SNAP Data QA/QC

Introduction

The stats_setup.R script prepares extracted AR4/CMIP3 and AR5/CMIP5 regional temperature and precipitation data for direct use.

Motivation

This script transforms and organizes output data from a parent script into a format that is easier to work with in certain frameworks.

Details

The data are extracted based on regional shapefiles and the full Alaska-Canada 2-km extent for each of SNAP’s five models, from both CMIP3 and CMIP5, and three emissions scenarios and RCPs, respectively. Data manipulation is performed to organize the inputs into more usefully formatted outputs.

Capabilities

This script uses parallel processing via the base parallel package and the function mclapply. It uses as many CPU cores as are available since it is parallelized across regions, which exceed the number of cores. It becomes serial beyond the number of cores simultaneously in use.

Limitations

This code expects all ten models from CMIP3 and CMIP5 combined have been processed by the parent script. The code does not make use of Rmpi or similar options so it cannot take advantage of multi-node cluster parallel processing.

R Code

Setup

Setup is minimal. Set working directory. List GCM files while making sure to avoid CRU 3.x files. SNAP’s ten combined CMIP3 and CMIP5 models are in a hardcoded list as in the parent script.

setwd("/workspace/UA/mfleonawicz/projects/SNAPQAQC/data/regional/stats")

library(data.table)

files <- list.files(pattern = "regions_stats.RData$")
files <- files[substr(files, 1, 3) != "CRU"]

models <- list(c("CCCMAcgcm31", "CCSM4"), c("GFDLcm21", "GFDLcm3"), c("MIROC32m", 
    "GISSe2-r"), c("MPIecham5", "IPSLcm5a-lr"), c("ukmoHADcm3", "MRIcgcm3"))

Load and compile data

dlist <- vector("list", length(files))
for (i in 1:length(files)) {
    load(files[i])
    m <- do.call(rbind, stats.out)
    n <- nrow(m)
    p <- length(stats.out)
    dlist[[i]] <- data.frame(Phase = rep(c("AR4", "AR5"), each = n/(4 * p)), 
        Scenario = rep(c("SRES B1", "SRES A1B", "SRES A2", "RCP 4.5", "RCP 6.0", 
            "RCP 8.5"), each = n/(12 * p)), Model = rep(models[[i]], each = n/(4 * 
            p)), Var = rep(c("Temperature", "Precipitation"), each = n/(2 * 
            p)), Location = rep(names(stats.out), each = n/p), m, stringsAsFactors = F)
    print(i)
}
d <- as.data.frame(rbindlist(dlist))
rm(stats.out, n, m, i, p, files, models, dlist)

Organize data frame and support objects

d[d$Var == "Temperature", 6:(ncol(d) - 1)] <- round(d[d$Var == "Temperature", 
    6:(ncol(d) - 1)], 1)
d[d$Var == "Precipitation", 6:(ncol(d) - 1)] <- round(d[d$Var == "Precipitation", 
    6:(ncol(d) - 1)])
d$Year <- results.years
rm(results.years)
d$Month <- month.abb
d$Month <- factor(d$Month, levels = month.abb)
d$Scenario <- factor(d$Scenario, levels = unique(d$Scenario))

phases <- unique(d$Phase)
years <- unique(d$Year)
decades <- years[years%%10 == 0]
d$Decade <- paste0(substr(d$Year, 1, 3), 0)
varnames <- unique(d$Var)
modnames <- lapply(1:length(phases), function(i, x, p) unique(x$Model[x$Phase == 
    p[i]]), x = d, p = phases)
scennames <- lapply(1:length(phases), function(i, x, p) unique(as.character(x$Scenario)[x$Phase == 
    p[i]]), x = d, p = phases)
stats.columns <- seq(which(names(d) == "Mean"), length.out = length(agg.stat.names))
stats.multiplier <- 1  # Multiply by 10 to shrink file size (doesn't seem to help with files this small)

Save output files

library(parallel)
f <- function(i, index, multiplier) {
    name.tmp <- as.character(unlist(region.names.out))[i]
    region.dat <- subset(d, Location == name.tmp)
    region.dat <- multiplier * unlist(region.dat[, index])
    names(region.dat) <- NULL
    grp <- rep(names(region.names.out), times = sapply(region.names.out, length))[i]
    dir.create(outDir <- file.path("../../final/region_files_GCM/stats", grp, 
        name.tmp), recursive = T, showWarnings = F)
    save(region.dat, file = file.path(outDir, "stats_climate.RData"))
    print(i)
}

mclapply(1:sum(sapply(region.names.out, length)), f, index = stats.columns, 
    multiplier = stats.multiplier, mc.cores = 32)

gcm.stats.df <- subset(d, Location == as.character(unlist(region.names.out))[1])
gcm.stats.df[, stats.columns] <- NA
gcm.stats.df$Location <- NA
stats.colnames <- names(gcm.stats.df)[stats.columns] <- agg.stat.colnames
rm(d, f)
save.image("../../final/meta.RData")