Previous models showed that using two sites to predict onto a third were less accurate than using randomly selected plots from across all sites.
There are four potential reasons for this.
Each site has a unique soil composition. Therefore predicting onto a third site is extrapolating past the information contained in the model.
Species may not be evenly distributed across the sites. Therefore models based on site data may not contain enough presences to create an accurate model.
Spatial autocorrelation may affect the ability to predict. There should be a increasing gradient in prediction accuracy as plots are chosen to minimise autocorrelation.
A linear discriminant analysis showed that the sites had a unique soil composition.
Most species didn’t occur evenly across sites.
A linear discriminant analysis based on soil variables showed that the plots clearly grouped by site. Site 1 had some overlap with sites 2 and 3, which makes sense as it was the middle plot.
Species tended to be widespread in only one or two sites. Possibly could test for evenness somehow. Chi squared?
Site 1 | Site 2 | Site 3 | Total | |
---|---|---|---|---|
R_burtoniae | 9 | 10 | 2 | 21 |
R_comptonii | 7 | 0 | 15 | 22 |
D_diversifolium | 6 | 12 | 16 | 34 |
A_delaetii | 21 | 16 | 18 | 55 |
A_fissum | 2 | 1 | 9 | 12 |
A_framesii | 8 | 4 | 0 | 12 |
C_spissum | 24 | 13 | 26 | 63 |
C_staminodiosum | 1 | 0 | 9 | 10 |
Dicrocaulon_sp | 10 | 16 | 3 | 29 |
Oophytum_sp | 1 | 11 | 2 | 14 |
Created seven models with decreasing amount of spatial autocorrelation. Red represents plots used in the models, blue shows plots used as test data.
AUCs
Species | Mod1 | Mod2 | Mod3 | Mod4 | Mod5 | Mod6 | Mod7 |
---|---|---|---|---|---|---|---|
R_burtoniae | 0.85 | 0.94 | 0.91 | 0.99 | 0.82 | 0.94 | 0.90 |
R_comptonii | 0.52 | NA | 0.71 | 0.94 | 0.95 | 0.93 | 0.93 |
D_diversifolium | 0.43 | 0.32 | 0.36 | 0.81 | 0.78 | 0.82 | 0.74 |
A_delaetii | 0.58 | 0.68 | 0.91 | 0.79 | 0.74 | 0.76 | 0.88 |
A_fissum | 0.42 | 0.08 | 0.41 | 0.57 | 0.57 | 0.29 | 0.29 |
A_framesii | NA | 0.53 | 0.66 | 0.55 | NA | 0.88 | 0.81 |
C_spissum | 0.49 | 0.40 | 0.36 | 0.66 | 0.68 | 0.87 | 0.63 |
C_staminodiosum | 0.72 | NA | 0.73 | 0.66 | 0.80 | 0.86 | 0.88 |
Dicrocaulon_sp | 0.49 | 0.64 | 0.36 | 0.81 | 0.81 | 0.77 | 0.85 |
Oophytum_sp | 0.66 | 0.65 | 0.92 | 0.78 | 0.72 | 0.58 | 0.68 |