i USING PYTHON AND XML FOR FLOOD PLAIN DELINEATION MODELING AND DYNAMIC INUNDATION ANALYSIS OF THE MISSOURI RIVER VALLEY IN HOLT COUNTY, MISSOURI A RESEARCH PAPER PRESENTED TO THE DEPARTMENT OF HUMANITIES AND SOCIAL SCIENCES IN CANDIDACY FOR THE DEGREE OF MASTER OF SCIENCE By JEFFREY K. HERZER NORTHWEST MISSOURI STATE UNIVERSITY MARYVILLE, MISSOURI November 1, 2016
103
Embed
USING PYTHON AND XML FOR FLOOD PLAIN DELINEATION …...This paper describes how ArcGIS and Python scripting were used to create a simple floodplain delineation model and to drive analysis
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
i
USING PYTHON AND XML FOR FLOOD PLAIN DELINEATION MODELING AND DYNAMIC
INUNDATION ANALYSIS OF THE MISSOURI RIVER VALLEY IN
HOLT COUNTY, MISSOURI
A RESEARCH PAPER PRESENTED TO
THE DEPARTMENT OF HUMANITIES AND SOCIAL SCIENCES
IN CANDIDACY FOR THE DEGREE OF
MASTER OF SCIENCE
By
JEFFREY K. HERZER
NORTHWEST MISSOURI STATE UNIVERSITY
MARYVILLE, MISSOURI
November 1, 2016
ii
Abstract
This paper describes how ArcGIS and Python scripting were used to create a simple
floodplain delineation model and to drive analysis on the model with real-time river gage
data retrieved from the USGS XML data feed. Model output is assessed against photographs
of the study area during severe flooding in 2011. Floodplain delineation models compare
water elevations against land elevations; flooded areas exist where water surface elevation
values exceed those of the terrain surface. The model's terrain surface was created from a
LiDAR-derived 1m Digital Elevation Map, while the water surface was created by drawing
an overlay of polygons representing elevation "offset zones”, with values based on the
average drop in terrain elevation upstream and downstream from the Rulo, Nebraska river
gage. Offset values were assigned to the terrain surface polygons within each offset zone.
The output of the model is a "Potential for Flooding" (PFlood) map -- called
"potential" because the model does not calculate "flood cell discontinuity", those areas that
flood or remain dry based on which levees remain intact. Model output provides highly
detailed assessments of which levees may remain above the water surface and the
inundation depth of those below the water surface at a given river gage value.
iii
Table of Contents
I. Introduction ................................................................................................................................ 1
II. Objective ....................................................................................................................................... 2
III. Study Area ................................................................................................................................... 2
IV. Research Background ............................................................................................................. 6
a. Floodplain Modeling................................................................................................ 6
b. River Gage and Flow Data .................................................................................. 10
1. XML ......................................................................................................................... 10
Figure 29: Elevation Values Selected for “Immersion Scenarios” ................................................... 36
Figure 30: Floodplain Immersion Scenarios with Elevation Offsets .............................................. 37
Figure 31: Floodplain Immersion Scenarios Without Elevation Offsets ...................................... 38
Figure 32: Floodplain Immersion Scenarios with Modified Elevation Offsets........................... 40
Figure 33: PFlood Projection vs. Landsat Image from 17 July 2011 (Source: USGS) .............. 41
Figure 34: Study Area Scenario Locations w/ Nearby Levee Breach Sites (insets) ................. 43
Figure 35: Mill Creek and Union Township Levee Districts (Source: USACE) ........................... 44
Figure 36: SCENARIO A, Map and Aerial View ....................................................................................... 47
Figure 37: SCENARIO A, Vicinity of Big Lake and US 159 between Rulo, NE and Fortescue,
MO .................................................................................................................................................................. 48
Figure 38: SCENARIO A, Big Lake pre-flood, view to the north 16 JUN 2011 (Max Stage:
Figure 10: River gage web map with dynamic content from NWIS data
Figure 11: Dynamic map, PHP back-end code
16
2. MassDOT Traffic Data Feed
The second proof-of-concept was a study of the Massachusetts Department of
Transportation’s (MassDOT) XML traffic data internet feed and porting the live data to
ArcGIS to support “real time” display shown in Figure 12. The data feed, seen in Figure 13,
describes location specific “events” such as construction or maintenance, with each
identified spatially using lat-long coordinates. XML data is arranged in a “tree” hierarchy
and the data of interest is four levels down from the “root”. The right margin shows the
“addresses” needed to access data from tree “branches”. The ElementTree XML Python API
(Python Software Foundation, n.d.), uses these to navigate through the tree hierarchy, read
parameters in the stream, save them to variables and write values into database fields
using a cursor. The processing algorithm described in Figure 14 describes how the XML
data was handled once it was parsed.
