04_ana_armed_conflicts.Rmd

---
title: "Data analysis: Armed conflicts locations"
author: "Linda Petutschnig"
date: "25 5 2023"
output:
  html_document:
    theme: paper
    highlight: default
---

This script performs the following tasks:

- Applies a hotspot classification algorithm using a hexagonal grid as atomic units, categorizing the hexagons based on the number of ACLED events during a specified period.
- Converts the hotspot classes (no pattern, former hotspot, emerging hotspot, intermittent hotspot, persistent hotspot) into numeric values for better integration with other risk indicators.
- Writes the results to the hard drive ("hotspots_3months.gpkg").
 
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

```{r load, message = FALSE, warning = FALSE}

# Loads necessary packages
library(sf) # to manipulate vector features
library(dplyr) # for tidy data manipulation
library(here) # to access data via relative paths
library(sfhotspot) # to perform hotspot analysis based on vector features

# Add aceld.api
# https://www.rdocumentation.org/packages/acled.api/versions/1.1.5
# Citing ACLED: https://developer.acleddata.com/dashboard/terms-of-use/
 
```

```{r emerging hotspots, message = FALSE, warning = FALSE}

# Reads the hexagon shapes from hard drive
aoi_shape <- st_read(here("data/geopackages","AOI_hex5.gpkg")) %>% 
  st_cast("POLYGON") 

# Renames the column geom to geometry
st_geometry(aoi_shape) <- "geometry"

# Reads the ACLED data downloaded in acc_armed_conflicts.Rmd
acled_17_20_aoi <- st_read(here("data/geopackages","acled_17_20_aoi.gpkg"))

# As of 11.07.23 the hotspots_classify function gave an error, because in the just loaded dataset, some edges have duplicate
# vertexes with other edges. This has not been a problem before. 
# Adding the bit below solves the issue. 
sf_use_s2(FALSE)

# Calls the hotspot function on the just downloaded ACLED data
# The hexagons are used as grid. For "period", 1,3,6 and 12 months were tested and visually compared. 
# 3 months were chosen for the analysis because it gave the most
# distinct pattern. Other periods may work as well. 
hotspots_3m <- hotspot_classify(acled_17_20_aoi, time = event_date, period = "3 months", grid = aoi_shape)

# Plots the resulting hotspots
autoplot(hotspots_3m)

# Writes a column with the hotspots classes into aoi_shape 
aoi_shape$hotspot <- hotspots_3m$hotspot_category

# The hotspot categories are changed from categorical to nominal values to enable integration with the rest of the indicators. 
aoi_shape$hotspot_num[hotspots_3m$hotspot_category == 'no pattern'] <- "0"
aoi_shape$hotspot_num[hotspots_3m$hotspot_category == 'former hotspot'] <- "20"
aoi_shape$hotspot_num[hotspots_3m$hotspot_category == 'emerging hotspot'] <- "50"
aoi_shape$hotspot_num[hotspots_3m$hotspot_category == 'intermittent hotspot'] <- "80"
aoi_shape$hotspot_num[hotspots_3m$hotspot_category == 'persistent hotspot'] <- "100"

# Converts the newly defined values to numeric format
aoi_shape$hotspot_num <- as.numeric(aoi_shape$hotspot_num)
                                    
# Converts the result back into a MULTIPOLYGON, so that it can be matched with the other indicators.
aoi_shape_multipolygon <- aoi_shape %>%
  group_by(h3_index) %>%
  summarise(geometry = st_combine(geometry),
            hotspot_num = first(hotspot_num)) %>%
  st_cast("MULTIPOLYGON")


# Creates the map object
tmap::tm_shape(aoi_shape_multipolygon) +
  tmap::tm_polygons(col = "hotspot_num")

#  Writes the results to hard drive
st_write(aoi_shape_multipolygon, here("data/geopackages","AOI_hex5_ind1.gpkg"), append = FALSE)

```