3. validation of mismatching locations

packages

library(plyr)

data

Dataset created to validate used locations by roe and red deer. For each of the five roe and red deer population 100 points were compared with a satellite layer (i.e., ground-truth). Hence, in total we did a validation of 500 points per species. For some locations we were not able to determine the ground-truth due to inclarity of the satellite layer. These locations have a missing value (NA) in the column ‘satellite’.

• population: unique identifier study area as defined in the eurodeer database
• animals_id: unique identifier of animal as defined in the eurodeer database
• latitude: latitude of mismatching location (in WGS84)
• longitude: latitude of mismatching location (in WGS84)
• yearx: the year of the observation point
• tcd: 0/1, 1 if the location is identified as forest by tcd, else 0
• clc: 0/1, 1 if the location is identified as forest by clc, else 0
• satellite: 0/1, 1 if the location is identified as forest by the ground-truth layer, else 0
• correct: tcd, if tcd is corresponding to the ground-truth, else clc

# load data 
roe <- read.csv("../data/3_validation_gps/roe.csv",header=T, sep=',') 
red <- read.csv("../data/3_validation_gps/red.csv",header=T, sep=',')

# change 0 and 1 to open and forest, respectively  
roe[which(roe$tcd==1),]$tcd <- 'forest'
roe[which(roe$tcd==0),]$tcd <- 'open'
red[which(red$tcd==1),]$tcd <- 'forest'
red[which(red$tcd==0),]$tcd <- 'open'
roe[which(roe$satellite==1),]$satellite <- 'forest'
roe[which(roe$satellite==0),]$satellite <- 'open'
red[which(red$satellite==1),]$satellite <- 'forest'
red[which(red$satellite==0),]$satellite <- 'open'
roe$punto <- NULL
red$punto <- NULL

head(roe)

##   population animals_id latitude longitude yearx    tcd clc satellite correct
## 1          1        784 45.99947  11.08740  2006 forest   0    forest     tcd
## 2          1        795 46.01295  11.08614  2005 forest   0      <NA>    <NA>
## 3          1        768 46.00790  11.04297  2006 forest   0      open     clc
## 4          1        800 46.06284  11.08369  2005 forest   0      open     clc
## 5          1        776 45.96659  10.95757  2006 forest   0    forest     tcd
## 6          1        784 45.99813  11.08744  2007 forest   0    forest     tcd

nrow(roe)

## [1] 500

head(red)

##   population animals_id latitude longitude yearx    tcd clc satellite correct
## 1          3        130 46.36950  10.68811  2005 forest   0    forest     tcd
## 2          3        137 46.36474  10.67057  2011 forest   0      open     clc
## 3          3        129 46.39063  10.67036  2007 forest   0    forest     tcd
## 4          3        135 46.47717  10.48563  2011   open   1      open     tcd
## 5          3        116 46.53655  10.28553  2010 forest   0    forest     tcd
## 6          3        116 46.53581  10.30131  2011 forest   0    forest     tcd

nrow(red)

## [1] 500

table 3b: validation of mismatching points

## confusion matrices 

#### roe ####

### absolute 
(roe_t <- table(roe[,c('tcd','satellite')]))

##         satellite
## tcd      forest open
##   forest    270   66
##   open       58   74

### proportion
(roe_p <- prop.table(roe_t))

##         satellite
## tcd         forest      open
##   forest 0.5769231 0.1410256
##   open   0.1239316 0.1581197

#### red ####

#### absolute 
(red_t <- table(red[,c('tcd','satellite')]))

##         satellite
## tcd      forest open
##   forest    228   43
##   open       46  158

#### proportion 
(red_p <- prop.table(red_t))

##         satellite
## tcd          forest       open
##   forest 0.48000000 0.09052632
##   open   0.09684211 0.33263158