Compute Functional Diversity Hill Indices

Sebastien Villeger

2024-02-26

About this tutorial

What is this tutorial about?


This tutorial explains how to compute the family of indices presented in Chao et al. (2019) using mFD.


Let’s load data and compute functional distance


The data set used to illustrate this tutorial is the fruits dataset based on 25 types of fruits (i.e. species) distributed in 10 fruits baskets (i.e. assemblages). Each fruit is characterized by six traits values summarized in the following table:

Trait name Trait measurement Trait type Number of classes Classes code Unit
Size Maximal diameter Ordinal 5 0-1 ; 1-3 ; 3-5 ; 5-10 ; 10-20 cm
Plant Growth form Categorical 4 tree; shrub; vine; forb NA
Climate Climatic niche Ordinal 3 temperate ; subtropical ; tropical NA
Seed Seed type Ordinal 3 none ; pip ; pit NA
Sugar Sugar Continuous NA NA g/kg
Use Use as food Fuzzy 3 raw ; pastry ; jam %


We load the three objects used to compute functional framework (for more explanations, see mFD General Workflow):


data("fruits_traits", package = "mFD")

knitr::kable(head(fruits_traits),
      caption = "Species x traits data frame based on the **fruits** dataset")
Species x traits data frame based on the fruits dataset
Size Plant Climate Seed Sugar Use.raw Use.pastry Use.jam
apple 5-10cm tree temperate pip 103.9 50 50 0
apricot 3-5cm tree temperate pit 92.4 40 10 50
banana 10-20cm tree tropical none 122.3 70 20 10
currant 0-1cm shrub temperate pip 73.7 10 10 80
blackberry 1-3cm shrub temperate pip 48.8 30 10 60
blueberry 0-1cm forb temperate pip 100.0 10 40 50


data("baskets_fruits_weights", package = "mFD")

knitr::kable(as.data.frame(baskets_fruits_weights[1:6, 1:6]), 
             caption = "Species x assemblages matrix based on the **fruits** dataset")
Species x assemblages matrix based on the fruits dataset
apple apricot banana currant blackberry blueberry
basket_1 400 0 100 0 0 0
basket_2 200 0 400 0 0 0
basket_3 200 0 500 0 0 0
basket_4 300 0 0 0 0 0
basket_5 200 0 0 0 0 0
basket_6 100 0 200 0 0 0


data("fruits_traits_cat", package = "mFD")
knitr::kable(head(fruits_traits_cat), 
             caption = "Traits types based on **fruits & baskets** dataset")
Traits types based on fruits & baskets dataset
trait_name trait_type fuzzy_name
Size O NA
Plant N NA
Climate O NA
Seed O NA
Sugar Q NA
Use.raw F Use


Then, we can compute functional distance using the mFD::funct.dist() function as follows:

USAGE

## [1] "Running w.type=equal on groups=c(Size)"
## [1] "Running w.type=equal on groups=c(Plant)"
## [1] "Running w.type=equal on groups=c(Climate)"
## [1] "Running w.type=equal on groups=c(Seed)"
## [1] "Running w.type=equal on groups=c(Sugar)"
## [1] "Running w.type=equal on groups=c(Use,Use,Use)"


Generalisation of Hill numbers for alpha functional diversity


The family of indices presented in Chao et al. (2019) allows computing FD based on pairwise distance between species and their weights in assemblages. This generalization of Hill numbers framework is based on two parameters:


Indices are expressed as effective number of functionally equally distinct species (or virtual functional groups) and could thus be directly compared to taxonomic Hill numbers (including species richness).


NOTE For more details about the properties of Hill numbers FD read Chao et al. (2019) and especially its Figures 1 & 2.


All these indices can be computed with the function mFD::alpha.fd.hill().


Here we start by comparing the ‘classical’ Rao’s quadratic entropy expressed in Hill numbers following Ricotta & Szeidl (2009) which is the special case with q = 2 and tau = "max".


