2_assign_grid.Rmd

---
title: "Assign grid to occurrences"
author:
- Damiano Oldoni
- Peter Desmet
date: "`r Sys.Date()`"
output:
  html_document:
    toc: true
    toc_depth: 3
    toc_float: true
    number_sections: true
---

In this document we transform the text file containing the occurrences of (alien) species for Europe into a sqlite database. Then, we filter on issues and occurrence status. This is a way to handle the critical huge amount of occurrences. Note: some of these steps could take long.

# Setup

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, warning = FALSE, message = FALSE)
```

Load libraries:

```{r load_libraries_assign_grid_eu}
library(tidyverse)      # To do datascience
library(tidylog)        # To provide feedback on dplyr functions
library(here)           # To find files
library(rgbif)          # To use GBIF services
library(glue)           # To write queries
library(sp)             # To work with geospatial data
```

# Get occurrence data

## Get CSV file from GBIF

We download the occurrences from GBIF, based on the key got in `download.Rmd`:

```{r get_occ_file_eu}
key <- "0036191-231002084531237"
zip_filename <- paste0(key, ".zip")
zip_path <- here::here("data", "raw", zip_filename)
if (!file.exists(zip_path)) {
  occ <- occ_download_get(
    key = key, 
    path = here::here("data", "raw")
  )
}
```

We unzip text file with occurrences as  `key_number` + `occurrence.txt` in `./data/raw`:

```{r unzip_csv_occs_eu}
occ_file <- paste(key, "occurrence.txt", sep = "_")
occ_path <- here::here("data", "raw", occ_file)

if (!file.exists(here::here("data", "raw", occ_file))) {
  unzip(zipfile = zip_path,
        files = "occurrence.txt",
        exdir = here::here("data", "raw"))
  file.rename(from = here::here("data", "raw", "occurrence.txt"),
              to = occ_path
  )
}
``` 

Name of columns:

```{r get_cols_occsfile_eu}
cols_occ_file <- read_delim(
  occ_path, "\t", n_max = 1,
  quote = ""
)
cols_occ_file <- names(cols_occ_file)
```

Number of columns present:

```{r n_cols_occ_file}
length(cols_occ_file)
```

## Define columns to select

We define a subset of columns we are interested to:

```{r columns_to_use_eu}
cols_to_use <- c("gbifID", 
                 "scientificName",
                 "kingdom",
                 "phylum",
                 "class",
                 "order",
                 "family",
                 "genus",
                 "specificEpithet",
                 "infraspecificEpithet",
                 "taxonRank",
                 "taxonomicStatus",
                 "datasetKey",
                 "basisOfRecord",
                 "occurrenceStatus",
                 "lastInterpreted",
                 "identificationVerificationStatus",
                 "hasCoordinate",
                 "hasGeospatialIssues",
                 "decimalLatitude",
                 "decimalLongitude",
                 "coordinateUncertaintyInMeters",
                 "coordinatePrecision",
                 "pointRadiusSpatialFit",
                 "verbatimCoordinateSystem",
                 "verbatimSRS",
                 "eventDate",
                 "startDayOfYear",
                 "endDayOfYear",
                 "year",
                 "month",
                 "day",
                 "verbatimEventDate",
                 "samplingProtocol",
                 "samplingEffort",
                 "issue",
                 "taxonKey",
                 "acceptedTaxonKey",
                 "kingdomKey",
                 "phylumKey",
                 "classKey",
                 "orderKey",
                 "familyKey",
                 "genusKey",
                 "subgenusKey",
                 "speciesKey",
                 "species")
```

Columns in occurrence file not present in the subset:

```{r cols_in_cols_to_use_not_present_in_cols_occ_db_eu}
cols_to_use[which(!cols_to_use %in% cols_occ_file)]
```

will be removed from the selection:

```{r remove_cols_not_in_cols_occ_db_eu}
cols_to_use <- cols_to_use[which(cols_to_use %in% cols_occ_file)]
```

Final number of columns to select:

```{r n_cols_to_use_eu}
length(cols_to_use)
```

## Read occurrence data

### Define column type specifications

The following columns should contain integers:

1. `*Key`, e.g. `taxonKey`, `speciesKey`
2. `*DayOfYear`: `startDayOfYear` and  `endDayOfYear`  
3. `year`
4. `month`
5. `day`

```{r define_col_integer_for_specific_columns_eu}
int_colnames <- 
  cols_to_use[str_detect(cols_to_use, "Key") & 
                !str_detect(cols_to_use, "datasetKey")]
int_colnames <- c(
  int_colnames,
  cols_to_use[str_detect(cols_to_use, "DayOfYear")],
  cols_to_use[cols_to_use == "year"],
  cols_to_use[cols_to_use == "month"],
  cols_to_use[cols_to_use == "day"]
)

int_cols <- 
  map(int_colnames, ~ col_integer()) %>% 
  setNames(int_colnames)
```

The following columns should contain real numbers:

1. `decimal*`: `decimalLatitude` and `decimalLongitude`
2. `coordinate*`: `coordinateUncertaintyInMeters` and `coordinatePrecision`
3. `pointRadiusSpatialFit`

