ERDC/CHL CHETN-IV-104

August 2015

Automated Method to Develop a Clark

Synthetic Unit Hydrograph within ArcGIS
by Michael Follum

PURPOSE: This Coastal and Hydraulics Engineering technical note (CHETN) describes an
automated method to develop a synthetic unit hydrograph of a small watershed using readily
available spatial data products, namely Digital Elevation Model (DEM), and land cover data. A
Python script is utilized within ArcGIS (Environmental Systems Research Institute 2011) to
quickly and efficiently determine how the spatial features of the watershed affect the runoff of
the basin and therefore the unit hydrograph at the outlet of the basin.

BACKGROUND: Rainfall-runoff response within a watershed is a core consideration of

hydrologists. The use of unit hydrographs as a way to analyze the rainfall-runoff response was
first proposed by Sherman (1932). Since then, many interpretations of the unit hydrograph have
been developed that assume runoff is routed through a series of linear reservoirs, including the
following: instantaneous unit hydrograph method (Nash 1957), geomorphic instantaneous unit
hydrograph method (Rodríguez‐Iturbe and Valdés 1979), and the Clark unit hydrograph (CUH)
method (Clark 1945). The effectiveness of the unit hydrograph is well evident in hydrologic
practice, with the popularity of the CUH method having been increased particularly due to its
inclusion in the U.S. Army Corps of Engineers (USACE) Hydrologic Engineering Center (HEC)
suite of watershed models, including HEC-1 (Feldman 1995) and the Hydrologic Modeling
System (Scharffenberg and Fleming 2006).

Within gauged basins, the unit hydrograph can be developed using observed storm events
(precipitation and runoff data required). In basins with little-to-no observed data, a synthetic unit
hydrograph can be developed. The CUH is a synthetic unit hydrograph that is developed using a
time-area method. To develop a hydrograph using the CUH method, three parameters are
required: time of concentration (Tc), storage attenuation coefficient (R), and a time-area
histogram (TAH) of the basin. Tc is a measure of the time that it takes for water to flow from the
most remote part of the basin to the outlet. Theoretically, from the time rainfall begins, all of the
runoff from the entire basin should be contributing to the outlet when the time reaches Tc. R
represents the effects of storage on the timing of the outflow from the basin. TAH is a means to
represent the area within the basin that drains to the outlet at given time intervals and is the focus
of this CHETN.

Previous methods have used GIS to help create a CUH for a basin (Labadie 2014; Maidment et
al. 1996; Maidment 1993). To help improve implementation times during emergency scenarios,
this document shows how a TAH can be automatically derived using spatially heterogeneous
parameters based on topographic (derived from DEM) and land cover characteristics. Because
the TAH is calculated based on routing times, Tc is intrinsically included within TAH. For the
purposes of this document, any delay in outflow due to R is also assumed to be included within
the flow routing scheme. A simple Python script was created that automates the generation of the

Approved for public release; distribution is unlimited.

ERDC/CHL CHETN-IV-104
August 2015

CUH. The Python script (Appendix A) runs within ArcGIS and generates the TAH by calling
ArcGIS functions and a simple executable, whose code is shown in Appendix B. The format of
this CHETN is to show the methodology on how the Python script works in a step-by-step
example. The purpose of this CHETN is to show how readily available spatial data can be used
within GIS to create a simple, yet accurate, spatially discretized CUH.

METHODS/EXAMPLE: The test location selected for this CHETN is the New River basin in
coastal North Carolina (Figure 1). The basin is approximately 211 km2 and ranges in elevation
between 0.55 m and 36.80 m. As presented in Table 1, the area is predominately evergreen
forest, cultivated crops, and woody wetlands. This test location was selected because it is near
Marine Corps Base Camp LeJeune, the site of a USACE test project aimed at deployment of
fast-response hydrologic models.

Figure 1. New River basin, NC.

