Interpretation of `mesma` probabilities >1

[michaeldorman]

According to the mesma documentation, returned values are expected where "each value representing the estimated presence probability of the endmember per pixel (0 to 1)". However, the resulting rasters in the provided example have values that are >1:

library(raster)
library(RStoolbox)
data(lsat)
em_names <- c("water", "land")
pts <- data.frame(class=em_names, cell = c(47916,5294))
em <- lsat[pts$cell]
rownames(em) <- em_names
probs <- mesma(lsat, em, method = "NNLS")
range(probs[[1]][], na.rm = TRUE)
## [1] 0.000000 1.024394
range(probs[[2]][], na.rm = TRUE)
## [1] 0.000000 1.543188

My question is on how to interpret this. Is this expected behavior? If so, should the values just be truncated so that any probability >1 becomes 1? If not, is this potentially a bug?

Thanks!

[Joy-Giovanni]

I also have the same question, In my current project values were not only >1, they were in thousands. @16EAGLE can you help?

[BillyBrowning1]

Was this ever solved? Same question here

[16EAGLE]

This has been solved with the most recent update, implementing a sum-to-one constraint on top of NNLS. See https://github.com/bleutner/RStoolbox/blob/master/NEWS.md#rstoolbox-041 and ?RStoolbox::mesma after updating to the most recent github version for details.