1. Figure 1. Map of Pennsylvania with Allegheny County highlighted.
GIS Final Project
Allegheny County, Pennsylvania
Armando Drain, Juliet Fielding, Matthew Vaughan
Introduction
Our storm was located in Allegheny County, Pennsylvania on 17 June 2008. Allegheny County is
located in southwest Pennsylvania. The main geographic features of the county that are significant to our
project are the Allegheny Mountains and Allegheny, Monongahela, and Ohio Rivers. These three rivers all
flow through Pennsylvania into Pittsburgh, Allegheny County’s largest city. Other cities include
McKeesport, Wexford, Lexington, Penn Hills, and Mount Lebanon. Due to the amount of waterways
flowing through Allegheny County, there is a relatively high risk of flooding when storms with large
amounts of rainfall pass through the area. Our storm was centered on the northern part of the county.
For this reason it did not seem to have a significant impact on Pittsburgh or cause large amounts of
flooding. There was rainfall for about twelve hours and damage was limited.
Objectives
The objective of our project was to examine a storm with rainfall, and produce maps that show
what areas have the highest probability of flooding throughout the county by creating a GIS model. This
enabled us to predict the effects of rainfall on the population and roads.
Software and Data Sources
1. ESRI ArcGIS
2. Microsoft Excel
3. Microsoft Access
4. SSURGO soil Data
5. US Census Bureau population data
6. NOAA National Weather Service
7. TIGER Data
8. National Land Cover Dataset (NLCD)
9. National Climate Data Center NEXRAD
Methods
Preparing for the Final Project
Leading up to the final project, we picked a county that had potential for large amounts of
flooding based on its geography. Then we had to determine that our selected county had completed soil
survey data and a documented flash flood that was in the access database “flash floods,” which was
provided to us. We originally selected the Allegheny County storm with the greatest amount of damage;
however it did not have complete data. Therefore, we picked the second largest storm with flash flooding
2. reported and confirmed its soil survey data was compete.
Next, we downloaded TIGER data for streams, roads, and
block groups from the Census Bureau web site, and added
to our base map in ArcGIS. The geodatabase was set in
UTM and included spatial reference, feature datasets, and
feature classes. All pre-collected data was brought into
ArcGIS and imported into the geodatabase.
The next step was to download NLCD land cover
and NED elevation raster datasets from the USGS National
Map Viewer. It was essential to isolate the data for
Allegheny County and clip all excess outside areas. The
new data once again had to be put into UTM for
compatibility with all preexisting layers. The land cover data was coded to show the various categories of
cover by color. County place names were gathered as well.
The first step of the final project was to collect SSURGO data about the various soil types within
the county by requesting the data from the US Department of Agriculture. The National Climatic Data
Center provided the data about our selected storm. The data was collected from the NEXRAD station
closest to our storm, which happened to be in our county.
Step 1: Raster Total Storm Precipitation Layer
Using the data collected from the National Climate Data Center NEXRAD we collected the total
storm precipitation of the chosen storm at its peak rainfall. After producing frames of the storm, we
exported the one where the storm was most intense as a vector shapefile.
Step 2: Rainfall Data Collection
The next step was to find the 1-year value for our storm. The 1-year value is the total amount of
rainfall over a specific period of time that is likely to occur once every year. To find our value, we
examined rainfall data by date on the NOAA’s National Weather Service website. Our storm’s rainfall
lasted for about twelve hours on 16-17 June 2008. Therefore, according to the rainfall probabilities table,
for Allegheny County, our 1-year value is 1.67. Once we had the rainfall value, we converted the soils
vector layer into raster, and using the combine tool combined it with the NLCD land cover raster layer.
With the combined soil and land cover raster, we created new fields as necessary and joined the curve
values (imported from Excel table) to the layer based on the unique code given to each soil and land cover
combination. This gave us a raster layer with curve values for every cell.
Step 3: SSURGO Soil Data
The raster SSURGO data was used to determine soil retention levels in each cell and assign them
numerical values indicating the amount of runoff they will likely produce. This data is important to our
Figure 2. Allegheny County, PA.
3. final product, because surfaces with lower retention
levels are more likely to produce higher amounts of
runoff.
S = (1000.0/CN) - 10.0
CN = curve numbers from USGS
(values 30-100), based on unique soil-
land cover code
Step 4: Hydrology Calculations
The calculated values were used to compute
the actual values of runoff in each cell based on the 1-
year storm value (1.67) to produce the Q layer.
However, because runoff does not start until the rainfall
total (P layer) is greater than .25, everything less was
reclassified to 0.
Q= (P-.2S)^2 / (P+0.8(1.67))
Step 5: Flow Accumulation
In order to determine the flow accumulation, we needed to complete three major steps. First,
we had to run the Fill tool on the NED elevation layer to assure that no pits existed to stop the calculation
of water flow downstream, and export the new layer. Second, we ran the Flow Direction tool to
determine the direction in which runoff would flow from each cell. Finally, we ran the Flow Accumulation
tool on the Flow Direction output to determine the total flow of runoff into each cell. The Q layer was
used as the weight when running the Flow Accumulation tool in order to simulate actual water amounts.