. Figure 12: Data display from the (MassDOT) XML traffic data feed
17
Figure 13: Massachusetts DOT Road Condition XML Feed
18
Figure 14: MassDOT XML Display, workflow algorithm
V. Data Sources
1m Bare Earth, LiDAR-derived Digital Elevation Model: Missouri Spatial Data Information Center http://www.msdis.missouri.edu/data/lidar/County_LiDAR/index.html USGS National Water Information System (NWIS) XML data feed: http://waterdata.usgs.gov/nwis
Set the workspace path name using the 'raw' string
Set the target feature class
Set the target field class path
Allow output results to be overwritten
Get user input to set the maximum number of event records to
be returned
Designate the URL to query and retrieve data from the XML
data source
Open the XML data URL
Parse the XML data
Load the parsed XML data into a dictionary object
Count the total number of event records
Count the total number of event parameters in the XML schema
Validation: if returnEvents > eventCount:Display Error
Message
List fields to be accessed
Initialize the cursor
Initialize an iterative counter LOOP: while i <
returnEvents: write retrieved data into variables
Read variables into an array for database entry
Increment the counter to read the next event
Insert returned records into the destination table
19
VI. Development
The utility is comprised of four component parts (Figure 15): 1) the Geodatabase; 2)
the Floodplain Delineation Model; 3) the XML Data Parser, and; 4) the Analysis Tool. Each
component is executed through a Python scripted tool in a Floodplain Delineation Toolbox.
a. The Geodatabase
The geodatabase initially contains two source files and three database tables. The
three source files are: a LiDAR-derived DEM (raster); a hand-drawn polygon shapefile
defining the area to be clipped from the DEM, and; a hand-drawn Water Elevation Offset
Zone polygon shapefile. The two tables hold data on river gage sites (gageSites) and
statistics (gageStatistics). Once the river gage site registered in the database, time stamped
data on gage height and discharge volume can be read from the XML feed and stored in the
gageStatistics table.
b. Floodplain Delineation Model
1. Modeling the Ground Surface
To meet the project goals, the ground surface layer had to be highly detailed, easily
generated and not overly cumbersome to process, and the system of levees had to be
represented with precision. A Triangulated Irregular Network (TIN) surface did not
provide sufficient detail over such a wide area, while the 1D and 2D approaches were
beyond the scope of the researcher’s ability. Where Choung’s (2014) task was to extract
levee data and generate polygons representing levee structures, the task in this project was
simply to ensure linear flood control structures of various sizes were identifiable. Two
methods were evaluated for this task, one using Slope Analysis and the other using the
Slope Analysis (Figure 16) produced excellent linear detail and the least extraneous
(non-linear) data in the range between 21 and 29 percent. The drawback is that Slope
Analysis does not produce discreet areas with individual elevation values needed to
compare against the water surface. The output of Raster to Polygon conversion (Figure 17)
is a highly detailed, continuous coverage of polygons representing discrete elevation zones
in one-meter increments. However, this method had one major drawback: the conversion
produced more than 3-million polygons, a fatal threat to the goals of easy data generation
and short processing times. Visually rendering the polygon data took at least two minutes
per redraw and simply opening the attribute table took more than 5 minutes. Operations
involving this data frequently crashed the computer, a Dell Precision T3610 64-bit
workstation with an Intel® Xeon® E5-1620 v2 @ 3.7GHz processor and 16 GB of RAM
using Windows 7. The issue was easily remedied by deleting polygons with shape areas (as
indicated in the Shape_Area field of the data table) below a minimum threshold. In one
instance, deleting polygons with shape areas of less than 50 square meters reduced the
total number of polygons from 3,164,578 to only 60,373 (Figure 18). The time needed to
draw polygons across the entire coverage was reduced to around four seconds. At small
scales, elevation polygons draw more quickly. Eliminating small polygons resulted in a
large number of no-data “holes” which could be considered as minute transitions in
elevation or areas of uncertainty in floodwater coverage. Raster to Polygon conversion was
the appropriate choice as it provided a highly detailed representative ground surface and
met the project objectives of fast analyses at both wide area and highly localized scales.
22
Figure 16: Ground surface model from slope analysis
23
Figure 17: Output of raster to polygon conversion
24
Figure 18: Raster to Polygon coverage - Number of Polygons vs. Polygon Minimum Area
2. Modeling the Water Surface
Water surface elevations are easily characterized across short distances with a flat
plane having a single elevation value. The problem becomes more complex across larger
areas, where surges caused by rainfall or changes in discharge from dams will cause
“ripples” on the plane. The floodplain delineation map in Error! Reference source not
found.Figure 19, developed from an Esri tutorial on Floodplain Delineation from LiDAR
Points (Esri, n.d.) has one-dimensional elevation cross sections or "surface profiles" with
multiple surfaces projected from center stream elevations. A process based on this
approach was used to develop a water profile in this project.
Among the factors considered: the points where the Missouri River meets Holt
County’s northern and southern boundaries are 38.8 air miles (62.4 km) apart but cover 56
25
linear river miles (90 km); between these points, the elevation drops from ~883 feet/269
meters to ~826 feet/252 meters above sea level; the rate of elevation drop is not constant
along the 250 river miles between Omaha, Nebraska and Kansas City, Missouri as seen in
Figure 20, though the total drop of 242 feet neatly averages to nearly 1 foot per mile (0.97),
and; the initial water surface created for use in analysis was based on a one-foot-per-mile
(or 1 meter per three mile) elevation drop rate for simplicity.
To recreate this elevation drop rate, rectangular polygon “strips” approximately 3
miles wide, each representing a 1-meter elevation offset, were arbitrarily drawn over the
area into a new feature class, using river mile markers as references (Figure 21). Positive
offset values were assigned to upstream zones and negative values were assigned to
downstream zones. The elevation offset values were assigned to ground surface polygons
within each zone using overlay analysis. A field named to hold calculated “PFlood” values
was added to the table to hold values for the calculated flood potential of each polygon.
