SPEI returns NA values for time periods < 5 years

[gingerlolipop]

I encountered an issue while calculating SPEI for time periods of < 5 years. When I tried to debug with simulation datasets, I find that SPEI only returns correct values when the number of years is >= 5. Specifically:

For 1, 2, or 3 years, SPEI returns all NAs.
For 4 years, it seems mostly correct, but I still observe NA values for January and February in all years.
In my examples, the frequency is set to 12 (monthly data), and the scale is set to 3. Below is the original code used for calculation:

years <- 3 # Try 1, 2, 3, or 4 years
n_months <- years * 12
tave <- runif(n_months, -5, 25)  # Simulate temperature data
ppt <- runif(n_months, 0, 200)   # Simulate precipitation data
latitude <- 40                   # Set latitude

pet_thor <- thornthwaite(tave, latitude)
wb <- ppt - pet_thor
spei_values <- spei(wb, scale = 3)$fitted