2. Advanced hydrograph separation

Introduction

The core of the grwat package is the advanced hydrograph separation method by Rets et al. (2022), which utilizes supplementary temperature and precipitation data to classify quickflow events by their genesis. The genetic types are rain, thaw and spring. The latter is a specific kind of thaw flood (sometimes called a freshet) that appears in cold climates at the beginning of the spring during the massive snow melting in a watershed. It is not uncommon that a spring flood is significantly larger than others during the year, and produces the major part of the annual runoff volume.

Throughout grwat package documentation a sample dataset spas containing the daily runoff data for Spas-Zagorye gauge on Protva river in Central European plane is used. The dataset is supplemented by meteorological variables (temperature and precipitation) obtained from CIRES-DOE (1880-1949) and ERA5 (1950-2021) data averaged inside gauge’s basin:

library(grwat)
data(spas) # example Spas-Zagorye data is included with grwat package
head(spas)
#> # A tibble: 6 × 4
#>   Date           Q   Temp  Prec
#>   <date>     <dbl>  <dbl> <dbl>
#> 1 1956-01-01  5.18  -6.46 0.453
#> 2 1956-01-02  5.18 -11.4  0.825
#> 3 1956-01-03  5.44 -10.7  0.26 
#> 4 1956-01-04  5.44  -8.05 0.397
#> 5 1956-01-05  5.44 -11.7  0.102
#> 6 1956-01-06  5.58 -20.1  0.032

This 4-column representation of the date, runoff, temperature and precipitation is a standardized form of a data frame required for advanced separation

Separation basics

Advanced separation is a formal analytical procedure that is performed in several stages (Rets et al. 2022):

A. Identify the start dates of water resources years. The start date is the first day of the spring flood.

B. For each water-resources year:

  1. Separate quickflow and baseflow. Identify the end date of the spring flood as the first day after the start without quickflow.
  2. Calculate the long periods of significant precipitation and low temperatures.
  3. Identify second-order rain floods during the spring flood.
  4. Calculate the first day of winter season.
  5. Attribute all quickflow between the last day of spring flood and the first day of winter season as the rain flow.
  6. Attribute all quickflow between the first day of winter and the last day of water-resources year as the thaw flow.

This algorithm is executed by gr_separate() function, which requires a 4-column data frame as specified earlier, and the list of parameters. The number parameters is quite big (\(39\) for the current version of the package), and the parameters depend on the regional climate and the size of the river basin. Currently the recommended parameters are available for some regions in the center of the East European Plane. You can use them as the starting point for experimentation. The regions are:

  1. northwest
  2. center
  3. south
  4. northeast
  5. volga
  6. oka
  7. southeast

To ease the exchange of the parameters, they are organized as list, as returned by gr_get_params(). You can pass either the number of region, or its name in reg argument:

params = gr_get_params(reg = 'south')
head(params)
#> $winmon
#> [1] 11
#> 
#> $grad1
#> [1] 2
#> 
#> $grad2
#> [1] 5
#> 
#> $gratio
#> [1] 350
#> 
#> $spmon1
#> [1] 3
#> 
#> $spmon2
#> [1] 5

params = gr_get_params(reg = 2)
head(params)
#> $winmon
#> [1] 11
#> 
#> $grad1
#> [1] 1.7
#> 
#> $grad2
#> [1] 5
#> 
#> $gratio
#> [1] 400
#> 
#> $spmon1
#> [1] 2
#> 
#> $spmon2
#> [1] 5

