Nothing
climdex2ecad()
and cost2ecad()
) have been added. They provide useful tools to convert data to the ECA&D format (the main format of input data in INQC). The conversion can be performed for the COST Home and the RClimDex formats. ./raw
used to store the data. Station file is always stations.txt
and as pass through readecad()
. stations.txt
must contain all stations to be processed. ECA&D seems to to have a comprehensive station file only subset if country sets are takenhomofolder
, so it is declared at first and does not need to be passed on from routine to routine. Implementing it progressively once it is completed, will remove this mentionreadecad()
changed to adapt the ever changing ECA&D formatstemperature()
starts adapted to cope with the fact that some ECA&D formats have 4 columns (i.e., not STAID), which is addedtxtnblend()
, which works or blended series. Here, no big efforts to isolate the "right" station: is is the same staid changing the variable name ... oh, man, I love blended series!!!readecad()
, implemented with readheader()
dostats()
, function which iterates QCSummary and provides metastatisticsjummps2()
improved to avoid very short series to cause error. If the series has less than 100 valid values, it is not checkedflatsun()
has been corrected. When ran in modonube
, the values returned in bad where referred to the sun file instead of the cloud file. Solved by replacing:
z<-merge(x,y,all.x=TRUE,all.y=FALSE)
with if(!modonube){z<-merge(x,y,all.x=TRUE,all.y=FALSE)}; if(modonube){z<-merge(x,y,all.x=FALSE,all.y=TRUE)}
listas()
reads station files as csv files. Realized this is not smart for the autodownloading function, but works for sent filesflat
allows to exclude sequences for a value, i.e., excluding 0, will allow to work flat
for precipitationselepe
: results of badfriki
were assigned to flat
in the log files. selepe
parametrization, inadvertently, it was -8000 instead of 8000retornoracha = 500
instead of retornoracha = 1000
in precip
(fine tuning with benchmark, need to evaluate false positives)rm(liston)
, which is executed at the end of the code. It has to be rm(liston,envir=GlobalEnv)
. This error was preventing the generation of new station lists and using previous ones. This caused txtn
to not perform as expected closetxtn
. Problems that mostly affected Baboon data, due to its peculiarities. Changes should not affect real data if this is not the case, revert to a February versionsunafterdark
: it compares sunshine duration to the maximum possible according to calendar and positionIQRoutliers
function to exclude values (e.g., 0 for precipitation)IQRoutliers
to the precip
routine, with very wide limits to avoid flagging excessive number of valuesconsolidator()
was only being applied to temperature!weirddate
that was flagging as bad the 29-th of February of leap yearslistas()
made resistant to small format changes in the input files (e.g. no header or different widths than the "official" specification)badfriki
was not working well. Replaced by newfriki
for pressure and, most likely, will be deprecated for other variables toojumps2
replaces jumps
. Can be ran in two modes: split by month and with thresholds computed from the distribution of the differences or with an absolute threshold. In both cases, jumps2
does not flag BOTH elements in a jump, only the one which is most likely the culpritQCConsolidated
, adds reference to this software and is "true" ECA&D format, with corrected widths and comma-separateddrywetlong
: the parameter sueco
is the threshold for dividing dry and wet. This is useful to label other binary sequences, e.g. for 0 radiation. Now it is <= and >, instead of < and >=txtn
improved:
a) closetxtn
is not used anymore (too slow) and the pair is located by staid
, eleid
and parid
. If sticking to ECA&D, safe enough.
b) once values are flagged as "tx <= tn", we look if any value in the pair is central and the other extreme to their monthly distributions. Then, only the extreme is flaggedinithome()
. Creates all the folders and updates stations listTG
does not exist, the call to temperature(element='TG')
should not break the code anymore)sundur()
had a parameter omitted (roundmax
) which was causing the code to break. Solvedroundprecip()
. It used to flag most of the values due to an error. This also made the code really slow. Solved. drywetlong()
resistant to precipitation series with almost no data (RR_SOUID102169.txt has only ONE valid value, the rest is NA!!!)roundprecip()
made resistant to series with no data different than 0. Returns NULL potpareto()
made resistant to cases were all values = 0 (e.g, RR_SOUID107781.txt, with < 800 values, and all them 0)drywetlong()
: there was a mistake in rle()
usage. Example: rle$values 1 2 2 5, it was interpreted as if the strike labeled with "5" was starting in position 10, when actually runs from 6 to 10!!!bad<-flat(x$value%%10,roundmax)
for some variables: roundmax
was not parameterizedtoomany
instead of toomanymonth
and toomanyyear
)consolidator
, a functions which creates a second version of the QCed data. Creates a replica of the original series, with the same name but with the QC column updated to 0 (OK)selepe
txtn()<-closetxtn()<-listas()<-distHaversine()
There were a few errors (hopefully, corrected) and now it is much faster just by limiting the station past as candidates to disthaversine to those in a radius of 1deg lat and 1deg lon from the candidate. For the whole ECA&D, the performance is down from 14 seconds to less than 2!
** Further improvements to the closesetation()
function: does not give an error if the tries to open a file listed but non-existingroundprecip
and a bad parameterization in precipip()
potpareto()
function to avoid problems with some matrices encountered in drywetlong
: myfit<-fpot(xx,threshold,std.err=FALSE)
drywetlong
to avoid problems with desertic locations with almost no rain eventsAny scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.