02 - Delayedflow Analysis in R tidyverse • delayedflow

This is a quick example how to use the delayedflow package in a R tidyverse environment:

here we using tidyverse syntax
the 4 catchments coming from different streamflow regimes in Switzerland (rain, hybrid, snow)

#devtools::install_github(modche/delayedflow)
library(delayedflow) #load package

Load data

df <- read_tsv('https://raw.githubusercontent.com/modche/science_contr/master/4ezg.txt', col_types = cols())

CDC: Calculation and plotting

First of all, CDCs for all catchments are generated. For this a delayed flow separation is performed for all block length between 1 and 180. The plot highlights the differences of the CDCs of four catchments.

df %>% 
    group_by(filename) %>% 
    summarise(cdc = dfi_n(q = q, n = 1:180)$dfi, .groups = "keep", n = 0:180 ) %>% 
    ungroup() %>% 
    ggplot(aes(x=n, y = cdc)) +
        geom_line(aes(colour = filename)) +
        scale_x_continuous(breaks = c(0:6 * 30)) +
        theme_bw(14) +
        labs(title = "",
             subtitle = "",
             caption = "", 
             colour = "CDCs",
             x = "Block length N (days)", y = "DFI (-)")+
        theme(legend.position = c(0.80, 0.70))

Breakpoint estimates

The breakpoints are the heart of the DFI analysis. For n breakpoints n+1 delayed (relative) contributions (sum = 1) can be estimated.

df2 <- df %>% 
    group_by(filename) %>% 
    summarise(cdc = dfi_n(q = q, n = 1:180)$dfi, .groups = "keep", n = 0:180 ) %>% 
    ungroup()

df2

## # A tibble: 724 × 3
##    filename     cdc     n
##    <chr>      <dbl> <int>
##  1 catchment1 1         0
##  2 catchment1 0.953     1
##  3 catchment1 0.895     2
##  4 catchment1 0.853     3
##  5 catchment1 0.816     4
##  6 catchment1 0.791     5
##  7 catchment1 0.764     6
##  8 catchment1 0.747     7
##  9 catchment1 0.727     8
## 10 catchment1 0.718     9
## # … with 714 more rows

bp_df <- df2 %>% 
        group_by(filename) %>% 
        summarize(bps = find_bps(cdc, n_bp = 3, bp_max = 65)$bps_position) %>% 
        mutate(id = names(bps)) %>% 
        ungroup() %>% 
        pivot_wider(names_from = "id", values_from = "bps")

## Calculating breakpoints...Done. 
## 
## Calculating breakpoints...Done. 
## 
## Calculating breakpoints...Done. 
## 
## Calculating breakpoints...Done.

Let’s have a look on the different breakpoints (in days) across the catchments.

bp_df

## # A tibble: 4 × 4
##   filename    bp_1  bp_2  bp_3
##   <chr>      <int> <int> <int>
## 1 catchment1     6    38    60
## 2 catchment2     2    11    47
## 3 catchment3     3    16    61
## 4 catchment4     2     8    30

Relative streamflow contributions

## Calculating breakpoints...Done. 
## 
## Calculating breakpoints...Done. 
## 
## Calculating breakpoints...Done. 
## 
## Calculating breakpoints...Done.

## # A tibble: 16 × 3
##    filename   rel_contr class  
##    <chr>          <dbl> <chr>  
##  1 catchment1    0.236  contr_1
##  2 catchment1    0.506  contr_2
##  3 catchment1    0.0736 contr_3
##  4 catchment1    0.184  contr_4
##  5 catchment2    0.184  contr_1
##  6 catchment2    0.134  contr_2
##  7 catchment2    0.157  contr_3
##  8 catchment2    0.525  contr_4
##  9 catchment3    0.488  contr_1
## 10 catchment3    0.233  contr_2
## 11 catchment3    0.148  contr_3
## 12 catchment3    0.130  contr_4
## 13 catchment4    0.566  contr_1
## 14 catchment4    0.213  contr_2
## 15 catchment4    0.114  contr_3
## 16 catchment4    0.107  contr_4

Plot relative contributions

Each contr can vary beetween 0 and 1. All contrs together add up to 1.

ggplot(data = contr_df) + 
    geom_col(aes(x = rel_contr, y = filename, fill = class), 
                    width = 0.50, 
                    colour = "black") +
    scale_fill_brewer(palette = "RdBu", 
                      direction = 1, 
                      guide = guide_legend(reverse = FALSE),
                      name = "Delay class") +
    scale_x_continuous(labels = scales::percent_format()) +
    theme_bw(14) +
    theme(legend.position = "bottom") +
    labs(title = "Delayed flow contributions",
     subtitle = "",
     caption = "", 
     x = "\nRelative contribution to streamflow", y = "")

Interpretation of the results

In this example Catchment #1 has a high proportion of intermediate contributions (i.e. here snow melt from a high alpine catchment). Catchment #2 has > 50% baseline contributions that means it is highly groundwater-dominated. Comparing Catchment #4 vs. Catchment #3 the first one has more short-delayed contributions also visible in the rapid drop in the CDC curve above.

02 - Delayedflow Analysis in R tidyverse