table_s2_bd_in_mpas_by_eez.Rmd

---
title: 'MPAs and biodiversity risk by EEZ'
author: "*Compiled on `r date()` by `r Sys.info()['user']`*"
output: 
  html_document:
    code_folding: hide
    toc: true
    toc_depth: 3
    toc_float: yes
    number_sections: true
    theme: cerulean
    highlight: haddock
    includes: 
      in_header: '~/github/src/templates/ohara_hdr.html'
  pdf_document:
    toc: true
---

``` {r setup, echo = TRUE, message = FALSE, warning = FALSE}

knitr::opts_chunk$set(fig.width = 6, fig.height = 4, fig.path = 'figs/',
                      echo = TRUE, message = FALSE, warning = FALSE)

library(raster)
library(data.table)

source('https://raw.githubusercontent.com/oharac/src/master/R/common.R')
  ### includes library(tidyverse); library(stringr); 
  ### dir_M points to ohi directory on Mazu; dir_O points to home dir on Mazu

dir_git <- '~/github/spp_risk_dists'

### goal specific folders and info
dir_data  <- file.path(dir_git, 'data')
dir_o_anx <- file.path(dir_O, 'git-annex/spp_risk_dists')

source(file.path(dir_git, '_setup/common_fxns.R'))

```

# Summary

Compare MPAs to biodiversity risk in table format.

# Methods

## Compare MPA coverage to biodiversity risk

MPA coverage will be compared to non-MPA areas within EEZs.  The .csv codes MPAs as their numeric IUCN category (i.e. Ia = Ib = 1, II = 2, ..., VI = 6); areas that are specifically designated as no-take (that are not already in categories Ia, Ib, and II) are coded as -1; areas that are designated in the WDPA database but do not fall into one of these categories are coded as 8.

``` {r set up mpa and values dataframe}

mean_rast     <- raster(file.path(dir_git, '_output', 'mean_risk_raster_comp.tif'))
eez_rast      <- raster(file.path(dir_git, '_spatial', 'eez_rast.tif'))
ocean_area_rast <- raster(file.path(dir_git, '_spatial', 'ocean_area_rast.tif'))

rgn_labels <- read_csv(file.path(dir_git, '_spatial/rgn_names.csv')) %>%
  mutate(r1_label = str_replace(r1_label, 'Americas', 'N. America'),
         r1_label = str_replace(r1_label, 'Latin Am.+', 'S. America'),
         r1_label = ifelse(str_detect(r2_label, 'Central America|Caribbean'), 
                           'C. America/Caribbean', r1_label),
         r1_label = ifelse(rgn_id == 135, 'N. America', r1_label)) %>%
  mutate(rgn_label = str_replace_all(rgn_label, 'Islands?', 'Is.'),
         rgn_label = str_replace_all(rgn_label, ' and ', ' & '),
         rgn_label = str_replace_all(rgn_label, ' the ', ' '),
         rgn_label = str_replace(rgn_label, 'R.union', 'Reunion'),
         rgn_label = str_replace_all(rgn_label, 'Saint', 'St.'),
         rgn_label = str_replace_all(rgn_label, 'North(ern)?', 'N.'),
         rgn_label = str_replace_all(rgn_label, 'South(ern)?', 'S.'),
         rgn_label = str_replace(rgn_label, 'Virgin Is. of United States', 'US Virgin Is.'),
         rgn_label = str_replace(rgn_label, 'Democratic Republic', 'Dem. Rep.'),
         rgn_label = str_replace(rgn_label, 'Territory', 'Terr.'))

risk_df <- data.frame(cell_id    = 1:length(values(mean_rast)),
                      ocean_area = values(ocean_area_rast),
                      eez        = values(eez_rast),
                      mean_risk  = values(mean_rast)) %>%
  left_join(rgn_labels, by = c('eez' = 'rgn_id')) %>%
  filter(!is.na(mean_risk))

