R语言批量转换.nc为.tif时输入文件匹配错误问题

我有500个.nc文件，每个文件包含PotEvap或pr变量的4年月度数据，对应variable（PotEvap/pr）、gcm（dry、middle等）、scenario（historical/rcp45等）的一种组合。目标是将每个.nc文件中的月度数据转换为单独的.tif文件，每个.nc需生成60个.tif。

当前程序能生成月度.tif，但输入.nc文件与输出.tif匹配错误：比如处理“pr middle rcp45”组合时，调用的却是PotEvap的.nc文件，输出文件名正确但数据源错误。

初始数据框结构 #

structure(list(variable = structure(c(1L, 2L, 1L, 2L), levels = c("PotEvap", 
"pr"), class = "factor"), gcm = structure(c(1L, 1L, 2L, 2L), levels = c("middle", 
"dry"), class = "factor"), scen = structure(c(1L, 1L, 1L, 1L), levels = "rcp45", class = "factor")), out.attrs = list(
    dim = c(121L, 12L, 2L, 2L, 1L), dimnames = list(Var1 = c("Var1=1950", 
    "Var1=1951", "Var1=1952", "Var1=1953", "Var1=1954", "Var1=1955", 
    "Var1=1956", "Var1=1957", "Var1=1958", "Var1=1959", "Var1=1960", 
    "Var1=1961", "Var1=1962", "Var1=1963", "Var1=1964", "Var1=1965", 
    "Var1=1966", "Var1=1967", "Var1=1968", "Var1=1969", "Var1=1970", 
    "Var1=1971", "Var1=1972", "Var1=1973", "Var1=1974", "Var1=1975", 
    "Var1=1976", "Var1=1977", "Var1=1978", "Var1=1979", "Var1=1980", 
    "Var1=1981", "Var1=1982", "Var1=1983", "Var1=1984", "Var1=1985", 
    "Var1=1986", "Var1=1987", "Var1=1988", "Var1=1989", "Var1=1990", 
    "Var1=1991", "Var1=1992", "Var1=1993", "Var1=1994", "Var1=1995", 
    "Var1=1996", "Var1=1997", "Var1=1998", "Var1=1999", "Var1=2000", 
    "Var1=2001", "Var1=2002", "Var1=2003", "Var1=2004", "Var1=2005", 
    "Var1=2006", "Var1=2007", "Var1=2008", "Var1=2009", "Var1=2010", 
    "Var1=2011", "Var1=2012", "Var1=2013", "Var1=2014", "Var1=2015", 
    "Var1=2016", "Var1=2017", "Var1=2018", "Var1=2019", "Var1=2020", 
    "Var1=2021", "Var1=2022", "Var1=2023", "Var1=2024", "Var1=2025", 
    "Var1=2026", "Var1=2027", "Var1=2028", "Var1=2029", "Var1=2030", 
    "Var1=2031", "Var1=2032", "Var1=2033", "Var1=2034", "Var1=2035", 
    "Var1=2036", "Var1=2037", "Var1=2038", "Var1=2039", "Var1=2040", 
    "Var1=2041", "Var1=2042", "Var1=2043", "Var1=2044", "Var1=2045", 
    "Var1=2046", "Var1=2047", "Var1=2048", "Var1=2049", "Var1=2050", 
    "Var1=2051", "Var1=2052", "Var1=2053", "Var1=2054", "Var1=2055", 
    "Var1=2056", "Var1=2057", "Var1=2058", "Var1=2059", "Var1=2060", 
    "Var1=2061", "Var1=2062", "Var1=2063", "Var1=2064", "Var1=2065", 
    "Var1=2066", "Var1=2067", "Var1=2068", "Var1=2069", "Var1=2070"
    ), Var2 = c("Var2= 1", "Var2= 2", "Var2= 3", "Var2= 4", "Var2= 5", 
    "Var2= 6", "Var2= 7", "Var2= 8", "Var2= 9", "Var2=10", "Var2=11", 
    "Var2=12"), Var3 = c("Var3=PotEvap", "Var3=pr"), Var4 = c("Var4=middle", 
    "Var4=dry"), Var5 = "Var5=rcp45")), row.names = c(1L, 1453L, 
2905L, 4357L), class = "data.frame")

