Severe weather events like floods and tornadoes have cost about 0.5 trillion USD to the US economy
According to the U.S. National Oceanic and Atmospheric Administration’s (NOAA) storm database for the last six decades US economy has lost 477 billion USD through severe weather events, i.e. the annual GDP of a country like Sweden. More than 15 thousand persons have lost their lives and some 141 thousands have been wounded. Most of the economic consequences of the natural disasters are property-related damages. However agricultural losses are either non-neglegable. Tornadoes, excessive heat, lightning and to the lesser extent floods are mostly responsable for fatalities and injuries. Whereas the primary causes of property damages are floods, hurricanes and tornadoes. Drought is obviously the main reason of the crop damage. Taking health and property damages together floods, tornadoes and hurricanes have had the most pronouncing devastating effect. Some care schould be taken, though, with respect to the conclusions presented above. More we drill down to the history of observations less data are reliable.More profound data cleaning work has to be undertaken in order to get more precise results.
options(warn=-1)
library(ggplot2)
# the library contains functions allowing to combine several ggplots
library(gridExtra) ## Loading required package: gridData processing
Brief descriptions of the data cleaning stage. Only columns nedeed for the analysis have been selected. Property and crop damage data are converted into numeric format by means of an auxillary multiplier conversion table. Any number in “exp” columns is treated as a power of 10 (like thousands,millions etc.). Any non-numeric or unrecognizable symbol is omitted (property/crop damage is taken as it is). Then the pdf description file is parsed so as to get standardized types of weather events. Finally storm data is aggregated using these newly created standardized types. For this purposes a function was written (shown below). It takes the types as they were recorded historically and translates them into standardized ones. Certainly, it’s far from perfect and the correspondence between recorded and standardized types is often very approximate.
1.Processing NOAA Storm Database
# auxillary dataframe necessary for conversion property and crop losses into numeric monetary units
correspondUnit<-data.frame(c("H","K","M","B"),c(10^2,10^3,10^6,10^9))
names(correspondUnit)<-c("MAG","VALUE")
correspondUnit$MAG<-as.character(correspondUnit$MAG)
#reading data file (after prelimenary analysis only 7 columns selected)
data <- read.csv("repdata-data-StormData.csv.bz2",nrow=10)
names(data)## [1] "STATE__" "BGN_DATE" "BGN_TIME" "TIME_ZONE" "COUNTY"
## [6] "COUNTYNAME" "STATE" "EVTYPE" "BGN_RANGE" "BGN_AZI"
## [11] "BGN_LOCATI" "END_DATE" "END_TIME" "COUNTY_END" "COUNTYENDN"
## [16] "END_RANGE" "END_AZI" "END_LOCATI" "LENGTH" "WIDTH"
## [21] "F" "MAG" "FATALITIES" "INJURIES" "PROPDMG"
## [26] "PROPDMGEXP" "CROPDMG" "CROPDMGEXP" "WFO" "STATEOFFIC"
## [31] "ZONENAMES" "LATITUDE" "LONGITUDE" "LATITUDE_E" "LONGITUDE_"
## [36] "REMARKS" "REFNUM"storm_data<-read.csv("repdata-data-StormData.csv.bz2",header = TRUE,colClasses=c(rep("NULL",7),"character",rep("NULL",14),"numeric","numeric","numeric","character","numeric","character",rep("NULL",9)))
#converting property and crop damages into USD
