NetCDF input/output?

[monocongo]

I would like to compare the results of this code with a Python implementation of scPDSI in order to validate results and determine if there are areas of possible improvement. It would be easiest for me to do this if I can use the same input files and then make a comparison of the outputs. The input format for the Python code is NetCDF. Is it possible to use NetCDF as input into this R code? I'm not adept with R so please forgive the rookie question -- hopefully, this is more straight-forward than I assume. Thanks in advance for any assistance.

[kongdd]

Hi @monocongo,

I have checked the output of this version scPDSI at here, which is exactly the same as the original cpp version (scPDSI2003) at both of monthly and weekly scale by using the example data in scPDSI2003.

By the way, you need to comment this line to make them same, https://github.com/kongdd/scPDSI/blob/master/src/pdsi_ext.cpp#L237.

[monocongo]

Thanks, @kongdd.

I apologize, I don't understand the gist of your comments. Are you essentially saying that the R implementation of scPDSI contained within this repository matches the results of the C++ version by Wells/Goddard/Hayes? If so then how is that relevant to this issue, i.e. comparing the results of the Python implementation of scPDSI against those produced by this package?

My intention is to use the same precipitation, temperature, and AWC inputs for the two implementations and then compare results in order to see 1) how much if any differences there are in the results of the two implementations, and 2) if there are discrepancies between the two then to determine which, if either, is closer to accurately modeling actual drought conditions.

When you say I need to comment a line to make them the same, what does this mean? The line referenced appears to call a Calibrate() method three times. How is that relevant to this issue?

[kongdd]

Sorry. I can't solve your issue.

1. Without clear benchmark, it is hard to say which version is better. I think what we can do is just close to the original version written by the author.
   If you need to make some improvement, you might need to illustrate its merits in a scientific paper.
2. In the original cpp version, it just Calibrate once.
   https://github.com/kongdd/scPDSI/blob/master/inst/scPDSI2003/bin/scpdsi.cpp#L1381

[monocongo]

In order to make a comparison, we can use the same input for both codes and then compare the results of the two packages. If one shows a better correspondence to actual drought events then we can consider that to be a better model/index, no? My initial question is regarding how can I use NetCDF as input to this code in order to perform this kind of comparison. The documentation for this package is not sufficient to make this determination, which is why I've asked for guidance as to how to use NetCDF as input to this package, so I can use it to compute scPDSI and then compare the results against those of my Python package.

[kongdd]

Ha ha, we need to push @Sibada to write more document.