```

``` {r set up no take vs risk df}

calc_mpa_areas <- function(df) {
  ### ensure MPA area is no larger than ocean area of the cell; and
  ### code NA as zero for means and weights calculations
  df %>%
    mutate(mpa = ifelse(is.na(mpa), 0, mpa),
           mpa = ifelse(mpa > ocean_area, ocean_area, mpa),
           non_mpa = ocean_area - mpa) %>%
    gather(mpa_status, area, mpa, non_mpa) %>%
    mutate(mpa_status = (mpa_status == 'mpa'))
}

mpa_df   <- read_csv(file.path(dir_git, '_spatial', 'wdpa_mpa_area.csv'),
                          col_types = 'dddd') %>%
  filter(wdpa_category <= 2) %>%
  select(-wdpa_category) %>%
  group_by(cell_id) %>%
  summarize(mpa = sum(prot_area_km2, na.rm = TRUE))

risk_mpa_df <- risk_df %>%
  left_join(mpa_df, by = 'cell_id') %>%
  calc_mpa_areas()

```

``` {r no take mpas, eval = TRUE}

rgn_no_take <- risk_mpa_df %>%
  filter(eez < 255 & eez != 213) %>%
  mutate(lbl = rgn_label, rgn = r1_label) %>%
  arrange(rgn, lbl) %>%
  mutate(lbl = forcats::fct_inorder(lbl))

prot_pct <- rgn_no_take %>%
  group_by(lbl) %>%
  summarize(pct_prot = sum(area * mpa_status) / sum(area) * 100,
            tot_area = sum(area))
### this is independent of range-rarity weighting

no_take_means <- rgn_no_take %>%
  group_by(mpa_status, lbl) %>%
  summarize(mu = sum(mean_risk * area) / sum(area)) %>%
  ungroup()
  
```

``` {r set up overall rr-weighted dataframe}

mean_rr_rast      <- raster(file.path(dir_git, '_output', 'mean_rr_risk_raster_comp.tif'))

risk_rr_df <- data.frame(cell_id    = 1:length(values(mean_rr_rast)),
                         ocean_area = values(ocean_area_rast),
                         eez        = values(eez_rast),
                         mean_risk  = values(mean_rr_rast)) %>%
  left_join(rgn_labels, by = c('eez' = 'rgn_id')) %>%
  filter(!is.na(mean_risk))

```

``` {r set up no take vs risk rr df}

risk_mpa_rr_df <- risk_rr_df %>%
  left_join(mpa_df, by = 'cell_id') %>%
  calc_mpa_areas()

```

``` {r rr-weighted protection level dataframes}

rgn_no_take_rr <- risk_mpa_rr_df %>%
  filter(eez < 255 & eez != 213) %>%
  mutate(lbl = rgn_label, rgn = r1_label) %>%
  arrange(rgn, lbl) %>%
  mutate(lbl = forcats::fct_inorder(lbl))

no_take_rr_means <- rgn_no_take_rr %>%
  group_by(mpa_status, lbl) %>%
  summarize(mu = sum(mean_risk * area) / sum(area)) %>%
  ungroup()

rgn_no_take_rr <- risk_mpa_rr_df %>%
  filter(eez < 255 & eez != 213) %>%
  mutate(lbl = rgn_label, rgn = r1_label) %>%
  arrange(rgn, lbl) %>%
  mutate(lbl = forcats::fct_inorder(lbl))

no_take_rr_means <- rgn_no_take_rr %>%
  group_by(mpa_status, lbl) %>%
  summarize(mu = sum(mean_risk * area) / sum(area)) %>%
  ungroup()
  
```

There seem to be two philosophies of MPA priority: place MPAs in impacted areas to reduce pressure and prevent further degradation, or place MPAs to protect pristine places to maintain their pristine state.  This suggests that we may see a bimodal distribution of mean risk within MPA cells relative to the distribution of mean risk in general.

