2 Downloading and preparing example data
This section describes how to download the relevant indicator data from the Nature Index Database. It is completely reproducible, and can be run by anyone with valid login credentials to the database. The downloaded data is then stored locally for use in the remainder of the code.
if(!require(NIcalc)){
devtools::install_github("NINAnor/NIcalc", build_vignettes = F)
}
library(NIcalc)First, you have to fill in your username (NINA email) and password.
myUser <- "user@nina.no" # insert NINA email
myPwd <- "" # secret passwordThen, choose which indicator(s) you want, use the NIcalc “importDatasetApi” function to retrieve data from the database, and save the dataset locally.
indicator <- c("Dikesoldogg",
"Jerv",
"Elg",
"Lomvi",
"Havørn",
"Lange")
for(i in indicator){
indicatorImport <- NULL
indicatorImport <- NIcalc::importDatasetApi(
username = myUser,
password = myPwd,
indic = i,
year = c("1990","2000","2010","2014","2019"))
assign(paste0(i, "_import"), indicatorImport)
}
path <- "P:/41201612_naturindeks_2021_2023_database_og_innsynslosning/temp/"
for(i in indicator){
temp <- get(paste0(i, "_import"))
saveRDS(temp, paste0(path, i, "_import.rds"))
}
for(i in indicator){
temp <- paste0(path, i, "_import.rds")
assign(i, readRDS(temp))
}Next, we need to assemble the data set. This shouldn’t be necessary since all the data is already present. But one thing we noticed was that for some indicators (e.g. jerv), the distribution family and parameters only appear after assembling.
# Spesify all of Norway incl the five regions, som NIunits:
myNIunits <- c(allArea = T, parts = T, counties = F)
# Include all BSunits (kommuner) irrespective of the proportion of the main ecosystems:
myPartOfTotal <- 0
for(i in indicator){
temp <- get(paste0(i, "_import"))
assemeble <- NULL
assemeble <- NIcalc::assembleNiObject(
inputData = temp,
predefNIunits = myNIunits,
partOfTotal = myPartOfTotal,
indexType = "thematic",
part = "ecosystem",
total = "total")
# I dont se the output changing if I for example chose total = marine. Perhaps 'part' and 'total' only becomes an issue if partOfTotal != 0.
assign(paste0(i, "_assemble"), assemeble)
}We then save the assembled data files.
for(i in indicator){
temp <- get(paste0(i, "_assemble"))
saveRDS(temp, paste0("data/", i, "_assemebled.rds"))
}If necessary, we load the datafiles back into R.
Since we are particularly concerned with visualizing uncertainty alongside means, we next need to simulate raw data and indicator values from the uncertainty distributions stored in the database.
myYears <- as.character(c(1990,2000,2010,2014,2019))
for(j in indicator){
print(j)
temp <- get(j)
temp2 <- get(paste0(j, "_import"))
temp_comb <- data.frame(NULL)
myMat2 <- NULL
myMat2_comb <- NULL
obstype <- NULL
obstype <- temp$referenceValues$distributionFamilyName
obstype[!is.na(obstype)] <- "tradObs"
obstype[is.na(obstype)] <- "customObs"
myMatr <- NIcalc::sampleObsMat(
ICunitId = temp$referenceValues$ICunitId,
value = temp$referenceValues$expectedValue,
distrib = temp$referenceValues$distributionFamilyName,
mu = temp$referenceValues$distParameter1,
sig = temp$referenceValues$distParameter2,
customDistribution = temp$referenceValues$customDistribution,
obsType = obstype,
nsim =1000
)
myMatr <- as.data.frame(myMatr)
myMatr <- myMatr %>%
tibble::add_column(.before=1,
ICunitID = row.names(myMatr))
myMatr <- myMatr %>%
tibble::add_column(.after = 1,
year = NA)
for(i in 1:length(myYears)){
print(i)
obs <- NULL
obs <- temp$indicatorValues[[i]]$distributionFamilyName
obs[!is.na(obs)] <- "tradObs"
obs[is.na(obs)] <- "customObs"
myMat <- NIcalc::sampleObsMat(
ICunitId = temp$indicatorValues[[i]]$ICunitId,
value = temp$indicatorValues[[i]]$expectedValue,
distrib = temp$indicatorValues[[i]]$distributionFamilyName,
mu = temp$indicatorValues[[i]]$distParameter1,
sig = temp$indicatorValues[[i]]$distParameter2,
customDistribution = temp$indicatorValues[[i]]$customDistribution,
obsType = obs,
nsim = 1000
)
myMat2 <- as.data.frame(myMat)
myMat2 <- myMat2 %>%
tibble::add_column(.before=1,
ICunitID = row.names(myMat))
myMat2 <- myMat2 %>%
tibble::add_column(.after = 1,
year = myYears[i])
myMat2_comb <- rbind(myMat2_comb, myMat2)
}
comb <- rbind(myMatr, myMat2_comb)
comb <- comb %>%
tibble::add_column(.after = 1,
ICunitName = temp2$ICunits$name[match(
comb$ICunitID, temp2$ICunits$id)])
comb2 <- comb[!is.na(comb$year),]
comb3 <- comb[is.na(comb$year),]
comb3$ref_mean <- rowMeans(comb3[,-c(1:3)])
combScaled <- comb2 %>%
tidyr::pivot_longer(cols = starts_with("V"))
combScaled <- combScaled %>%
tibble::add_column(ref = comb3$ref_mean[
match(combScaled$ICunitID, comb3$ICunitID)])
combScaled$scaledIndicator <- combScaled$value/combScaled$ref
combScaled <- dplyr::select(combScaled,
-name,
-value,
-ref)
assign(paste0(j, "_bootstrapped_raw"), comb)
assign(paste0(j, "_bootstrapped_scaled"), combScaled)
}Note that in this case, we also bootstrapped the reference values. These are coded as year = NA. This means that the simulation also accounts for reported uncertainty in reference values. For the purpose of this work, this makes sense as it leads to somewhat higher uncertainty in indicator values and hence a dataset that is more suitable to for testing out visualizations of uncertainty. However, up to and including the last iteration of the Nature Index (NI2020), uncertainty in reference values was NOT taken into account.
Finally, we save the simulated datasets.