Bird strike data
2024-08-20
Proportion of samples for which there were strikes
This document shows strike data for systematic observations only.
Data wrangling
I took the data frame and created a new variable for season. Basically, any observation before June is “spring” and any observation after June is “fall.”
bxb$season <- as.character(bxb$Date)
bxb$season [bxb$Date %in% 0:60000] <- "spring"
bxb$season [bxb$Date %in% 60001:123026] <- "fall"I created a table that shows the number of strikes per building, year, and season. Only the first 6 rows are shown as an example.
strike.table <- bxb %>% group_by(Building, Year, season) %>%
summarise(no.strikes=sum(Systematic))head(strike.table)## # A tibble: 6 × 4
## # Groups: Building, Year [6]
## Building Year season no.strikes
## <fct> <int> <chr> <int>
## 1 ACT 2019 spring 2
## 2 ACT 2020 spring 1
## 3 ACT 2021 spring 1
## 4 ACT 2022 spring 3
## 5 ACT 2023 spring 2
## 6 CFWC 2019 spring 0I created a table that shows the proportion of strikes per building, year, and season relative to sampling effort. Again, only the first 6 rows are shown as an example.
total.table <- bxb %>% group_by(Building, Year, season) %>%
summarise(total.obs=length(Building))total.table$proportion = c(strike.table$no.strikes / total.table$total.obs)
total.table$Year = factor(total.table$Year)
head(total.table)## # A tibble: 6 × 5
## # Groups: Building, Year [6]
## Building Year season total.obs proportion
## <fct> <fct> <chr> <int> <dbl>
## 1 ACT 2019 spring 29 0.0690
## 2 ACT 2020 spring 17 0.0588
## 3 ACT 2021 spring 29 0.0345
## 4 ACT 2022 spring 28 0.107
## 5 ACT 2023 spring 26 0.0769
## 6 CFWC 2019 spring 29 0Graphics
Boxplot: proportion of strikes per building.
This combines all years and seasons. Reeve and Sage have the highest.
total.table$Building = factor(total.table$Building)
#levels(total.table$Building)
boxplot(total.table$proportion ~ total.table$Building, xlab = "Building", ylab = "Proportion of samples")
A quick ANOVA shows that there are significant differences but I haven’t sorted out where the significance lies (though one can guess).
summary(aov(total.table$proportion ~ total.table$Building))## Df Sum Sq Mean Sq F value Pr(>F)
## total.table$Building 7 0.2087 0.029822 3.802 0.00412 **
## Residuals 32 0.2510 0.007844
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Scatterplot:
window strikes relative to sampling effort
The colors represent years and the shapes represent seasons.
ggplot(total.table,aes(x=Building,y=proportion, color = Year, shape = season))+
geom_jitter(width = 0.2, size = 3)+
labs(x="Building",y="Proportion of samples with strikes",title="Window strikes relative sampling effort")
separated out by year
Again, the colors represent years and the shapes represent seasons.
ggplot(total.table,aes(x=Building,y=proportion, color = Year, shape = season))+
geom_jitter(size = 2.5)+
labs(x="Building",y="Proportion of strikes",title="Window strikes as a proportion of sampling effort")+
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))+
facet_wrap(~Year)
Heatmaps:
The next three graphs are heat maps that might not be helpful but here they are. For each, the color represents proportion of samples that had strikes with the darker colors having a greater proportion.
strikes per building per year
The color represents proportion of samples that had strikes with the darker colors having a greater proportion.
ggplot(total.table, aes(Year, Building, fill= proportion)) +
geom_tile()+
scale_fill_gradient2(low="lavender", mid = "orchid", high="slateblue", guide="colorbar", midpoint = 0.25)
strikes per season per year
ggplot(total.table, aes(Year, season, fill= proportion)) +
geom_tile()+
scale_fill_gradient2(low="lavender", mid = "orchid", high="slateblue", guide="colorbar", midpoint = 0.25)
strikes per season per building
ggplot(total.table, aes(Building, season, fill= proportion)) +
geom_tile()+
scale_fill_gradient2(low="lavender", mid = "orchid", high="slateblue", guide="colorbar", midpoint = 0.25)
Strikes as a function of family
This section shows strike data by species for systematic and opportunistic observations. There are still quite a few unknowns that are not included but as I fill them in I’ll update the information. That said, you have plenty to present as it is.
