R/ogimet_hourly.R

In bczernecki/climate: Interface to Download Meteorological (and Hydrological) Datasets

#' Scrapping hourly meteorological (Synop) data from the Ogimet webpage
#'
#' Downloading hourly (meteorological) data from the Synop stations available in the https://www.ogimet.com/ repository
#'
#' @param date start and finish of date (e.g., date = c("2018-05-01","2018-07-01") ); By default last 30 days are taken
#' @param coords add geographical coordinates of the station (logical value TRUE or FALSE)
#' @param station WMO ID of meteorological station(s). Character or numeric vector
#' @param precip_split whether to split precipitation fields into 6/12/24h; default: TRUE
#' @param allow_failure logical - whether to proceed or stop on failure. By default set to TRUE (i.e. don't stop on error). 
#' For debugging purposes change to FALSE
#' @importFrom XML readHTMLTable
#' 
#' @export
#' 
#' @keywords internal
#' @returns data.frame with historical meteorological data for hourly time interval
#'
#' @examples 
#' \donttest{
#'   # downloading data for Poznan-Lawica, Poland
#'   poznan = ogimet_hourly(station = 12330, coords = TRUE)
#' }
#'

ogimet_hourly = function(date = c(Sys.Date() - 30, Sys.Date()), 
                         coords = FALSE,
                         station = 12330,
                         precip_split = TRUE, 
                         allow_failure = TRUE) {
  
  if (allow_failure) {
    tryCatch(ogimet_hourly_bp(date = date, coords = coords, 
                              station = station, 
                              precip_split = precip_split),
             error = function(e){
               message(paste("Problems with downloading data.",
                             "Run function with argument allow_failure = FALSE",
                             "to see more details"))})
  } else {
    ogimet_hourly_bp(date = date, coords = coords, 
                     station = station, 
                     precip_split = precip_split)
  }
}

#' @keywords Internal
#' @noRd
ogimet_hourly_bp = function(date = date,
                            coords = coords,
                            station = station,
                            precip_split = precip_split) {
  
  dates = seq.Date(min(as.Date(date)), max(as.Date(date)), by = "1 month") - 1
  dates = unique(c(dates, as.Date(max(date))))
  diff_dates = diff(dates)

  # initalizing empty data frame for storing results:
  data_station =
    data.frame(
      "Date" = character(),
      "hour" = character(),
      "TC" = character(),
      "TdC" = character(),
      "TmaxC" = character(),
      "TminC" = character(),
      "ddd" = character(),
      "ffkmh" = character(),
      "Gustkmh" = character(),
      "P0hPa" = character(),
      "PseahPa" = character(),
      "PTnd" = character(),
      "Precmm" = character(),
      "Nt" = character(),
      "Nh" = character(),
      "HKm" = character(),
      "InsoD1" = character(),
      "Viskm" = character(),
      "Snowcm" = character(),
      "WW" = character(),
      "W1" = character(),
      "W2" = character(),
      stringsAsFactors = FALSE
    )

  for (station_nr in station) {
    message(station_nr)
    # adding progress bar if at least 3 iterations are needed
    if (length(dates)*length(station) >= 3 ) pb = txtProgressBar(min = 0, max = length(dates)*length(station) - 1, style = 3)
    
    for (i in length(dates):1) {
      
      if (length(dates) >= 3 ) paste(setTxtProgressBar(pb, abs(length(dates) * length(station) - i)), "\n")
      
      year = format(dates[i], "%Y")
      month = format(dates[i], "%m")
      day = format(dates[i], "%d")
      ndays = as.numeric(diff_dates[i - 1])
      ndays = ifelse(ndays == 0, 1, ndays)
      ndays = sprintf("%02d", ndays)
      
      linkpl2 = paste("https://www.ogimet.com/cgi-bin/gsynres?ind=",
                      station_nr,
                      "&lang=en&decoded=yes&ndays=",
                      ndays,
                      "&ano=",
                      year,
                      "&mes=",
                      month,
                      "&day=",
                      day,
                      "&hora=23",
                      sep = "")
      if (month == "01") {
        linkpl2 = paste("http://ogimet.com/cgi-bin/gsynres?ind=",
                                       station_nr,
                                       "&lang=en&decoded=yes&ndays=31&ano=",
                                       year,
                                       "&mes=02&day=1&hora=00",
                                       sep = "")
      }
      
      temp = tempfile()
      test_url(linkpl2, temp)
      