```{r define_col_double_for_specific_columns_eu}
real_colnames <- cols_to_use[str_detect(cols_to_use, "decimal")]
real_colnames <- c(
  real_colnames,
  cols_to_use[str_detect(cols_to_use, "coordinate")],
  cols_to_use[cols_to_use == "pointRadiusSpatialFit"]
)

real_cols <- 
  map(real_colnames, ~ col_double()) %>% 
  setNames(real_colnames)
```

The other columns to select contain text:

```{r define_col_text_for_other_columns_eu}
char_colnames <- cols_to_use[!cols_to_use %in% real_colnames & 
                               !cols_to_use %in% int_colnames]
char_cols <- 
  map(char_colnames, ~ col_character()) %>% 
  setNames(char_colnames)
```

Final column specification:

```{r cols_type_to_use}
col_specs <- cols_only()
col_specs$cols <- c(char_cols, int_cols, real_cols)
col_specs
```

### Read data

Import occurrence data:

```{r read_occs_in_occ_eu}
occ_eu <- read_tsv(
  here::here("data", "raw", paste0(key, "_occurrence.txt")),
  na = "",
  quote = "",
  col_types = col_specs)
```

Number of occurrences

```{r n_occs_eu}
nrow(occ_eu)
```

Number of columns:

```{r n_cols_occ_eu}
ncol(occ_eu)
```

Preview:

```{r preview_occ_eu}
occ_eu %>% head()
```

# Filter data

## Filter on issues

Occurrences containing the following issues should be filtered out:

```{r issues_to_discard_occs_eu}
issues_to_discard <- c(
  "ZERO_COORDINATE",
  "COORDINATE_OUT_OF_RANGE", 
  "COORDINATE_INVALID"
)
names(issues_to_discard) <- issues_to_discard
issues_to_discard
```

Issues present in `occ_eu`:

```{r issues_in_data_occs_eu}
issues <-
  occ_eu %>% 
  distinct(issue) %>%
  separate(issue, into = "issues", sep = ";") %>%
  distinct() %>%
  arrange()
issues
```

Are there some `issues_to_discard` in `issues`?

```{r any_issue_to_discard_in_data_occs_eu}
any(issues_to_discard %in% issues$issues)
```

If yes, remove occurrences containing `issues_to_discard`:

```{r remove_occs_issues_to_discard_occs_eu}
if (any(issues_to_discard %in% issues$issues)) {
  occ_eu <- 
    map_dfc(issues_to_discard, 
        function(x) {
          str_detect(occ_eu$issue, x)
          }) %>% 
    mutate_all(funs(replace(., is.na(.), FALSE))) %>% 
    bind_cols(occ_eu) %>%
    filter_at(issues_to_discard, 
              all_vars(. == FALSE)) %>%
    select(-one_of(issues_to_discard))
}
```

## Filter on occurrence status

Occurrences with the following occurrence status should be filtered out:

```{r occurrenceStatus_to_discard_occs_eu}
occurrenceStatus_to_discard <- c(
  "absent",
  "excluded"
)
```

Occurrence status present in `occ_eu`:

```{r occ_status_in_data_occs_eu}
occurrenceStatus <-
  occ_eu %>% 
  distinct(occurrenceStatus) %>%
  distinct()
occurrenceStatus
```

Are there some `occurrenceStatus_to_discard` in `occurrenceStatus` (case insensitive search)?

```{r any_occurrenceStatus_to_discard_in_data_occs_eu}
any(occurrenceStatus_to_discard %in% 
      str_to_lower(occurrenceStatus$occurrenceStatus))
```

If yes, remove occurrences with `occurrenceStatus` equal to one of `occurrenceStatus_to_discard`:

```{r remove_occurrenceStatus_to_discard_occs_eu}
if (any(
  occurrenceStatus_to_discard %in% str_to_lower(occurrenceStatus$occurrenceStatus))) {
  occ_eu <- 
    occ_eu %>%
    filter(!str_to_lower(occurrenceStatus) %in% occurrenceStatus_to_discard)
}
```

## Filter on identification verification status

We won't take into account unverified observations:

```{r identificationVerificationStatus_to_discard_occs_eu}
identificationVerificationStatus_to_discard <- c(
  "unverified",
  "unvalidated",
  "not validated",
  "under validation",
  "not able to validate",
  "control could not be conclusive due to insufficient knowledge",
  "uncertain",
  "unconfirmed",
  "unconfirmed - not reviewed",
  "validation requested"
)
```

Identification verification status present in `occ_eu`:

```{r iidentification_status_in_data_occs_eu}
identificationVerificationStatus <-
  occ_eu %>% 
  group_by(identificationVerificationStatus) %>%
  count() %>%
  arrange(desc(n))
identificationVerificationStatus
```

Are there some `identificationVerificationStatus_to_discard` in `identificationVerificationStatus`?

