Did you ever get around to examining the LC11 method? If it helps you may also be interested in Frankignoul, C. 1999: A Cautionary Note on the Use of Statistical Atmospheric Models in the Middle Latitudes: Comments on “Decadal Variability in the North Pacific as Simulated by a Hybrid Coupled Model”, J. Clim., 12:1871-1872. Also, papers that cite Frankignoul 1999.

#####################################################################
# Ported IDL procedure: obs,jul,sst,sstmo,olr,osr
# Read 36-day mean and monthly mean observed data, and make outputs:
#      jul: Julian day, sst: SST, olr: LW flux, osr: SW flux
#
# See http://climateaudit.org/2011/09/25/lindzen-choi-2011/
# 
# 
# R function obs()
#
#  Parameters: none
#
#  Return:
#   The R - version of "obs()" returns a list containing two 
#   irregularily spaced time series named ts1 and ts2. Both
#   cover the same time span, but ts2 has more datapoints in
#   between.
#
#   $ts1  : sst,olr and osr, indexed by "jul" of type zoo
#   $ts2  : only sstmo, indexed by "julmo" of type zoo
#
#  Usage example: 
#   devAskNewPage(T)
#   data <- obs()
#   plot(data$ts1)
#   plot(data$ts2)
#
#  Bugs :
#   - Since the original version does not return julmo, but does return
#     sstmo, there might be a problem with the original script?
#   - Spectral darkening for single days might be due to a typo?
#   - Spectral darkening corrections do not cover the whole time range
#
#  Author:
#    Hugo M
#
#  License:
#    GPL as far as the port of the original IDL scripts to R is concerned. 
#    Original copyrights may belong to Lindzen and Choi
#
#####################################################################+



obs <- function()  {
  julday <- function(month,day,year) {
            julian(ISOdate(year=year, month=month,day=day,hour=0,tz="GMT"))
  }



  ref <- c(1,37,73,109,145,181,217,253,289,325) # DayOfYear for 36-day mean

# ---------------
# Read 36-day mean SST for 1985-1999
# 147 lines
# ---------------

  # Read data file  
  # 0=DayOfYear, 1=year, 2=Month, 3=Day, 4=SST
  dat           <- read.table('../obs/oisst_36_20n20s_1985_1999.txt',
                               col.names=c('DayOfYear','Year','Month','Day','SST'))
  sst36         <- dat$SST

  # Calculate anomaly
  # Note: Lindzen and Choi (2011) references the anomalies to the 
  #       36-day means for 1985-1989.
  for ( r in ref ) {
        idx        <- dat$DayOfYear == r
	midx       <- idx & dat$Year %in% 1985:1989
        sst36[idx] <- dat$SST[idx] - mean(dat$SST[midx],na.rm=T)
  }

  jul36 <- julday(dat$Month,dat$Day,dat$Year)


# ---------------
# Read monthly mean SST for 1985-2008
# ---------------

  # data example
  #   1  198501    26.767
  #   2  198502    26.999
  #   3  198503    27.270
  dat  <- read.fwf('../obs/amipsst_mon_20n20s_1985_2008.txt',
		   col.names=c('Order','Year','Month','SST'),
                   widths=c(4,-2,4,2,-4,6))


  # Calculate anomaly. Note:
  #   Lindzen and Choi (2011) references the anomalies to 
  #   the monthly means for 1985-1989.

  sstmo <- dat$SST
  for ( m in 1:12 ) {
        idx        <- dat$Month == m
	midx       <- idx & dat$Order %in% 1:60
        sstmo[idx] <- dat$SST[idx] - mean(dat$SST[midx],na.rm=T)
  }
 
  julmo <- as.numeric(julday(dat$Month,1,dat$Year))

  stopifnot(length(sst36) == 147)
  stopifnot(length(jul36) == 147)
  stopifnot(length(sstmo) >= 288)
  stopifnot(length(julmo) >= 288)

  # Combined 36-day and monthly mean for 1985-2008
  #   sst36: 1985.1-1999.12 (150 36-d mon) + 2000.1-2008.12 (108 mon)
  #   sstmo: Monthly mean for 1999.10-2008.12

  sst <- c(sst36,sstmo[178:288])

  # Combined Julian day
  #   jul36: 36-day mean for 1985.1-1999.9
  #   julmo: Monthly mean for 1999.10-2008.12
  jul <- c(jul36,julmo[178:288])

# ---------------
# Read 36-day mean ERBE flux for 1985-1999
# Nan values exist, 147 lines
# ---------------

  # Read data file
  dat <- read.table('../obs/erbe_s10n_36_20n20s_1985_1999.txt',
                    col.names=c('DayOfYear','Year','Month','Day','LW','SW'),
		    na.strings='-NaN')


  # Calculate anomaly
  # Lindzen and Choi (2011) references the anomalies to the 36-day means
  # for 1985-1989.

  olr36 <- dat$LW
  osr36 <- dat$SW

  for ( r in ref ) {
        idx        <- dat$DayOfYear == r
	midx       <- idx & dat$Year %in% 1985:1989
        olr36[idx] <- dat$LW[idx] - mean( dat$LW[midx],na.rm=T )
        osr36[idx] <- dat$SW[idx] - mean( dat$SW[midx],na.rm=T )
  }


