R- Plotting Seasonality

Question

I have a dataframe called 'madrid' that has the measurements of air pollutants along with the date, year, month, and day. There's data from 2001 to 2018

data.frame':    3808224 obs. of  15 variables:
 $ date   : Date, format: "2001-08-01" "2001-08-01" "2001-08-01" ...
 $ BEN    : num  NA 1.5 NA NA NA ...
 $ CO     : num  0.37 0.34 0.28 0.47 0.39 ...
 $ EBE    : num  NA 1.49 NA NA NA ...
 $ NMHC   : num  NA 0.07 NA NA NA ...
 $ NO_2   : num  58.4 56.2 50.7 69.8 22.8 ...
 $ O_3    : num  34.5 42.2 46.3 40.7 66.3 ...
 $ PM10   : num  105 100.6 100.1 69.8 75.2 ...
 $ SO_2   : num  6.34 8.11 7.85 6.46 8.8 ...
 $ TCH    : num  NA 1.24 NA NA NA ...
 $ TOL    : num  NA 10.8 NA NA NA ...
 $ station: int  28079001 28079035 28079003 28079004 28079039 28079006 28079007 28079009 28079038 28079011 ...
 $ year   : num  2001 2001 2001 2001 2001 ...
 $ month  : num  8 8 8 8 8 8 8 8 8 8 ...
 $ day    : int  1 1 1 1 1 1 1 1 1 1 ...

and I'm now trying to create a plot that shows the monthly total average of each pollutant. So, ideally one line/curve with the months plotted on the x-axis to see if there's seasonality. So far, I've done this (below) with dplyr:

madrid_season <- madrid %>%
  group_by(month, year) %>%
  summarise(BEN_mean = mean(BEN, na.rm = TRUE),
            CO_mean = mean(CO, na.rm = TRUE),
            EBE_mean = mean(EBE, na.rm = TRUE),
            NMHC_mean = mean(NMHC, na.rm = TRUE),
            NO_2_mean = mean(NO_2, na.rm = TRUE),
            O_3_mean = mean(O_3, na.rm = TRUE),
            PM10_mean = mean(PM10, na.rm = TRUE),
            SO_2_mean = mean(SO_2, na.rm = TRUE),
            TCH_mean = mean(TCH, na.rm = TRUE),
            TOL_mean = mean(TOL, na.rm = TRUE))

and then I sum the averages (excluding the month and year column):

madrid_season$tot <- rowSums(madrid_season[,2:11], na.rm = TRUE)

and then I try and plot:

p4 <- ggplot(madrid_season, aes(x=month,y=tot))+geom_point()+geom_line()

except this plots the total monthly average for every year, instead of for the single month for each year, so I end up with a wacky looking plot.

I think I need help using dplyr in this case (or open to other suggestions!)

How would you like to show the years and months on the plot? A x axis with all the years and months in chronological order? Superimposed plots for each year? One panel for each year? — user2332849, Mar 19 '20 at 23:54
@user2332849 Yeah ideally I'd like the years and months in chronological order if possible — dejsdukes, Mar 19 '20 at 23:58
Are you sure you want to sum all those air quality measurements? — Edward, Mar 19 '20 at 23:59
@Edward it was my initial idea, although I guess I should just plot them separately or with multiple lines on one plot — dejsdukes, Mar 20 '20 at 00:01
Put a `Date` class variable on the x-axis, like the first day of each month. Add in a column that's `as.Date(paste(year, month, "01", sep = "-"))` and put that on the x-axis. — Gregor Thomas, Mar 20 '20 at 00:04
@dejsdukes: Could you make your problem reproducible by sharing a sample of your data so others can help (please do not use `str()`, `head()` or screenshot)? You can use the [`reprex`](https://reprex.tidyverse.org/articles/articles/magic-reprex.html) and [`datapasta`](https://cran.r-project.org/web/packages/datapasta/vignettes/how-to-datapasta.html) packages to assist you with that. See also [Help me Help you](https://speakerdeck.com/jennybc/reprex-help-me-help-you?slide=5) & [How to make a great R reproducible example?](https://stackoverflow.com/q/5963269) — Tung, Mar 20 '20 at 01:17
@dejsdukes: fyi, you can use `summarize_all` or `summarize_if` instead of manually typing each variable in `summarize`. See here https://suzan.rbind.io/2018/04/dplyr-tutorial-4/ — Tung, Mar 20 '20 at 01:19

score 1 · Accepted Answer · answered Mar 20 '20 at 02:40

I wouldn't advise summing each monthly mean measurement. The measurement with the largest scale, here it looks like PM₁₀, would swamp all other measurements.

You can either plot each one individually to see the seasonality, or reshape into long format and use the measurement as a facet in the ggplot.

I'll use the mydata dataset from the openair package since you haven't provided any data. The variables are similar to yours.

library(openair)

str(mydata)
Classes ‘tbl_df’, ‘tbl’ and 'data.frame':   65533 obs. of  10 variables:
 $ date: POSIXct, format: "1998-01-01 00:00:00" "1998-01-01 01:00:00" ...
 $ ws  : num  0.6 2.16 2.76 2.16 2.4 3 3 3 3.36 3.96 ...
 $ wd  : int  280 230 190 170 180 190 140 170 170 170 ...
 $ nox : int  285 NA NA 493 468 264 171 195 137 113 ...
 $ no2 : int  39 NA NA 52 78 42 38 51 42 39 ...
 $ o3  : int  1 NA 3 3 2 0 0 0 1 2 ...
 $ pm10: int  29 37 34 35 34 16 11 12 12 12 ...
 $ so2 : num  4.72 NA 6.83 7.66 8.07 ...
 $ co  : num  3.37 NA 9.6 10.22 8.91 ...
 $ pm25: int  NA NA NA NA NA NA NA NA NA NA ...

library(dplyr)
library(lubridate)
library(ggplot2)

mydata %>%
  mutate(Year=factor(year(date)), Month=month(date, label = TRUE)) %>%
    group_by(Year, Month) %>%
  summarise_all(list(mean=mean), na.rm=TRUE) %>%
  ggplot(aes(y=o3_mean, x=Month, group=Year, col=Year)) +
  geom_line(lwd=1) + ylab(expression(O[3]))

So seasonality of O₃ is evident.

If you want to plot all measurements, then reshape to long form, and facet on the measurement.

library(tidyr)
mydata %>%
  mutate(Year=factor(year(date)), Month=month(date, label = TRUE)) %>%
  group_by(Year, Month) %>%
  summarise_all(list(mean=mean), na.rm=TRUE) %>%
  select(-date_mean) %>%
  pivot_longer(cols=ends_with("_mean"), 
               names_to = c("measurement", ".value"), 
               names_sep="_") %>%
  ggplot(aes(y=mean, x=Month, group=Year, col=Year)) +
  geom_line() +
  facet_wrap(~measurement, scales="free") +
  scale_x_discrete(labels=function(x) substr(x, 1, 1))

wow thank u! and yes I abandoned the idea to sum them shortly after posting lol — dejsdukes, Mar 20 '20 at 03:02

R- Plotting Seasonality

1 Answers1