Weather Events and their Impacts on Human Health and Economics

Data Subsetting

Since we’re only interested in:

• Event Type EVTYPE
• Fatalities FATALITIES
• Injuries INJURIES
• Damange to Property PROPDMGEPROPDMGEXP
• Damange to Crops CROPDMGECROPDMGEXP

dmg <- df[,c("EVTYPE", "FATALITIES", "INJURIES", "PROPDMG", "PROPDMGEXP", "CROPDMG", "CROPDMGEXP")]

Quantifying Data

Making the Data Visualization-ready

The dataset used some interesting notation to represent the amount of damage done to their crops and property. For example,

is represented in two columns as,

Let’s have a look at the "xEXP" values present in the dataset.

unique(dmg$PROPDMGEXP)

 [1] K M   B m + 0 5 6 ? 4 2 3 h 7 H - 1 8
Levels:  - ? + 0 1 2 3 4 5 6 7 8 B h H K m M

unique(dmg$CROPDMGEXP)

[1]   M K m B ? 0 k 2
Levels:  ? 0 2 B k K m M

Corresponding to each unique symbol, we need to assign a numeric value. Let’s create key-value pairs with our assigned exponents as values to facilitate our transition from this current state (shown here) to this.

1. Change all xEXP entries to uppercase.

cols <- c("PROPDMGEXP", "CROPDMGEXP")
dmg %<>% mutate_at(cols, toupper)
dmg <- as.data.table(dmg)

2. Map property and crop damage alphanumeric exponents to numeric values.

propDmgKey <-  c("\"\"" = 10^0,
                 "-" = 10^0, 
                 "+" = 10^0,
                 "0" = 10^0,
                 "1" = 10^1,
                 "2" = 10^2,
                 "3" = 10^3,
                 "4" = 10^4,
                 "5" = 10^5,
                 "6" = 10^6,
                 "7" = 10^7,
                 "8" = 10^8,
                 "9" = 10^9,
                 "H" = 10^2,
                 "K" = 10^3,
                 "M" = 10^6,
                 "B" = 10^9)

cropDmgKey <-  c("\"\"" = 10^0,
                "?" = 10^0, 
                "0" = 10^0,
                "K" = 10^3,
                "M" = 10^6,
                "B" = 10^9)

3. Replace the values in PROPDMGEXP with their corresponding numeric values.

dmg[, PROPDMGEXP := propDmgKey[as.character(dmg[,PROPDMGEXP])]]
dmg[is.na(PROPDMGEXP), PROPDMGEXP := 10^0 ]

dmg[, CROPDMGEXP := cropDmgKey[as.character(dmg[,CROPDMGEXP])] ]
dmg[is.na(CROPDMGEXP), CROPDMGEXP := 10^0 ]

NOTE: Using mutate_at along with a key-value fetcher function for this task didn’t work as intended and instead, copied the same values across all rows. This is why it was avoided in this chunk.

4. Use mutate to create columns for PropertyDamage, CropDamage and TotalDamage and get rid of the raw column data

dmg %<>% 
  mutate(PropertyDamage = PROPDMG*PROPDMGEXP, CropDamage = CROPDMG*CROPDMGEXP) %>% 
  mutate(TotalDamage = PropertyDamage+CropDamage, TotalCasualties = FATALITIES+INJURIES) %>% 
  select(-c("PROPDMG", "PROPDMGEXP", "CROPDMG", "CROPDMGEXP")) 

5. Preview the Dataframe before proceeding

dmg[0:100,]

Summary of Data

Let’s first summarize the data to find the total Economic and Manpower Damage of these Weather events

damageReport <- dmg %>% 
                  group_by(EVTYPE)  %>% 
                  summarize(PropertyDamage = sum(PropertyDamage),
                            CropDamage = sum(CropDamage),
                            TotalDamage = sum(TotalDamage),
                            Injuries = sum(INJURIES),
                            Fatalities = sum(FATALITIES),
                            TotalCasualties = sum(TotalCasualties))
damageReport

Prepping up the Data for Bar Plotting

Since ggplot works by plotting a set of variables grouped by a fill aesthetic, we need to melt our report data to make our report fully plot-ready. But first, to split it into its human and economic facets:

econLosses <- damageReport[ ,c("EVTYPE","PropertyDamage","CropDamage", "TotalDamage")]
humanLosses <- damageReport[, c("EVTYPE","Injuries", "Fatalities", "TotalCasualties")]

Then, to order by Total Losses

econLosses <- econLosses[order(econLosses$TotalDamage, decreasing = TRUE),]
humanLosses <- humanLosses[order(humanLosses$TotalCasualties, decreasing = TRUE),]

Take the Top 10 most costly events for each. The rest will merely clutter up our graph

econLosses <- econLosses[1:10,]
humanLosses <- humanLosses[1:10,]

Finally, melt the data

econLosses <- melt(econLosses)
humanLosses <- melt(humanLosses)

Plotting

# Specify Aesthetic mappings
plot <- ggplot(humanLosses, aes(x = reorder(EVTYPE, -value), y= value, fill= variable))

# Specify Bar Chart specs
plot = plot + geom_bar(stat = "identity", position = "dodge")

# Set x-axis label to blank
plot = plot + xlab(element_blank())

# Set y-axis label
plot = plot + ylab("Casualties") 

# Prevent clutter around the x-axis
plot = plot + theme(axis.text.x = element_text(angle=45, hjust=1), axis.title.x = element_blank())

# Set chart title and center it
plot = plot + ggtitle("Top 10 Deadliest Events") + theme(plot.title = element_text(hjust = 0.5))

plot

A Bar Plot of the Top 10 Events with the Most Reported Casualties

# Specify Aesthetic mappings
plot <- ggplot(econLosses, aes(x = reorder(EVTYPE, -value), y= value, fill= variable))

# Specify Bar Chart specs
plot = plot + geom_bar(stat = "identity", position = "dodge")

# Set y-axis label
plot = plot + ylab("Cost ($)") 

# Prevent clutter around the x-axis
plot = plot + theme(axis.text.x = element_text(angle=45, hjust=1), axis.title.x = element_blank())

# Set chart title and center it
plot = plot + ggtitle("Top 10 Costliest Events") + theme(plot.title = element_text(hjust = 0.5))

plot

A Bar Plot of the Top 10 Costliest Types of Weather Events

The data for Casualties is pretty clear on what the major contributor to weather-event related deaths is with Tornados taking a sizeable lead over the rest in both Injuries as well as Fatalities. The following four events are tied pretty evenly with each other while the events further down the list start appearing progressively insignificant next to one another.

On the other hand, the data on Economic Costs a steady progression a la Zipf’s Law with Floods still holding an indisputable lead over damages to Property and Crops with Typhoon and Tornados (no less) not very far behind. Interestingly, one may also notice that a few Events show extreme selectivity towards one type of Economic Resource. Upon closer inspection, however, it seems obvious why.