Re: [las_users] Any known issues with aggregating data sets with missing days?

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Hi Kacie,
This is the dataset you're talking about, is that correct?
http://thredds.jpl.nasa.gov/thredds/dodsC/GHRSST_NCDC-L4LRblend-GLOB-AVHRR_OI_TIMEAGG/aggregate__ghrsst_NCDC-L4LRblend-GLOB-AVHRR_OI.ncml

As you've found, Ferret running behind LAS will open the dataset and see the irregularly-spaced time axis that's in  your aggregation, and it'll go get data at the nearest time point for whatever date is requested from the UI.

Configuring LAS as if it's a regularly spaced axis does make the time menus a lot nicer. Here it means that the last date you show is Jun 9, 2009, whereas in fact the last timestep in the dataset is really Dec 7, 2009.  So that's a big downside of doing that.  There is no way to request dates at the end of the time series.  I think if you set the number of steps in the configuration to the number of days in the time range - as if all days were present - this would be fixed.  And it shouldn't cause trouble otherwise.  The UI presents the backend with date strings, and those are used to request the nearest time step in the aggregation.

If you've made an animation, you can play it one step at a time and check what you're seeing. Hit STOP then the +1 button to step through one frame at a time.  I haven't checked it but I think you'll be just skipping frames for the missing time steps.

I wonder as well whether some of the errors you've been asking about with Hovmuller plots, may be related to this.  I'll do some more testing and let you know.

Ansley

On 5/8/2013 2:25 PM, Shelton, Kacie E (398C) wrote:

Hi.

 

In order to make things easier for our users, we have taken data sets that have a few missing days, and instead of listing each individual file in the pull-down menu in LAS Date sector on the left, we have aggregated it into one reference – ie, in the XML configuration file for each dataset, we have the start date, the size (the number of granules), and the step (1 day), as if the dataset was not missing any days.  This causes LAS to have pull down menus for the month, day, year, which is much easier than having a huge list of each granule. 

 

Are there any known side effects for doing this, that would cause issues for LAS or the users?  We are trying to isolate and track down some errors, and are wondering if they may be associated with aggregating the files in this manner.

 

So far, we have seen that if a missing day is requested, LAS will automatically select the closest existing granule: for example, in GHRSST ODYSSEA, April 17^th 2008 (2008108) is missing, and asking LAS for it, LAS pulls up April 16^th, 2008.  Animate appears to skip the missing granules all together – but this is hard for us to tell, due to the NetCDF library issue (which means we can only load ~4-6 granules in Animate).

 

1)       Are there other side effects we have not noticed? 

2)      Does this cause any errors?

 

Thank you,

Kacie