Figure 19: Floodplain delineation map with projected water surface
26
Figure 20: Elevation Drops Between River Gages, Omaha to Kansas City
27
Figure 21: Water Elevation Offset Zones, Polygon Feature Class
a. The XML Feed
Figures 22, 23 and 24 show the XML data “tree” returned from the URL noted in
Figure 6, which requests data formatted in WaterML1.1, for site ID 06813500 (Rulo,
Nebraska) and figures for parameters 00060 (stream flow in cfs) and 00065 (river gage in
28
feet). Data is returned for each of the three data “branches” that extend from the root level;
they are referred to and represented in Python code as root[0], root[1] and root[2]. The
data we need may be contained within a tag, in an attribute within a tag or in the text
between opening and closing tags. The line of XML at the top of Figure 25 shows data
returned for the node at address root[1][0][1]; the following lines show how tag, attribute
and text within the node are retrieved. In the data tree figures, columns on the left margin
indicate the hierarchy of “addresses” used to retrieve data from branches or “nodes” of the
tree. For example, the site name “Missouri River at Rulo, NE” is located at address
root[1][0][0]. Some data is retrieved as deeply as five levels down from the root, as in the
case of site latitude at address root[1][0][3][0][0].
Figure 22: XML data tree from USGS river gage feed (Part 1)
29
Figure 23: XML data tree from USGS river gage feed (Part 2)
30
Figure 24: XML data tree from USGS river gage feed (Part 3)
31
Figure 25: Accessing parameters within XML nodes
b. Analysis
With the floodplain model and XML feed in place, the potential for flooding (PFlood)
in each individual ground surface polygon was calculated based on river gage values --
called “potential” because our model does not account for which flood control structures
protecting an area may or may not be intact. Figure 26 shows the calculations used to
determine PFlood:
Figure 26a: Water Surface Elevation asl at Zero Offset (WSEØ) was determined by
adding the river gage figure in feet to the datum elevation at Rulo, Nebraska (838.16
ft.), then converting the sum to meters.
Figure 26b: Zonal Water Surface Elevation for each zone (WSEz) is the sum of the
zero offset elevation (WSEØ) and a zone’s Elevation Offset value (ZOff).
Figure 26c: Flood Potential (PFlood) for individual polygons was determined by
comparing the polygon’s elevation (ElevPoly) to the Zonal Water Surface Elevations
(WSEz). The result is either a positive (not flooded) or negative (amount of
inundation) value.
32
Figure 26d: The formula used in Python code. The result is multiplied by -1 so that
negative values indicate elevations inundated by water and positive values indicate
elevations above water.
VII. Python Code
Five Python scripts comprising a “Floodplain Delineation” toolbox were developed
to execute the process. Figure 27 shows how individual tools relate to each other in a
systemic context. A schema-only layer package was developed so consistent symbolization
could be applied to model output.
a. Enter Gage Stations
USGS gage recording stations must be registered in the utility database before they
can be queried and their data used. The user must enter the gage’s 8-digit Site ID number,
its datum elevation and floodstage values in feet. The tool validates user input, queries the
USGS feed to confirm the ID is active, confirms data is available, retrieves fields and enters
them into the database.
Figure 26: PFlood (potential for flood) formulas
a. WSEØ = (Datum elevation + river gage) * 0.308
b. WSEz = WSEØ + ZOff
c. PFlood = ElevPoly – WSEz
d. PFlood = (WSEØ + [ZOff] - [ElevPoly]) * -1
33
b. Get XML Feed
With a Site ID as input, confirms the station is registered in the database; accesses
data by constructing and submitting a query URL; downloads and parses the XML data
stream to retrieve/process river gage, discharge and timestamp statistics; enters the data
into the database.
c. Build Floodplain Model
From three input files – a Digital Elevation Map (DEM) raster, clipping polygon and
“offset zones” source file -- produces a polygon elevation layer representing both land
elevations and adjusted water elevations with offsets to account for the general
downstream elevation drop. The average run time for this script on the development
workstation was 40-45 minutes.
d. Pflood from XML
Queries the database and retrieves the most recent recorded gage height in feet for
Rulo, Nebraska, Site ID 06813500; generates a floodplain scenario based on the returned
value.
e. PFlood Generate Scenario
Generates a floodplain scenario for Rulo, Nebraska based on a river gage height in
feet entered by the user.
A consistent color symbolization schema can be applied using a Layer Package
stored in the geodatabase and unpacked in ArcCatalog (Figure 28). As scenario iterations
are generated from either the “PFlood Generate Scenario” or “PFlood from XML” tools,
symbolization is updated when the floodplainModel feature class display is refreshed.
34
Figure 27: Workflows in the Floodplain Delineation Toolbox
Figure 28: The floodplainColors schema-only layer package file (.lpk)
35
VIII. Model Output
a. Symbolization
Symbolization is done with negative numbers representing inundation and positive
numbers representing land above the projected water surface. All values above ~2 meters
are classified as dry. Areas between 0 and 2 meters are considered as being at immediate
risk of inundation, with 0-1 meter areas symbolized in red and 1-2 meter areas symbolized
in yellow. Inundation is symbolized in shades of blue, with darker blues representing
greater inundation depths. In the example figures, inundation was symbolized in 3-meter
increments.
b. Offset and No-Offset
Floodplain delineation maps were generated for the four immersion scenarios listed
in Figure 29: the record crest set in 2011; a level one foot over Rulo’s 17-foot flood stage; a
typical annual low gage value, and; a random historic low water mark.
