The Impact of the 2011 Tōhoku tsunami on Albatross in Hawaii

A statisical analysis of the impact of the 2011 Tōhoku tsunami on albatross populations in Hawaii.

This is an image of Hōlanikū, also know as Kure Atoll, the northernmost landmass in the Hawaiian archipelago. It is located at 28.3925° N, 178.2936° W and is within the Papahānaumokuākea National Marine Monument.
Image Credit: Kure Atoll Conservancy

This statistical analysis was completed as an assignment for my Master’s program course, Environmental Data Science 222: Statistics for Environmental Data Science.

Research Question

Were albatross populations impacted by the 2011 Tōhoku tsunami event?

Researchers estimated the short term effects of the Tōhoku tsunami on albatross populations and concluded that the tsunami flooded between 26% - 52% of all Black-footed albatross nests and impacted more than 275,000 albatross nests throughout Papahānaumokuākea (Reynolds et al. 2017). This post will attempt to analyze and quantify the long term impact of the Tōhoku tsunami on two Hawaiian albatross species populations, the Laysan albatross and the Black-footed albatross.

Background

The Hawaiian archipelago is home to thousands of albatrosses. Albatrosses are an incredible bird species that have inspired authors, the fashion industry, and birders across the globe. They live to be over 65 years old, have the largest wingspan of any bird, are monogamous rearers, and have nest-site fidelity (meaning they return to the same nest location every year). Three species of albatross breed in Hawaii, the Laysan (Phoebastria immutabilis), Black-footed (Phoebastria nigripes), and Short-tailed (Phoebastria albatrus) albatross. Laysan albatrosses are listed as near threatened due to threats from climate change to their habitat and breeding grounds and long-line fishing operations (Arata, Sievert, and Naughton 2009). The IUCN Red List of Threatened Species assessed the Black-footed albatross as near threatened in 2020 (List 2017). The Short-tailed albatross almost went extinct in the early 1900s due to feather hunting and is currently listed as an endangered species in the United States (Alaska Region, n.d.).

Show code

data.frame("Common_name" = c("Black-footed albatross", "Laysan albatross", "Short-tailed albatross"),
           "Hawaiian_name" = c("Kaʻupu", "Mōlī", "Makalena|Kaʻupuakea"),
           "Species_code" = c("BFAL","LAAL", "STAL"),
           "Scientific_name" = c("Phoebastria nigripes", "Phoebastria immutabilis", "Phoebastria albatrus")) %>% 
  kable(caption = "Hawaiian Albatross Species Names") %>%
  kable_paper(full_width = FALSE) %>%
  kable_styling(latex_options = "striped",
                font_size = 15) %>% 
  column_spec(1, bold = T) %>%
  row_spec(0, bold = T, color = "black")

Table 1: Hawaiian Albatross Species Names

Common_name	Hawaiian_name	Species_code	Scientific_name
Black-footed albatross	Kaʻupu	BFAL	Phoebastria nigripes
Laysan albatross	Mōlī	LAAL	Phoebastria immutabilis
Short-tailed albatross	Makalena|Kaʻupuakea	STAL	Phoebastria albatrus

Adult and juvenile Laysan and Black-footed albatross on the coral runway of Hōlanikū (Kure Atoll).
Image Credit: J. Parish

Papahānaumokuākea, also known as the Northwestern Hawaiian Islands, is comprised of atolls, reefs, and pinnacles. It is home to 95% of the Black-footed albatross and 99% of the Laysan albatross global population (Arata, Sievert, and Naughton 2009). These low-lying islands are at extreme risk of inundation from tsunamis (Reynolds et al. 2017). On March 11, 2011, a 9.0 earthquake hit the Tōhoku region of Japan. The earthquake lasted over 6 minutes, creating a tsunami that impacted coastal areas and island nations throughout the Pacific region. Approximately 20,000 people lost their lives from the earthquake and resulting tsunami. The tsunami also killed or injured thousands of marine and terrestrial species. Many wildlife species found in the Papahānaumokuākea Marine National Monument (PMNM) were impacted by the Tōhoku tsunami.