property<-mapply(function(x,y){ifelse(grepl("[^0-9A-Za-z]",y)|y=="",x,ifelse(grepl("^[0-9]",y),x*(10^as.numeric(y)),x*correspondUnit[which(correspondUnit$MAG==toupper(y)),2]))},storm_data$PROPDMG,storm_data$PROPDMGEXP)
crop<-mapply(function(x,y){ifelse(grepl("[^0-9A-Za-z]",y)|y=="",x,ifelse(grepl("^[0-9]",y),x*(10^as.numeric(y)),x*correspondUnit[which(correspondUnit$MAG==toupper(y)),2]))},storm_data$CROPDMG,storm_data$CROPDMGEXP)
storm_data<-data.frame(storm_data,property,crop)
names(storm_data)[names(storm_data)=="property"]<-"PROPERTY"
names(storm_data)[names(storm_data)=="crop"]<-"CROP"
#some preparation of EVTYPE column for the next step
storm_data$EVTYPE<-as.character(storm_data$EVTYPE )
storm_data$EVTYPE<-sapply(storm_data$EVTYPE,function(x){toupper(x)})2.Processing Storm Data Documentation file
#reading and parsing pdf file and creating an character array of statndardized types of weather events
system("pdftotext 'repdata-peer2_doc-pd01016005curr.pdf' storm.txt -layout -f 2 -l 4")
storm_text<-readLines("storm.txt")
storm_text<-data.frame(storm_text)
names(storm_text)<-c("TYPE")
storm_text$TYPE<-as.character(storm_text$TYPE)
storm_text$TYPE<-sapply(storm_text$TYPE,function(x){gsub("^\\s+","",x)})
storm_text<-subset(storm_text,grepl("\\(M)?",storm_text$TYPE)|grepl("\\(Z)?",storm_text$TYPE)|grepl("\\(C)?",storm_text$TYPE),select=c(TYPE))
storm_text$TYPE<-sapply(storm_text$TYPE, function (x) { substr(x,(regexpr("[A-Z]",x))[1],(regexpr("\\(",x))[1]-1)})
storm_text$TYPE<-sapply(storm_text$TYPE,function(x){gsub(" $","",x)})3.Aggregating storm data across all types of events
#aggregation of the row data by the event types and excluding all-zero observations
ag_sd<-aggregate(cbind(FATALITIES,INJURIES,PROPERTY,CROP)~EVTYPE,data=storm_data,sum)
ag_sd<-subset(ag_sd,ag_sd$FATALITIES!=0|ag_sd$INJURIES!=0|ag_sd$PROPERTY!=0|ag_sd$CROP!=0)
#function "compare" makes the hidden work to create a new array with standardized event types
source("compare.R")
dd<-compare(ag_sd,storm_text)
ag_sd<-data.frame(ag_sd,dd$NEW)
names(ag_sd)[6]<-"EVTYPENEW"
#aggregation of data the by the standardized event types
ag_sd<-aggregate(cbind(FATALITIES,INJURIES,PROPERTY,CROP)~EVTYPENEW,data=ag_sd,sum)
ag_sd<-data.frame(ag_sd,(ag_sd$FATALITIES+ag_sd$INJURIES),(ag_sd$PROPERTY+ag_sd$CROP))
names(ag_sd)[c(6,7)]<-c("HEALTH","ECONOMIC")
ag_sd$EVTYPENEW<-as.character(ag_sd$EVTYPENEW)
ag_sd## EVTYPENEW FATALITIES INJURIES PROPERTY CROP
## 1 Astronomical Low Tide 0 0 320000 0
## 2 Avalanche 225 170 3721800 0
## 3 Blizzard 101 805 659813950 112060000
## 4 Coastal Flood 6 7 413482060 56000
## 5 Cold/Wind Chill 66 48 56554000 101142500
## 6 Dense Fog 81 1077 22829500 0
## 7 Dense Smoke 0 0 100000 0
## 8 Drought 2 4 1046106000 13972571780
## 9 Dust Devil 2 43 719130 0
## 10 Dust Storm 22 440 5619000 3600000
## 11 Excessive Heat 1920 6525 7753700 492402000
## 12 Extreme Cold/Wind Chill 496 427 79395400 1335623000
## 13 Flash Flood 1018 1785 17414731089 1437163150
## 14 Flood 530 6889 150423633386 10950975050
## 15 Freezing Fog 11 38 13504500 0
## 16 Frost/Freeze 8 35 19553200 1997061000
## 17 Funnel Cloud 0 3 194600 0
## 18 Hail 20 1466 16021881013 3111633873
## 19 Heat 1129 2544 12457050 407061500
## 20 Heavy Rain 103 306 3238944190 806505800
## 21 Heavy Snow 150 1122 982307750 134683100
## 22 High Surf 158 246 101475000 0