2
 ERDC/CHL CHETN-IV-104
August 2015

Table 1. NLCD 2006 identification number, classification, associated

overland flow conveyance (K) values adapted from McCuen (1998), and
runoff coefficient (Cr) values taken from the American Society of Civil
Engineers (ASCE) and Water Pollution Control Federation (WPCF) (1982).

The following is a step-by-step example on how a CUH can be easily derived at a point along the
river within the New River basin. The data used within the example are freely and readily
available nationwide, so the processes and methods shown can be implemented easily at another
location. An example of how flow can be calculated using the CUH method is also shown and
compared to observed data.

Step 1. Obtaining spatial data. The 1/3 arc second (~9.645 m) National Elevation Dataset
(Gesch et al. 2002) over the project area was obtained from the United States Geological Survey
(USGS) (http://nationalmap.gov/). A land cover map from the National Land Cover Database
2006 (NLCD 2006) (Fry et al. 2011) was obtained from the same website. Both datasets were
reprojected onto a UTM Zone 18N projection (the local UTM projection).

Step 2. Create flow direction and flow accumulation rasters; specify outlet. Flow
direction (FDIR) and flow accumulation (FAC) rasters were created over the entire area using
the “Hydrology Toolbox” within ArcGIS. For each cell in the domain, FDIR specifies the
direction of flow from a cell to its neighboring cell, and FAC indicates how many up-slope cells
hydrologically contribute to the cell. Based on the FAC raster, a point shapefile was created at
the outlet of the basin being examined. Figure 2 shows the FAC raster as well as the location of
the outlet selected. Steps 1 and 2 are typical processes carried out prior to any hydrologic
investigation and therefore were not included in the Python script. The remaining steps are
carried out within the Python script shown in Appendix A.

3
ERDC/CHL CHETN-IV-104
August 2015

Figure 2. New River FAC, outlet location, and watershed boundary.

Step 3. Creation of TAH. There are two products within this step: (1) a raster that represents the
time it takes each cell within the basin to drain to the outlet and (2) a TAH derived from the raster.
To complete this, several subtasks must be completed. These tasks are automated within a Python
script shown in Appendix A and a C++ code shown in Appendix B. Tasks 1–7 in the following list
provide a description of what is simulated within the Python script, while Tasks 8 and 9 provide a
brief description of how the two products are ultimately created using a C++ code.

Task 1 (Lines31–41): The watershed is delineated, and a watershed boundary is created (Figure 2).
The elevation and flow accumulation rasters are clipped to the watershed boundary. A stream grid
is created by assuming that any FAC cell that drains an area greater than 5 km2 (cell value of
53,744 at a 9.645 m resolution) is considered a stream.

Task 2 (Lines 44–48): The slope of each cell is calculated using the “Slope” function within
ArcGIS. The slopes are calculated for an area larger than the watershed boundary and then are
clipped to the watershed boundary. The clipped slope raster is initially calculated as a percentage
with all slopes less than 0.001% being assigned a value of 0.001%. The slope raster is then
converted into a decimal format raster (OutSlope_Dec). A third raster is also created to represent
the slope (decimal format) in only the channel cells (Slope_Chan).

Task 3 (Lines 51–58): The land cover raster is clipped to the watershed boundary and is set to
match the spatial extent and resolution of the DEM raster. Flow conveyance values (K) are then

4
 ERDC/CHL CHETN-IV-104
August 2015

mapped to a new raster (kfactor) based on the land cover classification and the associated K
value. Table 1 lists the values of K used in this project, which are based on McCuen (1998) and
Nicklow et al. (2006).

