HDF inputs

[Jauch]

No problems

[epelde.ane@gmail.com]

Hello again! I write here my new problem..

I have different stations with daily data (for a year) and 10min data(for the remaining period).

I´m using the following values:
MAX_TIME_SPAN (maximum span in seconds for a time serie value to be considered valid as data source): 86500 (because for the first year I have daily data)
OUTPUT_TIME : 0 600 (because I´m working with 10 min data so that I want to compute it with higher resolution than daily one)
MAX_DISTANCE : 50000 ( the problem with this keyword comes when it is so low, but it can be as high as we want, no? to be sure that we are considering the entire extension of the catchment)

you see sth strange here? because I get this error:

forrtl: severe (170): program exception - stack overflow

thanks!!

[Jauch]

Hello Ane

"Stack Overflow" happens when the program you are using try to allocate more memory than it's possible. Even if there is enough memory in the computer, sometimes this happens because the "place" where it is allocating has some kind of limitation by the operating system (windows).

Are you trying to use Fillmatrix?

If yes, I can try to compile a version that maybe will work (some compiling options can be changed and help with this kind of situation).

But, even if we manage to create the HDF, maybe the size of it will become to large to be used by the model...

[epelde.ane@gmail.com]

yes! i was using fillmatrix..
ok, so if you dont mind to compile the executable I´ll use it to see if this gets solved! thank you!
the keywords values seem ok, no?

[Jauch]

Hello Ane.

I sent you 4 executables for Fillmatrix.

One basic version, one with OpenMP (omp), one basic version 64 bits (x64) and one with OpenMP 64 bits (omp_x64).
The OpenMP versions maybe are faster than the others, and the 64 bits version can be used only in computers with a 64 bits operating system (Winfows 64 bis).

The Keywords seem to be ok.

Hope this will work

[epelde.ane@gmail.com]

Hello Eduardo!
Thank you a lot! I managed to run it.. but.. it is still runnig since yesterday Now it is almost 3GB and it still goes increasing.. So maybe as you said, it will be too big for mohid to read it. More or less which is the size it is able to cope with? Cause the other option would be to modify the data so that I work with hourly one..

[Jauch]

Hello Ane

Yes, 10 minutes interval, depending how many days you are trying to run, will create a huge file.
So, I think you have two options here.

1. Instead of using 10 minutes interval, use hourly interval, as you said.

2. To split both the meteorological HDF and the simulations.

The option you want to use will depend on how much work you are ok to do and how much detail you want in your simulation.
I would try first the aproach 1, because unless you are simulatin floods, is unlikely to need 10 minutes interval for meteorological data.

But you have to make some "math" before trying to interpolate your stations into an HDF, because I THINK that fillmatrix dont do that.
The 10 minutes data for each hour must be transformed. For T, RH, SR and W, you need to do the average, while for precipitation you have to sum.

Then with the hourly timeseries, you can create an hourly HDF, that MAYBE will be still to large, but we can see this than.
If this happens, than we have no other choice than split the simulation into more than one, with calculus continuation.

If this is necessary, I can help you to setup everything. I have a program that can automatically run the continuations and we can write a script to use with this program that will make the necessary interpolation on the fly, given the correct timeseries.

But I'll be able to help with this only next week.

[epelde.ane@gmail.com]

Thanks a lot!
I´ve created an hdf for hourly precipitation data from 3 stations, and it is 600 Mb. I think I will split the information in this file cause I have daily data for the first year (the warm up period). So that I´ll have daily data from 1 station for warming up, and then an hdf file with hourly data from 3 stations. And this will be jsut for precipitation, I will asume the same data for the rest atmospheric parameters in 3 stations so that the model go faster..
I hope it works and doesnt go very slow!

[Jauch]

I think this is a good approach

[epelde.ane@gmail.com]

Another one

Using fillmatrix I see that in the ms2 window all time instants defined inside the "start-end" period are shown, but then, I open the hdf with hdf viewer and I just see a maximum of 10000 time instants. This is a problem of the hdf viewer program isnt it? Due to the size I couldnt manage to open the file in mohid so I cant know if the file contains all the data, althought I guess it does..

[Jauch]

Hum...

I never tried to create an HDF with so many time instants...
I think you should split this file, so there are fewer time instants on each one and than run the simulations with continuation.

Anyway, if fillmatrix don't rise any error during the creation of the file, probably it is ok, and the 10.000 limit that you are seeing can be a limitation of HDFViewer...