Profile RUC-RAOB Statistics Change Details
- 26-Feb-2019 (production and beta)
- Corrected a slight error in the autocorrelation calculation (did
loop up to n_data, should have been n_data_n_gaps). Now r is
correctly unity. Also, limited the autocorrelation at 1 lag to be
0.95 so that error bars don't blow up for AC's very near unity.
Routine changed: error_library.pl.
- 19-Oct-2018 (beta)
- Changed a buffer from 200 bytes to 2000 bytes to account for very
long experiment names. Routines changed:
- 23-Jan-2018 (beta and production)
- Eliminated gaps in the profile plots, which occasionally occurred
when statistics from individual stations were plotted. Also
corected a null-pointer exception which occurred, apparently, when
errors are all zero due to insuficient data. Routines changed: DisplayFrame,
- 8-March-2017 (beta and production, on new server)
- The java applet itself now gets the list of models from
getModels.cgi on the server, and calculates default start and end
dates. This to facilitate a java web start version. Routines changed: StatsDisplay, getModels.cgi (new),
- 2-Jun-2016 (beta and production)
- in database sum_dR & sum2_dR = NULL at mb=0 for FIM which cause vertical profile fail to show.
Xue fixed this by add mb10> 0 in get_vert_stats.cgi
- 26-May-2016 (beta)
- Adjusted the data_getter script to not include data that are more
than three standard deviations from the mean at each level. Routines changed: error_library.pl (called by
- 15-Sept-2015 (beta and production)
- We now allow changing the curve titles in the legend when you
double-click a curve's legend. Routines changed:
makejar, DisplayFrame, gov.noaa.pmel.sgt.SGLabel,
gov.noaa.pmel.sgt.LineKey. Eliminated routine
- 10-Sept-2015 (beta)
- Changed the default space grid color to Color(220,220,220) from
lightGray, and set the Y-axis line to always be black. Routines changed: DisplayFrame,
- 31-Aug-2015 (beta and production)
- You can now double-click on a curve's legend, or on the grid lines
to bring up a dialog wherein you can change colors, etc. Routine changed: DisplayFrame.
- 24-Aug-2015 (beta and production)
- Changed default space grid color from black to lightGray. Routine changed: DisplayFrame.
- 6-Aug-2015 (beta and production)
- Some changes to accomodate automatic generation of plots done at
GSD. We changed the size of the plot so all legends can be seen,
and changed the size of the labels and legends.
- Also corrected an apparent error in java's dataURL creator that
caused the last part of the path to be lost.
- Routines changed: StatsDisplay,
- 23-Mar-2015 (beta and production)
- Pairwise differences fixed. Routines changed:
- Color coordination between each curve and its corresponding form
was corrected. Routines changed:
- 6-Mar-2015 (beta and production)
- Corrected an error in the legend for min and max dates when the
curves aren't matched. Routines changed:
- 24-Feb-2015 (beta and production)
- Put close_icon.gif in gov/noaa/fsl/frd/lib. Routines changed: CurveParameters.
- 20-Feb-2015 (beta and production)
- Fixed a lingering weakness in the 17-Feb-2015 error
calculations. We now correctly (I hope) account for gaps in the
sub timeseries that is used in calculating standard
error. (Formerly, the autocorrelation at 1 lag calculation was
inaccurate if there were gaps.) As expected, the change made only
a very slight difference--little more than roundoff error.
- Put some of the code into a new perl script: error_library.pl so
that get_vert_stats.pl isn't so long, and in hopeful preparation
for code sharing. Routines changed:
get_vert_stats.cgi, error_library.pl (new).
- 17-Feb-2015 (beta and production)
- Error bars are now optionally shown, See this explanation of EMB's standard error calculation.
- The option to show difference curves is now clearer, I hope. The
two options are a) to show all differences with respect to curve
1, or b) to show differences by pairs: 2 minus 1, 4 minus 3, etc.
- If you choose 'plot unmatched', you can get difference curves,
but not error bars. (The errors for the difference curves would
not necessarily be accurate.)
- To change the color or other properties of a curve, you now need
to use the 'Tree view' option below the plot, and chose the
'LineAttribute' for the curve or axis you want to
change. Previously, you could double click on the curve's
legend, or the axis. You can color the error bars and the curves separately.
- The java applet code was substantially reworked to allow the
plotting of error bars as separate 'layers', and to clean up the
behavior of the axes during zoom/unzoom operations.
- All error bars are determined by calculations on the server (in
get_vert_stats.cgi). However, overall errors for each curve that
are printed in the text display (which were never meaningful for
profiles anyway) are determined by the jave code:
gov.noaa.fsl.frd.lib.Stat. This java routine was corrected
to use the total number of non-missing data
points rather than the total number of data points in
calculating the standard error.
- Routines changed: get_vert_stats.cgi,
CurvesControl, DisplayFrame, gov.noaa.fsl.frd.lib.Stat,
- 15-October-2014 (beta and production)
- Some changes made to accommodate the pre-generation of plots,
while still using the same code as an applet for generating
interactive plots. (make_pngs.pl is run by ratchet's cron to
pre-generate plots, which are placed in the pngs/
subdirectory. Routines changed:
StatsDisplay, DisplayFrame, NewFrameThread.