Task 4 (Lines 61–63): The time it takes for water to flow overland within each cell
(timeoverland) is calculated based on an approximation of the Manning’s equation:

 1   
min   1 ft  1 0.5 1
timeoverland     K  S  0.0547   
C K  S 0.5
 60  sec   0.3048  m 

where K is taken from kfactor, S is the slope of the cell (decimal, taken from OutSlope_Dec), and
timeoverland is in min∙m-1. Values of K are approximations that incorporate channel/flow
geometry and roughness. It is also realistic to assume that K varies based on scale. A calibration
parameter (C) is included to manipulate the value of K for a given basin and is set to 0.3 for this
example, which effectively increases the time it takes for water to flow over each cell.

Task 5 (Lines 66–69): A raster representing the time of travel within each channel cell
(timechannel, min m-1) is calculated using the same approach as for overland flow, with the
exceptions that S is taken from Slope_Chan, and a static value of 25.0 is used for K. In reality,
the K value likely varies with each segment of stream, but for this simple example, it is left as a
constant for all channel cells. Similar to Task 4, a channel calibration parameter is also employed
to manipulate the value of K and is also set to 0.3 for this example.

Task 6 (Lines 72–74): A raster called traveltime is created that represents the travel times
(min∙m-1) of both the overland flow and channel flow. If a cell is classified as a channel, the
value from timechannel is used, otherwise the value is taken from timeoverland. After
calculation, the traveltime raster is converted into an ASCII format (traveltime.asc) for use in
Tasks 7–9.

Task 7 (Lines 77–78): Although the time-per-distance for water to travel over each cell has been
determined in Task 6, the path that water takes typically depends on topography and can be
approximated as the path it would follow along increasing cell values of the FAC. The
accumulated time for water to travel a hydraulically driven path from each cell to the outlet is
required to create a TAH. A function within ArcGIS that can complete this task does not exist.
Therefore, a program was written in C++ to complete this task and is shown in Appendix B.

Task 8 (Appendix B): A simple executable is called from the Python script. The executable was
developed to determine the time it takes for each cell to flow to the outlet based on the flow
pathway exhibited by the FAC raster and the time to flow across a given cell in Task 6. The
program routes the path water takes from each cell within the watershed to the outlet by
following a path of increasing FAC values. The time (min) it takes to travel across a single cell is
simply the distance across the cell (m) multiplied by the value within the traveltime.asc raster
(min∙m-1). The total time (min) it takes each cell to route to the outlet is the summation of the
times along the pathway. This summation of the time is recorded in the totaltime.asc raster and is
shown for the New River basin in Figure 3. To clarify, the totaltime.asc raster specifies the time,
in minutes, it takes for water to travel from that cell to the outlet of the basin.

5
ERDC/CHL CHETN-IV-104
August 2015

Figure 3. Time in minutes for each cell to drain to outlet (totaltime.asc).

Task 9 (Appendix B): A histogram based on the totaltime.asc raster shown in Task 8 can be
generated using a simple plotting program, such as Excel. Figure 4 shows such a histogram using
6 hr intervals, which is the 6 hr TAH.

Step 4. Flow Simulation. Two methods are available to determine flow based on a given
precipitation. For uniform precipitation, a unit hydrograph is easily generated using the S-curve
method and the TAH, where a standard unit hydrograph can be derived for a given time interval
of rainfall and used with the rainfall data, or a normalized unit hydrograph can be derived for use
in conjunction with peak flow calculations (Bras 1990; Brutsaert 2005). Figure 5 shows an S-
curve with a normalized unit hydrograph. The second method, which is used in this document,
uses spatial precipitation data in conjunction with the totaltime.asc raster to determine how long
the rain in each cell takes to get to the outlet. This second method takes a Lagrangian approach to
simulate flow because it follows rainfall through each cell to the outlet. Both methods described
assume the principle of superposition to simulate outflow.

6
 ERDC/CHL CHETN-IV-104
August 2015

Figure 4. Time-area histogram (TAH) for the New River basin.

Figure 5. 60 min Clark unit hydrograph (CUH) for the New River basin.

