SepHydro - Hydrograph Separation Tool

__CONTENTS__

- 1. About the tool
- 2. Background
- 3. Method Overview
- Lyne and Hollick Method
- Chapman Method
- Eckhardt Method
- Pettyjohn and Henning - Fixed Interval Method
- Pettyjohn and Henning -Sliding Interval Method
- Pettyjohn and Henning - Local Minimum Method
- TR-55 Method
- Szilagyi Method
- Boughton (AWBM 1993) Method
- Furey and Gupta Method
- Chapman and Maxwell (1996) Method
- 4. User Guide
- 5. Limitations
- 6. Terms of Use
- 7. References
- 8. Contact

1. About the tool

SepHydro is a baseflow (or hydrograph) separation web-based tool developed through a collaborative research effort between Canadian Rivers Institute (CRI), University of New Brunswick (UNB), Agriculture and Agri-Food Canada (AAFC) and Environment and Climate Change Canada (ECCC). SepHydro is a result of the research effort aimed at evaluating the effects of agricultural production systems on surface and groundwater water quality and on the quality of downgradient aquatic ecosystems.

The tool offers several customizable filtering algorithms for assessing surface runoff and groundwater contributions to streamflow, as well as various output and data visualization options. The web-based tool has been designed with a user-friendly interface and provides flexibility in the format of user-data input and output files through a streamlined process.

For detailed instructions on how to use this tool please refer to the User Guide section.

2. Background

Knowledge of

__baseflow contribution to streamflow__is important for understanding watershed scale hydrology, including groundwater-surface water interactions, impact of geology and landforms on baseflow, estimation of groundwater recharge rates, etc.
Baseflow separation methods can be used as supporting tools in many areas of environmental research, such as the assessment of the impact of agricultural practices, urbanization, climate change etc. on surface water and groundwater.

Conceptually,

__baseflow__is the portion of the streamflow that has a different source than surface runoff. Often times, baseflow is considered to represent the sum of both deep and shallow subsurface contributions to streamflow.
Recursive digital filters are one of the most popular methods used for baseflow separation. Digital filters, routinely used for signal analysis, allow for separation of “high” (i.e. surface runoff) and “low” (i.e. baseflow) frequency signals using a mathematical formulation. Each digital filter method is dependent on one or more filter parameters, which control the shape and proportion of baseflow.

One advantage of the digital filters is that they can be applied in multiple passes, and therefore the streamflow can be separated in more than two components. The components can then be attributed to various sources of streamflow such as surface runoff, delayed shallow flow, tile drain flow, deep groundwater flow, etc.

3. Method Overview

Currently, the tool offers a total of 11 methods for the purpose of baseflow separation. The methods are based on either digital filters or graphical techniques. Brief descriptions of each of the methods are included below.

3.1. Lyne and Hollick Method

This method was first introduced by Lyne and Hollick in 1979. It has been designed as a means of analyzing the time-dependent and slow stream responses to precipitation. The method assumes high frequency signals to represent direct runoff, while low frequency signals are associated with the baseflow component. The method calculates the surface runoff component which, together with the streamflow, is then used to derive the baseflow component

q_t = \alpha \times q_{t-1} + \frac{(1 + \alpha)}{2} \times (Q_t - Q_{t-1})

q – surface or quick runoff [m

^{3}/s];
Q - streamflow [m

^{3}/s];
t - the time (e.g. day) for which the surface runoff is calculated;

α - catchment constant [values between 0 and 1];

Lyne V., Hollick M. (1979) - Stochastic time-variable rainfall-runoff modelling. Institute of Engineers Australia National Conference. Publ. 79/10, pp. 89-93.

