondly, labor costs associated with these types of edits needed to be determined to assist
PNWHF partner organizations in their planning of future workloads. NHD updates
resulting from the two pilot areas have been used to develop cost estimates for labor
associated with these updates.
The project team incorporated the use of the NHD GeoEdit toolset for this
evaluation. This toolset has been developed by the U.S. Geological Survey (USGS) to
support NHD updates although its effectiveness on large edits with significant changes in
stream geometry has been under question. The LiDAR-based delineations for Panther
Creek and Ashland have these characteristics so the project team evaluated the use of the
NHD GeoEdit toolset to support these NHD edits. The Ashland Creek HU is the pilot area
used in the NHD integration discussion as it went through the entire update and review
process required by the NHD and the PNWHF protocols.
17
Figure 6. Perennial (dark blue) and intermittent (light blue) stream channels derived using the D8
hydrology model within the Ashland Creek 12-digit hydrologic unit..
18
Evaluation and Issues
The LiDAR hydrologic algorithms produce candidate stream lines on the basis of
flow direction that is derived from the elevation and slope grids. They do not take into
account important natural and anthropogenic features that can affect the flow of water.
Performing an NHD update that is based on these LiDAR sources relies on a variety of
ancillary data including current imagery, existing NHD, Digital Raster Graphics (DRG), etc.
This information is vital in resolving ambiguous data issues resulting from use of the LiDAR
delineations. New information not reflected on the DRG or USGS topographic map, such as
dam spillways, road and ditch intersections can assist the editor. The most recent NAIP
(National Agriculture Imagery Program) infrared imagery in Oregon is an excellent
resource that is available to resolve problems.
The issue of stream periodicity, e.g. perennial vs. intermittent, generated many
discussions by the project team. Some members wanted to incorporate other data such as
precipitation, soils, or stream gage data to help in this determination. In the end the USFS
and BLM hydrologists were adamant that GIS editors should not determine the periodicity.
Instead, the USFS hydrologists provided flow accumulation thresholds (thirty acres for
perennial and fifteen acres for intermittent streams) in the Ashland Creek area for defining
periodicity. On the other hand, the BLM hydrologist requested points along the stream to
symbolize different flow accumulation thresholds to help determine inception points and
periodicity during his review.
In general, there is a need to preserve NHD reach codes1 because the reach codes
may have important information tied to them (e.g., information related to Clean Water Act,
sections 303(d) and 305(b)). Even though the states of Washington and Oregon have not
tied their Clean Water Act reporting requirements to the NHD in the past, the
Environmental Protection Agency (EPA) has integrated these reports with the NHD reach
codes in their national database as well as their public web sites. It should be noted that
1 See the EPA web site http://www.epa.gov/waters/about/geography.html for a good explanation of the reach code and how impaired waters (as defined by the Clean Water Act) are tied to the specific reach codes.
State of Washington – Department of Ecology is in the process of migrating its Clean Water
Act data to the NHD.
It is critical to preserve these reach codes wherever possible, and this is the
rationale behind the instructions given to the editors for preserving reach codes whenever
a stream possessed a valid GNIS name or was depicted on the medium resolution
(1:100,000 scale) NHD. The medium resolution NHD dataset is currently housed in a
separate database from the high resolution NHD.
The NHD Geoeditor utilizes a Task Assistant to control the edit workflow while
revising the NHD. To preserve the existing GNIS names and reach codes, the editor chooses
REPLACE NHD FLOWLINE under NHD FLOWLINE TOOLS. For those streams that do not
have a valid GNIS name or do not appear in the medium resolution NHD, it is recommended
to use the ADD NHD FLOWLINE IMPORT NHD FLOWLINE GEOMETRY. Figure 9 depicts
the named streams in the original NHD, which represents 39% of the streams.
Using LiDAR to derive waterbodies such as lakes or two dimensional rivers was
outside the scope of this project. However, we did use the new NAIP imagery to update the
shape of one reservoir in Ashland Creek. The existing NHD geometry was incorrect at
Reeder Reservoir and artificial paths had to be moved. The recommendation to preserve
the artificial paths in NHD waterbodies is important because aquatic habitat or water
quality information may be tied to existing artificial paths and preserving these paths
minimizes event migration costs.
20
Figure 7. Ashland Creek Watershed – the light blue lines represent streams with GNIS names and
make up approximately 39% of the total stream length in the original NHD dataset. These would be
edited with REPLACE NHD FLOWLINE tool from the Task Assistant.