7
ERDC/CHL CHETN-IV-104
August 2015

TESTING: Two peak flows from the New River near Gum Branch streamgage occurred
between 15 August and 15 September 2011 when multiple large rain events occurred over the
study area. This time period will serve to show how the CUH method described can be used to
simulate flow.

Next-Generation Radar (NEXRAD) precipitation data in shapefile format were obtained from
the National Weather Service (http://www.srh.noaa.gov/ridge2/RFC_Precip/). The data were
reprojected, clipped to the area of interest, and converted into raster format using the inverse
distance weighted (IDW) function within ArcGIS. Flow data from the New River near Gum
Branch streamgage were obtained from the USGS (http://waterdata.usgs.gov/nwis/rt).

Using a simple programming script (not shown), the effective precipitation (Peff) was applied to
the totaltime.asc raster to generate a hydraulic response from each hour of precipitation. By
combining these hourly responses through the assumption of superposition, a simulated outflow
hydrograph is created. Peff represents the fraction of precipitation that contributes to immediate
runoff, and is calculated as

Peff  P  C r

where P is the precipitation over an hour time period for a given cell from the NEXRAD data
and Cr is the runoff coefficient (Table 1). Values of Cr vary widely and are typically calibrated
parameters. The values of Cr used in this project are based on land cover classifications and
ASCE and WPCF (1982). Developed areas are based on residential and neighborhood values,
while all other areas use a value of 0.15, which is the approximate average of the values for
lawns, unimproved areas, and parks.

The time-series of hourly precipitation (volume, m3) applied to the watershed is shown with the
simulated (Qsim) and observed flow (Qobs) in Figure 6. Qsim for the first event (26 August–4
September) peaked approximately 7 hr before Qobs and overestimated the peak by approximately
5.51 m3 s-1. Qsim peaked approximately 9 hr after Qobs during the second event (7–10 September)
and underestimated the peak by approximately 4.41 m3 s-1. The duration of both events matched
reasonably well when compared to Qobs. This highlights potential areas of improvement in both
the inclusion of baseflow estimation and the methods used to calculate the time to flow across
each cell. For both events, Table 2 shows the simulated and observed peak flows, mean bias
error (MBE), root mean squared error (RMSE), and Pearson’s correlation coefficient (R).
0.5
 1 n   n
2
 n  Qsim  Qobs  RMSE   n1  Qsim  Qobs  
MBE                               
   
 i 1   i 1 

Table 2. Flow event statistics (all in units of m3 s-1 except R,

which is 0-1).
Peak Qobs Peak Qsim MBE RMSE R
Flow Event 1 30.30 35.81 -1.73 3.77 0.95
Flow Event 2 10.34 5.93 -1.14 2.19 0.84

8
 ERDC/CHL CHETN-IV-104
August 2015

Figure 6. Test events: Precipitation (solid black line, top), Observed Flow (solid black line, bottom), and
Simulated Flow (dotted black line).

CONCLUSIONS: This CHETN describes in a step-by-step procedure how to automatically

develop a time-area unit hydrograph based only on elevation and land cover data. The CHETN
also provides the programming code, which can be used by anyone with low-to-intermediate
ArcGIS and programming skills. It was also shown that the time-area unit hydrograph methods
can in turn be used to accurately simulate outflow from a small basin when precipitation data is
known. Through an example, distributed rainfall data was applied to the generated synthetic unit
hydrograph to produce accurate results. Although the methods presented are simple, further
refinement can be introduced into the framework to improve accuracy in time-to-drain estimates
and effective precipitation estimation. Although this project was developed to assist in a specific
military need, the methods may be applicable to other modeling frameworks that use a unit
hydrograph method.

POINTS OF CONTACT: This CHETN was prepared by the Military Engineering Program
through the Austere Entry Assessment project. The POC for technical inquiries is Mike Follum
([email protected]). This technical note should be referenced as follows:

M. Follum. 2015. Automated method to develop a Clark synthetic unit

hydrograph within ArcGIS. ERDC/CHL CHETN-IV-104. Vicksburg, MS: U.S.
Army Engineer Research and Development Center. An electronic copy of this
CHETN is available from http://chl.erdc.usace.army.mil/chetn.