3.2. Chapman Method

The Chapman method was introduced in 1991 as a response to Lyne-Hollick algorithm's results, which incorrectly provided a constant streamflow or baseflow even when direct runoff has ceased.

b_t = \frac{(3 \times \alpha - 1)}{(3 - \alpha)} \times b_{t-1} + \frac{(1 - \alpha)}{(3 - \alpha)} \times (Q_t + Q_{t-1})

b - baseflow [m

^{3}/s];
Q - streamflow [m

^{3}/s];
t - the time (e.g. day) for which the baseflow is calculated;

α - hydrological recession constant [values between 0 and 1];

Chapman T. (1991) - Comment on evaluation of automated techniques for base flow and recession analyses, by RJ Nathan and TA McMahon. Water Resources Research, 27(7), pp. 1783-1784.

3.3. Eckhardt Method

In 2004, Eckhardt presented a new two-parameter algorithm for baseflow separation that was inspired by the previous one-parameter filters.

b_t = \frac{(1-BFI_{max}) \times \alpha \times b_{t-1} + (1-\alpha) \times BFI_{max} \times Q_t}{1-\alpha \times BFI_{max}}

b - baseflow [m

^{3}/s];
Q - streamflow [m

^{3}/s];
t - the time (e.g. day) for which the baseflow is calculated;

α – groundwater recession constant [values between 0 and 1];

BFI

_{max}- long-term ratio of baseflow to total streamflow [values between 0 and 1];
Eckhardt K. (2005) - How to construct recursive digital filters for baseflow separation. Hydrol. Process. 19, pp. 507–515.

3.4. Pettyjohn and Henning - Fixed Interval Method

One of the three methods developed by Pettyjohn and Henning in 1979. The method systematically draws connecting lines between the low points of the streamflow hydrograph to determine the baseflow hydrograph.
The low points of the streamflow hydrograph are determined using a fixed window of specified width (i.e. equal to a set number of streamflow readings in the source dataset). All baseflow values in a given interval or window are set to the minimum streamflow value in that respective interval.
The window width is the nearest integer between 3 and 11 that is equal to 2N, where N is empirically determined as follows:

N = A^{0.2}

N - number of days after which runoff ceases;

A - drainage area [km

^{2}]*;
* the original equation uses [sq mi], however the tool requires [km

^{2}] and converts the drainage area to the proper units internally;
Pettyjohn W.A., Henning R. (1979) - Preliminary estimate of ground-water recharge rates, related streamflow and water quality in Ohio: Ohio State University Water Resources Center Project Completion Report Number 552, p. 323.

3.5. Pettyjohn and Henning - Sliding Interval Method

The low points of the streamflow hydrograph are determined using a moving interval or sliding window of specified width (i.e. equal to a set number of streamflow readings in the source dataset). The baseflow value in the middle of each given inverval or window is set to the minimum streamflow value in that respective interval.
The window width is the nearest integer between 3 and 11 that is equal to 2N, where N is empirically determined as follows:

N = A^{0.2}

N - number of days after which runoff ceases;

A - drainage area [km

^{2}]*;
* the original equation uses [sq mi], however the tool requires [km

^{2}] and converts the drainage area to the proper units internally;
Pettyjohn W.A., Henning R. (1979) - Preliminary estimate of ground-water recharge rates, related streamflow and water quality in Ohio: Ohio State University Water Resources Center Project Completion Report Number 552, p. 323.

3.6. Pettyjohn and Henning - Local Minimum Method

The low points of the streamflow hydrograph are determined using a moving interval or sliding window of specified width (i.e. equal to a set number of streamflow readings in the source dataset). Each streamflow value is checked whether it is the lowest in a given interval or not. If it is, then it is considered a local minimum and is joined with adjacent local minimums using a straight line. The resulting line, across the whole dataset, represents the baseflow hydrograph.
The window width is the nearest integer between 3 and 11 that is equal to 2N, where N is empirically determined as follows:

N = A^{0.2}

N - number of days after which runoff ceases;

A - drainage area [km

^{2}]*;
* the original equation uses [sq mi], however the tool requires [km

^{2}] and converts the drainage area to the proper units internally;
Pettyjohn W.A., Henning R. (1979) - Preliminary estimate of ground-water recharge rates, related streamflow and water quality in Ohio: Ohio State University Water Resources Center Project Completion Report Number 552, p. 323.