To ease the understanding of the parameters, grwat contains the helper gr_help_params() function, which describes the meaning of each:

gr_help_params()
#> # A tibble: 39 × 12
#>        N name_old name   example desc    units formula comments  pics  desc_rus 
#>    <dbl> <chr>    <chr>  <chr>   <chr>   <chr> <chr>   <chr>     <lgl> <chr>    
#>  1     1 mome     winmon 11      the fi… month  <NA>   "not use… NA    месяц, с…
#>  2     2 grad     grad1  1.5     rate o… %      <NA>   "if chan… NA    интенсив…
#>  3     3 grad1    grad2  2       rate o… %     "\"\""  "\"\""    NA    интенсив…
#>  4     4 kdQgr1   gratio 150     maximu… %      <NA>   "can be … NA    максимал…
#>  5     5 polmon1  spmon1 2       the ea… month  <NA>   "can be … NA    cамый ра…
#>  6     6 polmon2  spmon2 5       the la… month  <NA>   "can be … NA    самый по…
#>  7     7 polkol1  sptri… 15      amount… days   <NA>    <NA>     NA    количест…
#>  8     8 polkol2  spris… 25      amount… days   <NA>    <NA>     NA    количест…
#>  9     9 polkol3  spdays 30      amount… days   <NA>   "can be … NA    количест…
#> 10    10 polgrad1 sprise 10      mean r… %      <NA>    <NA>     NA    значения…
#> # … with 29 more rows, and 2 more variables: role_1 <chr>, role_2 <chr>

You can tweak the parameters just by changing their values in the list:

params$sprise = 12
params$gratio = 500

It is quite hard to predict how effective will be parameters for the particular basin. Therefore, the search for optimal values is experimental work. After a preliminary parameters are set, you can separate the hydrograph by gr_separate():

# separate
sep = gr_separate(spas, params)
#> grwat: data frame is correct
#> grwat: parameters list and types are OK
head(sep)
#> # A tibble: 6 × 11
#>   Date           Q   Temp  Prec Qbase Quick Qspri Qrain Qthaw Season  Year
#>   <date>     <dbl>  <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>  <int> <int>
#> 1 1956-01-01  5.18  -6.46 0.453    NA    NA    NA    NA    NA     NA    NA
#> 2 1956-01-02  5.18 -11.4  0.825    NA    NA    NA    NA    NA     NA    NA
#> 3 1956-01-03  5.44 -10.7  0.26     NA    NA    NA    NA    NA     NA    NA
#> 4 1956-01-04  5.44  -8.05 0.397    NA    NA    NA    NA    NA     NA    NA
#> 5 1956-01-05  5.44 -11.7  0.102    NA    NA    NA    NA    NA     NA    NA
#> 6 1956-01-06  5.58 -20.1  0.032    NA    NA    NA    NA    NA     NA    NA

Resulting data frame is enriched with information about different kinds of flow. To evaluate the results, you can use separation plots provided by gr_plot_sep():

# One year
gr_plot_sep(sep, 1978) 


# Two years
gr_plot_sep(sep, c(1978, 2014)) 


# Four years in a matrix layout
gr_plot_sep(sep, 1988:1991, layout = matrix(1:4, nrow = 2, byrow = TRUE)) 
#> Warning: Removed 1 rows containing missing values (position_stack).
#> Removed 1 rows containing missing values (position_stack).

Tweaking of the parameters

Sometimes global parameters do not work, and you need to tweak the values for selected years. For this you use a list of parameter lists instead of the one list. In such case the number of parameter lists must be equal to the number of water-resources years in a runoff data. Instead of constructing such list manually, you can use gr_separate() in debug = TRUE mode. If this mode is activated, grwat will return two additional attributes:

The attributes are extracted via a base R function attributes():

# Debug mode gives access to additional information
sep_debug = gr_separate(spas, 
                        params = gr_get_params(reg = 'center'), 
                        debug = TRUE)
#> grwat: data frame is correct
#> grwat: parameters list and types are OK
#> Warning in gr_separate(spas, params = gr_get_params(reg = "center"), debug =
#> TRUE): grwat: 1974 years were not separated. Check the input data for possible
#> errors. Use gr_get_gaps() and gr_fill_gaps() functions to detect and fill
#> missing data.
#> Warning in gr_separate(spas, params = gr_get_params(reg = "center"), debug =
#> TRUE): grwat: 2002, 2014, 2019 years were processed with jittered parameters