出错的循环代码 #

foreach(k = iter(nc.vals1, by = "row")) %dopar% {
  require(raster)
  require(stringr)
  require(ncdf4)
  maca.dir <- "C:/Postdoc/ExposureProj/Climdata1"
  out.dir = "C:/Postdoc/ExposureProj/raw_tifs/"
  
  var.cur <- paste0(k$variable)
  gcm.cur <- paste0(k$gcm)
  scen.cur <- paste0(k$scen)
  
  cat("Processing:", var.cur, gcm.cur, scen.cur, "\n")
  
  dir.name <- paste0(var.cur, gcm.cur, scen.cur)
  dir.cur <- list.files(path = maca.dir, pattern = dir.name, full.names = FALSE)
  dat.dir <- paste0(maca.dir, "/", dir.cur)
  nc.files <- list.files(path = dat.dir, full.names = TRUE)
  
  for (f in nc.files) {
    cat("Processing file:", f, "\n")
    dat.stack <- stack(f)
    dat.stack <- raster::shift(dat.stack, dx = -360)
    crs(dat.stack) <- "+proj=longlat +ellps=WGS84 +datum=WGS84 +no_defs"
    
    num.layers <- seq(1:dim(dat.stack)[3])
    
    for (n in num.layers) {
      dat.raster <- dat.stack[[n]]
      yr.mo <- paste0(substr(names(dat.raster), 2, 5), "_", substr(names(dat.raster), 7, 8))
      
      # Correct the order of variables in the file name
      fname <- paste0(out.dir, var.cur, "_", gcm.cur, "_", scen.cur, "_", yr.mo, ".tif")
      
      cat("Writing to file:", fname, "\n")
      writeRaster(dat.raster, fname, drivername = "GTiff", overwrite = FALSE)
    }
  }
}

错误原因 #

核心问题出在文件夹匹配逻辑：

1. dir.name <- paste0(var.cur, gcm.cur, scen.cur)生成的字符串无分隔符，比如prmiddlercp45会和PotEvapmiddlercp45部分匹配（公共子串middlercp45），导致list.files返回错误文件夹。
2. list.files默认是部分匹配，而非精确匹配，进一步放大了匹配错误的概率。

修复方案 #

1. 精确匹配文件夹 #

用下划线分隔变量名，同时用正则表达式的^和$确保完全匹配文件夹名。

2. 增加校验逻辑 #

如果未找到对应文件夹或NC文件，直接跳过并报错，避免处理错误数据源。

修复后的代码 #

foreach(k = iter(nc.vals1, by = "row")) %dopar% {
  # 提前加载依赖包，避免循环内重复加载
  library(raster)
  library(stringr)
  library(ncdf4)
  
  maca.dir <- "C:/Postdoc/ExposureProj/Climdata1"
  out.dir = "C:/Postdoc/ExposureProj/raw_tifs/"
  
  # 转换因子为字符，避免因子编码问题
  var.cur <- as.character(k$variable)
  gcm.cur <- as.character(k$gcm)
  scen.cur <- as.character(k$scen)
  
  cat("Processing:", var.cur, gcm.cur, scen.cur, "\n")
  
  # 正则精确匹配文件夹名，下划线分隔变量
  dir.name <- paste0("^", var.cur, "_", gcm.cur, "_", scen.cur, "$")
  dir.cur <- list.files(path = maca.dir, pattern = dir.name, full.names = FALSE)
  
  # 校验是否找到目标文件夹
  if(length(dir.cur) == 0){
    cat("ERROR: No folder found for", var.cur, gcm.cur, scen.cur, "\n")
    next
  }
  
  # 用file.path拼接路径，适配不同系统
  dat.dir <- file.path(maca.dir, dir.cur)
  nc.files <- list.files(path = dat.dir, full.names = TRUE)
  
  # 校验是否找到NC文件
  if(length(nc.files) == 0){
    cat("ERROR: No NC files found in", dat.dir, "\n")
    next
  }
  
  for (f in nc.files) {
    cat("Processing file:", f, "\n")
    dat.stack <- stack(f)
    dat.stack <- raster::shift(dat.stack, dx = -360)
    crs(dat.stack) <- "+proj=longlat +ellps=WGS84 +datum=WGS84 +no_defs"
    
    # 用nlayers替代dim，更符合raster包规范
    num.layers <- seq_len(nlayers(dat.stack))
    
    for (n in num.layers) {
      dat.raster <- dat.stack[[n]]
      yr.mo <- paste0(substr(names(dat.raster), 2, 5), "_", substr(names(dat.raster), 7, 8))
      
      fname <- file.path(out.dir, paste0(var.cur, "_", gcm.cur, "_", scen.cur, "_", yr.mo, ".tif"))
      
      cat("Writing to file:", fname, "\n")
      writeRaster(dat.raster, fname, drivername = "GTiff", overwrite = FALSE)
    }
  }
}

额外优化点 #

• 用as.character()转换因子变量，避免因子转字符串时出现整数编码问题
• 用file.path()拼接路径，适配Windows/Linux/macOS的路径分隔符
• 用nlayers()替代dim(dat.stack)[3]，更符合raster包的API设计
• 增加文件夹和文件存在性校验，减少无意义的后续处理

内容的提问来源于stack exchange，提问作者Beardedant