1 Background

We had a bit or rock n roll on Monday night. So the question naturally arose: are earthquakes more likely to occur at night?.

You might think not, I couldn’t possibly comment.

Luckily in Aotearoa New Zealand GeoNet makes answering the question very easy. Just download the most recent year’s worth of earthquake records and off you go…

2 Data

We downloaded all earthquake records from 2018-09-01 to 2019-08-13. This is a grand total of 20498 records.

Not all the events are actually earthquakes. As Table 2.1 shows, there are explosions, quarry blasts etc.

t <- cube(dt, j = list("min magnitude" = min(magnitude),
                       "mean magnitude" = mean(magnitude),
                       "max magnitude" = max(magnitude), 
                       "n Obs" = .N), by = c("eventtype"))

## Warning in min(magnitude): no non-missing arguments to min; returning Inf

## Warning in max(magnitude): no non-missing arguments to max; returning -Inf

kableExtra::kable(t[!is.na(eventtype)], digits = 2,
                  caption = "Mean magnitude and number of events") %>%
  kable_styling()

dt <- dt[eventtype %like% "earthquake"]
dt <- dt[, hms := hms::as.hms(origintime)]
dt <- dt[, halfHour := hms::trunc_hms(hms, 60*30)]

As we can see there are some (4) records with magnitude < 0?

3 Q1: Are earthquakes more likely to occur at night?

To answer this we plot all earthquakes where magnitude > 0.

plotDT <- dt[magnitude > 0, .(nObs = .N,
                 meanMag = mean(magnitude)), keyby = .(halfHour)]

myCaption <- paste0("Source: NZ GeoNet (https://quakesearch.geonet.org.nz/)\nAll earthquakes ",
                    min(lubridate::date(dt$origintime)), " to ", max(lubridate::date(dt$origintime))
                    )

p <- ggplot2::ggplot(plotDT, aes(x = halfHour, 
                                 y = nObs, 
                                 colour = round(meanMag,1)
                                 )
                     ) +
  geom_point() +
  labs(x = "Time of Day (half hours)",
       y = "Number of Observations",
       caption = paste0(myCaption,
                        "\nAll magnitude > 0 (", nrow(dt[magnitude > 0]), ")")) +
  scale_color_continuous(low = "green", high = "red", name = "Mean magnitude") +
  theme(legend.position="bottom")

p

Figure 3.1: Number of earthquakes by time of day

Figure 3.1 suggests more occur (or at least are detected) at night. Really? But there is a little clue in the colour of the points - the mean magnitude is higher during the day. This seems unlikely.

4 Q2: Are stronger earthquakes more likely to occur during the day?

To answer this we plot all earthquakes where magnitude > 3.

plotDT <- dt[magnitude > 3, .(nObs = .N,
                 meanMag = mean(magnitude)), keyby = .(halfHour)]

p <- ggplot2::ggplot(plotDT, aes(x = halfHour, 
                                 y = nObs, 
                                 colour = round(meanMag,1)
                                 )
                     ) +
  geom_point() +
  labs(x = "Time of Day (half hours)",
       y = "Number of Observations",
       caption = paste0(myCaption,
                        "\nAll magnitude > 3 (", nrow(dt[magnitude > 3]), ")")
                        ) +
  scale_color_continuous(low = "green", high = "red", name = "Mean magnitude") +
  theme(legend.position="bottom")

p

Figure 4.1: Number of earthquakes by time of day (magnitude > 3)

This suggests that it is just easier to detect lower magnitude earthquakes at night when people are asleep, there are no trucks on the roads etc.

5 Q3: Are stronger earthquakes more likely during the day?

Just to check:

p <- ggplot2::ggplot(dt[magnitude > 0], aes(x = halfHour, y = magnitude, group = halfHour)) +
  geom_boxplot() +
  labs(x = "Time of Day",
       y = "Magnitude",
       caption = paste0(myCaption,
                        "\nAll magnitude > 0 (", nrow(dt[magnitude > 0]), ")")
                        ) +
  theme(legend.position="bottom")
p

Figure 5.1: Box plot of magnitude of earthquakes by time of day

No they are not.

