In smwesten-usgs/autobeale-R: autobeale: code to calculate loads by means of Beale's Ratio Estimator.

library(autobeale)

discharge_file     <- system.file( "extdata", "ROCKFLOW.DAT", package="autobeale")
concentration_file <- system.file( "extdata", "ROCKNO23.DAT", package="autobeale")

discharge_data     <- read.table( discharge_file, header=TRUE )
concentration_data <- read.table( concentration_file, header=TRUE )

discharge_conc_df  <- NULL
discharge_conc_df  <- dplyr::left_join(discharge_data, concentration_data, by=c("date" = "date") )

discharge_conc_df$discharge <- cfs_to_cms( discharge_conc_df$discharge )
names( discharge_conc_df )  <- c("date","discharge_cms","concentration_mg_L")

discharge_conc_df$date <- lubridate::ymd( discharge_conc_df$date )

Example #2: Beale Ratio Estimator with data partitioned into four stratums

In addition to the beale function are a number of R6 classes aimed at making it easier to examine various data partitioning strategies. Strategies may be ranked by their impact on the root mean square of error associated with the load estimate. In this example, a genetic algorithm package is wired up to assist in the search for an optimum partitioning strategy, with optimimum being defined as minimization of the root mean square of error associated with the overall loading estimate, similar to the original AutoBeale.

In order to use these classes, real R date classes (POSIX) must be supplied along with the discharge and concentration data. Another of Hadley Wickham's library packages, lubridate, makes this easy:

library(autobeale)
library(lubridate)
library(mcga)

The discharge_conc_df dataframe now looks like this:

knitr::kable(head(discharge_conc_df, 10))

Use of the support classes strata and partitions

First we creata a strata object of the appropriate size (4 stratums):

number_of_stratums <- 4
mystrata <- strata$new( number_of_stratums, discharge_conc_df )

The strata class creates an underlying partitions object to hold information about the dates dividing the stratums. At this point, the partitions themselves are created but uninitialized. We don't normally have to deal with the partitions directly.

We can create the partitions outside of the strata object as a demonstration. For example, if the date fraction vector has values of `c( 0.25, 0.5, 0.75 ):

mypartitions <- partitions$new( start_date=ymd( "1997-01-01"), end_date=ymd("1997-12-31"), num_partitions=3 )
mypartitions$update_fractions( c( 0.25, 0.5, 0.75 ) )
mydates <- mypartitions$get_date_values()
print( mydates )

Now that we have the strata and partitions objects created and populated, we can rearrange the stratums and calculate loads, and summarize the results. Functions in strata and partitions objects will take care of subsetting the data appropriately and run the Beale ratio estimator on each data subset.

mystrata$rearrange_stratums( mypartitions$get_date_values() )
mystrata$calc_loads()
myresults_df <- mystrata$summarize_results()

knitr::kable( myresults_df, digits=2 )

smwesten-usgs/autobeale-R documentation built on May 30, 2019, 5:04 a.m.