I would like to calculate SPEI using the R package developed by the authors. There seems to me to be two major reasons why I can't use the R package (I'm currently trying to figure out SPEI in Matlab). I'm computing SPEI for a city over 17 years at 30-meter resolution. Currently, I have all of my precipitation, potential evapotranspiration, and water balance data for the time series. Since I want the 3-month SPEI I also summed the water balances in three month chunks (April+May+June, May+June+July, etc.)
First, my data has a lot of NaNs, and the R package doesn't accept any NaNs. I can't delete my NaNs because the location of each data point in the matrix represents its location in space (lat/long). Is there any way around this? I need to know specifically which lat/long pixel an SPEI value corresponds to.
Second, my data is relatively high resolution. I have lots of pixels for each month in my 17-year study, however, it looks to me like the R package only accepts one data point per month, since each new row represents the following month. I was thinking of putting all my data points for each month into a vector along each row, particularly since the R package computes SPEI for each column individually, and I would like SPEI on a pixel-per-pixel basis. However, each row would be over 3 million points long, and I think I read somewhere that this is far beyond what the R package is able to handle.
Is there any way around these two issues so that I can use the R package? I'm having a lot of difficulty figuring out how to calculate SPEI in Matlab (trying to figure out probability-weighted moments). Thanks for any feedback.
https://cran.r-project.org/web/packages/SPEI/SPEI.pdf
https://cran.r-project.org/web/packages/SPEI/SPEI.pdf
It is difficult to provide useful help without a minimal example reproducing the problem (data and code). Can you post that ?
Regarding missing data, have you considered using the na.rm argument (setting it to TRUE) ? That's the most obvious solution I can think of.
All other issues you are raising can be addressed fairly easily with standard functions available in package raster (used to build SPEI). This package is very efficient to manage very large rasters. It might take some time, though.
You might be able to impute missing values, for example using the mice package.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R Statistical Package