To analyze the utility of elevation offsets, PFlood values in the first map sequence
(Figure 30) were calculated using offsets while values in second map sequence (Figure 31)
were not. On a wide scale (Macroanalysis), both sequences show how the potential for
flooding spreads from south to north as river gages increase. The maps indicate areas in
the extreme south would be at constant risk of overflow were it not for the levees that line
the river banks. This would be consistent with the general nature of the Missouri River and
its morphology as a controlled waterway. Flood control structures were built to keep the
river confined to its channel and to restrain its natural tendency for frequent overflow.
From here, the map sets begin to diverge. The Offset maps tend to be more “forgiving” on
the upper end, showing a more gradual south-to-north overflow at higher river gage
36
heights, but greater flood depths in lower terrain elevations at the southern end of the
extent. The No-Offset maps show not only a faster spread of flood potential, but potential
immersion over a far greater area at higher river gage heights. The exception is in the very
low water scenario, where the No-Offset map shows significantly less flood potential. The
breaklines between zone offset areas are also more apparent.
Figure 29: Elevation Values Selected for “Immersion Scenarios”
Gage (ft) Elev (ft) Elev (m) Notes
1.46 839.6 255.9 Random recorded low water level 12/19/1989
10 848.2 258.5 Typical annual low water value
18 856.2 261.0 One foot above floodstage level of 17 feet
27.26 865.4 263.8 Record high level 6/27/2011
37
Figure 30: Floodplain Immersion Scenarios with Elevation Offsets
38
Figure 31: Floodplain Immersion Scenarios Without Elevation Offsets
39
c. Modified Offset
Offset and No-Offset values were similar in that the terrain elevation drop was
evenly distributed -- by design in the first case, and by being ignored in the second. To
further investigate these differences, a third “Modified Offset” sequence map was produced
(Figure 32), using values more consistent with the per-mile average elevation drops
between river gages, diagrammed in Figure 20. The average 0.6 foot-per-mile drop
between the Brownville, Nebraska and Rulo gages is the most gradual while the 1.71 per-
mile average between Rulo and St. Joseph, Missouri is the steepest. With 3-mile wide offset
ones, values in increments of 1.8 feet (0.55m) were assigned to Zones 1 through 5, with
values in increments of 5.13 feet (1.56m) assigned to Zones -1 through -4.
The Modified Offset maps would seem to do a better job at the lowest river stage
scenario, with practically no threat of immersion anywhere, the outline of Big Lake left
visible at center left, and low elevations adjacent to the Squaw Creek National Wildlife
Refuge visible east of Big Lake. A comparison of a PFlood projection with actual conditions
on January 17, 2011 (Figure 33) shows immersion at the northern end of the extent, unlike
the initial offset map. Again, while these inundation potential maps are not meant to be
hydrologically accurate, it would seem the Modified Offset values do a better job of creating
projections, therefore this is the data set we will use in Microanalysis. The feature class
Zones_Modified has been hardcoded into the Build Floodplain Model script.
40
Figure 32: Floodplain Immersion Scenarios with Modified Elevation Offsets
41
Figure 33: PFlood Projection vs. Landsat Image from 17 July 2011 (Source: USGS)
42
IX. Analysis Scenarios
Scenarios covering four locations within the Study Area (Figure 34) were created
from the modified elevation offsets using river gages in Figure 29. As note on the maps,
river gages are referred to in feet while PFlood figures are measured in meters. In each
three-panel simulation, panel a. represents a river gage of 10 feet, panel b. represents 18
feet (1 foot over floodstage), panel c. represents 27.26 feet. Model output is compared
against conditions observed and photographed in 2011 to look for similarities and to
determine what kinds of conclusions might reasonably be drawn from the flood delineation
model.
Scenarios A and B are within a 17,366-acre leveed area owned by the Union
Township Levee District (Figure 35, bottom) described by the State of Missouri as “a
mainline levee and the first line of defense for much of northwestern Holt County” (Office
of Missouri Governor Jay Nixon, 2012). A levee at Lower Cotter Bend, just downstream
from the mouth of the Tarkio River at River Mile 507 in the northwestern corner of the
district was breached by high water in 2010 and the gap of 1100 feet/335 meters was not
repaired until 2012 (U.S. Army Corps of Engineers, 2012). Scenarios C and D occurred as
the result of a levee breach at the mouth of Mill Creek at approximately River Mile 515.5,
within the Corning Conservation Area. The breach was at the southern edge and just
outside of the federally constructed and non-federally operated L-536-550 Turkey Crk LB,
Rock Crk LB, Mo Riv LB, & Mill Crk RB Levee System (Mill Crk RB Levee System, Figure 35,
top) as described by the U.S. Army Corps of Engineers (2014).
