Handling Raster Data for Hydrologic Applications

Prepared by
Venkatesh Merwade
Lyles School of Civil Engineering, Purdue University
vmerwade@purdue.edu

January 2019

Objective

The objective of this exercise is to learn how to handle raster data in ArcGIS and
understand its properties such as data type, coordinate system, horizontal resolution and
vertical accuracy in hydrologic applications.

Learning outcomes

1) Adding and visualizing raster data in ArcGIS

Input Data

You are provided with one raster dataset, which is the Digital Elevation Model (DEM),
and several vector datasets. These data are available on blackboard as lab2.zip inside the
lab2 folder. The data are also available at:
ftp://ftp.ecn.purdue.edu/vmerwade/download/data/lab2.zip. Unzip the lab2.zip file in
your working folder (keep your folder name and structure simple) and you will see two
sub-folders named ned and vector. The ned folder contains the National Elevation
Dataset (raster layer) for a region in Indiana. The vector folder contains four shapefiles
named point, Tippecanoe, Wabash and watersheds. You will learn more about these
layers during the tutorial.

Opening and Saving ArcMap Documents

Open ArcMap as a blank new map by going to Programs…. Save the ArcMap document
as lab2.mxd by going to FileSave As..

1
Make sure you save the document at the same location where you have unzipped the
input data.

Adding Raster Data to ArcMap Document

Add any type of data to an ArcMap document by either going to FileAdd data or by
clicking the Add Data button . Add ned data to the map document. If you are asked
whether to build pyramids, select Yes. Think about why pyramids are built when you are
working with raster datasets. If needed, use ArcGIS Help or any online resource.

Exploring Properties of Raster Data

Lets explore the properties of the elevation data. Right click on ned, and select Properties.
In the Layer Properties window, select the Source tab.

A raster is organized as a matrix of square cells with each cell having a single value. In
the case of the elevation data, each cell will have an elevation value to represent the
average elevation for that cell. It is very important to know the size of this cell so we
know the “horizontal resolution” of a given dataset. The horizontal resolution of a raster

2
is typically referred by using multiple names, including grid size, cell size, or simply
resolution. Cell size (x,y) gives the spatial resolution of the data. Because the ned data is
not projected, the cell size that you see here is in angular units (degrees). This is one way
to tell whether your data is projected to a coordinate system or not. In geographic
coordinates, the data will have x,y cell size in degrees. The size of the matrix (number of
rows and columns) and the cell size defines the geographic extent of the dataset. Note
the cell size of the given data in angular units (arc seconds). Do you know how to convert
arc seconds to meters. What will be the approximate size of this DEM in meters.

The next key property in a raster dataset is the data type, which is provided as Pixel Type
in the layer properties window. The pixel type for the ned dataset is floating point. What
is a floating point data type? What other datatypes can be associated with a raster layer?

The extent gives the (x,y) coordinates of the four corners of the raster dataset. Spatial
reference gives the information of the geographic/projected coordinate system for the
dataset. Finally, the statistics gives some basic statistics (mean, minimum and maximum
values) of the raster dataset.

Projecting Raster Data

Assuming you now know the difference between geographic coordinates and projected
coordinates, lets assign some projected coordinates to the ned dataset, and explore the
changes to the dataset. If you do not know the difference between geographic and
projected coordinates, there is some reading provided on blackboard. Alternatively, you
can use ArcGIS help or any online resource to understand this difference. To project any
dataset, you will use the Arc Toolbox . Click on the Arc Toolbox. In the Arc Toolbox,
select Data Management ToolsProjections and TransformationsRasterProject
Raster

3
In the Project Raster window, select ned for Input Raster. Name the output raster as
ned_prj, and save it at the same location where your other information for this lab is
stored. For the output coordinate system, click the button next to it, and select Projected
Coordinate SystemUTMNAD 1983 NAD_1983_Zone_16_N. Change the output
cell size to 30 for both x and y, leave all the other default options unchanged, and Click
OK. Why are we using NAD 1983 UTM Zone 16 for this data. Remember, Tippecanoe
County is located in Indiana. Explore UTM zones within the context of projected
coordinate system.