```{r any_identificationStatus_to_discard_in_data_occs_eu}
any(identificationVerificationStatus_to_discard %in% str_to_lower(identificationVerificationStatus$identificationVerificationStatus))
```

If yes, remove occurrences with `identificationVerificationStatus` equal to one of `identificationVerificationStatus_to_discard`:

```{r remove_identificationStatus_to_discard_occs_eu}
if (any(
  identificationVerificationStatus_to_discard %in% str_to_lower(identificationVerificationStatus$identificationVerificationStatus))) {
  occ_eu <- 
    occ_eu %>%
    filter(!str_to_lower(identificationVerificationStatus) %in% identificationVerificationStatus_to_discard)
}
```

## Overview and control filtered data table

Number of occurrences left:

```{r nrow_occ_eu_after_filter}
nrow(occ_eu)
```

Preview:

```{r preview_occ_eu_after_filtering}
occ_eu %>% head()
```

# Assign grid

We assign grid to occurrences.

## Get geographic coordinates and coordinate uncertainty

Number of occurrences per each value of `coordinateUncertaintyInMeters`:

```{r n_occ_per_uncertainty_occs_eu}
occ_eu %>%
  group_by(coordinateUncertaintyInMeters) %>%
  count() %>%
  arrange(desc(n))
```

We assign 1000 meters to occurrences without uncertainty or occurrences with zero uncertainty:

```{r assign_fix_uncertainty_occs_eu}
occ_eu <- 
  occ_eu %>%
  mutate(
    coordinateUncertaintyInMeters = 
      if_else(is.na(coordinateUncertaintyInMeters) | 
                coordinateUncertaintyInMeters == 0,
              1000.0,
              coordinateUncertaintyInMeters)
)
```

We save geographic coordinates, `decimalLatitude` and `decimalLongitude` and coordinate uncertainty, `coordinateUncertaintyInMeters` as a new data.frame, `geodata_df`:

```{r geodata_df_occs_eu}
geodata_df <-
  occ_eu %>%
  select(decimalLatitude, 
         decimalLongitude, 
         coordinateUncertaintyInMeters)
nrow_geodata_df <- nrow(geodata_df)
```


## Project geographic coordinates

We project latitude and longitude by using the projection of the grid. We transform GBIF data which have coordinate reference system equal to EPSG code 4326 to Lambert projection with EPSG code 3035:

```{r transform_to_3035_occs_eu}
coordinates(geodata_df) <- ~decimalLongitude+decimalLatitude
proj4string(geodata_df) <- CRS("+init=epsg:4326")
geodata_df <- spTransform(geodata_df, CRS("+init=epsg:3035"))
colnames(geodata_df@coords) <- c("x", "y")
```

## Assign occurrence within uncertainty circle

Set random number generator seed (this helps reproducibility). We use the unique identifier of the reserved [Zenodo dataset's DOI](https://doi.org/10.5281/zenodo.10058400) which the occurrence cube will be published to:

```{r set_seed}
set.seed(10058400)
```

Assign the occurrence randomly within the circle with radius equal to `coordinateUncertaintyInMeters`:

```{r assign_pts_in_circle_occs_eu}
geodata_df@data <- 
  geodata_df@data %>% 
  mutate(random_angle = runif(nrow_geodata_df, 0, 2*pi))
geodata_df@data <- 
  geodata_df@data %>%
  mutate(random_r = sqrt(runif(
    nrow_geodata_df, 0, 1)) * coordinateUncertaintyInMeters)
geodata_df@data <- 
  geodata_df@data %>%
  mutate(x = geodata_df@coords[, "x"],
         y = geodata_df@coords[, "y"])
geodata_df@data <- 
  geodata_df@data %>%
  mutate(x = x + random_r * cos(random_angle),
         y = y + random_r * sin(random_angle))
geodata_df@data <- 
  geodata_df@data %>%
  select(-c(random_angle, random_r)) %>%
  select(x, y, coordinateUncertaintyInMeters)
```

Preview:

```{r preview_geodata_df_data_occs_eu}
geodata_df@data %>% head(n = 10)
```

Where `x` and `y` are the new coordinates while in `@coords` we keep track of the original coordinates:

```{r preview_geodata_df_coords_occs_eu}
geodata_df@coords[1:10,]
```

## Assign occurrences to grid cells

We assign each occurrence to a grid cell. 

```{r assign_grid_occs_eu}
geodata_df@data <- 
  geodata_df@data %>%
  mutate(eea_cell_code = paste0("1km", 
                                  "E", floor(x/1000), 
                                  "N", floor(y/1000)))
```
Preview:

```{r preview_gridcell_occs_eu}
geodata_df@data %>% head(n = 10)
```

We can now add the column `eea_cell_code` to `occ_eu`:

```{r add_eaa_cell_code_to_occ_eu}
occ_eu$eea_cell_code <- geodata_df@data$eea_cell_code
```

Preview:

```{r preview_with_eaa_cell_code_occs_eu}
occ_eu %>% head()
```

# Save data

Save occurrences with added EAA cell code in  `/data/interim`:

```{r save_occ_eu_after_filtering}
write_tsv(occ_eu, 
          here::here("data", "interim", paste0(key, "_occurrence.tsv")),
          na = "")
```