6 Data Annex

Original data as loaded from https://quakesearch.geonet.org.nz/csv?bbox=163.95996,-49.18170,182.63672,-32.28713&startdate=2018-09-01&enddate=2019-08-13:

skimr::skim(origDT) # original data

## Skim summary statistics
##  n obs: 20498 
##  n variables: 21 
## 
## ── Variable type:character ───────────────────────────────────────────────────────────────────────────────────────
##          variable missing complete     n min max empty n_unique
##         depthtype   16390     4108 20498  13  17     0        2
##        earthmodel      33    20465 20498   3   8     0        4
##  evaluationmethod      33    20465 20498   6   9     0        2
##    evaluationmode       0    20498 20498   6   9     0        2
##  evaluationstatus    3229    17269 20498   9  11     0        2
##         eventtype    3161    17337 20498   9  27     0        6
##     magnitudetype       0    20498 20498   1   6     0        5
##          publicid       0    20498 20498  11  11     0    20498
## 
## ── Variable type:numeric ─────────────────────────────────────────────────────────────────────────────────────────
##               variable missing complete     n     mean     sd        p0
##           azimuthalgap       0    20498 20498 139.05   66.85     0     
##                  depth       0    20498 20498  42.09   56.41     0     
##               latitude       0    20498 20498 -39.7     2.16   -48.91  
##              longitude       0    20498 20498 170.28   42.41  -180     
##              magnitude       0    20498 20498   2.08    0.71    -0.15  
##  magnitudestationcount       0    20498 20498   9.6     8.83     1     
##   magnitudeuncertainty       0    20498 20498   0.0034  0.033    0     
##        minimumdistance       0    20498 20498   0.33    0.6      0     
##            originerror       0    20498 20498   0.53    0.34     0.0051
##         usedphasecount       0    20498 20498  22.86   10.88     0     
##       usedstationcount       0    20498 20498  16.73    9.04     0     
##      p25    p50    p75   p100     hist
##   87.22  123.61 180.26 353.55 ▁▇▇▅▃▂▁▁
##    7.32   19.94  42.51 671.26 ▇▁▁▁▁▁▁▁
##  -40.87  -39.34 -38.39 -32.29 ▁▁▁▃▇▅▁▁
##  174.44  175.85 176.9  180    ▁▁▁▁▁▁▁▇
##    1.66    2.04   2.42   6.7  ▁▃▇▃▁▁▁▁
##    6       7     11    151    ▇▁▁▁▁▁▁▁
##    0       0      0      1.06 ▇▁▁▁▁▁▁▁
##    0.085   0.15   0.31   5.95 ▇▁▁▁▁▁▁▁
##    0.33    0.45   0.61  10.42 ▇▁▁▁▁▁▁▁
##   16      20     26    132    ▃▇▁▁▁▁▁▁
##   11      14     19    119    ▇▆▁▁▁▁▁▁
## 
## ── Variable type:POSIXct ─────────────────────────────────────────────────────────────────────────────────────────
##          variable missing complete     n        min        max     median
##  modificationtime       0    20498 20498 2018-09-01 2019-08-14 2019-03-16
##        origintime       0    20498 20498 2018-09-01 2019-08-12 2019-03-15
##  n_unique
##     20497
##     20496

7 R packages used

• base R (R Core Team 2016)
• data.table (Dowle et al. 2015) - for data munching
• drake (Landau 2019) - for data pre-loading and caching
• ggplot2 (Wickham 2009) - for plots
• hms (Müller 2018) - for time munching
• kableExtra (Zhu 2018) - for nice tables
• lubridate (Grolemund and Wickham 2011) - for dateTime munching
• readr (Wickham, Hester, and Francois 2016) - for data download & parsing
• skimr (Arino de la Rubia et al. 2017) - for data description