The Shaky Isles

Quake data

The data included here is provided by GeoNet New Zealand. For more details see https://www.geonet.org.nz/.

eq <- read_csv(url)

The dataset includes details on 260,997 observations between the start of 2010 and the end of 2020, not all of which are relevant to the current analysis. It includes measurements originating from outside New Zealand, as well as seismic events which aren’t strictly earthquake-related, such as volcanic eruptions and quarry blasting11 Somewhat unsettlingly, the dataset also includes ‘nuclear explosion’ as a seismic event - a measurement originating from North Korean underground testing..

eq <- eq %>%
        filter(magnitude > 0,
         longitude < 180 & longitude > 160,
         latitude < -29)

eq$date <- ymd_hms(eq$origintime)
eq$month <- factor(month(eq$date), 
            labels = c("Jan","Feb","Mar","Apr","May","Jun",
                       "Jul","Aug","Sep","Oct","Nov","Dec"))
eq$year <- factor(year(eq$date))

After a quick clean to remove rogue data points22 Removed entries include those with less than zero magnitude, or that lie outside of the immediate geographic surroundings of New Zealand., two subsets are created for further analysis - one with all readings that can be directly attributed to earthquakes, and another with only strong earthquakes.

eq_all <- eq %>%
  filter(eventtype == "earthquake")

eq_strong <- eq %>%
  filter(eventtype == "earthquake",
         magnitude >= 6)

What follows is a quick breakdown of the entire dataset by event type, including average magnitude and depth (in kms):

eq$eventtype <- str_to_sentence(eq$eventtype)

eq %>%
  group_by(eventtype) %>%
  summarise(Count = n(), 
            Magnitude = round(mean(magnitude),1), 
            Depth = round(mean(depth),1)) %>%
  knitr::kable()

eventtype	Count	Magnitude	Depth
Debris avalanche	2	2.6	0.0
Duplicate	1	2.2	12.2
Earthquake	142178	2.4	42.6
Experimental explosion	4	1.9	0.0
Explosion	43	1.2	0.3
Induced earthquake	1	2.2	5.0
Landslide	4	2.7	0.0
Not locatable	197	1.9	56.1
Nuclear explosion	1	1.1	50.2
Other	6	1.2	4.4
Outside of network interest	1444	3.8	189.0
Quarry blast	575	2.2	0.3
Snow avalanche	4	2.3	0.0
Volcanic eruption	3	1.6	0.7
NA	106011	2.0	39.8

As the table suggests, discarding measurements not directly related to earthquakes will leave us with 142,178 records.

Magnitude

The vast majority of earthquakes will pass by unremarked on by anyone other than seismic specialists. From year to year, the average magnitude of a quake hovers around 2 to 2½, with only outliers breaking through the 5 richter mark.

It’s hard to discern the impact that major earthquakes have on the distributions. Major quakes in 2010, 2011 and 2016 brought large waves of aftershocks in their wake, but don’t seem to have reshaped the plot to a significant degree.

eq_all %>%
  ggplot(aes(x = year, y = magnitude)) +
  geom_boxplot(outlier.alpha = 0.1) +
  theme_minimal() +
  labs(x = "", y = "Magnitude (Richter scale)") +
  theme(axis.ticks.x = element_blank())

Distribution in time

The major earthquakes and their accompanying wave of aftershocks are much more easily discernible when viewed over time. Three jump out of the data: the Darfield quake of 2010, Christchurch in 2011, and Kaikoura in 2016.

The persistence of aftershocks seems to be related to the depth of the initial quake33 Caution should be exercised when leaping to conclusions in light of the range of factors involved - the Kaikoura quake in particular has been described as ‘the most complex quake ever studied,’ involving over twenty-five faults rupturing simultaneously.. The Christchurch earthquake of 2011 was relatively shallow at a depth of 5km, and saw a sustained peak of quakes for two months afterward. 2016’s quake in Kaikoura on the other hand was three times as deep, and saw a strong peak in quake numbers before dropping off quickly.

