In mundl/smires: Science and Management of Intermittent Rivers and Ephemeral Streams

generated with r R.version.string and smires version r packageVersion("smires")

knitr::opts_chunk$set(echo = TRUE)

Most IRES metrics are constructed in a similar way.

1. Derive a binary time series from (daily) discharge observations.
2. Aggregate on an annual basis.
3. Aggregate every year to obtain a single metric per state.

These steps are explained in detail for the mean annual number of no-flow days. First of all we need to load the R-package smires, assuming it is already installed on our machine.

library(smires)

Derive a binary time series

Given our daily discharge time series $Q$ we derive a binary time series of flow and no-flow states with respect to a threshold $Q_0$. This transforms our initial time series

ampneyBrook

with r nrow(ampneyBrook) days of observations (rows) of a single continuous variable (column discharge can contain values from 0 to Inf) into a binary time series (column state is a factor with levels flow and no-flow)

binary <- char_binary(ampneyBrook, major = 91)
head(binary, 13)

with r nrow(binary) rows, r max(binary$spell) spells/events (rows) and r ncol(binary) -1 variables (columns). Most of these newly created variables like major, minor and group are used for grouping. To compute the mean annual number of no-flow days our main variable of interest is of course duration.

Aggregate on an annual basis

As there can be spells spanning multiple years (e.g. spell no. 3) or several spells/events per year (e.g. spell no. 4 & 6 in year 1989), we need to aggregate the durations on an annual basis for all combinations of year (variable major) and state. To do so, in general we have to supply three arguments to the function smires():

• the variable we want to aggregate (in this case duration is implicitly used),
• the interval we want to aggregate over (here hydrological years, major = 91) and
• the function used to summarize all the durations (fun_major = sum).

annual <- char_binary(ampneyBrook, major = 91, fun_major = sum)
annual

We receive the total spell durations for every year and flow state. This table has to be twice as long as there are years in our record, with zeros imputed when the corresponding spell was missing in a given year.

Aggregate every year

The mean annual number of flow/no-flow days is obtained by aggregating once more, this time only over the variable state. Here the mean() is used as the aggregation function and is specified via fun_total = mean.

totals <- char_binary(ampneyBrook, major = 91, fun_major = sum, fun_total = mean)
totals

This yields us the approximately r round(totals$duration[totals$state == "no-flow"], 1) days for the mean annual number of no-flow days and on average around r round(totals$duration[totals$state == "flow"], 1) flow days per years.

Summary

In order to facilitate the computation of every possible metric the aggregation intervals and the aggregation functions can be supplied as arguments to the function smries().

binary <- char_binary(ampneyBrook, major = 91)
annual <- char_binary(ampneyBrook, major = 91, fun_major = sum)
totals <- char_binary(ampneyBrook, major = 91, fun_major = sum, fun_total = mean)

Measures of locations are computed if we choose summarizing functions like mean(), median(), quantile() or similar. For measures of variability one could choose for example sd(), var(), IQR(), the coefficient of variation cv() or the mean absolute deviation mad(). In case we are interested in the distribution of our annual statistics, we simply omit the final aggregation function.

For a few common metrics wrapper functions exist. Take for example the function MAN() which is just a tiny wrapper around smires() with the appropriate aggregation functions along with a few options controlling how the result is printed.

MAN

Applying this wrapper function to the same data set as before yields the same value for the mean annual number of no-flow days.

MAN(ampneyBrook, major = 91)

mundl/smires documentation built on May 23, 2019, 8:22 a.m.