Dear community
Basis:
We further developed the standard InVEST pollination model (python script) to make possible the input of spatially explicit species data, e.g. instead of a guild table we used a raster map –> “raster” version.
Vision: we do not want 1 Mio InVEST models for each raster, but one InVEST model automatically running at the raster level.
Problem:
The transformation of the normalization is tricky
Standard invest
invest/src/natcap/invest/pollination.py at main · natcap/invest · GitHub : Line 1249
# * species_abundance[species] (string->float)
total_relative_abundance = guild_df[_RELATIVE_SPECIES_ABUNDANCE_FIELD].sum()
result[‘species_abundance’] = {}
for species in result[‘species_list’]:
result[‘species_abundance’][species] = (
guild_df[_RELATIVE_SPECIES_ABUNDANCE_FIELD][species] /
total_relative_abundance)
The exact transformation of the standard InVEST to the “raster” version would mean that the relative abundance of each raster is calculated by using it’s individual total relative abundance. However this was technically not possible in the pre-processing data normalization version. I just used one total relative abundance for each cell for simplification (after discussions with experts we agree that some rasters will be under- and some others overestimated) – namely the total relative abundance from the guild table. However I see that this approach is not accurate. We got some quite unrealistic values > 0.8 (many values over 0.4, so not just exceptions.
Idea
-
At the moment, we are working with the sum of the average “relative abundance for each apple species” for the normalization.
-
As we have many grid points, the sum for single species is the better options than the average.
-
Here my plan was to work with the sum “relative abundance for each apple species” and then calculating the total relative abundance.
-
For total relative abundance we take the sum of the “sum values for each species values” for normalization (or the average of the “sum” values”). I suppose we need “sum”, to take account the many grid points.
Questions:
(1) Dou you agree about my conclusion of unrealistic values of 0.8, see also statitiscs.
(2) What do you think about my idea, and the difference between taking the sum of individual “average” values and the sum of individual “sum” values for normalization (simplified).
statistics
$average
[1] 0.2086073
$median
[1] 0.1804454
$standard_deviation
[1] 0.1343401
$min
[1] 0
$max
[1] 0.8561345
$first_quartile
25%
0.1110333
$third_quartile
75%
0.2965108
Kind regards
Sibylle