# ---------------
# Read monthly mean CERES flux for 2000-2008
#  -999 values exist
# ---------------

  # data example:
  # 183  200003   254.748    89.582
  # 184  200004   253.510    88.463
  # 185  200005   257.127    87.050
  # 186  200006   258.042    89.873
  dat <- read.fwf('../obs/ceres_es4_mon_20n20s_2000_2008.txt',
                  widths=c(4,-2,4,2,-3,7,-3,7),
                  col.names=c('Order','Year','Month','LW','SW'))

  dat$jd <- julday(dat$Month,1,dat$Year)

  dat$SW[dat$SW == -999] <- NA
  dat$LW[dat$LW == -999] <- NA

  # CERES shortwave flux scaling due to spectral darkening 
  # (Mattews et al. 2005 SPIE) 


  warning("please check if setting values for spectral darkening for single days")
  warning("Spectral darkening corrections do not cover the whole time range")

  idx <- dat$jd == julday(03,01,2000)
  dat$SW[idx] <- dat$SW[idx] * 1.003
  
  idx <- dat$jd >= julday(05,01,2000) & dat$jd <= julday(09,01,2000) 
  dat$SW[idx] <- dat$SW[idx] * 1.001

  idx <- dat$jd == julday(10,01,2000)
  dat$SW[idx] <- dat$SW[idx] * 1.002

  idx <- dat$jd >= julday(11,01,2000) & dat$jd <= julday(02,01,2001) 
  dat$SW[idx] <- dat$SW[idx] * 1.003

  idx <- dat$jd == julday(04,01,2001) 
  dat$SW[idx] <- dat$SW[idx] * 1.006 

  idx <- dat$jd >= julday(05,01,2001) & dat$jd <= julday(08,01,2001) 
  dat$SW[idx] <- dat$SW[idx] * 1.007 

  idx <- dat$jd == julday(09,01,2001) 
  dat$SW[idx] <- dat$SW[idx] * 1.008 

  idx <- dat$jd >= julday(10,01,2001) & dat$jd <= julday(02,01,2002) 
  dat$SW[idx] <- dat$SW[idx] * 1.007 

  idx <- dat$jd == julday(03,01,2002) 
  dat$SW[idx] <- dat$SW[idx] * 1.009 

  idx <- dat$jd >= julday(04,01,2002) 
  dat$SW[idx] <- dat$SW[idx] * 1.011 



  # Calculate anomaly
  # Lindzen and Choi (2011) references the anomalies to the monthly means
  # for 2000-2008. (??? /hm)
  warning("dat[1:60] does not span 2000-2008, but mar 2000 - feb 2005")

  olrmo <- dat$LW
  osrmo <- dat$SW
  for ( m in 1:12 )  {
        idx        <- dat$Month == m 
	midx       <- idx & dat$Year %in% 2000:2005 & dat$Order <= 242 # 03/2000 - 02/2005
        olrmo[idx] <- dat$LW[idx] - mean( dat$LW[midx],na.rm=T )
        osrmo[idx] <- dat$SW[idx] - mean( dat$SW[midx],na.rm=T )
  }

  # Combined 36-day and monthly mean for 1985-2008

  stopifnot(length(olr36) == 147)
  stopifnot(length(olrmo) == 258-153+1)

  # 1985.1-1999.12 (150 36-d mon) + 2000.1-2008.12 (108 mon)
  olr          <- replicate(258,as.numeric(NA))
  olr[1:147]   <- olr36 # 36-day mean for 1985.1-1999.8
  olr[153:258] <- olrmo # Monthly mean SST for 2000.3-2008.12

  # 1985.1-1999.12 (150 36-d mon) + 2000.1-2008.12 (108 mon)
  osr          <- replicate(258,as.numeric(NA))
  osr[1:147]   <- osr36 # 36-day mean for 1985.1-1999.8
  osr[153:258] <- osrmo # Monthly mean SST for 2000.3-2008.12


  require(zoo)

  warning("length(sstmo) == ",length(sstmo), " is greater than length(jul) == ", length(jul))

  as.date <- function(jd) as.Date(jd, origin=as.POSIXct("1970-01-01", tz="GMT"))
  list(ts1 = zoo(data.frame(sst=sst,olr=olr,osr=osr),as.date(jul)),
       ts2 = zoo(data.frame(sstmo=sstmo),as.date(julmo)))
 
}

If you suss the other probs please post when you update.

seems like you have stale version of textoidl which uses textable. Maybe wget had produced something along “textoidl.pro.1″ because another textoidl.pro version already existed on your disk? Anyway, simply remove all calls to textoidl and replace the call by plain strings (like ‘W/m^2′), since textoidl doesn’t work as expected anyway.

The rejects appear a bit strange. I tested the patch locally before uploading it. Did you use the patch from gist.github.com?

Re sourcecode tags, probably because I used angle brackets instead of squared brackets.

Btw, I’ve much beter version ready, which does also account for a severe GDL bug ( “PLOTS destroys plot layout with !P.MULTI – ID: 3323178″), fixes the color scheme and also prints the currently skipped plots. I’ll post this as soon as I’ve figured out why XTICKFORMAT isn’t working as expected.

NASA has a page devoted to IDL and floating point issues:
http://idlastro.gsfc.nasa.gov/idl_html_help/Accuracy_and_Floating_Point_Operations.html

It includes code for checking accuracy and errors.

Some remaining issues in textoidl.

% LOADCT: Loading table STD GAMMA-II
% TEXTOIDL: Function not found: TEXTABLE
% Error occurred at: TEXTOIDL           129 textoidl.pro
%                    MAIN                39 main.pro

thanks for all your efforts.