3.7. TR-55 Method

TR-55 is a method developed by USDA starting with 1975 and provides a number of techniques used for small watersheds, especially urbanized watersheds.
The presented method predicts the peak rate of runoff as well as the total volume.

q_t = \frac{(P_t - Ia)^{2}}{(P_t - Ia + S)}

S = \frac{1000}{CN} - 10

Ia = 0.2 \times S

q - surface runoff [m

^{3}/s];
P - precipitation [mm/d]*;

t - the time (e.g. day) for which the surface runoff is calculated;

CN - runoff curve number;

S - potential maximum retention after runoff begins [in**; calculated based on CN];

Ia - initial abstraction (i.e. sum of all losses before runoff begins) [in**; calculated based on CN];

* the original equation uses [in/d], however the tool requires [mm/d] and converts the precipitation to the proper units internally;

** the original equation uses [in], however the tool requires [mm] and converts the parameters to the proper units internally;

** the original equation uses [in], however the tool requires [mm] and converts the parameters to the proper units internally;

Cronshey R. et. al (1986) - Urban Hydrology for Small Watersheds, Technical Release 55, June 1986, United States Department of Agriculture.

3.8. Szilagyi Method

The short-time solution defines constants a and b as follows:

For the long-time solution, constants a and b are defined as follows:

Introduced in 1997, this method estimates the value of the baseflow maximum as well as the baseflow recession hydrograph by fitting the analytical solutions of the Boussinesq equation to the observed discharge values for individual flood events.
The method provides both a short-term and a long-term solution, as follows:

b_t = (Q_t^{1-b} - (1-b) \times a \times d)^{\frac{1}{1-b}}

b - baseflow [m

^{3}/s];
Q - streamflow [m

^{3}/s];
t - the time (e.g. day) for which the baseflow is calculated;

d - average interval length between adjacent discharge measurements [h];

a – catchment parameter (see below);

b – catchment parameter (see below);

The short-time solution defines constants a and b as follows:

a_{short} = \frac{1.133}{K \times \phi \times D^{3} \times L^{2}}

b_{short} = 3

For the long-time solution, constants a and b are defined as follows:

a_{long} = \frac{4.804 \times K^{\frac{1}{2}} \times L}{\phi \times A^{\frac{3}{2}}}

b_{long} = \frac{3}{2}

A - catchment area [km

^{2}];
D - acquifer depth [m];

L - length of contributing channels [km];

K – saturated hydraulic conductivity [m/s];

Φ - drainable porosity or specific yield [values between 0 and 1];

Szilagyi J., Parlange M. B. (1997) - Baseflow separation based on analytical solutions of the Boussinesq equation, Journal of Hydrology 204 (1998), pp. 251-260.

3.9. Boughton (AWBM 1993) Method

Developed by Walter Boughton in 1993, together with the widespread AWBM rainfall-runoff model. The method is using groundwater levels for baseflow separation and essentially works as a one-pass filter.

b_t = \frac{k}{1 + C} \times b_{t-1} + \frac{C}{1 + C} \times Q_t

b - baseflow [m

^{3}/s];
Q - streamflow [m

^{3}/s];
t - the time (e.g. day) for which the baseflow is calculated;

k – groundwater recession constant [values between 0 and 1];

C - shape constant [positive values];

Boughton W.C. (1993) - A hydrograph-based model for estimating water yield of ungauged catchments. Institute of Engineers Australia National Conference. Publ. 93/14, pp. 317-324.

3.10. Furey and Gupta Method

This filter, introduced in 2001, is based on a mass balance equation for baseflow through a hillside, and its construction is founded on a physical-statistical theory of low streamflows developed by Furey and Gupta.

b_t = (1 - \gamma) \times b_{t-1} + \gamma \times \frac{c3}{c1} \times (Q_{t-d-1} - b_{t-d-1})

b - baseflow [m

^{3}/s];
Q - streamflow [m

^{3}/s];
t - the time (e.g. day) for which the baseflow is calculated;

γ – recession constant [values between 0 and 1];

c

_{1}– ratio of overland flow to precipitation;
c

_{3}– ratio of groundwater recharge to precipitation;
d - time-delay between precipitation and groundwater recharge [d]*;

* d is assumed to be equal to zero, however the user can change this value during method selection;

Gupta V. K., Furey P. R. (2001) - A physically based ﬁlter for separating base ﬂow from streamﬂow time series, Water Resources Research, Vol. 27, No. 11, November 2001, pp. 2709-2722.