Once ned_prj is added to the map document, open its properties window, and look at the
source tab to see what has changed. What changes do you see in the properties of ned_prj
compared to ned? How did the coordinates for your data change after doing the
projection? What do some of the new terms such as false easting, false northing, central
meridian, etc. associated with the projected spatial reference mean. Why do you think this
is useful?

Clipping Raster Data

Most often the data we get is in square or rectangular tile (tiles). When we get multiple
tiles, we have to mosaic the tiles to create a single raster (this should be done before
projecting). In this case, we have a single tile so we do not have to mosaic. Most often
even with or without mosaicing, the spatial extent of a DEM or any other dataset is
greater than our area of interest. In such case, to avoid the additional computational

4
burden from extra data, we clip the data to match our area of interest. We call this process
as “clipping” or intersection.

Using the Add Data button, browse to the vector folder and add Tippecanoe.shp to the
map document. This shapefile gives you the boundary of the Tippecanoe County. Note
the coordinate system of the Tipppecanoe boundary shapefile.

Project the Tippecanoe shapefile to the same coordinate system as ned_prj. For projecting
ned, you used RasterProject Raster in ArcToolbox. In this case, you will use
Projections and TransformationsProject. Remember Tippecanoe boundary is a vector
feature layer and ned is a raster layer. Name the projected feature as tipp_prj.shp to save
it as shapefile in the same vector folder.

We are interested in getting the elevation data just for the Tippecanoe County. First we
will clip the DEM to match with the county boundary. To clip ned_prj, you will be using
the Arc Toolbox. Go to Data Management ToolsRasterRaster ProcessingClip. In
the Clip window, select ned_prj as the input and tipp_prj.shp for the output extent. Check
the box for using input features for clipping the geometry. Name the output raster as
ned_tipp, and click OK. For some reason, if Clip does not work for you, try the
Extraction tool in Spatial Analyst.

Note the maximum, minimum and average elevation in the Tippecanoe County? (Hint:
look at properties)

5
Extracting values from a raster data set for points, lines and polygons

Most often, hydrologists are interested in extracting information for a given location or
area from a raster. To extract information at one or multiple points, you can use Arc
Toolbox. Add points.shp to your map document from the data folder. Again check if this
dataset is projected. If not, go ahead and project it to match its projection with other
projected datasets. Name the new point shapfile as point_prj.shp.

The point shapefile contains the location of Lafayette and West Lafayette in the
Tippecanoe County in the form of points. Open the attribute table of points and look at its
attributes. We would like to know the elevation at these points. One way to do this is use
the identifier button, and click on one of the points. Initially you may only see the
attributes of the point shapefile, and not the elevation. Change the settings of the identify
window to see the information for all layers (shown below) and hit the identify button
again.

You will then see the elevation at this point next to ned_tipp in the identify window as
shown below.

6
This is a manual way of extracting information for a point from a raster. When you have
multiple points, it is tedious to click on each point to get the elevation from the
underlying DEM.A better way is to use the ArcToolbox. In Arc Toolbox, select Spatial
Analyst ToolsExtractionExtract Values to Points. In the next window, provide the
point_prj.shp as input features, ned_tipp as input raster and save the output features in
points_z.shp. Click OK.

After the process is complete, points_z.shp will be added to the map document. Open the
attribute table of points_z.shp. You will see the extracted values in a field named
RASTERVALU. Note the elevation (in meter) of points corresponding to Lafayette and
West Lafayette in the DEM.

