R/alignment_functions.R

In afsc-gap-products/gapctd: Process CTD data from bottom trawl surveys

#' Correlation between two channels (R workflow)
#' 
#' Calculate correlation coefficient between two channels (e.g. temperature and conductivity)
#' 
#' @param x oce object
#' @param exclude_flag Logical denoting whether bad scans (flag < 0) should be excluded.
#' @param c1 Name of first channel as a character vector. Default = "temperature"
#' @param c2 Name of second channel as a character vector. Default = "conductivity"
#' @param min_pressure Minimum pressure to use for correlations
#' @param cor_method Correlation method to use as a character vector (see ?cor)
#' @return Correlation coefficient (1L numeric)
#' @export
#' @author Sean Rohan

channel_correlation <- function(x, 
                                exclude_flag = TRUE, 
                                c1 = "temperature", 
                                c2 = "conductivity", 
                                min_pressure = 4, 
                                cor_method = "pearson") {
  
  if(exclude_flag) {
    T_vec <- x@data[[c1]][x@data$flag >= 0 & x@data$pressure >= min_pressure]
    C_vec <- x@data[[c2]][x@data$flag >= 0 & x@data$pressure >= min_pressure]
  }
  
  if(!exclude_flag) {
    T_vec <- x@data[[c1]][x@data$pressure >= min_pressure]
    C_vec <- x@data[[c2]][x@data$pressure >= min_pressure]
  }
  
  cor_coef <- try(cor(T_vec,
                      C_vec,
                      use = "complete.obs",
                      method = cor_method), silent = TRUE)
  
  if(class(cor_coef) == "try-error") {
    cor_coef <- 0
  }
  
  return(cor_coef)
}


#' Estimate channel alignment parameter (R workflow)
#'
#' Estimate optimal alignment parameter for a channel (e.g. "temperature") within a range of candidate alignment offsets. The 'best' alignment is taken to be the one that maximizes the correlation between channels.
#' 
#' @param x oce object
#' @param cast_direction Cast direction as a character vector ("downcast" or "upcast"). Passed to gapctd::slowdown()
#' @param variable Variable to be aligned ("temperature")
#' @param offsets Numerical vector of offsets to evaluate, in seconds.
#' @param cor_method Correlation method, passed to gapctd::channel_correlation().
#' @param ... Additional arguments passed to gapctd::channel_correlation()
#' @return A named numeric vector containing the estimated optimal offset ("best_offset") and correlation for the offset ("best_corr")
#' @export
#' @author Sean Rohan

optim_align_par <- function(x, cast_direction, variable = "temperature", offsets = seq(-1,1, 0.01), cor_method = "pearson", ...) {

  offset_corr <- numeric(length = length(offsets))
  
  out <- list()
  
  for(ii in 1:length(offsets)) {
    
    if(variable == "temperature") {
      offset_corr[ii] <- x |>
        gapctd:::align_var(variables = variable, offset = offsets[ii], interp_method = "linear") |>
        # gapctd:::slowdown(min_speed = 0.1, window = 5, cast_direction = cast_direction) |>
        gapctd:::channel_correlation(exclude_flag = TRUE, 
                                     min_pressure = 1.01, 
                                     cor_method = cor_method, 
                                     c1 = "temperature", 
                                     c2 = "conductivity")
    }
    
    if(variable == "oxygen") {
      
    }

  }
  
  out[[variable]] <- c(offset = offsets[which.max(offset_corr)[1]],
           corr = offset_corr[which.max(offset_corr)[1]])
  return(out)
  
}


#' Calculate alignment correlation for fixed parameters
#' 
#' @param x oce object
#' @param align_pars A list object with alignment parameters for a variable, e.g., list(temperature = -0.5)
#' @param cor_var A vector of variable names that channels being aligned should be compared with (e.g., corr_var = "conductivity")
#' @param cor_method Correlation method, passed to gapctd::channel_correlation().
#' @param cast_direction Cast direction as a character vector ("downcast" or "upcast"). Passed to gapctd::slowdown()
#' @export
#' @author Sean Rohan

fixed_alignment <- function(x, align_pars, cor_var = "conductivity", cor_method = "pearson", cast_direction) {
  align_list <- list()
  for(jj in 1:length(align_pars)) {
    
    offset <- align_pars[[names(align_pars[jj])]]
    
    channel_corr <- x |>
      gapctd:::align_var(variables = names(align_pars[jj]), 
                         offset = offset, 
                         interp_method = "linear") |>
      gapctd:::slowdown(min_speed = 0.1, window = 5, cast_direction = cast_direction) |>
      gapctd:::channel_correlation(exclude_flag = TRUE, 
                                   min_pressure = 4, 
                                   cor_method = cor_method, 
                                   c1 = names(align_pars[jj]), 
                                   c2 = cor_var[jj])
    
    align_list[[names(align_pars[jj])]] <- c("offset" = offset, "corr" = channel_corr)
  }
  
  return(align_list)
}