CS_09.Rmd

---
title: Tracking Hurricanes!
subtitle: Analyze historical storm data from the NOAA API
week: 9
type: Case Study
reading:
   - Overview of the [International Best Track Archive for Climate Stewardship (IBTrACS)](https://www.ncdc.noaa.gov/ibtracs/index.php?name=ibtracs-data-access)
   - Performing [Spatial Joins with sf](https://r-spatial.github.io/sf/reference/st_join.html)
tasks:
   - Write a .Rmd script to perform the following tasks
   - Use an API to access NOAA Storm data over the web
   - Intersect the storms with US states to quantify how many storms in the database have hit each state.
---

```{r, echo=FALSE, message=FALSE, results='hide', purl=FALSE}
source("functions.R")
source("knitr_header.R")
```

# Reading

```{r reading,results='asis',echo=F}
md_bullet(rmarkdown::metadata$reading)
```

# Tasks

```{r tasks,results='asis',echo=F}
md_bullet(rmarkdown::metadata$tasks)
```

## Libraries & Data

```{r message=F,warning=FALSE}
library(sf)
library(tidyverse)
library(ggmap)
library(rnoaa)
library(spData)
data(world)
data(us_states)
```

## Objective

In this case study you will download storm track data from NOAA, make a summary plot, and quantify how many storms have hit each of the United States.  This will require you to use a spatial join (`st_join`).

### Your goal
```{r download_data, echo=F, purl=F, warning=F}
storms <- storm_shp(basin = "NA")
storm_data <- read_sf(storms$path)
```

```{r, echo=F, purl=F, warning=F}
storms <- storm_data%>%
   filter(year>=1950)%>%
   mutate_if(is.numeric, 
             function(x) ifelse(x==-999.0,NA,x))%>% #-999 to NA
   mutate(decade=floor(year/10)*10) #add decade

region=st_bbox(storms)
```
Your desired figure looks something like the following:
```{r, echo=F, purl=F, warning=F, fig.width=15, fig.height=15}
map1=ggplot() +
  geom_sf(data=world,
          inherit.aes = F,size=.1,
          fill="grey",colour="black")+
  facet_wrap(~decade)+
  stat_bin2d(data=storms,
             aes(y=st_coordinates(storms)[,2],
                 x=st_coordinates(storms)[,1]),bins=100)+
  scale_fill_distiller(palette="YlOrRd",
                       trans="log",
                       direction=-1,
                       breaks = c(1,10,100,1000))+
  coord_sf(ylim=region[c(2,4)],
           xlim=region[c(1,3)])+
  labs(x="",y="")
map1
```


Calculate a table of the five states that have experienced the most storms.
```{r, echo=F, purl=F, warning=F, message=F}
library(knitr);library(kableExtra)
states<- st_transform(us_states,st_crs(storms))

storm_states <- st_join(storms, states, 
                join = st_intersects, 
                left = F)

storm_states%>%
   group_by(NAME)%>%
   summarize(storms=length(unique(Name)))%>%
   st_set_geometry(NULL)%>%
   arrange(desc(storms))%>%
   slice(1:5)%>%
   kable()%>%
   kable_styling()
```


<div class="well">
<button data-toggle="collapse" class="btn btn-primary btn-sm round" data-target="#demo1">Show Hints</button>
<div id="demo1" class="collapse">

## Steps

1. Use the API to Download storm data
   * Use `storm_shp()` for `basin = "NA"`
   * Read the points in with `storm_shp_read()`
   * Convert to `sf` format with `st_as_sf()`
2. Wrangle the data
   * Filter to storms 1950-present with `filter()`
   * Use `mutate_if()` to convert `-999.0` to `NA` in all numeric columns with the following command from the `dplyr` package: `mutate_if(is.numeric,` `function(x) ifelse(x==-999.0,NA,x))`
   * Use the following command to add a column for decade: `mutate(decade=(floor(year/10)*10))`
   * Use `st_bbox()` to identify the bounding box of the storm data and save this as an object called `region`.
3.  Make the first plot
   * Use `ggplot()` to plot the `world` polygon layer and add the following:
   * add `facet_wrap(~decade)` to create a panel for each decade
   * add `stat_bin2d(data=storms,` `aes(y=st_coordinates(storms)[,2],` `x=st_coordinates(storms)[,1]),bins=100)`
   * use 
`scale_fill_distiller(palette="YlOrRd",` 
`trans="log",` 
`direction=-1,`
`breaks = c(1,10,100,1000))` to set the color ramp
   * use `coord_sf(ylim=region[c(2,4)],`
   `xlim=region[c(1,3)])` to crop the plot to the region.
4. Calculate table of the five states with most storms.
   * use `st_transform` to reproject `us_states` to the reference system of the `storms` object (you can extract a CRS from a sf object with `st_crs(storms)`
   * Perform a spatial join between the storm database and the states object with: `storm_states <- st_join(storms, states, `
   `join = st_intersects,left = F)`.  This will 'add` the state to any storm that was recorded within that state.
   * Use `group_by(NAME)` to group the next step by US state (beware that there is `NAME` for name of state and `Name` for name of storm.  storm_states
   * use `summarize(storms=length(unique(Name)))` to count how many unique storms occurred in each state.
   * use `arrange(desc(storms))` to sort by the number of storms in each state
   * use `slice(1:5)` to keep only the top 5 states
```


</div>
</div>

---

<div class="extraswell">
<button data-toggle="collapse" class="btn btn-link" data-target="#extras">
Extra time? Try this...
</button>
<div id="extras" class="collapse">


Try to replicate the following graphic using the data you transformed above.

```{r, echo=F, purl=F, fig.height=10,fig.width=10}
storm_states_year <- storm_states%>%
   group_by(NAME,year)%>%
   summarize(storms=length(unique(Name)))

ggplot(storm_states_year,
#       aes(y=fct_reorder(NAME,storms),
       aes(y=NAME,
           x=year,
           fill=storms))+
   geom_tile()+
   scale_fill_viridis_c(name="# Storms")+
   ylab("State")
```

Can you sort the rows (states) in order of storm frequency (instead of alphabetical?
</div>
</div>