Rank histogram (or Talagrand Diagram) of precipitation data

Question

Given a dataset of lines containing 6 wind forecast values plus 1 observed (actual) value in each like:

 FCT1   FCT2     FCT3   FCT4   FCT5     FCT6   OBSERVED
-3.17   3.51    -5.71   1.37   -0.22   -0.65    -2.38   
-2.7    2.21    -0.71   2.73   -0.33   -2.62    -1.38   
-1.2    3.15    -4.17   3.33   -0.48   -1.65    -2.30   
...
-3.0    3.50    -1.79   3.37   -0.18   -0.62    -2.32   

To make a rank histogram (or Talagrand Diagram), I understand that I need to loop through the lines, sort the forecast values for each one and assume that the ordered values (six, in this case) are the inner limits of each bin on the diagram. 6 limits generates 7 bins. Then, I need to take the corresponding observed value and increase the bin it fits (the bin that its range contains the observed value). I need to do it for every row, so each row has its limits. It has to do with What PDF should be fit to a rank histogram?. I think it is not a simple histogram built via hist() in R. Am I wrong?

How about a precipitation forecast data? Like:

 FCT1   FCT2   FCT3   FCT4   FCT5   FCT6   OBSERVED
  0       0     0.1    0     0        0        0
  0       0     0      0     0.02     0        0
  0       0.1   0      0     0        0        3

What am I supposed to do to know the bin that has the range that fits the observed value like 0, for example? How can I build 7 bins from this forecast data?

Answer 1

I found the answer in this pdf:

When two or more forecasts have same value (most commonly 0), random selection is made for which bin receives the count.

The formal source can be found in this pdf.

Also, the function verifRankHist() from the R package ensembleBMA, does this job.