Family x building
I created a table that shows the number of strikes per family and building
FxBtable = table(spp$Family, spp$Building)
FxBtable##
## ACT Albee CCE&D CFWC Clow FDL Halsey Harrington Polk Prkgramp
## Alcedinidae 0 0 0 0 0 1 0 0 0 0
## Bombycillidae 0 0 0 0 0 1 0 0 0 0
## Cardinalidae 0 0 0 0 0 0 0 0 0 0
## Certhiidae 0 0 0 0 0 0 1 0 0 0
## Columbidae 1 0 0 0 1 0 0 0 0 0
## Emberizidae 0 0 0 0 0 1 0 0 0 0
## Fringillidae 0 0 0 2 1 0 0 0 1 0
## Mimidae 1 0 0 1 0 0 0 0 0 0
## Paridae 0 0 0 0 0 0 0 0 0 0
## Parulidae 2 0 0 1 1 0 0 0 0 1
## Passerelidae 0 0 0 2 1 0 1 0 0 0
## Passeridae 0 0 0 0 0 0 0 0 1 0
## Phasianidae 0 1 0 0 0 0 0 0 0 0
## Picidae 0 0 0 0 0 0 0 1 0 0
## Regulidae 0 0 0 0 2 1 2 0 0 0
## Sittidae 0 0 0 0 0 0 0 0 0 0
## Sturnidae 0 0 0 0 0 0 0 0 0 0
## Trochilidae 0 0 0 0 0 0 0 0 0 0
## Turdidae 3 0 1 0 1 1 2 0 0 0
## Vireonidae 0 0 0 0 0 0 1 0 0 0
##
## Radford Reeve Sage SRWC UNKNOWN
## Alcedinidae 0 0 0 0 0
## Bombycillidae 0 0 0 0 0
## Cardinalidae 0 1 0 0 0
## Certhiidae 0 0 0 0 0
## Columbidae 0 0 0 0 0
## Emberizidae 0 1 3 0 0
## Fringillidae 0 0 1 2 0
## Mimidae 0 0 1 0 0
## Paridae 0 0 1 0 0
## Parulidae 0 7 11 0 0
## Passerelidae 0 2 2 0 0
## Passeridae 0 1 0 0 0
## Phasianidae 0 0 0 0 0
## Picidae 1 1 0 0 0
## Regulidae 0 4 3 0 1
## Sittidae 0 0 1 0 0
## Sturnidae 0 2 0 1 0
## Trochilidae 0 0 1 0 0
## Turdidae 0 4 2 1 0
## Vireonidae 0 1 0 0 0I then created a vector of the counts from the table.
FxB.counts = as.integer(table(spp$Family, spp$Building))
#FxB.counts
#length(FxB.counts)
# FxB.counts goes through all families for one buildingI created a family vector that corresponded to the counts…
FxB.family = rep(c(levels(spp$Family)), times=nlevels(spp$Building))
FxB.family = factor(FxB.family)
#FxB.family
#length(FxB.family)and a building vector that corresponded to the counts.
FxB.building = rep(c(levels(spp$Building)), each=nlevels(spp$Family))
FxB.building = factor(FxB.building)
#FxB.building
#length(FxB.building)I combined the new counts, family, and building vectors into one data frame called FxB.
FxB = data.frame(FxB.family, FxB.counts, FxB.building)
#head(FxB)ggplot(FxB, aes(FxB.family, FxB.building, fill= FxB.counts)) +
geom_tile()+
theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))+
scale_fill_gradient2(low="lavender", mid = "orchid", high="slateblue", guide="colorbar", midpoint = (max(FxB.counts)/2))+
labs(x="Familes",y="Buildings",title="Number of strikes per building per Family")
Family x year
Using the same methods as above, I created a data frame that allows me to map strikes per family per year. I don’t show the code here to save space. Please ask if you want to see it.
Family x season
Again, I don’t show the code here to save space. Please ask if you want to see it.