eq_all_grouped <- eq_all %>%
  group_by(month, year) %>%
  summarise(total = n())

eq_all_grouped %>%
  ggplot(aes(year, month, fill = total)) +
  geom_tile() +
  theme(panel.background = element_blank()) +
  scale_y_discrete(limits = rev(levels(eq_all_grouped$month))) +
  scale_fill_viridis(labels=function(x) 
    format(x, big.mark = ",", 
           scientific = FALSE)) +
  labs(x = "Year", y = "Month", fill = "Quakes")

table(eq_all$month, eq_all$year) %>%
  knitr::kable()

	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020
Jan	1839	1928	890	551	373	582	617	900	449	1491	1628
Feb	1914	3027	320	480	177	394	579	761	457	1603	1567
Mar	1789	2824	349	459	250	613	728	1393	497	2012	1603
Apr	1799	2070	469	471	313	492	698	1256	389	1940	1653
May	1898	2028	396	473	304	599	638	1370	376	2229	1976
Jun	1554	1976	367	436	330	431	646	1053	436	2379	1773
Jul	2031	1447	504	1369	353	502	585	774	422	1754	1530
Aug	1618	1669	424	1196	379	597	576	714	481	1574	1734
Sep	3888	1636	547	576	394	452	1391	662	505	1574	1381
Oct	2360	1486	502	539	378	552	670	770	629	1626	1588
Nov	2055	1555	524	457	435	587	4015	734	450	1402	1497
Dec	1752	2210	567	408	573	533	1026	574	1257	1442	1524

Geographic distribution

People look upon the natural world as if all motions of the past had set the stage for us and were now frozen. To imagine that the turmoil is in the past and somehow we are now in a more stable time seems to be a psychological need. … [But] the time we’re in is just as active as the past. The time between events is long only with respect to a human lifetime.

Longitude and latitude records for each observation allow them to be projected onto a map of the country, and grouped by location.

The fault systems that run the length of the country can be discerned in the location of these clusters. Large groups of quakes are evident in Fiordland, following the Southern Alps northward, before crossing over into Canterbury, and into Marlborough beyond. Clusters in the North Island follow a more or less direct line from Wellington to the Bay of Plenty.

Centres distant from faultlines experience a comparatively tiny number of events - despite being located on a volcanic field, Auckland saw less than 200 quakes of any size in a decade, and Dunedin little over half that.

nz_long <- 174.8860
nz_lat <- -41.2749

leaflet() %>%
  setView(lng = nz_long, lat = nz_lat, zoom = 5) %>%
  addProviderTiles("Esri.WorldStreetMap") %>%
  addMarkers(data = eq_all, clusterOptions = markerClusterOptions())

Strong earthquakes

The Richter scale defines a strong quake as greater than 6 in magnitude, of which there are approximately 100 to 150 globally each year. In everyday terms, a ‘strong’ earthquake is one expected to cause widespread damage to structures. The Christchurch quake of 2011 measured 6.3, levelling modern buildings and damaging others beyond repair.

The mapping also highlights the largely forgotten 2016 Te Araroa earthquake, which took place more than a hundred kilometers off the coast of the East Cape. Despite a cluster of three major quakes as well as a significant number of strong aftershocks, no injury or major damage was reported.

pal <- colorNumeric(palette = "plasma",
                    domain = eq_strong$magnitude)

leaflet() %>%
  setView(lng = nz_long, lat = nz_lat, zoom = 5) %>%
  addProviderTiles("Esri.WorldStreetMap") %>%
  addCircles(
    data = eq_strong,
    radius = sqrt(10 ^ eq_strong$magnitude) * 10,
    color = ~ pal(eq_strong$magnitude),
    popup = paste0(
      "<strong>Date: </strong>",
      eq_strong$date, "<br>",
      "<strong>Magnitude: </strong>",
      round(eq_strong$magnitude, 2), "<br>",
      "<strong>Depth (km): </strong>",
      round(eq_strong$depth, 0)),
    popupOptions = popupOptions(closeonclick = TRUE)) %>%
  addLegend(
    data = eq_strong, "bottomright",
    pal = pal, values = ~ magnitude,
    title = "Magnitude", opacity = 0.8)