It has been more than a "few days," but the issues identified have been addressed. If you are "upgrading" to this version (wee_trend 1.1.0,) you
have to do an uninstall and re-install as described in README.md. There are a couple of items I want to mention. First, this program is not part of the weewx project and is not a supported wee_weeex utility. I'll try to fix any errors or omissions as best I can, but Tom et al. are not involved with wee_trend. Second, most problems encountered are quality control, of which only the user is aware. wee_trend plots are "interesting," but few to no weewx users have 20+ years of NOAA data. Also, as amateur weather observers, we typically do not maintain strict, consistent industry standards with our hardware. We upgrade or change weather stations, move their locations occasionally, etc. For example, my weather station was lowered from the meteorological standard of 10 meters to 8.5 meters above ground in 2019. This created a bias toward lower wind speeds. wee_trend can't know this. It also can't know if a bird built a nest in your rain gauge for several months. These things may make the plots appear to show trends that are not statistically significant. To paraphrase Tom, I wrote wee_trend during the summer of 2022 to learn a bit about Python/git/GitHub, and because it was hot and muggy in Nova Scotia. As before, it's available here: https://github.com/ve1dx/wee_trend - Paul VE1DX On Thursday, September 29, 2022 at 6:06:51 PM UTC-3 [email protected] wrote: > Awesome, thanks Paul! And please don’t consider anything as criticism. > Just testing. =) > > > > Thanks for your work on this, it’s pretty cool! I’m curious what we’ll see > when Tom runs it with all his data. =P > > > > *From:* [email protected] <[email protected]> *On > Behalf Of *Paul Dunphy > *Sent:* Thursday, September 29, 2022 1:45 PM > *To:* weewx-development <[email protected]> > *Subject:* Re: [weewx-development] Non-metric NOAA file? > > > > Thanks for the feedback and testing the program with your data. It works > as intended, but that doesn't mean it can't be better. Consider it between > alpha and beta testing now. > > I am dropping entire months if any column has more missing data points > than allowed by the tolerance parameter. In your case, you could get around > that by specifying a higher -t value. However, I think a better option > would be for me to refactor the code to drop the bad columns with missing > data. > > There's nothing wrong with your temperature, precipitation, etc., data for > that month. I'll re-work it to do as described above. Meanwhile, try it > with a -t 4 or -t 5 and see if that keeps the month. It may take me a few > days. I'll let you know when I have it ready. > > On Thursday, September 29, 2022 at 1:55:11 PM UTC-3 [email protected] > wrote: > > Thanks Paul. Just did that, and indeed, message is more verbose, but might > still need some clarity > > For example: > Processing month December > Inspecting month 2016-12 > > > Tolerance for missing days is 0 > > 2016-12 has 27 days of complete data > Data for 2016-12 incomplete, dropping it > > > > But looking at my NOAA report: > https://www.staze.org/weewx/tabular.html?report=NOAA/NOAA-2016-12.txt > > It's missing some wind data, I assume because my anemometer must have been > encased in ice/snow. Is the script dropping that whole month from ALL > plots, or just the Wind plot (in this case)? If the former, then why? If > the latter, then maybe adding some language about which data is incomplete > and indicating it's just being dropping from that plot. Basically, every > December is being flagged for that exact reason. > > > > Thanks, this is pretty cool. > > > > On Thursday, September 29, 2022 at 5:50:10 AM UTC-7 Paul Dunphy wrote: > > I reworded the diagnostic messages. Suggest you uninstall wee_trend (pip > uninstall wee_trend), download the new version, and re-install it. Run it > with 'wee_trend -V 1' to give a verbose set of diagnostics. The reason for > the dropped months will then become more obvious. You can adjust the > tolerance with the -t parameter to make it process incomplete months with > up to 25 days of missing data, but this quality control issue requires the > user's judgment. Interpreting the results is beyond the scope of wee_trend. > > On Wednesday, September 28, 2022 at 4:18:49 PM UTC-3 Paul Dunphy wrote: > > Thanks for testing and the feedback. It can't process 2022-12 because we > have no data for that month. The tolerance set to 0 will drop any month > with even 1 day missing. 0 is the default. If you re-run it with -t 10 or > the like, it will process incomplete months, and the 38 dropped months > should be less. > > The output of interest is the plots. If you see plots with a data point > after the current month, that's an impossible situation, which means there > is a defect in my program. > > The wording of the verbose messages may be confusing. Maybe it should say > "Inspecting" instead of "Processing," and I will look into that. > > The processing time of 46 seconds you see on a Pi 4 is consistent with > mine (43 seconds.) It will always generate 120 plots, but the number of > months it has to work on will depend on how many years of data you have. > > On Wednesday, September 28, 2022 at 3:25:18 PM UTC-3 [email protected] > wrote: > > I had to install liblapack-dev (and liblapack3) and libatlas3-base-dev on > my Pi to get this to work. (just sudo apt-get install liblapack-dev > libatlas3-base-dev) > > > > Took my Pi4 46 seconds to process everything, though for some reason it > dropped 38 of the months claiming missing days... also tried running months > that don't exist yet (Working on month 2022-12; Tolerance for missing days > is 0; Data for 2022-12 incomplete, dropping it). =) > > > > Any idea why it thinks some days are missing in the reports? > > On Wednesday, September 28, 2022 at 10:42:32 AM UTC-7 Paul Dunphy wrote: > > > > I believe I accidentally replied to Vince only. Here's what it > generates. Well, 120 of them if you run it in batch mode. Many are > irrelevant because of the geographical location, and most are just "noise" > because there isn't enough data. However, they are "interesting." I'd be > more interested in feedback if the download/install works on other systems > in addition to the ones I tried. > > > > - Paul VE1DX > > > > On Wednesday, September 28, 2022 at 12:23:49 PM UTC-3 [email protected] > wrote: > > Lets see some example plots !!!! > > (and good luck with Fiona) > > -- > > You received this message because you are subscribed to the Google Groups > "weewx-development" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > > To view this discussion on the web visit > https://groups.google.com/d/msgid/weewx-development/d9f99aad-030d-4769-9be6-9b830f7be97dn%40googlegroups.com > > <https://groups.google.com/d/msgid/weewx-development/d9f99aad-030d-4769-9be6-9b830f7be97dn%40googlegroups.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "weewx-development" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/weewx-development/eb22f54a-975d-420d-aafe-545fdcee81aan%40googlegroups.com.
