Performs a flow volume frequency analysis on annual statistics from a daily streamflow data set. Defaults to a low flow frequency analysis using annual minimums. Set use_max = TRUE for annual high flow frequency analyses. Calculates statistics from all values, unless specified. Function will calculate using all values in 'Values' column (no grouped analysis). Analysis methodology replicates that from HEC-SSP. Returns a list of tibbles and plots.

compute_annual_frequencies(
  data,
  dates = Date,
  values = Value,
  station_number,
  roll_days = c(1, 3, 7, 30),
  roll_align = "right",
  use_max = FALSE,
  use_log = FALSE,
  prob_plot_position = c("weibull", "median", "hazen"),
  prob_scale_points = c(0.9999, 0.999, 0.99, 0.9, 0.5, 0.2, 0.1, 0.02, 0.01, 0.001,
    1e-04),
  fit_distr = c("PIII", "weibull"),
  fit_distr_method = ifelse(fit_distr == "PIII", "MOM", "MLE"),
  fit_quantiles = c(0.975, 0.99, 0.98, 0.95, 0.9, 0.8, 0.5, 0.2, 0.1, 0.05, 0.01),
  plot_curve = TRUE,
  water_year_start = 1,
  start_year,
  end_year,
  exclude_years,
  months = 1:12,
  complete_years = FALSE,
  ignore_missing = FALSE,
  allowed_missing = ifelse(ignore_missing, 100, 0)
)

Arguments

data

A data frame of daily data that contains columns of dates and flow values. Groupings and the groups argument are not used for this function (i.e. station numbers). Leave blank or set to NULL if using station_number argument.

dates

Name of column in data that contains dates formatted YYYY-MM-DD. Only required if dates column name is not 'Date' (default). Leave blank or set to NULL if using station_number argument.

values

Name of column in data that contains numeric flow values, in units of cubic metres per second. Only required if values column name is not 'Value' (default). Leave blank if using station_number argument.

station_number

Character string vector of seven digit Water Survey of Canada station numbers (e.g. "08NM116") of which to extract daily streamflow data from a HYDAT database. Requires tidyhydat package and a HYDAT database. Leave blank if using data argument.

roll_days

Numeric value of the number of days to apply a rolling mean. Default 1.

roll_align

Character string identifying the direction of the rolling mean from the specified date, either by the first ('left'), last ('right'), or middle ('center') day of the rolling n-day group of observations. Default 'right'.

use_max

Logical value to indicate using maximums rather than the minimums for analysis. Default FALSE.

use_log

Logical value to indicate log-scale transforming of flow data before analysis. Default FALSE.

prob_plot_position

Character string indicating the plotting positions used in the frequency plots, one of 'weibull', 'median', or 'hazen'. Points are plotted against (i-a)/(n+1-a-b) where i is the rank of the value; n is the sample size and a and b are defined as: (a=0, b=0) for Weibull plotting positions; (a=.2; b=.3) for Median plotting positions; and (a=.5; b=.5) for Hazen plotting positions. Default 'weibull'.

prob_scale_points

Numeric vector of probabilities to be plotted along the X axis in the frequency plot. Inverse of return period. Default c(.9999, .999, .99, .9, .5, .2, .1, .02, .01, .001, .0001).

fit_distr

Character string identifying the distribution to fit annual data, one of 'PIII' (Log Pearson Type III) or 'weibull' (Weibull) distributions. Default 'PIII'.

fit_distr_method

Character string identifying the method used to fit the distribution, one of 'MOM' (method of moments) or 'MLE' (maximum likelihood estimation). Selected as 'MOM' if fit_distr ='PIII' (default) or 'MLE' if fit_distr = 'weibull'.

fit_quantiles

Numeric vector of quantiles to be estimated from the fitted distribution. Default c(.975, .99, .98, .95, .90, .80, .50, .20, .10, .05, .01).

plot_curve

Logical value to indicate plotting the computed curve on the probability plot. Default TRUE.

water_year_start

Numeric value indicating the month (1 through 12) of the start of water year for analysis. Default 1.

start_year

Numeric value of the first year to consider for analysis. Leave blank or set well before start date (i.e. 1800) to use from the first year of the source data.

end_year

Numeric value of the last year to consider for analysis. Leave blank or set well after end date (i.e. 2100) to use up to the last year of the source data.

exclude_years

Numeric vector of years to exclude from analysis. Leave blank or set to NULL to include all years.

months

Numeric vector of months to include in analysis. For example, 3 for March, 6:8 for Jun-Aug or c(10:12,1) for first four months (Oct-Jan) when water_year_start = 10 (Oct). Default summarizes all months (1:12).

complete_years

Logical values indicating whether to include only years with complete data in analysis. Default FALSE.

ignore_missing

Logical value indicating whether dates with missing values should be included in the calculation. If TRUE then a statistic will be calculated regardless of missing dates. If FALSE then only those statistics from time periods with no missing dates will be returned. Default FALSE.

allowed_missing

Numeric value between 0 and 100 indicating the percentage of missing dates allowed to be included to calculate a statistic (0 to 100 percent). If 'ignore_missing = FALSE' then it defaults to 0 (zero missing dates allowed), if 'ignore_missing = TRUE' then it defaults to 100 (any missing dates allowed); consistent with ignore_missing usage. Supersedes ignore_missing when used.

Value

A list with the following elements:

Freq_Analysis_Data

Data frame with computed annual summary statistics used in analysis.

Freq_Plot_Data

Data frame with co-ordinates used in frequency plot.

Freq_Plot

ggplot2 object with frequency plot.

Freq_Fitting

List of fitted objects from fitdistrplus.

Freq_Fitted_Quantiles

Data frame with fitted quantiles.

Examples

if (FALSE) {

# Working examples (see arguments for further analysis options):

# Compute an annual frequency analysis using default arguments
results <- compute_annual_frequencies(station_number = "08NM116",
                                      start_year = 1980,
                                      end_year = 2010)
                           
# Compute an annual frequency analysis using default arguments (as listed)
results <- compute_annual_frequencies(station_number = "08NM116",
                                      roll_days = c(1,3,7,30),
                                      start_year = 1980,
                                      end_year = 2010,
                                      prob_plot_position = "weibull",
                                      prob_scale_points = c(.9999, .999, .99, .9, .5, 
                                      .2, .1, .02, .01, .001, .0001),
                                      fit_distr = "PIII",
                                      fit_distr_method = "MOM")
                                      
# Compute a 7-day annual frequency analysis with "median" plotting positions
# and fitting the data to a weibull distribution (not default PIII)
results <- compute_annual_frequencies(station_number = "08NM116",
                                      roll_days = 7,
                                      start_year = 1980,
                                      end_year = 2010,
                                      prob_plot_position = "median",
                                      fit_distr = "weibull")
               
}