Scaling API-first – The story of a global engineering organization
Characterizing regional dikes with LiDAR using ArcGIS 10 by Grontmij
1.
2. 2
For those seeking presentations on #LiDAR analysis based on
point clouds; you will not find it in this presentation.
This presentation is about the analysis of raster data, derived
from LiDAR data, that has been created specifically for the
Water Boards in the Netherlands (bare earth, 0.5 m resolution).
3. 3
About two thirds of the Netherlands is vulnerable to flooding
1953 North Sea flood; 9% of total Dutch farmland flooded, 1835 people
killed, 30000 animals drowned, 47300 buildings damaged
Long history of reclamation of marshes and fenland, resulting in
some 3,000 polders
There are about 14.000 km of regional dikes in the Netherlands
Sources:
Wikipedia, Flood control in the Netherlands
Wikipedia, North Sea flood of 1953
Wikipedia, Polders and the Netherlands
Deltawerken online
Waternet, Flood control and protection
4. 4
Determination of dike top height for five-year legal assessment
Determination of dike strength on representative profiles
Mapping of primary and regional dikes and embankments
Doing that manually is a time consuming effort …
… the alternative is using GIS and LiDAR data
5. 5
(Up-to-date Height Model of the Netherlands)
AHN is funded by the Directorate General for
Public Works and Water Management of the
Dutch Government and the 26 Regional Water
Boards
The main purpose of AHN is to get a highly
detailed representation of the bare earth.
Objects (houses, trees, etc.) are filtered.
Available
Spring 2012
Source:
Spring 2013
AHN website, availability of AHN2 (Dutch)
6. 6
Managing more than 600 km of regional dike in the Northern
part of NL and about 66 km of sea dikes.
7. 7
Pre-process AHN2 data
Correct location of the dike based on AHN2
Divide line representing the dike in parts of 100m
Determine the lowest point per 100m (H100Min)
Create cross profile AHN2 on each H100Min
Analyze standard profiles
Determine maximum width Bmax
Determine the design profile
Iteration to correct profiles
8. 8
Necessary to filter outliers
Choose filter small to avoid puling the initial dike track away from
the water -> use a Focal mean 3x3
Fill up small areas with NoData
Unfiltered Filtered
9. 9
Step size interval on centerline (10m)
Maximum search tolerance (5m)
Evaluate several methods: Without step size and search tolerance
Search highest point in original AHN2
Search highest point in filtered AHN2
Determine highest point weighted by distance from
centerline and using a minimal increment (e.g., 2cm/m)
Correct original centerline to create a new centerline based
on the elevation data.
10. 10
Influence of rising “hinterland” at small dikes, pulling the line
away from the actual dike 1
Small outliers can create large errors 2
Resulting centerline is less gradually 3
1
2
3
11. 11
Influence of rising “hinterland” at small dikes remains 1
Small outliers are removed 2
Resulting centerline is a bit more gradually 3
1
2
3
12. 12
Influence of rising “hinterland” at small dikes is less apparent 1
Result looks better, but… creates new inconsistencies 2
2 1
13. 13
Alternative: determine the maximum deviation of two
consecutive points (exclude red point, see 1 )
Incorrect position of the centerline will have consequences
1
14. 14
The corrected centerline is divided into parts of 100m. When a
part is smaller than 50m it is added to the previous part.
Gather statistics per part
(min, max, mean, etc)
15. 15
For each part the location of the lowest point H100Min is
determined
16. 16
On each H100Min location a cross profile is drawn (100m)
Step size (precision) is pixel size of AHN2 (0.5m)
Store as XYZ and dZ (=distance vs. elevation) lines
17. 17
Distance d (horizontally) versus Z (vertically)
Z (* factor 5)
L R
Distance from H100min (d)
H100min
First NoData L and R from H100min
Classic example of a centerline
that doesn’t follow the highest
part of the dike and causes the
H100min to be located to low.
18. 18
4 standard trapezoid profiles are fitted at the highest position
centered underneath the H100min point
Exaggeration Z-axis is factor 5
4 standard trapezoid profiles defined by widths 1m and 2m versus slopes 1:1.5 and 1:2
19. 19
The highest trapezoid profile is selected. When more than 1
profile have the highest position, the widest is chosen.
If more than 1 option remains, the one with the slightest
inclination is chosen.
Exaggeration Z-axis is factor 5
20. 20
Try to extent the best standard trapezoid profile by
determining the maximum width Bmax at that height.
Exaggeration Z-axis is factor 5
21. 21
Over 6000 locations have been processed
In about 1300 locations the results were different than expected,
due to errors in the location of the original dike line
In the second part of the project, code was developed to correct
these situations
Dike is here
Dike is NOT here
22. 22
Does not have to be centered at H100min
Should not cross NoData (probably water) After
Should limit search to 5 meter L and R iterations
H100min
Last point
Last point before NoData
before NoData on right side
on left side
Standard trapezoid
L1 before iteration
23. 23
All 4 standard trapezoids should be evaluated
Best option is selected
24. 24
Bmax is determined for new location of best trapezoid
H100min Max search tolerance 5m
Bmax of best
standard
trapezoid
Best standard trapezoid
after iteration
25. 25
Z
3D view of a location: X
Y
First result
before correction
Second result
after correction
with better position
26. 26
Water Board Noorderzijlvest is using these results to fill their
database, fieldwork would have required many months more.
Relatively large datasets can be used for this analysis (a raster
dataset of 40Gb was used, but can be much bigger too)
Process is flexible and can be used for any Water Board
Generation of centerline should be done implementing the
techniques developed for iteration of the standard trapezoids
Centerline should have correct position or the results will not
be reliable
ArcGIS 10.1 beta 2 was used for this analysis, but ArcGIS 10.0 or 9.3.x can be used as well.