43
Figure 34: Study Area Scenario Locations w/ Nearby Levee Breach Sites (insets)
44
Figure 35: Mill Creek and Union Township Levee Districts (Source: USACE)
45
a. SCENARIO A: Rulo/Big Lake
This area is immediately east of Rulo, Nebraska along US Highway 159 and includes
all of Big Lake. Missouri Highway 111 runs north from US 159 and along the eastern edge
of Big Lake (Figure 36, 37). Figure 37a shows how low the general area is and why flood
protection is required. Much of the area is within 2 meters of flooding at a low river gage. A
description of flood impacts from the National Weather Service’s Advanced Hydrologic
Prediction Service (National Weather Service, 2015) notes levees protecting Big Lake begin
to overtop at a river gage of 26 feet, and; at 27.26 feet, the record historic crest reached in
2011, “significant flooding will encompass a very large area”. In Figure 37c, the entire area
is inundated with nearly all flood control structures under water. Figure 38 shows how Big
Lake is dry at a river gage of 24.12. Figure 39 shows conditions five days later, after the
water has risen and weakened levees failed, the area is flooded at river gage 26.51.
The 2011 Flood overtopped Big Lake, flooded homes along its banks (Figure 38, 39)
and washed out the Burlington Northern-Santa Fe (BNSF) railroad tracks running adjacent
to US 159 (Figure 40, 41, 42). In Figure 40c, the flood pattern along a curve of US 159 west
of Big Lake closely greatly resembles that observed on 21 June (Figure 41) with the river
gage at 26.51; as in the PFlood map, the railroad tracks along the northern edge of the
curve are above water at the west end while the roadway running parallel to the south is
immersed.
Figure 42 confirms deep immersion at the south end of Big Lake as seen in Figure
43, 44, and 45. By the time the photos were taken in October, the destroyed railroad tracks
had been replaced. US 159, the roadway furthest south, has been washed out and destroyed
46
by a large scour hole. High water marks of approximately 2 meters on buildings in Figure
45 compare with the 3-meter immersion depth projected by the model.
47
Figure 36: SCENARIO A, Map and Aerial View
48
Figure 37: SCENARIO A, Vicinity of Big Lake and US 159 between Rulo, NE and
Fortescue, MO
49
Figure 38: SCENARIO A, Big Lake pre-flood, view to the north 16 JUN 2011 (Max Stage:
24.12)
Figure 39: SCENARIO A, Big Lake inundated, view to the north 21 JUN 2011 (Max Stage
26.51)
50
Figure 40: SCENARIO A, US 159 between Rulo and Big Lake
51
Figure 41: SCENARIO A, water over US 159, view to the west 21 JUN 2011 (Max Stage
26.51)
52
Figure 42: SCENARIO A, roadway and rail bed, south shore of Big Lake
53
Figure 43: SCENARIO A, roadway and rail bed, south shore of Big Lake, view to the
north 14 OCT 2011 (Max Stage 14.19)
Figure 44: SCENARIO A, US 159 between Rulo and Big Lake, post-flood, view to the west 14 OCT 2011 (Max Stage 14.19)
54
Figure 45: SCENARIO A, Google Streets images from 2013 show high water marks
55
b. SCENARIO B: Missouri Highway 111/118 (Fred Guthrie Site)
Located approximately two miles north of Big Lake at the junction of Missouri
Highways 111 and 118 and approximately 5 miles/8 km southwest of the levee breach at
River Mile 507. Missouri State Highway Patrol Trooper Fred Guthrie and his K-9 Reed
were swept away by fast currents where Missouri 111 runs west from the junction (see In
Memoriam on the last page). The intersection is circled in Figure 46a and is seen while
submerged in Figure 47 and post-flood in Figure 48. The huge scour hole that formed at
this location, seen in the Figure 48 photo, is also visible in Figure 49a.
The degree of flooding at the Guthrie site was greatly similar to levels seen in Figure
49 . Figure 49b clearly shows Missouri 111 trailing off into high water as seen in Figure 50,
a photo taken hours after Trooper Guthrie and Reed had disappeared. Water depth was
enough to allow motorized, shallow draft boats to operate freely (Figure 51). At the
111/118 intersection, Figure 52 shows roadways to be up to 2 meters above water level, in
the same locations seen in the model. Figure 53 and Figure 54, covering the area just north
of the Guthrie site, show the potential for flooding caused by breaks in agricultural levees.
Areas south and east of the levee break in the Union Township system are covered in
Figure 55; the levee break occurred at the upper left corner of the mapped area.