USAGE

baskets_FD2max <- mFD::alpha.fd.hill(
  asb_sp_w = baskets_fruits_weights, 
  sp_dist  = fruits_gower, 
  tau      = "max", 
  q        = 2)


Then, we can compute Hill numbers FD of order 2 computed with tau = "mean" and q = 2 as recommended in Chao et al. (2019)


USAGE

baskets_FD2mean <- mFD::alpha.fd.hill(
  asb_sp_w = baskets_fruits_weights, 
  sp_dist  = fruits_gower, 
  tau      = "mean", 
  q        = 2)


We can now compare these two metrics:

round(cbind(FD2max  = baskets_FD2max$"asb_FD_Hill"[ , 1], 
            FD2mean = baskets_FD2mean$"asb_FD_Hill"[ , 1]), 2)
##           FD2max FD2mean
## basket_1    1.50    2.62
## basket_2    1.83    3.97
## basket_3    1.86    4.10
## basket_4    1.27    1.72
## basket_5    1.30    1.85
## basket_6    1.74    3.73
## basket_7    1.82    3.94
## basket_8    1.40    2.16
## basket_9    1.53    2.75
## basket_10   1.53    3.01


We can see that FD computed with tau = "max" is less variable (ranging from 1.50 to only 1.86) than FD computed with tau = "min" (ranging from 1.72 to 4.10) illustrating its higher sensitivity to functional differences between species.


NB Note that even with q = 0, weights of species are still accounted for by FD. Hence, if the goal is to compute a richness-like index (i.e. accounting only for distance between species present), function mFD::alpha.fd.hill() should be applied to species occurrence data (coded as 0/1, previously computed using sp.tr.summary) so that all species have the same weight). Species occurrence data can be retrieve through the mFD::asb.sp.summary() function:


USAGE

# Retrieve species occurrences data:
baskets_summary    <- mFD::asb.sp.summary(baskets_fruits_weights)
baskets_fruits_occ <- baskets_summary$"asb_sp_occ"

head(baskets_fruits_occ)
##          apple apricot banana currant blackberry blueberry cherry grape
## basket_1     1       0      1       0          0         0      1     0
## basket_2     1       0      1       0          0         0      1     0
## basket_3     1       0      1       0          0         0      1     0
## basket_4     1       0      0       0          0         0      0     0
## basket_5     1       0      0       0          0         0      0     0
## basket_6     1       0      1       0          0         0      0     0
##          grapefruit kiwifruit lemon lime litchi mango melon orange
## basket_1          0         0     1    0      0     0     1      0
## basket_2          0         0     1    0      0     0     1      0
## basket_3          0         0     1    0      0     0     1      0
## basket_4          0         1     1    0      0     0     0      1
## basket_5          0         1     1    0      0     0     0      1
## basket_6          0         0     0    1      1     1     0      1
##          passion_fruit peach pear pineapple plum raspberry strawberry tangerine
## basket_1             1     0    1         0    0         0          1         0
## basket_2             1     0    1         0    0         0          1         0
## basket_3             1     0    1         0    0         0          1         0
## basket_4             0     1    1         0    1         0          0         1
## basket_5             0     1    1         0    1         0          0         1
## basket_6             0     0    0         1    0         0          0         0
##          water_melon
## basket_1           0
## basket_2           0
## basket_3           0
## basket_4           0
## basket_5           0
## basket_6           1
# Compute alpha FD Hill with q = 0:
baskets_FD0mean <- mFD::alpha.fd.hill(
  asb_sp_w = baskets_fruits_occ, 
  sp_dist  = fruits_gower, 
  tau      = "mean", 
  q        = 0)

round(baskets_FD0mean$"asb_FD_Hill", 2)
##           FD_q0
## basket_1   4.73
## basket_2   4.73
## basket_3   4.73
## basket_4   1.93
## basket_5   1.93
## basket_6   4.57
## basket_7   4.57
## basket_8   3.67
## basket_9   3.67
## basket_10  3.52


We can see that baskets with same composition of fruits species have same FD values (e.g basket_1, basket_2 and basket_3)


Generalisation of Hill numbers for beta functional diversity


Framework of Chao et al. (2019) also allows computing beta-diversity, with 2 framework similar to Jaccard and Sorensen ones for taxonomic diversity. The mFD:beta.fd.hill() function computes these indices.