REFERENCES
American Society of Civil Engineers (ASCE) and Water Pollution Control Federation (WPCF) (Joint Committee). 1982.
Design and construction of sanitary and storm sewers. ASCE (American Society of Civil Engineers) Manual on
Engineering Practice No. 37 and WPCF (Water Pollution Control Federation) Manual of Practice No. 9., 5th ed.

9
ERDC/CHL CHETN-IV-104
August 2015

Bras, R. L. 1990. Hydrology: An introduction to hydrologic science. Reading: Addison-Wesley.

Brutsaert, W. 2005. Hydrology: An introduction. New York: Cambridge University Press.
Clark, C. 1945. Storage and the unit hydrograph. Transactions of the American Society of Civil Engineers
110(1):1419–1446.
Environmental Systems Research Institute. 2011. ArcGIS Desktop: Release 10. Redlands, CA: Environmental
Systems Research Institute.
Feldman, A. 1995. HEC-1 flood hydrograph package. Computer Models of Watershed Hydrology 119:150.
Fry, J. A., G. Xian, S. Jin, J. Dewitz, C. Homer, Y. Limin, C. Barnes, N. Herold, and J. Wickham. 2011. Completion
of the 2006 national land cover database for the conterminous United States. Photogrammetric Engineering and
Remote Sensing 77(9):858–864.
Gesch, D., M. Oimoen, S. Greenlee, C. Nelson, M. Steuck, and D. Tyler., 2002. The national elevation dataset.
Photogrammetric Engineering and Remote Sensing 68(1):5–32.
Labadie, J. 2014. Lecture Handouts for CIVE 577: GIS in civil and environmental engineering, Lecture 8: Time-
area unit hydrographs using GIS. Ft. Collins, CO: Department of Civil and Environmental Engineering,
Colorado State University,
Maidment, D., F. Olivera, A. Calver, A. Eatherall, and W. Fraczek., 1996. Unit hydrograph derived from a spatially
distributed velocity field. Hydrological Processes 10(6):831–844.
Maidment, D. R. 1993. Developing a spatially distributed unit hydrograph by using GIS. In Applications of
Gepographic Information Systems in Hydrology and Water Resources Management, ed. K. Kovar and H.
Nachtnebel. IAHS 211:181–192.
McCuen, R. H. 1998. Hydrologic analysis and design. New Jersey: Prentice-Hall.
Nash, J. 1957. The form of the instantaneous unit hydrograph. In General Assembly of Toronto, Vol. III Surface
Waters, Prevision, Evaporation. IAHS 45(3):114–121.
Nicklow, J. W., P. F. Boulos, M. K. Muleta. 2006. Comprehensive urban hydrologic modeling handbook for
engineers and planners. MWH Soft, Incorporated.
Rodríguez‐Iturbe, I., J. B. Valdés. 1979. The geomorphologic structure of hydrologic response. Water Resources
Research 15(6):1409–1420.
Scharffenberg, W. A., and M. J. Fleming. 2006. Hydrologic modeling system HEC-HMS: User's manual. Davis,
CA: U.S. Army Corps of Engineers, Hydrologic Engineering Center.
Sherman, L. K. 1932. Streamflow from rainfall by the unit-graph method. Engineering News Record 108:501–505.

10
 ERDC/CHL CHETN-IV-104
August 2015

Appendix A: Python Program Code Executed through ArcGIS

11
ERDC/CHL CHETN-IV-104
August 2015

Appendix B: C++ Code to Compute TAH

NOTE: The contents of this technical note are not to be used for advertising, publication,
or promotional purposes. Citation of trade names does not constitute an official
endorsement or approval of the use of such products.

12