## 23 High Wind 293 1507 6003453043 686301900
## 24 Hurricane/Typhoon 135 1333 85356410010 5516117800
## 25 Ice Storm 103 2369 3959592360 5022114300
## 26 Lake-Effect Snow 0 0 40182000 0
## 27 Lakeshore Flood 0 0 7570000 0
## 28 Lightning 817 5232 930542430 12092090
## 29 Marine Hail 0 0 4000 0
## 30 Marine High Wind 49 59 47967591510 855000
## 31 Marine Strong Wind 15 22 438330 0
## 32 Marine Thunderstorm Wind 10 26 436400 50000
## 33 Other 93 86 328520600 163061400
## 34 Rip Current 577 529 163000 0
## 35 Seiche 0 0 980000 0
## 36 Sleet 2 0 2000000 0
## 37 Storm Tide 0 0 9425150 0
## 38 Strong Wind 139 395 191658740 70813500
## 39 Thunderstorm Wind 731 9464 11149296450 1207144288
## 40 Tornado 5658 91364 58552152426 417461520
## 41 Tropical Depression 0 0 1737000 0
## 42 Tropical Storm 66 383 7714390550 694896000
## 43 Tsunami 34 131 144062000 20000
## 44 Volcanic Ash 0 0 500000 0
## 45 Waterspout 6 72 60730200 0
## 46 Wildfire 90 1608 8501628500 403281630
## 47 Winter Storm 217 1353 6748997251 32444000
## 48 Winter Weather 62 615 27310500 15000000
## HEALTH ECONOMIC
## 1 0 320000
## 2 395 3721800
## 3 906 771873950
## 4 13 413538060
## 5 114 157696500
## 6 1158 22829500
## 7 0 100000
## 8 6 15018677780
## 9 45 719130
## 10 462 9219000
## 11 8445 500155700
## 12 923 1415018400
## 13 2803 18851894239
## 14 7419 161374608436
## 15 49 13504500
## 16 43 2016614200
## 17 3 194600
## 18 1486 19133514886
## 19 3673 419518550
## 20 409 4045449990
## 21 1272 1116990850
## 22 404 101475000
## 23 1800 6689754943
## 24 1468 90872527810
## 25 2472 8981706660
## 26 0 40182000
## 27 0 7570000
## 28 6049 942634520
## 29 0 4000
## 30 108 47968446510
## 31 37 438330
## 32 36 486400
## 33 179 491582000
## 34 1106 163000
## 35 0 980000
## 36 2 2000000
## 37 0 9425150
## 38 534 262472240
## 39 10195 12356440738
## 40 97022 58969613946
## 41 0 1737000
## 42 449 8409286550
## 43 165 144082000
## 44 0 500000
## 45 78 60730200
## 46 1698 8904910130
## 47 1570 6781441251
## 48 677 42310500Results
1.Across the United States, which types of events are most harmful with respect to population health?
ag_sd<-ag_sd[order(-ag_sd$FATALITIES),]
health_chart1<-ggplot(head(ag_sd,5), aes(x=reorder(EVTYPENEW,FATALITIES), y=FATALITIES)) +geom_bar(stat='identity',fill="#4A4A4A",position = 'dodge') +coord_flip()+ylab("Persons killed")+xlab("Types of events")+ggtitle("TOP FIVE DEADLIEST WEATHER EVENTS")
ag_sd<-ag_sd[order(-ag_sd$INJURIES),]
health_chart2<-ggplot(head(ag_sd,5), aes(x=reorder(EVTYPENEW,INJURIES), y=INJURIES)) +geom_bar(stat='identity',fill="#4A4A4A",position = 'dodge') +coord_flip()+ylab("Persons injured")+xlab("Types of events")+ggtitle("TOP FIVE WEATHER EVENTS CAUSING MOST INJURIES")
ag_sd<-ag_sd[order(-ag_sd$HEALTH),]
health_chart3<-ggplot(head(ag_sd,5), aes(x=reorder(EVTYPENEW,HEALTH), y=HEALTH)) +geom_bar(stat='identity',fill="#4A4A4A",position = 'dodge') +coord_flip()+ylab("Persons killed or injuried")+xlab("Types of events")+ggtitle("TOP FIVE MOST HARMFUL WEATHER EVENTS (FATALITIES AND INJURIES COMBINED)")
health_chart<- arrangeGrob(health_chart1, health_chart2, health_chart3, ncol=2,main = textGrob("\n1.WEATHER EVENTS WITH MOST SEVERE HEALTH IMPACT",gp = gpar(fontsize=25)))
health_chart
Tornadoes, floods and heat have lethal effect. These events alongside with thunderstorm wind are also primary causes of injuries.