One useful way of thinking about the differences in the stream data derived from
LiDAR vs. the NHD is to compare the stream length divided by the area of the watershed. In
the Ashland Creek watershed, for example, this ratio is 80.93 km/15,786 acres for the
original NHD and 140.24 km/15,786 acres for the LiDAR-based stream network.2 Other
statistical information such as perennial and intermittent stream length from before and
after the LiDAR Ashland Creek update is given in Table 3.
2 The reason for mixing units is the original stream length in the NHDFlowline attribute table is in kilometers and the Hydrologic Units are traditionally defined in acres.
21
Table 3. Stream length information from before and after the LiDAR update for Ashland
Creek.
Ashland Creek NHD Original LiDAR Percentage change
Total Stream Length
(km)
80.93 140.24 +73 %
Acres 15786 15786
Perennial streams
(km)
72.34 100.50 + 39%
Intermittent streams
(km)
8.59 39.74 + 462 %
GNIS streams (km) 31.56 32.13 + 2%
LiDAR-based Stewardship Process
BLM and the Forest Service hydrologists at the districts and forest levels act as data
stewards for NHD streams and lakes data. Consequently, they must play an integral part in
any process that updates the surface waters on their agency’s lands. The workflow
proposed by this pilot project includes at least two reviews by these data stewards (Figure
8). In addition, the LiDAR update process for the NHD has to go through the existing
stewardship review process of the PNWHFP partnership.
22
Figure 8. Flowchart illustrating the interactions between the spatial analysts, GIS editor, and NHD
data steward leading to the submission of LiDAR-based streams to the NHD.
1. Spatial analyst
delineates stream network
from LiDAR, generates
supporting data layers
(e.g., hillshade and
contours), and provides
required metadata
documentation
2. GIS editor reviews
stream network, records
comments and initial edits,
and prepares data layers
for NHD data steward.
3. NHD data steward
reviews GIS editor’s
comments and initial edits,
annotates areas of
concern, and makes
recommendations.
4. GIS editor modifies
stream network based on
NHD data steward’s
review. Interaction
between GIS editor and
NHD data steward
repeated, as necessary.
5. After NHD data
steward’s approval, GIS
editor finalizes metadata
documents before
preparing NHD update.
6. Following NHD
stewardship procedures,
any affected parties are
notified of pending update,
and LiDAR-based streams
are submitted to the USGS.
23
Labor Cost Estimates
The cost estimate to update the NHD is presented in hours rather than dollars as
each agency’s personnel or contractor costs may vary. It is important to note that the NHD
integration and update was performed by a highly skilled NHD GIS editor. The total
editing hours required to update the NHD in Ashland Creek were 30 hours/15,786 acres. If
we extrapolate this time to the average size of a 10 digit Hydrologic Unit (formerly 5th Field
HUC) in Oregon which is approximately 117,650 acres, then it would take the editor
approximately 220 hours/117,650 acres to accomplish this work. However, some
efficiencies of scale may be gained by working in an entire 10-digit HU. On the other hand,
in an agricultural area that contains many lakes and rivers that require artificial paths, the
editing time could increase. Therefore, it is recommended that a complete 10-digit HU be
updated in Phase 2 of the project to obtain more realistic metrics of the time involved in
updating the NHD with LiDAR-derived data.
OVERALL CONCLUSIONS AND RECOMMENDATIONS
Automatically deriving acceptable LiDAR-based stream networks using traditional
spatial analyst tools proved to be difficult and time consuming. The project team made
contact with several other agencies using these tools and these projects all reported similar
problems. Many of these agencies abandoned an automated approach and performed
head-up digitizing from a LiDAR-based shaded relief. These difficulties forced the project
team to investigate a toolset based on the principles of terrain analysis, TauDEM. This
toolset proved to be effective in overcoming most, but not all problems encountered with
the out-of-the-box spatial analyst tools. Therefore, it is recommended that future efforts to
delineate stream networks using LiDAR-based DEMs in areas further evaluate the TauDEM
tool-set. Spatial analyst tools using the D8 flow direction model may generate an
acceptable network, but requires interactive editing before a stream network meeting NHD
requirements is generated.
24
Integrating the LiDAR stream network requires the use of ancillary data such as the
NAIP imagery, DRGs, and the original NHD. Special emphasis is placed on preserving
streams with existing GNIS names even though this could make the editing process more
expensive. The time required to integrate these stream data into the NHD appears to be
high enough to justify a cost-benefit analysis before embarking on an update of one or
more 10-digit HUs. Stewardship review and communication with other PNWHF partners
are critical components for a successful NHD update based upon LiDAR.