      # run only if downloaded file is valid
      if (!is.na(file.size(temp)) & (file.size(temp) > 100)) {
        #a = getURL(linkpl2)
        a = readHTMLTable(temp, stringsAsFactors = FALSE)
        unlink(temp)
        b = a[[length(a)]]
        
        if (is.null(b)) {
          warning(paste0("Wrong station ID: ", station_nr, 
                         " You can check station ID at https://ogimet.com/display_stations.php?lang=en&tipo=AND&isyn=&oaci=&nombre=&estado=&Send=Send"))
          return(data_station)
        } 

        cols = gsub("[^A-Za-z0-9]", "", as.character(lapply(b[1, ], as.character), stringsAsFactors = FALSE))
        if (ncol(b) <= length(cols)) {
          colnames(b) = cols
          colnames(b) = c("Date", "hour", colnames(b)[2:(ncol(b) - 1)]) # workaround for adding hour which is wrongly recognized
        } else {
          stop("Could not detect proper column names. Please investigate the URL manually.")
        }
      
        b = b[-1, ]
        b["station_ID"] =  station_nr
        # to avoid gtools::smartbind function or similar from another package..
        if (ncol(data_station) >= ncol(b)) {
          b[setdiff(names(data_station), names(b))] = NA # adding missing columns
          data_station = rbind(data_station, b)  # joining data
  
        } else { # when b have more columns then data_station
          if (nrow(data_station) == 0) {
            data_station = b
          } else {
            # adding missing columns
            data_station = merge(b, data_station, all = TRUE) # joining data
          }
        }
  
          if (coords) {
          coord = a[[1]][2, 1]
          data_station["Lon"] = get_coord_from_string(coord, "Longitude")
          data_station["Lat"] = get_coord_from_string(coord, "Latitude")
        }
        
      } # end of checking for empty files / problems with connection
    } # end of looping for dates
  }# end of looping for stations
  
  if (nrow(data_station) > 0) {
    data_station = data_station[!duplicated(data_station), ]
    # converting character to proper field representation:
    # get rid off "---" standing for missing/blank fields:
    data_station[which(data_station == "--" | data_station == "---" | data_station == "----" | data_station == "-----", arr.ind = TRUE)] = NA
    # changing time..
    data_station$Date = strptime(paste(data_station$Date, data_station$hour), "%m/%d/%Y %H:%M", tz = 'UTC')
    data_station$hour = NULL
    # other columns to numeric:
    columns = c("TC", "TdC", "ffkmh",  "Gustkmh", "P0hPa", "PseahPa", "PTnd", "Nt", "Nh",
                "HKm", "InsoD1", "Viskm", "Snowcm", "station_ID")
    columns = colnames(data_station)[(colnames(data_station) %in% columns)]
    suppressWarnings(data_station[, columns] <-
      as.data.frame(sapply(data_station[,columns], as.numeric)))
  
    # changing order of columns and removing blank records:
    if (coords) {
      ord1 = c("station_ID", "Lon", "Lat", "Date", "TC")
      ord1 = c(ord1, setdiff(names(data_station), c("station_ID", "Lon", "Lat", "Date", "TC")))
      ord1 = ord1[!(ord1 %in% c("WW", "W1","W2","W3"))]
      data_station = data_station[, ord1]
    } else {
      ord1 = c("station_ID", "Date", "TC")
      ord1 = c(ord1, setdiff(names(data_station), c("station_ID", "Date", "TC")))
      ord1 = ord1[!(ord1 %in% c("WW", "W1","W2","W3"))]
      data_station = data_station[, ord1]
    }
    # setdiff(names(df), c("station_ID", "Date", "TC"))
  
  
    # splitting precipitation into 6-12-24 hours from a default string in the Precmm column:
    if (precip_split) {
      if (all(is.na(data_station$Precmm))) {
        data_station$pr6 = NA
        data_station$pr12 = NA
        data_station$pr24 = NA
      } else {
      data_station$pr6 = precip_split(data_station$Precmm, pattern = "/6")
      data_station$pr12 = precip_split(data_station$Precmm, pattern = "/12")
      data_station$pr24 = precip_split(data_station$Precmm, pattern = "/24")
      }
    }
  
    # clipping to interesting period as we're downloading slightly more than needed:
    data_station = data_station[which(as.Date(data_station$Date) >= as.Date(min(date)) & as.Date(data_station$Date) <= as.Date(max(date))), ]
  
    } # end of checking whether object is empty

  data_station = unique(data_station)
  rownames(data_station) = NULL
  return(data_station)
}