- 29-Aug-2014 (production)
- All beta changes moved to production.
- 13-Aug-2014 (beta)
- Added support for persistence forecasts (negative forecast
lengths). Routines changed:
- 9-Oct-2013 (beta)
- For individual RAOBs (as opposed to pre-defined regions), we now
generate data at every 10mb, and at specific surface
levels. (Rather than calculating a 'delta_mb', we simply plot all
the mb levels that come from the sql query -- and this new
strategy is used for the pre-defined regions as well, although the
pre-defined regions only have data every 50mb.) Routines changed: get_vert_stats.cgi
- 20-May-2013 (beta)
- Found and corrected an error when the stat 'model' is used. It was
previously calculated for T and RH as ob + bias. But since bias is
defined as 'ob minus model', to get the model value correctly, we
must use ob - bias. This is now corrected. Routine changed: get_vert_stats.cgi
- 1-Apr-2013 (beta, no fooling)
- Corrected units on the plots for RH and RHob. Routine changed: DisplayFrame.
- Made the web page a little narrower for ease of reading on mac
laptops. Routines changed: CurveParameters.
- 28-Mar-2013 (beta)
model an editable field, so the user can enter
the model name directly. (Saves finding it in the long list.)
Routine changed: CurveParameters.
- Added a
show text button. When requested, a text
window comes up showing the data and error characteristics of each
curve. Routines changed: DisplayFrame,
- Extended the possible forecast projections out to 240 hours to
accommodate the FIM and related models. Routines
- 26-Mar-2013 (beta)
- Added checking for null values when matching events, so that hours
with missing RHobT (or other) values wouldn't falsely count as
matched events. Routine changed:
- 22-Mar-2013 (beta)
- Changed 'humidity' variable to 'RH' and added 'RHobT' (Model RH
w.r.t, observed T) as a variable for some models.
- Added a warning message whenever there's an error reading a curve,
so folks don't get the 'wait cursor of death'.
- Cleaned up less-often-used stats such as for individual RAOB
sites, and RAOB and model averages.
- Routines changed: get_vert_stats.cgi, DisplayFrame,
- 27-Feb-2013 (beta)
- Brought back the zero line when differences are displayed (it
vanished due to a bug). Also,
changed the order of colors, so the zero line shows in gray when
two curves are differences. Also, linked the colors of the curves
with those of the curve-parameter forms. Routine
changed: get_vert_stats.cgi, DisplayFrame,
- 15-Feb-2013 (beta)
- Corrected a bug that caused the displayed max date to be
determined solely by the first curve's date, rather than the union
of all curves. Routine changed:
- 11-Oct-2012 (beta)
- We now allow a comma-delimited series of RAOB wmoid's to be
entered in the 'Region' field. Routine changed:
- 4-Oct-2012 (beta)
- The returned date ranges are now the actual min and max dates
returned by the query. Previously the dates were simply the
requested min and max dates. Routines changed:
- 24-Sep-2012 (beta)
- Cloud cover ("cvr") option added. If you select "cloudy",
statistics will be generated by comparing RAOBS only with model
soundings that have cloud hydrometers at some level. But the
comparison will be done at all levels. Clear is the
converse. WARNING hydrometeor values are not
loaded into the database for all models. If the number of 'cloudy'
comparisons is zero, this is probably one of those models.
- Average RAOB T and RH and average model T and RH are now
- Routines changed:
- 13-Sep-2012 (beta)
- Matching is done via a joint mySQL query, rather than in
perl. This is a more robust and accurate way of matching.
Also, we send out the scaled results, so we can deal better with
large values of N (the number of observations).
- Added additional regions (15-18).
- Routine changed: get_vert_stats.cgi,
- 5-Mar-2012 (beta)
- The default curve weight is now HEAVY, rather than SOLID, per
Stan's request. Routine changed
- 22-Feb-2012 (beta)
- Adjusted the web page to correctly list models that have tables
that end in "_Areg0" as well as "_reg0". The tables that end in
"_Areg0" (or other region numbers) include average model wind
speed and average RAOB wind speed, so wind speed bias can be
calculated. Routine changed:
- 9-Feb-2012 (beta)
- Added region '14' -- the HRRR domain -- so comparisons between the
HRRR and models in larger regions will be accurate. (HRRR does not
include 14 active RAOBs that are in the RUC domain.)
- The regions are now provided as a parameter to the applet, rather
than hardwired in the code.
- The forecast_lens ("projections") are now provided as a parameter
to the applet, rather than hardwired in the code.
- Routine changed: StatsDisplay, CurveParameters,
- 5-Jan-2012 (production and beta)
- The upper limit of the drop-down box for years now updates
automatically to the current year +1. Routine
changed: CurveParameters, lib.JYearBox.
- 17-Nov-2010 (production and beta)
- The list of models is now generated by looking at the database,
for tables that match "ruc_ua_sums2.*_reg0". If a table of that
name exists, the model name implicit in the table name is included
in the list of models. Routine changed:
- 4-Nov-2010 (production)
- All beta changes moved to production.