Analysis Plan

This statistical analysis tested the hypothesis if the Tōhoku tsunami impacted albatross populations in Hawaii. The table below outlines the phases of the analysis.

\[H_{0}: There\ was\ no\ impact\ of\ the\ Tōhoku\ tsunami\ on\ albatross\ populations\ in\ Hawaii.\] \[H_{1}: There\ was\ an\ impact\ of\ the\ Tōhoku\ tsunami\ on\ albatross\ populations\ in\ Hawaii.\]

Show code

data.frame("Phase" = c(1:5),
           "Description" = c("Identify research question",
                             "Collect data",
                             "Visualize data",
                             "Conduct regression analysis",
                             "Conclusion & Future Research")) %>% 
  kable(caption = "Tōhoku Tsunami Impact Analysis Plan Outline") %>%
  kable_paper(full_width = FALSE) %>%
  kable_styling(latex_options = "striped",
                font_size = 15) %>% 
  column_spec(1, bold = T) %>%
  row_spec(0, bold = T, color = "black")

Table 2: Tōhoku Tsunami Impact Analysis Plan Outline

Phase	Description
1	Identify research question
2	Collect data
3	Visualize data
4	Conduct regression analysis
5	Conclusion & Future Research

Note: The Short-tailed albatross was not included in this analysis as there is only one breeding pair in Hawaii.

Collect Data

After researching several data sources for albatross populations, I retrieved banding data for both Laysan and Black-footed albatross from the USGS Bird Banding Laboratory (BBL). The BBL has data on bird species for the past 100 years. The data contains information about an individual bird’s sex, age, health condition, and coordinates where the bird was banded. To access data from the BBL, it requires establishing an account on the USGS Bird Banding Lab Bander Portal website. Once you submit a data request, files will be available for download within 24 - 48 hours. I requested data for Black-footed and Laysan albatrosses in Hawaii between the years 1996 and 2020. Albatrosses reach reproductive maturity at nine years old on average. Data starting in 2019-2020 would reflect the cohort born in 2011 returning to nest and produce first young.

Author and a colleague banding a juvenile Black-footed albatross.
Image Credit: E. Opie

Visualize Data

Upon completion of data tidying and transformation, the next phase of analysis is to visualize the data.

Distribution Map

The initial visualization mapped the banding data on to a map of Hawaii using the ggmap() package. The banding data point distribution accurately reflects the known species distribution of both species of albatross in Hawaii. This map shows most albatross populations reside within the Papahānaumokuākea MNM. The only two main Hawaiian Islands where albatross nest are Kauai and Oahu. Due to the extensive length of the Hawaiian archipelago compared to the small size of the Northwestern Hawaiian Islands, each atoll is outlined with a grey box and the Hawaiian place name is indicated above the parameter box. The division between the Main Hawaiian Islands and the Papahanaumokuakea National Marine Monument is indicated by a dashed line on the map visualization.

Show code

albie_map <- ggmap(hawaii_map) +
  geom_point(aes(x = lon_dd, y = lat_dd, color = species_name),
             data = albie_band,
             size = 0.5,
             alpha = 0.25) +
  geom_vline(xintercept = -163, linetype="dashed", color = "#0f173d") +
  annotate('text', x = -170, y = 23, 
           label = 'Papahanumokuakea Marine National Monument', color = 'black', size = 3, fontface = 'italic') +
    annotate('text', x = -156.8, y = 23, 
           label = 'Main Hawaiian Islands', color = 'black', size = 3, fontface = 'italic') +
  scale_color_manual(name = "Albatross species",
                     values = c("#9E7E8C", "#39ACB1"),
                     labels = c("Black-footed Albatross", "Laysan Albatross"),
                     guide = guide_legend(override.aes = list(size = 3, alpha = 1))) +
  annotate("rect", xmin = -162.13, xmax = -161.78, ymin = 22.94, ymax = 23.22, 
           color = "#728A72", fill = "white", alpha = 0.2) +
  annotate('text', x = -161.7, y = 23.3, 
           label = 'Nihoa', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -164.856, xmax = -164.55, ymin = 23.42, ymax = 23.74, 
           color = "#728A72", fill = "white", alpha = 0.2) +
  annotate('text', x = -164.45, y = 23.8, 
           label = 'Mokumanamana', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -166.54, xmax = -165.88, ymin = 23.45, ymax = 24.11, 
           color = "#728A72", fill = "white", alpha = 0.0) +
  annotate('text', x = -166.54, y = 24.3, 
           label = 'Lalo', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -168.19, xmax = -167.83, ymin = 24.82, ymax = 25.16, 
           color = "#728A72", fill = "white", alpha = 0.2) +
  annotate('text', x = -167.8, y = 25.3, 
           label = 'Onunui', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -170.78, xmax = -170.46, ymin = 25.33, ymax = 25.68, 
           color = "#728A72", fill = "white", alpha = 0.2) +
  annotate('text', x = -170.39, y = 25.8, 
           label = 'Kamokuokamohoalii', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -171.95, xmax = -171.45, ymin = 25.53, ymax = 26, 
           color = "#728A72", fill = "white", alpha = 0.0) +
  annotate('text', x = -172, y = 26.2, 
           label = 'Kamole', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -174.21, xmax = -173.8, ymin = 25.84, ymax = 26.24, 
           color = "#728A72", fill = "white", alpha = 0.2) +
  annotate('text', x = -173.85, y = 26.5, 
           label = 'Kapou', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -175.47, xmax = -174.9, ymin = 27.22, ymax = 27.61, 
           color = "#728A72", fill = "white", alpha = 0.2) +
  annotate('text', x = -174.8, y = 27.7, 
           label = 'Kamole', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -176.16, xmax = -175.53, ymin = 27.57, ymax = 28.14, 
           color = "#728A72", fill = "white", alpha = 0.0) +
  annotate('text', x = -175.97, y = 28.35, 
           label = 'Kuaihelani', color = 'black', size = 2.3, hjust = 0) +
  annotate("rect", xmin = -177.19, xmax = -177.6, ymin = 28.1, ymax = 28.4, 
           color = "#728A72", fill = "white", alpha = 0.0) +
  annotate('text', x = -177.55, y = 28.65, 
           label = 'Holaniku', color = 'black', size = 2.3, hjust = 0) +
  annotate('text', x = -177.5, y = 19.5, 
           label = 'Created: 2021 J. Parish\nData Source: USGS Bird Banding Lab', 
           color = 'black', size = 1.5, hjust = 0) +
  labs(title = "Figure 1: Banded Albatross Species in the Hawaiian Archipelago",
        subtitle = "1996 - 2020",
        x = "Longitude",
        y = "Latitude") +
  theme(legend.title = element_text(size=12, color = "black", face="bold"),
           legend.justification=c(0,1), 
           legend.position=c(0.72, 0.98),
           legend.background = element_blank(),
           legend.key = element_blank())
albie_map