NB Note that total weight of assemblages is affecting computation of functional beta-diversity. Hence if it is does not reflect an ecological pattern (e.g. rather difference in sampling effort), it is recommended to apply mFD::beta.fd.hill() to relative weight of species in assemblages.


# retrieve total weight per basket:
baskets_summary$"asb_tot_w"
##  basket_1  basket_2  basket_3  basket_4  basket_5  basket_6  basket_7  basket_8 
##      2000      2000      2000      2000      2000      2000      2000      2000 
##  basket_9 basket_10 
##      2000      2000
# Here baskets all contain 2000g of fruits, we illustrate how to compute...
# relative weights using the output of asb.sp.summary:

baskets_fruits_relw <- baskets_fruits_weights / baskets_summary$"asb_tot_w"
apply(baskets_fruits_relw, 1, sum)
##  basket_1  basket_2  basket_3  basket_4  basket_5  basket_6  basket_7  basket_8 
##         1         1         1         1         1         1         1         1 
##  basket_9 basket_10 
##         1         1


Now we can compute functional beta-diversity of order q = 2 (with tau = "mean" for higher sensitivity) with Jaccard-type index:


USAGE

# Compute index:
baskets_betaq2 <- mFD::beta.fd.hill(
  asb_sp_w  = baskets_fruits_relw, 
  sp_dist   = fruits_gower, 
  q         = 2,
  tau       = "mean", 
  beta_type = "Jaccard")

# Then use the mFD::dist.to.df function to ease visualizing result
mFD::dist.to.df(list_dist = list("FDq2" = baskets_betaq2$"beta_fd_q"$"q2"))
##          x1        x2        FDq2
## 1  basket_1  basket_2 0.058982325
## 2  basket_1  basket_3 0.078716397
## 3  basket_1  basket_4 0.029573623
## 4  basket_1  basket_5 0.027059789
## 5  basket_1  basket_6 0.484115290
## 6  basket_1  basket_7 0.292594562
## 7  basket_1  basket_8 0.290545545
## 8  basket_1  basket_9 0.185475113
## 9  basket_1 basket_10 0.011136995
## 10 basket_2  basket_3 0.004420448
## 11 basket_2  basket_4 0.161833512
## 12 basket_2  basket_5 0.162571972
## 13 basket_2  basket_6 0.260541701
## 14 basket_2  basket_7 0.097053161
## 15 basket_2  basket_8 0.294504888
## 16 basket_2  basket_9 0.225897615
## 17 basket_2 basket_10 0.058298760
## 18 basket_3  basket_4 0.172877455
## 19 basket_3  basket_5 0.178123024
## 20 basket_3  basket_6 0.207102590
## 21 basket_3  basket_7 0.081951839
## 22 basket_3  basket_8 0.308336365
## 23 basket_3  basket_9 0.241482168
## 24 basket_3 basket_10 0.082649928
## 25 basket_4  basket_5 0.001049851
## 26 basket_4  basket_6 0.511165067
## 27 basket_4  basket_7 0.421141181
## 28 basket_4  basket_8 0.482330219
## 29 basket_4  basket_9 0.342926459
## 30 basket_4 basket_10 0.050817451
## 31 basket_5  basket_6 0.532084800
## 32 basket_5  basket_7 0.438841544
## 33 basket_5  basket_8 0.496554512
## 34 basket_5  basket_9 0.336894275
## 35 basket_5 basket_10 0.044052657
## 36 basket_6  basket_7 0.068382884
## 37 basket_6  basket_8 0.759422492
## 38 basket_6  basket_9 0.680332414
## 39 basket_6 basket_10 0.453325136
## 40 basket_7  basket_8 0.478528108
## 41 basket_7  basket_9 0.431531941
## 42 basket_7 basket_10 0.265928889
## 43 basket_8  basket_9 0.020812705
## 44 basket_8 basket_10 0.345652088
## 45 basket_9 basket_10 0.219780509

We can see that basket 1 is similar (beta < 0.1) to baskets 2,3,4,5,10 and that it is the most dissimilar to basket 8 (beta > 0.5). Baskets 4 and 5 are highly dissimilar (beta > 0.8) to basket 8.


References