I am trying to evaluate my model using the timeseries plots in AMET. Everything is working fine I am just hoping for an explanation of how AMET compares sites and concentations. For example I have one timeseries plot for the entire AQS network in Utah.
The models max is around 60 however when I use the same model data for a different timeseries plot comparing one site near salt lake the models max is around 100-110.
I am assuming that AMET averages the hourly data between sites based on the time series explanation in the user guide." * Creates a time series plot. With multiple sites; the sites are time averaged to create a single plot. Also plots the bias and error between the obs and model" but the “time averaged” part confused me. Any documentation that deals with this would also be helpful!
First, thanks for including the plots. Very helpful.
Yeah, I see that is confusing. It really should say “spatial averaged” and not “time averaged”. I’ll make that correction in the next documentation update. Unless you specify some kind of time averaging (see below), multiple sites would just be spatial averaged. Sorry for creating confusion.
Note that it is possible to time average the time series plot (which is probably what I was thinking when I wrote that in the documentation). For example, you could create a time series that plots monthly average values using sub-monthly data. You may already be aware of that, but in case you were not.
Let me know if you have any more questions about that. Thanks.
Thanks for the info!
Would the time averaging be similar to the a daily time series vs an hourly time series?
I’m not sure I follow your question. Are you referring to the difference in plotting something like hourly PM2.5 vs. daily PM2.5?
The time averaging simply segregates the incoming data into the desired time average. Meaning, if your data are daily and you request monthly average, the daily data are segregated by month and then the values are computed using the monthly segregated data. Note that the calculations are still done on the individual model/ob pairs and then those values are averaged versus calculating the average values first and then doing the calculations (e.g. bias). So, the method I use retains the spatial correlation along with the temporal correlation.
I was just trying to draw an example of what you said “Note that it is possible to time average the time series plot (which is probably what I was thinking when I wrote that in the documentation). For example, you could create a time series that plots monthly average values using sub-monthly data. You may already be aware of that, but in case you were not.”