3.11. Chapman and Maxwell (1996) Method

The Chapman and Maxwell digital filter, introduced in 1996, can be viewed as a simplified version of the Boughton filter from 1993.

b_t = \frac{1}{2 - k} \times b_{t-1} + \frac{1 - k}{2 - k} \times Q_t

b - baseflow [m

^{3}/s];
Q - streamflow [m

^{3}/s];
t - the time (e.g. day) for which the baseflow is calculated;

k - recession constant [values between 0 and 1];

Chapman, T. G., Maxwell, A. I. (1996) - Baseflow separation - comparison of numerical methods with tracer experiments, in Hydrol. and Water Resour. Symp., Institution of Engineers Australia, Hobart. pp. 539-545.

4. User Guide

*SepHydro*is a web application that performs baseflow separation of streamflow using a multitude of methods. The tool provides tabular and graphical representations of the streamflow and baseflow data, as well as representative statistics.

On the left, a contextual menu provides access to the various components of the tool. At the top of the page, a

__progress bar__displays the current status of the analysis.
The three steps required for the use of this tool (I. Load Streamflow Data; II. Perform Baseflow Separation; III. Investigate Results and Export Data) are presented in detail below.

I. Load Streamflow Data

The first step in any analysis is to choose a streamflow time series to be used for baseflow separation. This can be done using the

__Upload User Data__menu entry under the__SOURCE DATA__menu section.
Alternatively, the user can use the sample data set provided by using the

__Load Sample Data__menu entry. The data set contains one year of daily streamflow and precipitation records.
A streamflow dataset is a collection of flow records (i.e. discharge) over a period of time. The dataset should contain at least two columns: a

__time series__(e.g. daily, monthly, etc.) and a flow__data series__. The tool can also accomodate datasets with multiple columns (e.g. with__precipitation data series__). In this case, the user has to select the columns representing the time series and the flow and precipitation records, respectively.
The tool accepts source data sets in

__Excel__(xls, xslx) and__Comma Separated File__(csv) formats. The tool can accommodate missing data or erroneous data entries, by removing the respective lines, but we recommend the user to verify the integrity of the source data before uploading.
Once a streamflow dataset is loaded to the tool, a series of options become available in the

__SOURCE DATA__menu section. These allow for investigation of the streamflow data using a tabular format (Table View) as well as for visualization of the stream hydrograph through a graphical representation (Graphical View). The Graphical View allows for inspection of the full data set or of a reduced subset by changing the beginning and the ending values of the time interval that is displayed. Statistics for both the entire dataset and the selected subset are shown in both the__Table View__and the__Graphical View__.
II. Perform Baseflow Separation

Baseflow Separation can be performed once the streamflow data is loaded.

The

__Analyze__menu entry becomes available once the streamflow data is loaded. By following this link, the user is presented with a set of methods that can be used for baseflow separation.
Multiple methods can be used at the same time by clicking the checkboxes next to their names. For each selected method, parameters can be configured using the input boxes.

A

__preview__is also available. The preview is generated only for a limited interval around a local maximum in the data. More in depth visualization is available once the analysis is executed and completed.
The algorithm is forcing baseflow to be equal to streamflow (b = Q) for the instances when the calulated baseflow is larger than the streamflow. The number of such values in the resulting dataset is recorded and can be used for evaluating the parameter selection and overall method reliability.