- Fixed the repainting so that deleted CurveParameters truly
disappear. Routine changed: CurvesControl.
- 2-Nov-2010 (beta)
- Corrected order of colors in curves. Routines
- Made the zero line HEAVY so it shows better. Routines
- 28-Oct-2010 (beta)
- Added 18h forecasts to the list of available 'proj'. Routine(s) changed: CurveParameters.
- 22-Oct-2010 (beta)
- Can now eliminate any curves by clicking on the 'x' in the upper
right of each curve's parameter specification. (Previously, you
could only eliminate the last curve.) Also, compensated for a
bug in Java 6 update 22 that breaks JButtons with html tags in
their titles. Routine(s) changed: DisplayFrame,
CurveParameters, CurvesControl, StatsDisplay, SimpleLine2,
- 15-Feb-2010 (beta)
- If difference curves are not asked for, the zero
curve is not put out. Routines changed:
- 12-Feb-2010 (beta)
- Eliminated those pesky little alerts that come up when no data are
available. Routines changed: DisplayFrame.
- 27-Jan-2010 (production)
- All beta changes moved to production
- 25-Jan-2010 (beta)
- Difference curves are now optional. Routines
changed: get_vert_stats.cgi, CurvesControl.
- 24-July-2009 (production and beta)
- Added 'plot pairwise' option, in which difference curves are put
out for each pair of curves, as is the case with the time series
plots. Routines changed: get_vert_stats.cgi,
- 20-Jan-2009 (production and beta)
- All beta changes moved to production.
- Re-arranged order of fields to use screen space better.
- We only put out difference curves when 'plot matching' is chosen.
- Routines changed: DisplayFrame, CurveParameters,
- 23-Sept-2008 (beta)
- We flag as missing any level that has fewer than 10% of the number
of obs found at the level with the maximum number of obs. This is
done to avoid spurious data at very high and low levels, where
only a few RAOBs have observations. Routine
changed: DisplayFrame, get_vert_stats.cgi.
- 16-June-2008 (production and beta)
- Region short names, rather than region numbers, now appear on the
output plots. Also, Antarctic region added as an option, although we do not yet
save daily sums for the Antarctic. Routines
changed: CurveParameters, DisplayFrame, get_vert_stats.cgi.
- A difference curve is no longer generated when 'bias' is the
statistic plotted. Routine changed:
- 3-June-2008 (beta and production)
- The sign of the bias statistic has been changed: it is now 'model
minus RAOB', to match what is done at NCEP and by RTVS. Routines changed: CurveParameters, get_vert_stats.cgi.
- When two or more curves are plotted, a 'zero' curve is also
plotted. Routine changed: get_vert_stats.cgi.
- 5-May-2008 (production and beta)
- Order of curve legends is reversed. First curve plotted is at the
bottom, subsequent curves are higher up. This was done so that
difference curves can more easily be clipped off screen
captures. Routine changed: DisplayFrame.
- 31-March-2008 (production and beta)
- All beta changes moved to production, except retro runs.
- Curve matching fixed. (Previously, when 'all times' was chosen
as the valid time, and one model was missing at 0
or 12 UTC (but not both), the matching failed
to eliminate the unmatched date and hour. This didn't effect
matching when 0 UTC or 12 UTC was chosen as the valid time.)
Routine changed: get_vert_stats.cgi.
- Projections 24 and 36 were added. Routine
- 17-August-2007 (beta)
- When two or more curves are plotted, difference
curves (with respect to the first curve) are also plotted. I also
changed the sign of the difference curves: they are now curve_n
minus curve_0. So, if curve_n has larger errors, the difference
curve will be positive. Routines changed:
- 3-May-2007 (beta and production)
- Matching runs is now an option, rather than required. If curves
have different dates, you can now use the 'plot (un)matched'
button to force the curves to be plotted. Use the 'plot matched'
button to force the curves to show data from the same valid
times. Routines changed: CurvesControl,
- 23-April-2007 (beta and production)
- statistics choices clarified a bit in the choice window
- region '5' added as an option (for some retro cases only)
- regions '3' and '4' clarified in the choice window
- 'close plots' button added.
- Routines changed: CurvesControl, DisplayFrame,
- 28-March-2007 (beta and production)
- 12 hr forecasts added as an option. Routine changed:
- 1-March-2007 (beta and production)
- All profiles are automatically matched so that
they include the same valid times.
- For the region, you can enter a RAOB WMO ID.
- Under statistics, added 's-bias' (speed bias), which only applies
to the wind variable for individual RAOBs.
- Removed the 'n-in-average' statistic.
- Added the left-hand RUC menu for consistency with other RUC
- 30-Jan-2007 (beta and production)
- Now use the ruc_ua_sums2 database, which uses daily running sums
on 50 mb intervals to produce plots much faster.
- 8-Jan-2007 (beta and production)
- For the dev2 cycle, added 'regions' 3 and 4: GPS RAOBs and non-GPS RAOBs
- 20-Oct-2006 (production version 0.1)
- Adapted this from the time-series viewer.
Last modified: Tue Feb 26 23:33:07 GMT 2019