Since only two small 12-digit HUs on the west side of the Cascades were included in
the first phase of this pilot, the team recommends updating a complete 10-digit HU to
better calibrate the costs and technical processes. In addition, the team recommends
picking a HU with a greater percentage of private ownership so that an institutional
process can be defined to review the new hydrographic data that is derived from LiDAR.
Not all areas or jurisdictions have on-the-ground experts available to review updates like
the Forest Service, BLM, and the Oregon Department of Forestry. Possible HUs should
contain flatter, agricultural areas or a greater percentage of suburban development.
Another consideration for future work is that the HU contain larger number of lakes and
wide streams with artificial paths to not only better estimate the time involved editing, but
to help clarify the cost of any event migration required from a LiDAR update of the NHD.
And finally, this project should be placed into the context of national efforts that
utilize LiDAR to update the NHD. The national NHD steering committee is interested in
defining procedures and estimating costs to integrate LiDAR-derived stream networks into
the NHD. It is imperative that the PNWHF communicate their results to this group as well
as being able to learn from other initiatives that are taking place around the nation.
25
LITERATURE CITED
ESRI. 2008. ArcGIS 9.3. Environmental Systems Research Institue, Inc. (ESRI), Redlands, CA, USA. Garbrecht, J. and Martz, L.W. 2000. Digital elevation model issues in water resources modeling. In: Maidment, D. and Djokic, D., eds. Hydrologic and Hydraulic Modeling Support with Geographic Information Systems. Redlands, California: Environmental Systems Research Institute, Inc.: 1-27. Lefsky, M.A., Cohen, W.B., Parker, G.G., and Harding, D.J. 2002. Lidar remote sensing for ecosystem studies. BioScience. 52: 19-30. Lewis, D. 2006. Seeing landslides with LiDAR. Cascadia. 4: 1,3,8. Osborn, K., List, J., Gesch, D., Crowe, J., Merrill, G., Constance, E., Mauck, J., Lund, C., Caruso, V., and Kosovich, J. 2001. National Digital Elevation Program (NDEP). In: Maune, D.F., ed. Digital Elevation Model Technologies and Applications: The DEM Users Manual. Bethesda, Maryland: American Society for Photogrammetry and Remote Sensing: 83-120. Saunders, W. 2000. Preparation of DEMs for use in environmental modeling analysis. In: Maidment, D. and Djokic, D., eds. Hydrologic and Hydraulic Modeling Support with Geographic Information Systems. Redlands, California: Environmental Systems Research Institute, Inc.: 30-51. Tarboton, D.G., Bras, R.L., and Rodriguez-Iturbe, I. 1991. On the extraction of channel networks from digital elevation data. Hydrological Processes. 5:81-100. Tarboton, D.G. 2009. Terrain Analysis Using Digital Elevation Models (TauDEM) 4.0. http://hydrology.neng.usu.edu/taudem/
Passalacqua, P., Do Trung, T., Foufoula-Georgiou, E., Sapiro, G., and Dietrich, W.E. 2010. A geometric framework for channel network extraction from LiDAR: Nonlinear diffusion and geodesic paths. Journal of Geophysical Research 115:
F01002, doi:10.1029/2009JF001254 Poppenga, S.K., Worstell, B.B., Stoker, J.M., and Greenlee, S.K. 2009. Comparison of surface flow features from LiDAR-derived digital elevation models with historical elevation and hydrography data for Minnehaha County, South Dakota: U.S. Geological Survey Scientific Investigations Report 2009–5065, 24 p. Tarboton, D.G. 1997. A new method for the determination of flow directions and upslope areas in grid digital elevation models. Water Resources Research 33: 309-319. Tveite, H. 1999. An accuracy assessment method for geographical line data sets based on buffering. International Journal of Geographical Information Science 13: 27-47. Wilson J.P. and Gallant J.C. 2000. Digital terrain analysis. In: Wilson J.P., Gallant J.C., eds. Terrain analysis: principles and applications. New York: John Wiley and Sons 1-27.
27
ACRONYMS
BLM – Bureau of Land Management
DEM – Digital Elevation Model
GIS – Geographic Information System
HU – Hydrologic Unit
LiDAR - Light Detection and Ranging
NAIP - National Agriculture Imagery Program
NED – National Elevation Dataset
NHD – National Hydrography Dataset
ODF – Oregon Department of Forestry
OWRD – Oregon Water Resources Department
PNWHF – Pacific Northwest Hydrography Framework
TauDEM – Terrain Analysis Using Digital Elevation Models