Once the method selection and parameter configuration is final, the user can proceed to analyse / perform baseflow separation by clicking the

__Run Analysis__button at the bottom of the page.
III. Investigate Results and Export Data

Following completion of the data analysis / baseflow separation, the tool redirects to the

__Baseflow Hydrograph__page, where the user can visually investigate the results. Alternatively, a__Table View__is available through the menu entry with the same name. These entries can be found under the__OUTPUT DATA__menu section, which becomes visible only when data analysis is complete.
The

__Graphical View__allows the user to choose the desired results for review, as well as a custom viewing interval, by using drop-down lists. Several statistics, including the baseflow contribution to streamflow are shown on this page. A more comprehensive list of statistics is available under the__Table View__section.
When multiple methods were used concurrently for baseflow separation, a menu entry named

__Compare Methods__also becomes available. By following this link, the user can configure and view comparison hydrographs between the methods, as well as representative statistics and graphical representations related to the results.
The data set resulted from the analysis (i.e. output dataset), as well as the graphical representations and statistics relevant for the analysis, can be downloaded by following the specific links in the left-hand menu and in the

__Table View__,__Graphical View__and__Compare Methods__pages.
5. Limitations

a) For large data sets (i.e. over 3 000 data points) we do not recommend using all filters simultaneously. Rather, try to use only up to three baseflow separation methods at a time;

b) The optimum number of filters that can be used simultaneously can be determined by trial;

c) If the analysis cannot be conducted even when only one filter is selected, we recommend splitting the input data set into smaller files

6. Terms of Use

*SepHydro*can be used freely.

The tool can be referenced as follows: Environment and Climate Change Canada, 2016, SepHydro - Hydrograph Separation Tool. Also, we kindly ask you to provide the web address of the current website.

The authors do not assume any responsibility for the tool's operation, output, interpretation, or use of results.

For more information you may wish to contact the authors. See the Contact section for details.

7. References
Pettyjohn W.A., Henning R. (1979) - Preliminary estimate of ground-water recharge rates, related streamflow and water quality in Ohio: Ohio State University Water Resources Center Project Completion Report Number 552, p. 323.

Lyne V., Hollick M. (1979) - Stochastic time-variable rainfall-runoff modelling. Institute of Engineers Australia National Conference. Publ. 79/10, pp. 89-93.

Chapman T. (1991) - Comment on evaluation of automated techniques for base flow and recession analyses, by RJ Nathan and TA McMahon. Water Resources Research, 27(7), pp. 1783-1784.

Eckhardt K. (2005) - How to construct recursive digital filters for baseflow separation. Hydrol. Process. 19, pp. 507–515.

Cronshey R. et. al (1986) - Urban Hydrology for Small Watersheds, Technical Release 55, June 1986, United States Department of Agriculture.

Szilagyi J., Parlange M. B. (1997) - Baseflow separation based on analytical solutions of the Boussinesq equation, Journal of Hydrology 204 (1998), pp. 251-260.

Boughton W.C. (1993) - A hydrograph-based model for estimating water yield of ungauged catchments. Institute of Engineers Australia National Conference. Publ. 93/14, pp. 317-324.

Gupta V. K., Furey P. R. (2001) - A physically based ﬁlter for separating base ﬂow from streamﬂow time series, Water Resources Research, Vol. 27, No. 11, November 2001, pp. 2709-2722.

Chapman, T. G., Maxwell, A. I. (1996) - Baseflow separation - comparison of numerical methods with tracer experiments, in Hydrol. and Water Resour. Symp., Institution of Engineers Australia, Hobart. pp. 539-545.

8. Contact

Serban Danielescu, Ph.D.

Research Scientist | Chercheur scientifique

Environment and Climate Change Canada | Environnement et Changements Climatiques Canada

Agriculture and Agri-Food Canada | Agriculture et Agroalimentaire Canada

Agriculture and Agri-Food Canada | Agriculture et Agroalimentaire Canada

Fredericton Research and Development Centre | Centre de recherche et développement de Fredericton

850 Lincoln Rd., Fredericton, NB, E3B 4Z7

Telephone/Téléphone: 506-460-4468

Facsimile/Télécopieur: 506-460-4377

850 Lincoln Rd., Fredericton, NB, E3B 4Z7

Telephone/Téléphone: 506-460-4468

Facsimile/Télécopieur: 506-460-4377