2.Across the United States, which types of events have the greatest economic consequences?
ag_sd<-ag_sd[order(-ag_sd$PROPERTY),]
econ_chart1<-ggplot(head(ag_sd,5), aes(x=reorder(EVTYPENEW,PROPERTY), y=PROPERTY/10^9)) +geom_bar(stat='identity',fill="#4A4A4A") +coord_flip()+ylab("Billion USD")+xlab("Types of events")+ggtitle("TOP FIVE COSTLIEST WEATHER EVENTS (PROPERTY DAMAGE)")
ag_sd<-ag_sd[order(-ag_sd$CROP),]
econ_chart2<-ggplot(head(ag_sd,5), aes(x=reorder(EVTYPENEW,CROP), y=CROP/10^9)) +geom_bar(stat='identity',fill="#4A4A4A") +coord_flip()+ylab("Billion USD")+xlab("Types of events")+ggtitle("TOP FIVE COSTLIEST WEATHER EVENTS(AGRICULTIRUAL DAMAGE)")
ag_sd<-ag_sd[order(-ag_sd$ECONOMIC),]
econ_chart3<-ggplot(head(ag_sd,5), aes(x=reorder(EVTYPENEW,ECONOMIC), y=ECONOMIC/10^9)) +geom_bar(stat='identity',fill="#4A4A4A") +coord_flip()+ylab("Billion USD")+xlab("Types of events")+ggtitle("TOP FIVE COSTLIEST WEATHER EVENTS(PROPERTY AND AGRICULTURE COMBINED)")
econ_chart<- arrangeGrob(econ_chart1, econ_chart2, econ_chart3, ncol=2,main = textGrob("\n2.WEATHER EVENTS WITH MOST SEVERE ECONOMIC IMPACT",gp = gpar(fontsize=25)))
econ_chart 
Floods, tornadoes and hurricanes produce the most severe property damages. Draught and floods are main causes of the crop losses.
3.Across the United States, which types of events produce most devastating effect on economy and population together?
ag_sd<-ag_sd[order(ag_sd$EVTYPENEW),]
mat<-data.matrix(ag_sd[c(6,7)],)
rownames(mat)<-c(ag_sd$EVTYPE)
my_palette <- colorRampPalette(c("#F4A460", "#8B3626"))(n = 50)
hm <- heatmap(mat, Rowv=NA, Colv=NA, col = my_palette, scale="column", cexCol=1.2, main="What types of weather events we have suffered of most?") 
Tornadoes are most devastating if we combine health and economic damages.