# a vector of years with jittered params
jit = attributes(sep_debug)$jittered
print(jit)
#> [1] 2002 2014 2019

# actual params used for each year
parlist = attributes(sep_debug)$params
partab = do.call(dplyr::bind_rows, parlist) # View as table
head(partab)
#> # A tibble: 6 × 39
#>   winmon grad1 grad2 gratio spmon1 spmon2 sprisedays1 sprisedays2 spdays sprise
#>    <dbl> <dbl> <dbl>  <dbl>  <dbl>  <dbl>       <dbl>       <dbl>  <dbl>  <dbl>
#> 1     11   1.7     5    400      2      5           8          10     30     10
#> 2     11   1.7     5    400      2      5           8          10     30     10
#> 3     11   1.7     5    400      2      5           8          10     30     10
#> 4     11   1.7     5    400      2      5           8          10     30     10
#> 5     11   1.7     5    400      2      5           8          10     30     10
#> 6     11   1.7     5    400      2      5           8          10     30     10
#> # … with 29 more variables: spratio <dbl>, sprecdays <dbl>, spcomp <dbl>,
#> #   precdays <dbl>, frostdays <dbl>, windays <dbl>, floodprec <dbl>,
#> #   floodtemp <dbl>, frosttemp <dbl>, wintemp <dbl>, signratio1 <dbl>,
#> #   signratio2 <dbl>, floodratio <dbl>, gaplen <dbl>, snowtemp <dbl>,
#> #   gradabs <dbl>, mntmode <dbl>, mntgrad <dbl>, mntavgdays <dbl>,
#> #   mntratiodays <dbl>, mntratio <dbl>, niter <dbl>, a <dbl>, k <dbl>, C <dbl>,
#> #   aq <dbl>, padding <dbl>, passes <dbl>, filter <chr>

After the list of parameters is extracted, any of those can be referenced by the character string of the water-resources year. For example, if you want to apply the parameters of one tweaked your globally, the following will work:

# extract and tweak parameters for selected year
p = parlist[['2014']]
p$grad1 = 1
p$grad2 = 2.5

# use tweaked parameters for all years
sep_debug = gr_separate(spas, params = p, debug = TRUE)
#> grwat: data frame is correct
#> grwat: parameters list and types are OK
#> Warning in gr_separate(spas, params = p, debug = TRUE): grwat: 1974 years were
#> not separated. Check the input data for possible errors. Use gr_get_gaps() and
#> gr_fill_gaps() functions to detect and fill missing data.
#> Warning in gr_separate(spas, params = p, debug = TRUE): grwat: 2002, 2019 years
#> were processed with jittered parameters

# Visualize
gr_plot_sep(sep_debug, c(1978, 2014)) 
#> Warning: Removed 3 rows containing missing values (position_stack).

However, the most powerful strategy is to keep the nested list structure and change the parameters individually for different years. If you want to set some parameter for multiple years, then use gr_set_param():

# actual params used for each year
parlist = attributes(sep_debug)$params

# tweak parameters for selected year
parlist[['2014']]$grad1 = 3
parlist[['2014']]$grad2 = 6

# set the sprecdays parameter for multiple years
parlist = gr_set_param(parlist, sprecdays, 
                       years = c(1978, 1999:2015), 
                       value = 15)

# set the spcomp parameter for all years
parlist = gr_set_param(parlist, spcomp, value = 2.5)

# use the list of parameters for separation
sep_debug = gr_separate(spas, params = parlist, debug = TRUE)
#> grwat: data frame is correct
#> grwat: parameters list and types are OK
#> Warning in gr_separate(spas, params = parlist, debug = TRUE): grwat: 1974 years
#> were not separated. Check the input data for possible errors. Use gr_get_gaps()
#> and gr_fill_gaps() functions to detect and fill missing data.

# Visualize
gr_plot_sep(sep_debug, c(1978, 2014))
#> Warning: Removed 2 rows containing missing values (position_stack).

References

Rets, E. P., M. B. Kireeva, T. E. Samsonov, N. N. Ezerova, A. V. Gorbarenko, and N. L. Frolova. 2022. “Algorithm Grwat for Automated Hydrograph Separation by B. I. Kudelins Method: Problems and Perspectives.” Water Resources 49 (1): 23–37.