Step 6: Flow Accumulation Surface
To create a surface displaying the predicted flow accumulation, we plugged the provided
equations into the raster calculator with the actual rainfall totals from our raster storm precipitation layer
rather than the 1-year value (1.67). The same steps were taken to create the storm flow accumulation as
for the 1-year flow accumulation.
Step 7: Reclassify
The next step was to create a flow accumulation layer that predicts water accumulation in each
cell. To limit the actual flow accumulations to streams or flood-created streams, we reclassified all flow
accumulation values less than 1000 to 0, and all other values were reclassified to 1. We then multiplied
the flow accumulation raster by this multiplier layer using the Raster Calculator to remove the undesired
values. This was done for both the actual storm and one-year value layer.
Figure 3. Storm Rainfall 16-17 June 2008.
4. Step 8: Accumulation Percentage
To compute the storm’s percentage of 1-year flow accumulation at all locations, we used the
raster calculator to create a new layer by dividing the storm flow accumulation layer by the 1-year flow
accumulation layer.
Step 9: Reclassify
Before proceeding, we reclassified all values below 100% as 0. These values represented
locations where flow was less than expected and there was no probability of flooding. Using the storm
percentage flow accumulation data, we then reclassified all values greater than 100% on a scale of 1-4
using equal intervals to indicate the probability of flooding within each cell. 1 represented the lowest
probability of flooding, and 4 the highest.
Step 10: Flooding Severity
In this step, we needed to account for flooding outside of the banks of rivers and streams. To
accomplish this we created a series of buffers using the Expand tool. The final buffer expressed the
varying levels of flooding with 2 expanded by 1 cell, 3 expanded by 2 cells, and 4 expanded by 3 cells. The
buffers were developed by using the Expand tool in the following series:
Flood Buffer 1: Expand, 2, 3, 4 by 1 (Green/little to no flooding)
Flood Buffer 2: Expand 3, 4 by 1 (Yellow/moderate flooding)
Flood Buffer 3 (Final): Expand 4 by 1 (Red/most severe flooding)
The streams were coded in red, orange, yellow, and green to indicate the severity of flooding in each cell.
As expected, the most severe flooding occurred in the northern region of the county.
Step 11: Tabulate Area Tool
The block groups UTM layer was used as the input into the Tabulate Area tool, with the flood
buffer as the feature class data input to calculate the area within block groups affected by flooding. The
block groups UTM table needed an area field in order to determine what area of each block group had
been affected by the storm. We joined the block groups UTM table to the Area Affected table to create
one table capable of computing data about flood damage within each block group. We then sorted the
new table to determine which block groups were affected and to what degree, and exported the new
data. The affected area in each block group then needed to be divided by the total area of the block group
to determine the percent that was affected. Based on their percent they were classified as none, low,
moderate, high, and very high flood percentage. To calculate which block groups fell into each category,
the Field Calculator was used. To display the extent to which the population was affected in each block
group, five additional fields were calculated with the same scale (none-very high). Once again all fields
were calculated using the Field Calculator, and subtracting the various levels of damage from the total
population, and placing all remaining citizens into the “no flood” field.
5. To determine if the roads in each block group were affected by flooding, a spatial join was run on
the “roadsUTM” layer to determine the kilometers of road within each block group. Next, the unique sum
values were added to the table. In newly created fields, we once again used the none-very high scale to
determine to what extent any roads were affected by flooding. They were calculated with the block group
flood percentage and new road length (within block groups) fields in the Field Calculator.
Conclusion
In conclusion, we were able to produce a series of tables and maps that clearly depict the effect
our storm had on Allegheny County. Our maps show where the storm took place, its varying levels of
severity, and which areas felt the greatest effects of flow accumulation/flash flooding. These tables and
maps could be used by Allegheny County to predict future flood patterns.
12. Storm Q Layer:
Runoff (in.)
NLCD Land
Cover
Storm Flow
Accumulation
1-Year Flow
Accumulation
Combined Soil and
Land Cover with
Curve Values
S Layer: Soil
Moisture
Retention
SSURGO Soil
Data
NCDC NEXRAD
Storm Rainfall (P)
US Census
Population
Data
TIGER Data
NED 1/3”
Elevation
USGS Curve
Values from
Table
Join
Raster Calc: S = (1000/CN) - 10
Combine
Raster Calc: Q = (P – 0.2S)2 / (P + 0.8S) NCDC NEXRAD 1-
Year Value (P)
1-Year Q Layer:
Runoff (in.)
Fill
Flow Direction: Q as Weight
Storm’s Percentage of
1-Year Flow
Accumulation
Reclassify
Raster Calc: Divide
Expand
Flood Buffer
Reclassify
Reclassify
Flooding by Block
Groups
Population and
Roads Affected
Flow Chart of Final Project Process
Join