In order to conduct an analysis on the albatross count data, I created a new column in the albatross banding data frame for the total count for albatrosses for each year between 1996 and 2020.

Show code

band_count <- albie_band %>% 
  group_by(species, year) %>% 
  summarise(total = sum(count))

Scatterplot

To conduct analysis on the albatross banding data, it is necessary to create a total count of albatross banded for each year from 1996 to 2020. Once a total count was summed, I used a line and point plot to visualize albatross count annually and added a line to indicate with the Tōhoku tsunami occurred (2011). This plot does suggest that there may be a negative effect from the Tōhoku tsunami on albatross populations as the counts drop significantly after 2011. It also shows that the number of banded Black-footed albatrosses declined more than Laysan albatrosses after the tsunami. This trend may reflect the data that Reynolds et al. found as Black-footed albatrosses’ nest along coastal areas whereas Laysan albatrosses tend to nest more inland or at higher elevations on islands (Reynolds et al. 2017).

Show code

#plot banding count data

albie_count_plot <- ggplot(band_count, aes(x = year, y = total, group = species)) +
  geom_point(aes(color = species,
                 shape = species)) +
  geom_line(aes(color = species)) +
  scale_color_manual(name = "Albatross species",
                     values = c("#9E7E8C", "#39ACB1"),
                     labels = c("Black-footed Albatross (BFAL)", "Laysan Albatross (LAAL)"),
                     guide = guide_legend(override.aes = list(shape = c(19, 17)))) +
  geom_vline(xintercept = 2011,
             linetype = "solid",
             color = "goldenrod1",
             size = 2) +
  guides(shape = FALSE) +
  annotate("text",
           label = "Tōhoku tsunami",
           x = 2014,
           y = 4700,
           color = "black",
           size = 4) +
  labs(title = "Figure 2: Hawaii Albatross Count Based On Band Data",
       subtitle = "Data source: USGS Bird Banding Lab",
       x = "Year",
       y = "Albatross Banded",
       color = "Species") +
  theme_minimal() +
  theme(legend.background = element_blank(),
        legend.position = "bottom")  
  
albie_count_plot

Population Means

I calculated the mean of banded albatross species between 1996 and 2020 as well as comparing the population means of each species pre- and post-tsunami event. These means indicate that more black-foot albatross is banded than Laysan albatross in Hawaii. This is an interesting data point as the total Laysan albatross population is larger than the Black-footed albatross population (VanderWerf 2012). The Laysan albatross actually holds the honor of having the largest population of all albatross species in the world (out of 21 species). Reviewing the two population mean tables also indicated that the number of banded albatrosses declined after the Tōhoku tsunami.