### Mean risk

For the EEZ-level analysis, generate a table with total area, pct protected area, mean protected/unprotected, 

```{r}

library(sf)
rgn_shp <- read_sf(file.path(dir_M, 'git-annex', 'globalprep/spatial/v2017', 
                             'regions_2017_update.shp'))
ocean_area_from_shp <- rgn_shp %>%
  as.data.frame() %>%
  filter(rgn_type == 'eez') %>%
  select(rgn_id, rgn_name, area_km2) %>%
  left_join(rgn_labels, by = c('rgn_id')) %>%
  select(rgn_label, area_km2)

rgn_no_take_sum <- rgn_no_take %>%
  mutate(mpa_status = ifelse(mpa_status, 'mpa', 'nonmpa')) %>%
  group_by(lbl, rgn, mpa_status) %>%
  summarize(mean_risk = sum(mean_risk * area) / sum(area)) %>%
  ungroup() %>%
  spread(mpa_status, mean_risk)

table_sum <- rgn_no_take_sum %>%
  full_join(prot_pct, by = 'lbl') %>%
  mutate(mpa = ifelse(is.na(mpa), 0, mpa)) %>%
  full_join(ocean_area_from_shp, by = c('lbl' = 'rgn_label')) %>%
  mutate(risk_diff = mpa - nonmpa) %>%
  arrange(risk_diff)

table_format <- table_sum %>%
  filter(pct_prot > 0.5 & tot_area > 1000) %>%
  mutate(pct_prot  = round(pct_prot, 1),
         area_1000km2 = formatC(area_km2 / 1000, big.mark = ',', format = 'f', digits = 1),
         mpa       = round(mpa, 3), 
         nonmpa    = round(nonmpa, 3), 
         risk_diff = round(risk_diff, 3)) %>%
  select(`Region` = lbl,
         `Georegion` = rgn,
         `Area (x10³ km²)` = area_1000km2,
         `Pct marine reserve` = pct_prot,
         `μ₁` = mpa,
         `μ₀` = nonmpa,
         `Risk differential<br>(μ₁ - μ₀)` = risk_diff)

write_csv(table_format, 'ms_figures/table_s2_risk_diff_by_eez.csv')
knitr::kable(table_format, 
             caption = 'Mean risk inside and outside marine reserves, for EEZs greater than 
                        1000 km² and reserves totalling more than 0.5% of EEZ.  Negative 
                        risk differential indicates preference for protecting areas of 
                        low biodiversity risk within EEZ.')
```

```{r}

rgn_no_take_rr_sum <- rgn_no_take_rr %>%
  mutate(mpa_status = ifelse(mpa_status, 'mpa', 'nonmpa')) %>%
  group_by(lbl, rgn, mpa_status) %>%
  summarize(mean_risk = sum(mean_risk * area) / sum(area)) %>%
  ungroup() %>%
  spread(mpa_status, mean_risk)

table_rr_sum <- rgn_no_take_rr_sum %>%
  full_join(prot_pct, by = 'lbl') %>%
  mutate(mpa = ifelse(is.na(mpa), 0, mpa)) %>%
  full_join(ocean_area_from_shp, by = c('lbl' = 'rgn_label')) %>%
  mutate(risk_diff = mpa - nonmpa) %>%
  arrange(risk_diff)

table_rr_format <- table_rr_sum %>%
  filter(pct_prot > 0.5 & tot_area > 1000) %>%
  mutate(pct_prot  = round(pct_prot, 1),
         area_1000km2 = formatC(area_km2 / 1000, big.mark = ',', format = 'f', digits = 1),
         mpa       = round(mpa, 3), 
         nonmpa    = round(nonmpa, 3), 
         risk_diff = round(risk_diff, 3)) %>%
  select(`Region` = lbl,
         `Georegion` = rgn,
         `Area (x10³ km²)` = area_1000km2,
         `Pct marine reserve` = pct_prot,
         `μ₁` = mpa,
         `μ₀` = nonmpa,
         `Risk differential<br>(μ₁ - μ₀)` = risk_diff)

write_csv(table_rr_format, 'ms_figures/table_s3_risk_diff_rr_by_eez.csv')
knitr::kable(table_rr_format, 
             caption = 'Mean risk inside and outside marine reserves, for EEZs greater than 
                        1000 km² and reserves totalling more than 0.5% of EEZ.  Negative 
                        risk differential indicates preference for protecting areas of 
                        low biodiversity risk within EEZ.')
```