Groundwater Toolbox Tutorial

Updated Functionality for Analysis of Time‐Series Data

October 1, 2016 (Version 1.2 release)

This document describes updated functionality for the Groundwater (GW) Toolbox to analyze
time‐series data. The updates (1) allow users to work with base‐flow and runoff time series
generated by use of the “Interactive” mode of hydrograph separation, (2) provide
enhancements to the existing Kendall Tau trends functionality, and (3) allow subsetting,
splitting, filtering, and other manipulations of a time series. Each of the updates is described
using streamflow data available for the USGS Beaver River near Guymon, Oklahoma,
streamgage (07232500). Streamflow was monitored at the site between October 1, 1937, and
September 30, 1993. Wahl and Tortorelli (1997, U.S. Geological Survey Water‐Resources
Investigations Report 96‐4303) have shown that streamflow and base flow at the site sharply
decreased during the monitoring period in response to groundwater development in the basin.
The declines are illustrated by the following streamflow hydrograph for the period of record:

1
Working with Calculated Base‐Flow and Runoff Time Series

The GW Toolbox was used to calculate daily time series of base flow and runoff at the Beaver
River streamgage for the period of record using the PART, HYSEP‐FIXED, and BFI Standard
approaches (see USGS Techniques and Methods 3‐B10, which is distributed with the GW
Toolbox, for a description of the hydrograph‐separation methods, as well as the tutorial on
“New and Updated Functionality for Hydrograph Separation with the Groundwater Toolbox”).
Three time series were generated for each of the hydrograph‐separation methods: daily base
flow (‘BF_’), base‐flow percentage (‘BFPct_’), and runoff (‘RO_’). These time series are now
available for graphing and other analyses within the Toolbox, as illustrated by the following
data‐selection dialog box and the examples in the remainder of this tutorial:

Note that after closing the current Toolbox session, the generated time series are not saved
with the project, as is done for the streamflow data. However, the base‐flow and runoff time
series are available in the Daily.csv files that are generated as part of the hydrograph‐
separation process, and can be imported back into the Toolbox from the csv files (see the
tutorial on ‘Scripting_to_import_Excel_data.pdf’).

2
Kendall Tau Trends Functionality

The GW Toolbox includes the Kendall Tau statistic to test for trends in an annual time‐series
data set. The functionality is available within the “Analysis>USGS Surface Water Statistics
(SWSTAT)>Trend” option:

In this example, annual trends in observed streamflow and base flow calculated by the BFI
Standard method will be analyzed for several user defined N‐day periods. The first step is to
select the streamflow and BFI base‐flow records:

3
Hitting “Ok,” brings up the Trends dialog box:

4
As shown on the dialog box, the user can choose to analyze annual trends for either high‐flow
or low‐flow conditions for defined numbers of days (N‐day). The default N‐day options can be
modified by adding (+ button) or subtracting (‐ button) values in the entry box. For this analysis,
we’ll determine trends in high flows for all of the default days, and modify the Year/Season
Boundaries to October 1 through September 30 for consistency with the two time series:

5
After hitting ‘Trend List’ the following results display the Kendall Tau calculations, including the
Kendall Tau statistic (‘KENTAU’), p‐level (‘KENPLV’), and median slope (‘KENSLPL’) for the
streamflow and base‐flow time series:

6
As shown by the results, there are statistically significant declines in streamflow and base flow
for all of the N‐day periods, which is consistent with the findings of Wahl and Tortorelli (1997).
The 365‐day (‘H365’) results calculated by the program for streamflow are nearly identical to
those given in Wahl and Tortorelli (1997, p. 28); the results for base flow are slightly different
from those presented in Wahl and Tortorelli (1997), most likely due to the modification made
to the BFI algorithm for implementation in the GW Toolbox (see USGS TM 3‐B10, p. 3).

Statistically significant declines in streamflow and base flow also are evident for the low‐flow N‐
day calculations (see below). Also note that the trend calculations for the 365‐day low flows
(‘L365’) are equivalent to those for the 365‐day high flows because of the use of the same flow
statistics for both computations.

One final note concerning trends functionality in the Toolbox: The “Data Tree” option under
“Analysis” includes 1‐day high‐flow (H001) and 7‐day low‐flow (L007) calculations for the
selected time series, as illustrated for the streamflow data for the Beaver River streamgage:

7
The 1‐day high‐flow values are exactly equal to those calculated by the “Trends” functionality
because the annual analysis period is the same (October 1 through September 30). Note,
however, that the “Data Tree” calculations for the 7‐day low‐flow values use an annual date
range of April 1 through March 31, which results in different calculations for the 7‐day low‐flow
values from those made with the “Trends” functionality above.

“Subset, Split, or Filter Selected Data” Functionality

A new functionality has been developed to subset, split, or filter time‐series data that have
been selected for graphing or analysis. The functionality is implemented by checking the
subsetting option in the ‘Select Data’ dialog boxes, such as for graphing:

8
which results in the following dialog box:

A few examples are provided to illustrate how the functionality can be used.

In the first example, streamflow data for the Beaver River are subsetted by a particular month,
in this case, October, for graphing. This is done in the ‘Subset By Date’ tab, in which the user
checks the ‘Apply month/day range to each year’ button and modifies the data ranges to
include only days in October:

After then choosing the ‘Time Series’ and ‘Flow/Duration’ graphing options, the following two
plots are generated:

9
10
The next example illustrates how the streamflow data can be split into 5‐year increments to
evaluate how flow‐duration conditions changed at the streamgage over the period January 1,
1938, through December 31, 1992. First, after selecting the ‘Subset, Split, or Filter Selected
Data’ option within the graphing dialog box, the user specifies the date range of the analysis
within the ‘Subset By Date’ tab:

and then selects the ‘Seasons’ tab, which results in the following dialog box:

11
For this analysis, the user checks the ‘Split Into Seasons’ button, selects ‘Calendar Year’ under
‘Type of seasons to split into:’, highlights the years of interest in blue (which in this case have
been pre‐loaded for 1938‐1992), checks the ‘Separate time series for each group of __ seasons’
button, and then specifies ‘5’ for the number of seasons:

12
The resulting ‘Time Series’ and ‘Flow/Duration’ graphs for these selection criteria are:

13
The flow‐duration curves for the 5‐year intervals clearly demonstrate a decline in streamflow
over the 55‐year period, with those for the earliest time periods showing streamflows greater
than 0.1 cubic feet per second more than approximately 95‐percent of the time and those for
the last two time periods (1983‐1987; 1988‐1992) showing flows below 0.2 cubic feet per
second at the streamgage more than 75 percent of the time.

A third example illustrates a trend analysis of the base‐flow data calculated by the BFI Standard
method for the 6‐month period May‐October each year (a typical irrigation season in parts of
the United States, such as the Central Valley of California), for the period 1938‐1992. After
selecting the “Trend” functionality (“Analysis>USGS Surface Water Statistics
(SWSTAT)>Trend”), the 6‐month period and date range are specified in the “Subset by Date”
tab:

14
Additional information is then specified in the “Seasons” tab:

15
After hitting “Ok,” the user is directed to the “Trend” dialog box. The May 1 through October 31
season boundaries have been pre‐loaded, but the user needs to specify a “Custom” “Start Year”
of 1938 and “End Year” of 1992.

Hitting “Trend List” results in the following high‐flow trends analyses:

As with the previous analyses, there are statistically significant declines in base flow during the
6‐month season May‐October during the period 1938‐1992.

16
Additional details for the analysis can be shown by selecting “List” within the “Analysis” menu
option of the trends dialog box:

17