56
Figure 46: SCENARIO B, Map and Aerial View; Fred Guthrie Site circled
57
Figure 47: SCENARIO B, Looking south towards Big Lake from the Guthrie site 01 AUG
2011(Max Stage 24.03)
Figure 48: SCENARIO B, Scour hole at the Guthrie site (r. center) and sand deposition
along Highway 111 (l. center), view south towards Big Lake 14 OCT 2011 (Max Stage 14.19)
58
Figure 49: SCENARIO B, Fred Guthrie Site at Junction of Highway 111 and 118
59
Figure 50: SCENARIO B, Looking west towards the Guthrie site with water over
Highway 111 01 AUG 2011 (Max Stage 24.03)
Figure 51: SCENARIO B, Water is deep enough for shallow draft boats to operate freely
around the Guthrie site 01 AUG 2011 (Max Stage 24.03)
60
Figure 52: SCENARIO B, Level of immersion at Jct Missouri 111/118 on the morning of
7 AUG 2011, view to the south (top) and to the west (center and bottom)
61
Figure 53: SCENARIO B, potential for flooding caused by breaks in agricultural levees
62
Figure 54: SCENARIO B, The flood’s areal coverage grows as agricultural levees are
breached inland and individual flood cells are filled 21 JUN 2011 (Max Stage 26.51)
63
Figure 55: SCENARIO B, areas east and south of the Union Township levee breach
64
c. SCENARIO C: Tarkio River Southwest of Corning, Missouri
The Tarkio River, or the “Big Tarkio”, flows onto the flood plain from the hills north
and east of Corning, Missouri (Figure 56) and is channelized with high levees on either side
to where it meets the Missouri River southwest of Craig, Mo. These levees not only protect
a large area including the town of Craig from Missouri River overflows, they also keep the
Tarkio River from spilling onto the flood plain, On July 7, 2011, the Tarkio River
contributed to an already disastrous situation in Holt County after rising more than 15
feet/4.5 meters in 5 hours and spilling through a levee breach created two weeks earlier
(Norvell, 2011).
Figure 57 shows the Tarkio River levee remaining largely above water even in the
worst case scenario (Figure 57c). Photos in Figure 58 confirm the levees as formidable
structures and the size of the channel between them. In high water conditions, the river
itself is clearly defined between the levees in Figure 59 with the levees themselves
remaining well above the potential flood level. This is confirmed in Figure 60, a photo taken
on 21 June when the river stage was 26.51 feet. In our map developed from thousands of
polygons, it is possible to examine elevation differences at very high scale. Figure 61 shows
minute elevation differences on the levee that might be potential locations for overtopping
at higher water levels. We can even estimate the size of the indicated gap at 12.5 meters.
65
Figure 56: SCENARIO C, Map and Aerial View (Google Maps, Google Earth)
66
Figure 57: SCENARIO C, The leveed Tarkio River
67
Error! Reference source not found.
Figure 58: SCENARIO C, Tarkio River, view to the north, as overflow begins to seep through the levees 21 JUN 2011 (Max Stage 26.51) and post-flood 14 OCT 2011 (Max Stage 14.19)
68
Figure 59: SCENARIO C, The leveed Tarkio River and railroad bridge at County Road 125
69
Figure 60: SCENARIO C, Tarkio River and railroad bridge at County Road 125 21 JUN 2011 (Max Stage 26.51)
70
Figure 61: SCENARIO C, Minute elevation differences seen at very high map scales
71
d. SCENARIO D: Corning, Missouri and Corning Conservation Area
The area shown in Figure 62 ncludes: the site of the Mill Creek levee breach on the
Missouri River, visible as a sand deposition plume at the upper left of the diagrams in
Figure 63; the town of Corning, Mo, two miles east, and; the Interstate 29 interchange and
mainline.
The extreme flood scenario in Figure 63c shows a water depth of 0-3m in the town
of Corning. This would seem to be confirmed by photos in Figure 64 and Figure 65 which
show flood conditions and post-flood high water marks on buildings. The conditions
projected at the Corning Interchange and along the I-29 mainline in Figure 66 match
conditions documented in Figure 67 and Figure 68.
Figure 69 clearly indicates the I-29 mainline remains well above the flood waters. It
compares well to the actual flood levels in Figure 70, Figure 71, and Figure 72. In addition,
documented flood conditions at the Corning Interchange (Figure 70) closely match an
extremely high scale projection of the area in Figure 71 down to the very small patch of
land remaining above water.
Error! Reference source not found.
Figure 62: SCENARIO D, Map and Aerial View (Google Maps, Google Earth)
72
Figure 63: SCENARIO D, Missouri River Levee Breach to Corning, MO and Interstate 29
73
Figure 64: SCENARIO D, Flooding in Corning, MO 23 JUN 2011 (Max Stage 26.96)
74
Error!
Reference source not found.
75
Figure 65: SCENARIO D, Corning, MO, high water marks of around 1 meter 11 JAN 2012
Figure 66: SCENARIO D, Corning, MO.
76
Figure 67: SCENARIO D, Looking southwest from the I-29 Corning interchange to the
town of Corning and the Missouri River 8 JUL 2011 (Max Stage 25.81)
Figure 68: SCENARIO D, High water failed to top the Interstate 29 mainline south of the
Corning interchange, view to the north 23 JUN 2011 (Max Stage 26.96)
77
Figure 69: SCENARIO D, Interstate 29 and the Corning Interchange
78
Figure 70: SCENARIO D, I-29 Corning Interchange, view to the east 23 JUN 2011 (Max
Stage 26.96)
Figure 71: SCENARIO D, Corning Interchange flood map, extremely high scale; island of
land noted by arrow, compare to feature in Figure 70.
79
Figure 72: SCENARIO D, Green vegetation marks areas that stayed dry along I-29 just
south of the Corning interchange, looking south 14 OCT 2011 (Max Stage 14.19)
80
X. Conclusions and Further Development
The major objectives of this project were successfully met: a simple floodplain
delineation model was created with Python script; data tables were populated with and
analysis was driven by “live” XML river gage data; the model quickly and efficiently
returned precise results, viewable from low to very high scales, across a wide area.