Function Compare.R
compare <- function(ar1,ar2 ) {
OLD<-ar1\(EVTYPE NEW<-rep("NA",nrow(ar1)) ar3<-data.frame(OLD,NEW) ar3\)NEW<-as.character(ar3\(NEW) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl(".*URBAN.*SMALL*",x),"FLOOD",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl(".*HURRICANE.*|.*TYPHOON.*",x),"HURRICANE/TYPHOON",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("^AVAL",x),"AVALANCHE",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("^HYP",x),"HEAT",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("WINTR",x),gsub("WINTR*","WINTER WEATHER",x),x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl(".*FIRE.*",x),"WILDFIRE",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("^IC.*ROAD.*",x),"FROST/FREEZE",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("HVY",x),gsub("HVY","HEAVY",x),x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("*.BURST|GUST.*",x),"THUNDERSTORM WINDS",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl(".*FROST.*|.*FREEZE.*",x),"FROST/FREEZE",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl(".*EXTREME|EXTENDED.*COLD|CHILL.*",x),"EXTREME COLD/WIND CHILL",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl(".*GLAZE.*",x),"ICE STORM",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("SHOWER",x),gsub("SHOWER","RAIN",x),x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("RAINFALL",x),"HEAVY RAIN",x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("PRECIPITATION",x),gsub("PRECIPITATION","RAIN",x),x)}) ar1\)EVTYPE<-sapply(ar1\(EVTYPE,function(x){ifelse(grepl("LAKE EFFECT",x),gsub("LAKE EFFECT","LAKE-EFFECT",x),x)}) ar1\)EVTYPE<-sapply(ar1$EVTYPE,function(x){ifelse(grepl(“LAKE FLOOD”,x),gsub(“LAKE FLOOD”,“LAKESHORE FLOOD”,x),x)})
for (i in 1:nrow(ar2)) {
for (j in 1:nrow(ar1)) {
if (ar3$NEW[j]=="NA")
{
if (grepl(paste0("^",ar2$TYPE[i]),ar1$EVTYPE[j],ignore.case=TRUE))
{
ar3$NEW[j]<-ar2$TYPE[i]
}
}
} } for (i in 1:nrow(ar2)) { for (j in 1:nrow(ar1)) {
if (ar3$NEW[j]=="NA")
{
if (grepl(ar2$TYPE[i],ar1$EVTYPE[j],ignore.case=TRUE))
{
ar3$NEW[j]<-ar2$TYPE[i]
}
} } } ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.FREEZING.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Freezing Fog”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*HEAVY SEAS.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Marine High Wind",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.HIGH SEAS.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Marine High Wind”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*HIGH WATER.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Marine High Wind",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.ROUGH SEAS.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Marine High Wind”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*PRECIP.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Heavy Rain",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.EROSION.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Coastal Flood”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*DUST.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Dust Storm",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.FOG.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Dense Fog”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl("^COLD",ar3\)OLD[i]),ar3\(NEW[i]<-"Cold/Wind Chill",ar3\)NEW[i])})
ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.THU.|.TU.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Thunderstorm Wind”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*TSTM.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Thunderstorm Wind",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“[^HEAVY]?.SNOW“,ar3\(OLD[i]),ar3\)NEW[i]<-”Heavy Snow“,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*RAIN.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Heavy Rain",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==”NA“&grepl(”.STORM.“,ar3\(OLD[i]),ar3\)NEW[i]<-”Marine High Wind“,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*WIND.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Strong Wind",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==”NA“&grepl(”.ICE.“,ar3\(OLD[i]),ar3\)NEW[i]<-”Ice Storm“,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*RISING WATER*",ar3\)OLD[i]),ar3\(NEW[i]<-"Flood",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==”NA“&grepl(”.TORN.“,ar3\(OLD[i]),ar3\)NEW[i]<-”Tornado“,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*SURF.*",ar3\)OLD[i]),ar3\(NEW[i]<-"High Surf",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==”NA“&grepl(”.LIG.ING.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Lightning”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*TIDE.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Storm Tide",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.WAVE.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Tsunami”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*COLD.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Cold/Wind Chill",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.COOL.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Cold/Wind Chill”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*MIX.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Heavy Rain",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.SURGE.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Marine High Wind”,ar3\(NEW[i])}) ar3\)NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\(NEW[i]=="NA"&grepl(".*SWELL.*",ar3\)OLD[i]),ar3\(NEW[i]<-"Marine Strong Wind",ar3\)NEW[i])}) ar3\(NEW<-sapply(1:nrow(ar3),function(i){ifelse(ar3\)NEW[i]==“NA”&grepl(“.FLD.”,ar3\(OLD[i]),ar3\)NEW[i]<-“Flood”,ar3\(NEW[i])}) ar3\)NEW<-sapply(ar3$NEW,function(x){ifelse(x==“NA”,x<-“Other”,x)}) return (ar3) }