Hi AMET experts,
I noticed when I used run_timeseries.csh to plot time series for so many sites (300-400 sites), the script only plotted data for about 10-14 days despite my data is available for 5 months. When I reduced the number of sites to 100-150 sites then the time series plots showed results for entire 5 months.
Have anyone experienced the same issue and figured out a fix for this? I am using AMETv1/4
Good morning Trang.
Thanks for using AMET. I’m pretty sure I know the issue.
If you look in your scripts_analysis directory for your project ID, there is a subdirectory input_files icluding one for the timeseries script. I.e., $AMETBASE/scripts_analysis/MyProjectID/input_files/timeseries.input
See this setting.
maxrec ← 10E4
This limits the amount of data extracted from the database. It was originally in place for web applications so a user did not accidently try to pull 10 years of data the database as an example. I see this is 10E4 by default.
Question. Are you using setenv AMET_GROUPSITES TRUE ???
If so, 300-400 sites would definitely pull way more than 10E4 samples and therefore cut the timeseries short. I would increase to 10E6 to be safe. I actually just increased this default level to 10E6 for the next release because of this issue.
Now, we had it lower because this GROUPSITES setting was more for smaller groups of sites like in a state or near a city. It was never designed to do large domains although it can be done. In my view, it loses some meaning when you average many sites over large areas because biases at sites tend to compensate with opposite bias at other sites, so the timeseries look really good. Just a word of caution when interpretation the results. This is where spatial statistics can help. Anyhow, try this and see if my hunch is correct.
Yes your trick works beautifully for me. Thank you so much, Rob. I am glad I asked, to get just a simple trick but works! It would have taken me days to figure it out without asking experienced AMET experts.
My domain is not very big but it is in New York State area so there are quite intense MADIS network there.
Your comment on interpreting the results are also very helpful for me! I haven’t thought about that. Thank you so much! It is nice to learn how each type of plot could complement each other for a good interpretation.