What you just did was extracting elevation for points. In hydrologic applications,
elevations also play role in defining the profile of river bed, river cross-sections, river
banks and other lateral features. Most lines are 2D, and when you associate elevations
with a 2D line, it becomes a 3D line. Add Wabash.shp from the vector folder to your map

7
document. This shapefile describes a short reach of the Wabash river going through
Tippecanoe County. Is Wabash.shp projected? Open the attribute table for Wabash.shp to
look at its Shape attribute. It should say polyline.

We will now convert the Wabash 2D line into a 3D line. This can be accomplished by
using the interpolate shape tool. In the Arc Toolbox, select 3D Analyst
ToolsFunctional SurfaceInterpolate Shape. Use ned_tipp as the input surface,
Wabash.shp as input features, and name the output features as Wabash_z.shp. Leave the
other default options unchanged, and click OK

After the process is complete, you will see a new shapefile Wabash_z.shp added to the
map document. Open the attribute table of Wabash_z.shp, and look at its shape field.
How is it different compared to the shape field in Wabash.shp? What do you think is the
difference between a “Polyline” shape and a “PolylineZ” or “PolylineMZ” shape.

Select the line from Wabash_z.shp, and use the profile graph tool in 3D analyst (Make
sure 3D Analyst Tooblar is added and the 3D Analyst extension is enabled) to see that the
elevations from the DEM are now transferred to the line in Wabash_z.shp to create the
river profile (3D line).

8
The profile graph tool will plot the elevation profile of the just created Wabash_z line as
shown below.

You can follow the same procedure (using Interpolate Shape tool) to convert points, lines
and polygons into surfaces.

Now you know how to extract values from a surface to a line or point. In the case of
polygons, hydrologists are generally interested in computing average properties for
different variables such as watershed slope, average watershed elevation, and average
curve number. This task can be accomplished by using the zonal statistics tool. Add
watersheds.shp to your map document. Is this layer projected? If not, you should know
the drill by now.

What we are going to do now is to compute average elevation for each feature in the
watershed layer by using a DEM. In the Arc toolbox, select Spatial Analyst
ToolsZonalZonal Statistics as Table.

9
Input watersheds.shp for input features. Zone field is the field that has some sort of
unique identifier for each polygon so that you can use this identifier to link the computed
values to the polygons in the input features. In this case, HU_NAME is unique for each
polygon. The input raster is ned_prj, and save the output table as watershed_statistics in
your working folder. In the statistics type box, you can pick the statistic that you are
interested in, or you can compute them all. In this case, we will compute all. Click OK.

The data table that you just created is not linked to the shapefile. Link this table to the
shapefile, and create a map that shows different colors for all watershed based on the
average elevation. <Hint: use the Join property of the watersheds layer to join the
watersheds features with the watersheds_statistics table. Use HU_NAME as the common
identifier>. What is the average elevation of the following watersheds: Indian Creek,
Buck Creek and Sugar Creek-Little Sugar Creek in Tippecanoe County?

Computing vertical accuracy of digital elevation models

A digital elevation model gives elevation values for each cell. While a finer resolution
(small cell size) DEM is desirable, it is more important to have a DEM that is also
accurate. That means the elevation that you get for each cell is closer to the ground truth
as much as possible. This property is generally referred to as vertical accuracy. If a DEM
has a vertical accuracy of 0.5m, the average difference between the DEM and the actual
topography is around 0.5m. The more data you have for actual topography, the more
confidence you have on the vertical accuracy. In this exercise, you will only learn the
process of calculating the vertical accuracy by using the two points we have for Lafayette
and West Lafayette in the points shapefile.

The process is relatively straightforward. You need to have points where you have
observed ground elevations. In our case, that information is stored in the ELEVATION
field (values are in feet) in the point feature class. Next, you need to extract the elevation
values from the DEM for these locations. This is already done when you created the
point_z shapefile. All you need to do is compute the root mean square error (RMSE)
between the two data series, and this will give you the vertical accuracy of the DEM. You
can find the equation for RMSE online.

Congratulations!! You successfully finished this tutorial!!

10