Avenues for further development of this application include:
1) Use on more sophisticated flood models: The model used in this study is based on
elevations alone and does not purport to be hydrologically accurate. Industrial-grade
flood models can be studied to determine where and how live data or other XML inputs
could be used.
2) Flood area/flood cell discontinuity: Our method for determining a flood’s areal
coverage considers only comparative elevations between terrain and a projected water
surface, and does not consider containment within flood control structures. A method
for analyzing flood cell contiguity would be useful, though the model has proven useful
enough without this capability.
3) As part of the research for this project, some linear structures were extracted manually
at individual elevations for demarcation from the rest of the polygons. The process was
deemed too time consuming for this effort. Steinfeld, et al. (2013) studied three semi-
automated GIS and traditional visual interpretation techniques for detecting
earthworks. The color symbolization used in this project to emphasize polygons within
2 meters of inundation is a workaround for this issue.
81
4) Temporal applications: Flood events are comprised of incidents that occur geospatially
and have consequences over time. For example, the length of time a flood control
structure has been exposed to standing water may signal a loss of structural integrity
and impending breaches; the amount of time land is inundated may affect soil health
and plant life recovery. Timestamped entries in the database support temporal
analysis.
5) Mobile data gathering technology: There are number of field data collection
applications such as Collector for ArcGIS, which run on tablet devices such as
iPhones/iPads. One such application, MotionX-GPS® by Fullpower Technologies, Inc.,
was evaluated as part of this study as a way of providing location-specific data. Motion-
X records tracks and waypoints as .GPX (GPs eXchange) files, an open-source XML
format described in the ArcGIS help file as the standard for saving the results from a
GPS receiver (Herzer, 2012). The track in Figure 73 was recorded during the search for
Trooper Fred Guthrie and “Reed” as the aircraft orbited the main search area. GPX files
can either be parsed with Python or imported directly into ArcGIS using the Conversion
> From GPS > GPX to Features tool.
6) Changes to source data: The U.S. Geological Survey has outlined changes to USGS water
data offered online, scheduled to be implemented state-by-state over a 7-month period
that began in late summer 2016 (U.S. Geological Survey, 2016). Some nationwide
changes were rolled out in July 2016. These changes may impact users with advanced
applications that use water data and bookmarked URLs. The list includes replacing data
descriptors with time-series identifiers, changes in URLs for time-series graphs, in tab-
82
separated (RDB) output and in some less common time-series parameter codes, e.g. for
stream velocity.
Figure 73: Flight track constructed from GPX files
* * * * *
83
In Memoriam: Trooper Frederick F. Guthrie and K-9 Reed
Missouri State Highway Patrol
On August 1, 2011, Trooper Frederick F. Guthrie Jr., and his Patrol K-9 Reed were assigned to Missouri River flood duty. They were working in the area of Big Lake on Missouri Highway 118 at Missouri Highway 111 in Holt County, Missouri, when they were apparently swept away by swift flood water.
At approximately 6:25 p.m. on Tuesday August 2, 2011, K-9 Reed was located in swift moving flood water approximately 100 yards from where Trooper Guthrie's patrol truck and boat were located.
The search for Trooper Guthrie continued for months. On January 12, 2012, his body was recovered under approximately three-and-a-half feet of packed sand and silt, after a brush pile was removed during an excavation process. The recovery site was south of where K-9 Reed was found months earlier, near the original search site.
Trooper Guthrie, 46, was the 30th member of the Missouri State Highway Patrol to make the ultimate sacrifice while serving and protecting the citizens of Missouri. K-9 Reed was a five-year veteran with the Patrol.
This study is dedicated to their memory. Source: Missouri State Highway Patrol
84
Acknowledgements
For my wife Jeannine, who has supported and endured my every career move and
bright idea utterly without complaint.
With thanks to my friend and colleague Sgt. Kevin G. Haywood Sr., Missouri State
Highway Patrol and to my Highway Patrol family.
Ground truth photographs were taken by the author working as a member of the
Missouri State Highway Patrol Communications Division during and flying in a Highway
Patrol aircraft piloted by Sgt. Kevin G. Haywood, Sr. Reconnaissance flights to look for levee
breaches and flooding between Iatan, Missouri and the Iowa state line began in early June
2011. Information was shared with the Missouri State Emergency Management Agency
(SEMA) and local law enforcement agencies. The author also participated in the months
long search for Highway Patrol Trooper Fred F. Guthrie and his K-9 “Reed”, who were
swept away by flood water north of Big Lake on August 1, 2011.
85
APPENDICE
90
91
95
99
103
106
107
References
British Hydrological Society, 2014. The Science of Hydrology. [Online] Available at: http://www.hydrology.org.uk/science_of_hydrology.php [Accessed 2 May 2016].
Choung, Y., 2014. Mapping Levees Using LiDAR Data and Multispectral Orthoimages in the Nakdong River Basins, South Korea. Remote Sensing, Volume 6, pp. 8696-8717.
Esri, 2008. XML Schema of the Geodatabase. [Online] Available at: http://downloads.esri.com/support/whitepapers/ao_/XML_Schema_of_Geodatabase.pdf [Accessed 7 May 2016].
