R/phenology_sequential.R

In fruclimadapt: Evaluation Tools for Assessing Climate Adaptation of Fruit Tree Species

#' Prediction of phenological stages using a sequential model
#'
#'
#' The function predicts phenological phases for a climate series
#' from daily chill and heat requirements and daily chill and forcing heat 
#' data. The sequential model used in the function considers that chilling 
#' and heat have independent effects. It consists of an accumulation of chill 
#' up to the plant requirement, followed by heat up to forcing requirement, 
#' with no overlap between both phases. The function is independent of the
#' method used to calculate chill and forcing heat, so that chill can
#' be supplied as chill hours, chill units or chill portions (recommended, 
#' particularly for warm climates or in climate change studies), forcing heat 
#' accumulation can be supplied either as GDD or GDH. The function allows
#' predicting several stages (or the same for different cultivars), by supplying
#' a dataframe in which each row contains chill and heat requirements for a
#' phenological stage.
#'
#' @param GDH_day a dataframe with daily chilling and forcing accumulation.
#' It must contain the columns Year, Month, Day, DOY, Chill, GD.
#' @param Reqs a dataframe in which each row contains the chilling and forcing
#' heat requirements of one phenological stage. It must contain the columns 
#' Creq (for chilling) and Freq (for forcing heat).
#' @param Start_chill parameter indicating the day of the year when chill
#' accumulation is supposed to start.
#' @return dataframe with the predicted dates of chilling requirement fulfillment
#' and date of occurrence of each phenological stage defined in Reqs. Columns are
#' Creq and Freq (chilling and forcing heat requirements for the phenological stage),
#' Season, Creq_Year and Creq_DOY (year and day of the year in which chill 
#' requirements are fulfilled), Freq_Year and Freq_DOY (year and day of the year
#' of occurrence the phenological stage). 
#' 
#' @author Carlos Miranda, \email{carlos.miranda@@unavarra.es}
#' @examples
#' data(Tudela_DW)
#' data(Bigtop_reqs)
#' library(magrittr)
#' library(dplyr)
#' library(lubridate)
#' # Select the first two seasons in Tudela_DW example dataset
#' Tudela_Sel <- Tudela_DW %>% 
#'        filter (Tudela_DW$Year<=2002)
#' # Generate hourly temperatures from the example dataset
#' Tudela_HT <- hourly_temps(Tudela_Sel,42.13132)
#' # Calculate chill as chill portions, starting on DOY 305
#' Chill <- chill_portions(Tudela_HT,305)
#' # Calculate forcing heat as growing degree hours (GDH) with the linear model,
#' # using base temperature 4.7 C and no upper thresholds
#' GDH <- GDH_linear(Tudela_HT,4.7,999,999)
#' # Combine Chill and GDH values in a dataframe with a format compatible with
#' # the function phenology_sequential
#' Tudela_CH <- merge(Chill,GDH) %>%
#'   select(Date, Year, Month, Day, DOY, Chill,GDH) %>%
#'     rename(GD=GDH)
#' # Obtain the predicted dates using the example dataset with requirements
#' Phenology_BT <- phenology_sequential(Tudela_CH, Bigtop_reqs, 305)
#' 
#' @export phenology_sequential
#' @import magrittr dplyr 
#' @importFrom lubridate make_date

phenology_sequential <- function(GDH_day,Reqs,Start_chill){
  Seasons <- unique(GDH_day$Year)
  Phendates_cols <- c("Creq","Freq","Season","Creq_Year","Creq_DOY","Freq_Year","Freq_DOY")
  Phendates_pred <-data.frame(matrix(ncol=7, nrow=0, byrow=FALSE)) 
  colnames(Phendates_pred) <- Phendates_cols
  Phendates_pred<-Phendates_pred%>% mutate(Season = as.character(Season))
  for(i in 1:nrow(Reqs)) {
    Sel_reqs <- slice(Reqs, i)
    Creq <- as.numeric(Sel_reqs[1])
    Freq <- as.numeric(Sel_reqs[2])
    for(sea in 1:(length(Seasons)-1)){
      Anno = as.numeric(Seasons[sea])+1
      if (Start_chill>=305){
        Data_fil <- GDH_day %>% filter(
          (GDH_day$Year == Anno-1 & GDH_day$DOY>= Start_chill) |
            (GDH_day$Year == Anno))
      } else {
        Data_fil <- GDH_day %>% filter(
          (GDH_day$Year == Anno & GDH_day$DOY>= Start_chill))
      }
      Data_fil <- arrange(Data_fil, Data_fil$Year, Data_fil$DOY) %>%
        filter(Chill>=Creq)
      Data_fil2 <- Data_fil %>%
        mutate(GDacu=cumsum(GD)) %>%
        select("GDacu") %>%
        unlist(use.names=FALSE)
      Dop <-which(Data_fil2>=Freq)[1]
      Sel_Creq <- select(slice(Data_fil, 1), Year, DOY) %>% rename(Creq_Year=Year, Creq_DOY=DOY)
      Sel_Freq <- select(slice(Data_fil, Dop), Year, DOY) %>% rename(Freq_Year=Year, Freq_DOY=DOY)
      if (nrow(Sel_Freq)==0){
        newRow <- data.frame(Freq_Year= Anno, Freq_DOY = -99)
        Sel_Freq <- rbind(Sel_Freq , newRow)
      }
      Sel_case <- bind_cols(Sel_reqs, Sel_Creq, Sel_Freq) %>%
        mutate(Season=paste(Anno-1,"-",Anno)) %>%
        select(Creq, Freq, Season, everything())
      Phendates_pred <- bind_rows(Phendates_pred, Sel_case)
    }
  }
  return(Phendates_pred)
}