Show code

pop_params_table <- knitr::kable(pop_params_summary,
                                 digits = 0,
                                 col.names = c('Albatross Species', 'Population Mean', 'Population Median', 'Population Max'),
                                 align = "lccc",
                                 caption = "Hawaii Albatross Population Summary 1996 - 2021") %>%
  kable_paper(full_width = FALSE) %>%
  kable_styling(latex_options = "striped",
                font_size = 15) %>% 
  column_spec(1, bold = T) %>%
  row_spec(0, bold = T, color = "black") #%>% 
  # save_kable(here("images/pop_params.png"))

pop_params_table 

Table 3: Hawaii Albatross Population Summary 1996 - 2021

Albatross Species	Population Mean	Population Median	Population Max
BFAL	2003	2008	4740
LAAL	1396	902	3447

Show code

pop_mean_table <- knitr::kable(pop_mean,
                               col.names = c('Species', 'Pre_Pop_Mean', 'Post_Pop_Mean'),
                               caption = "Hawaii Albatross Count Means Pre- & Post- Tōhoku Tsunami") %>% 
  kable_paper(full_width = FALSE) %>%
  kable_styling(latex_options = "striped",
                font_size = 12) %>% 
  column_spec(1, bold = T) %>%
  row_spec(0, bold = T, color = "black")
pop_mean_table

(#tab:species means)Hawaii Albatross Count Means Pre- & Post- Tōhoku Tsunami

Species	Pre_Pop_Mean	Post_Pop_Mean
BFAL	3627.11	517.8
LAAL	2778.89	455.9

Conduct and Interpret Regression Analysis

To determine the distribution of the count data for both albatross species, I created Q-Q plots, or probability plots. The regression analysis used on the band count data was linear regression.

QQ Plot

The QQ plots for both the Black-footed and Laysan albatross show that the distribution is kurtosis, or they have heavy tails. Neither count data for the albatross species have a normal distribution based on the QQ plots.

Show code

bfal_qqplot <- ggplot(bfal_count) +
  geom_qq(aes(sample = total),
          color = "#9E7E8C",
          size = 3) +
  geom_qq_line(aes(sample = total),
            color = "grey") +
  xlab("Normal distribution quantiles") +
  ylab("Sample quantiles") +
  labs(title = "Figure 3: Black-footed Albatross (BFAL) QQ Plot") +
  theme_minimal() +
  theme(line = element_blank(),
        panel.grid = element_blank(),
        strip.text = element_blank(),
        axis.text.x = element_text(size = 10),
        axis.text.y = element_text(size = 10),
        legend.position = "none")
bfal_qqplot 

Show code

laal_qqplot <- ggplot(laal_count) +
  geom_qq(aes(sample = total),
          color = "#39ACB1",
          shape = 17,
          size = 3) +
  geom_qq_line(aes(sample = total),
            color = "grey") +
  xlab("Normal distribution quantiles") +
  ylab("Sample quantiles") +
  labs(title = "Figure 4: Laysan Albatross (LAAL) QQ Plot") +
  theme_minimal() +
  theme(line = element_blank(),
        panel.grid = element_blank(),
        strip.text = element_blank(),
        axis.text.x = element_text(size = 10),
        axis.text.y = element_text(size = 10),
        legend.position = "none")

laal_qqplot 

Simple Linear Regression

The first regression I conducted on the albatross banding data was a simple linear regression.

\[ \text{Albatross count}_i = \beta_0 + \beta_1 \text{tsunami event} + \varepsilon_i \]

Show code

tab_model(post_tsunami_simple_mod ,
          pred.labels = c("Intercept (mean albatross count pre-tsunami)", "Avg. banded albatross count post-tsunami"),
          string.ci = "Conf. Int (95%)",
          string.p = "P-value",
          title = "Table 4: Simple Linear Regression Model for Banded Albatross in Hawaii",
          digits = 4)

Table 4: Simple Linear Regression Model for Banded Albatross in Hawaii

	total
Predictors	Estimates	Conf. Int (95%)	P-value
Intercept (mean albatross count pre-tsunami)	2616.9667	2274.2624 – 2959.6710	<0.001
Avg. banded albatross count post-tsunami	-2169.0121	-2695.8899 – -1642.1343	<0.001
Observations	52
R2 / R2 adjusted	0.578 / 0.569

The result of the simple regression shows that the intercept, 2616, is the mean number of banded albatrosses in Hawaii prior to the Tōhoku tsunami event in 2011. The second predictor, or β1, is the mean banded albatrosses less than the intercept for any year after the tsunami. The mean albatross count is 447 for any year after the tsunami, which is a marked difference from the average of 2617 prior to the tsunami. There is a 5% chance that the average banded albatross will be between 2,2274 and 2,959 prior to the Tōhoku tsunami. The confidence interval post-tsunami shows that the mean will be -2,695 and -1,642 of the total albatrosses. Because the p-value is smaller than 0.05, the null hypothesis that there was no impact from the Tōhoku tsunami on albatross counts is rejected. The data provides convincing evidence that there is a negative difference in mean number of albatrosses post-tsunami. There is a statistically significant difference, at the 5% significance level, in the count of albatrosses in Hawaii. Based on the adjusted R-squared, this model explains 57% of the variation in the albatross count data around its mean.

