READMEDoc

This is a convenient compilation of daily minimum/maximum temperature, precipation for the period 1970 - 2011 and in compact R binary format. Monthly average minimum/maximum temperatures and monthly precipitation totals are also included. The basic source is the GHCN data base although this has been cleaned by some by the mica group in processing. Surprisingly the variables tmin and tmax are the basic measurements for most stations. The daily mean is typically taken to be (tmin+ tmax)/2.

NOTE Due to different missing value patterns tmin and tmax have different numbers of stations. For convenience the monthly data has removed the stations that are not common.

Daily temperatures and Precipitation

so it possible to view the extent of these data without downloading the entire volume.

gunzip tmaxR.tar.gz
tar -xvf tmaxR.tar

This will create a directory tmaxR of about 500 files. Monthly data files have the name tmaxYYYYMM indicating the year and month and includes all the stations in the same order as tmaxIndex. So tmax197503 has the daily values for all stations and for March, 1975.

To keep the index and data together the index file is also included as the last file (in alpha order) in this directory.

Loading each of these data files into R results in an R object, a matrix, with name tmaxYYYYMM where rows index stations and columns index the days of the month. The names the of columns are dates as character strings but in the default format to be converted by the as.Date function into a date Object .

NOTE: tmin and tmax datasets have different numbers of stations, use the match function on the station ids to reconcile these to get common tmin, tmax values. The precipitation station network is substantially difference from the surface temperature network.

Monthly average station values

By "monthly average daily minimum temp" it is meant: find the minimum temperature for each day and then take the average of these for the month. Temperatures are reported in degrees C. These data consist of about 17K stations but typically there are about 9K reporting in a given month. If a station has more than 15 days missing in a month I have recoded the temperatures to missing (NA in R). The R binary data set file can be downloaded from:

To access this data set in R set your working directory to the same directory as the tempMonthly.rda or in the example below include the full pathname to this file.

> load("tempMonthly.rda")
> ls()
[1] "indexMonthly" "mn"
[3] "time"         "tmaxMonthly"
[5] "tminMonthly"  "yr"

So tminMonthly[5222,12, 1] is the value for station #5222 (actually Boulder) for DEC 1970

This array is organized so that if you collapse on 2nd and 3rd dimensions you will get a monthly time series for a station. E.g. c( tminMonthly[5222, ,] ) times series of tmin for Boulder

Quick examples for daily data

#Set the working directory to  include  the tmaxR directory from the tarball. 

load("tmaxR/tmax197503.rda")
dim(tmax197503)
[1] 17596    31 

#plot of daily max for March 15, 1975

load("tmaxR/tmaxIndex.rda") loc<- cbind( tmaxIndex\(lon, tmaxIndex\)lat) colnames( tmax197503 ) #double check dates are right Y<- tmax197503[,15] ind<- !is.na( Y) library(fields) quilt.plot( loc[ind,], Y[ind]) ```

Assembling a times series for a station(s) from the monthly files

fileNames<- dir("tmaxR")
fileNames<- fileNames[-505]
# omit last one (it is the index file!)
load("tmaxR/tmaxIndex.rda") # load index file
print( tmaxIndex[4500,]) # a station

Read the whole daily tmax files into R this will take some memory -- but works on my macbook air(!)

for(fName in fileNames){
cat(fName, fill=TRUE)
load(paste0("tmaxR/",fName))
}
# remove the .rda from names
objectNames<- substring(fileNames,1,10)
# loop through all months and accumulate the
# station values
temp<- NULL
for( dName in objectNames){
cat(dName, fill=TRUE)
# select stations 4500 and 4503 this
# of course  can be modified to 
# grab a more interesting subset of stations
nextMonth<-(get(dName))[c(4500,4503),]
temp<- cbind( temp, nextMonth)
}
time<- as.Date( colnames( temp)) # names of the columns are date strings
matplot( time, t(temp), type="l",lty=1)
title( paste("station 4500 and 4503 daily tmax record") )

Example with prcp

load("prcpR/prcpIndex.rda") loc<- cbind( prcpIndex\(lon, prcpIndex\)lat) colnames( prcp197503 ) #double check dates are right Y<- prcp197503[,15] ind<- !is.na( Y) library(fields) quilt.plot( loc[ind,], Y[ind])

Quick examples for monthly data

#Set the working directory to be the one including this file
load( "tempMonthly.rda")

these are all the station locations

library( fields)
loc<- cbind( indexMonthly$lon, indexMonthly$lat)

ind<- !is.na(tminMonthly[,6,1])
quilt.plot(loc[ind,] ,  (tminMonthly[ind,6,1] + tmaxMonthly[ind,6,1])/2)
map( "world", add=TRUE)
title("1970 June  daily average (tmin+tmax)/2")

average for 1975

ann1975<- apply( tminMonthly[,,6], 1, FUN="mean", na.rm=TRUE)
ind2<- !is.na(ann1975)
quilt.plot(loc[ind2,] ,  ann1975[ind2] )
title( "Tmin  june average")

number of stations reporting by month and by year

good<-  apply( !is.na(tminMonthly), c(2,3), FUN="sum")
image.plot( 1:12, 1970:2011, good)

plot( time, c( good), xlab="years", ylab="stations reporting")

find the closest station to Boulder,CO coordinates.

distBoulder<- rdist.earth( cbind( -105.2, 40), loc)
indexBoulder<- which.min( distBoulder)

t2<- c(tmaxMonthly[indexBoulder , ,] )
t1<- c(tminMonthly[indexBoulder , ,])

plot( time,  (t1+t2)/2,  type="l" )
title("Boulder, CO monthly mean temperatures")

plot( mn, (t1+t2)/2, xlab="month")
title("Boulder seasonality")

Daily Met Surface observations for the NARCCAP region and time period.