Esri, 2011. Arc Hydro Tools 2.0 Tutorial. [Online] Available at: http://downloads.esri.com/archydro/archydro/Tutorial/Doc/Arc%20Hydro%20Tools%202.0%20-%20Tutorial.pdf [Accessed 1 April 2016].
Esri, n.d. Floodplain delineation from lidar points. [Online] Available at: http://desktop.arcgis.com/en/arcmap/10.3/manage-data/las-dataset/floodplain-modeling-using-lidar-in-arcgis.htm [Accessed 25 April 2016].
Herzer, J., 2012. Georeferencing on a Budget: Using iPhone and Freeware Apps to Geotag a Mountain of Image Files. [Online] Available at: http://www.directionsmag.com/entry/georeferencing-on-a-budget-using-iphone-and-freeware-apps-to-geotag-a-/270266 [Accessed 12 July 2016].
Horritt, M. & Bates, P., 2001. Predicting floodplain inundation: raster-based modelling versus the finite-element approach. HYDROLOGICAL PROCESSES, Volume 15, p. 825–842.
Kahn, B., 2011. Missouri River Flood Drama Likely Took Direction from La Niña. [Online] Available at: http://www.climatewatch.noaa.gov/article/2011/missouri-river-flood-drama-likely-took-direction-from-la-nina [Accessed 13 February 2012].
KISTERS North America, Inc., 2016. WaterML2: A Global Standard for Hydrological Time Series. [Online] Available at: http://www.waterml2.org/ [Accessed 30 June 2016].
Larson, L., 1996. The Great USA Flood of 1993. [Online] Available at: http://www.nwrfc.noaa.gov/floods/papers/oh_2/great.htm [Accessed 31 October 2016].
108
Mahoney, M., 2011. Holt County Missouri Officials Accuse Corps of “Devastation by Design”, with 2011 Flooding. [Online] Available at: https://20poundsofheadlines.wordpress.com/2011/08/15/10799/ [Accessed 2 February 2013].
National Weather Service, 2012. Service Assessment Report, The Missouri/Souris River Floods of May-August 2011 [pdf].. [Online] Available at: http://www.nws.noaa.gov/os/assessments/pdfs/Missouri_floods11.pdf [Accessed 30 January 2013].
National Weather Service, 2015. Advanced Hydrologic Prediction Service. [Online] Available at: http://water.weather.gov/ahps2/hydrograph.php?gage=ruln1&wfo=oax [Accessed 7 August 2016].
Norvell, K., 2011. Big Tarkio up 15 feet in 5 hours. [Online] Available at: http://www.newspressnow.com/news/article_871a7069-fba8-5e0a-aceb-103ae7af1246.html [Accessed 3 October 2016].
Office of Missouri Governor Jay Nixon, 2012. Gov. Nixon announces more than $3.3 million to assist seven levee districts along the Missouri River to rebuild. [Online] Available at: https://governor.mo.gov/news/archive/gov-nixon-announces-more-33-million-assist-seven-levee-districts-along-missouri-river [Accessed 26 September 2016].
Python Software Foundation, n.d. The ElementTree XML API. [Online] Available at: https://docs.python.org/2/library/xml.etree.elementtree.html [Accessed 11 December 2015].
Steinfeld, C. et al., 2013. Semi-automated GIS techniques for detecting floodplain earthworks. Hydrological Processes, 27(4), pp. 579-591.
Swift, B., 2014. Comparison and Utilization of 1D, 2D, and 3D Hydraulic Models on a Complex Diversion Structure. [Online] Available at: http://www.floods.org/Files/Conf2015_ppts/G7_Swift.pdf [Accessed 2 May 2016].
U.S. Army Corps of Engineers, Hydrologic Engineering Center, n.d. HEC-GeoRAS. [Online] Available at: http://www.hec.usace.army.mil/software/hec-georas/ [Accessed 3 May 2016].
U.S. Army Corps of Engineers, 2012. 2011 Missouri River Levee Rehab Update. [Online] Available at: http://www.nwk.usace.army.mil/Portals/29/docs/emergencymanagement/leveerehab/100412/Union_Township_Template.pdf [Accessed 27 September 2016].
U.S. Army Corps of Engineers, 2014. L-536-550 PI Executive Summary. [Online] Available at: http://nld.usace.army.mil/egis/NLDE_PROD.media_api.download?p_media_id=472900114
109
7&p_submission_id=2396 [Accessed 29 September 2016].
U.S. Army Corps of Engineers, n.d. Missouri River Recovery Program. [Online] Available at: http://moriverrecovery.usace.army.mil/ [Accessed 5 August` 2016].
U.S. Geological Survey, 2016. Water Data for the Nation changes. [Online] Available at: https://help.waterdata.usgs.gov/news/061016 [Accessed 9 September 2016].
W3 Schools, n.d. JSON Tutorial. [Online] Available at: http://www.w3schools.com/json/default.asp [Accessed 8 September 2016].
W3 Schools, n.d. XML Namespaces. [Online] Available at: http://www.w3schools.com/xml/xml_namespaces.asp [Accessed 8 September 2016].
W3Schools , n.d. Introduction to XML. [Online] Available at: http://www.w3schools.com/xml/xml_whatis.asp [Accessed 11 December 2015].
W3Schools, n.d. PHP 5 Introduction. [Online] Available at: http://www.w3schools.com/php/php_intro.asp [Accessed 11 December 2015].