Multiple Linear Regression

The second regression I conducted on the albatross banding data was a multiple linear regression with an interaction term of the tsunami event on each albatross species.

\[ \text{Albatross count}_i = \beta_0 + \beta_1 \text{tsunami event} + \beta_2 \text{albatross species} + \beta_3\text{tsunami event:species} + \varepsilon_i \]

Show code

tab_model(interaction_model,
          pred.labels = c("Intercept (mean BFAL count pre-tsunami)", "Mean BFAL count post-tsunami", "Mean LAAL count pre-tsunami", "Mean LAAL count compared to BFAL count post-tsunami"),
          string.ci = "Conf. Int (95%)",
          string.p = "P-value",
          title = "Table 5: Multiple Linear Regression Model Results for Banded Albatross in Hawaii",
          digits = 4)

Table 5: Multiple Linear Regression Model Results for Banded Albatross in Hawaii

	total
Predictors	Estimates	Conf. Int (95%)	P-value
Intercept (mean BFAL count pre-tsunami)	3125.1333	2676.1144 – 3574.1523	<0.001
Mean BFAL count post-tsunami	-2653.4061	-3343.7333 – -1963.0788	<0.001
Mean LAAL count pre-tsunami	-1016.3333	-1651.3420 – -381.3246	0.002
Mean LAAL count compared to BFAL count post-tsunami	968.7879	-7.4823 – 1945.0580	0.052
Observations	52
R2 / R2 adjusted	0.653 / 0.631

The result of the simple regression shows that the intercept, 3125, is the mean number of banded Black-footed albatrosses in Hawaii prior to the Tōhoku tsunami event in 2011. The second predictor, or β1, is the mean banded Black-footed albatrosses less than the intercept for any year after the tsunami. The mean count Black-footed albatross for any year after the tsunami is 472. This is a significant decline from the mean count of 3125 for any year before the tsunami. The β2 is the mean banded Laysan albatrosses less than the mean number of Black-footed albatross before the tsunami. The mean Laysan albatross count prior to the tsunami was 2,109. The β3 is the mean Laysan albatrosses count post-tsunami compared to Black-footed albatrosses’ post-tsunami. This is also the interaction term. There are, on average, 969 more Laysan albatrosses than Black-footed albatrosses for any year after the Tōhoku tsunami.

Since the p-values are smaller than 0.05, the null hypothesis that there was no impact from the Tōhoku tsunami on albatross counts is rejected. The data provides convincing evidence that there is a negative difference in mean number of both Black-footed and Laysan albatrosses post-tsunami. There is a statistically significant difference, at the 5% significance level, in the count of albatross in Hawaii. The p-value is greater than 0.05 when comparing the mean Laysan albatross count post-tsunami to the Black-footed albatross. This predictor may indicate that Laysan albatross counts were not as significantly impacted by the tsunami as Black-footed albatross. Based on the adjusted R-squared, this model explains 63% of the variation in the albatross count data around its mean.

Conclusion and Future Research

In conclusion, the linear regression model predicts that there are significant negative relationships between the independent variable (years before and after tsunami) and the dependent variable (count of banded albatross). It was also found that a multiple regression with an interaction term based on the tsunami event impact of each albatross species is a better model fit. This analysis suggests that the Tōhoku tsunami had a negative long term impact on albatross populations in Hawaii.

For future analysis on the impact of the Tōhoku tsunami, I would like to research what other variables may be influencing the number of albatrosses being banded by biologist in Hawaii each year. I have a suspicion that there was a reduced banding effort throughout the archipelago since the tsunami, especially in 2020 due to the COVID-19 pandemic. I am unsure why there would be a reduction in banding effort immediately post-tsunami and prior to the pandemic, but this may be a significant influence on the count data used for this analysis. I would also like to contact researchers in Hawaii to determine if there is more comprehensive census data for albatross.

Two Black-footed albatross practicing their mating dances and calls, while a Laysan albatross sits quietly on a nest in the background. The Black-footed albatross on the right has been banded with two types of bands (color and incoloy metal). The Black-footed albatross on the left has not been banded.
Image Credit: I. Nimz

Corrections

If you see mistakes or want to suggest changes, please create an issue on the source repository.

Source Code

Find the source code for this blog post here.