2.1 Spatial visualization

In this post, we assess the global trade regarding vaccine supply sources and international commerce using UN’s Comtrade data.

# The Comtrade data previously acquired
load(paste0(prodClass,'/comtrade_vaccine.RData'))
dt_list<-dt_list[c('2021','2020','2019','2018','2017','2016','2015',
                   '2010','2005','2000','1995','1990')]

In this .md, we assess the UN Comtrade’s vaccine data acquired previously by exploring spatial data analysis.

The database does not distinguish the type of vaccines, nor are they deployed regarding the infections against which they are designed.

The variable selected for the analysis is the total exports in values for vaccines, as the net weight values were not present for all countries in the returned data. Also, the records for net weight can be presented in the database in different identified quantities, comprising different mass presentations (Kg, thousands of kilograms) to count in the amount of units, which makes them not easily comparable.

Some inconsistencies in the data, such as missingness and mismatch among exports and imports because of methodological limitations, are known, but the Comtrade database is described as already treated for major inconsistencies.

Also, as the database is about international trade, it does not account for local consumption and does not reflect the actual production estimates.

Spatial data visualization: points and regions in the map

Comtrade’s data does not comprise spatial data per se, but it can be converted to spatial data and conduct visualization and analysis easily, by attributing coordinates using the country-related (Reporter and Partner) variables. There are some specific packages for spatial data analysis. And certainly, ggplot2 is not the mainstream tool for such analysis. But as the major and the largest R package for visualization, it has some tools available to ease some basic spatial data visualization.

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(ggplot2)
library(gtable)
library(gridExtra)

## 
## Attaching package: 'gridExtra'

## The following object is masked from 'package:dplyr':
## 
##     combine

library(maps)
library(viridis)

## Loading required package: viridisLite

## 
## Attaching package: 'viridis'

## The following object is masked from 'package:maps':
## 
##     unemp

library(classInt)
library(treemapify)

Preparing the data

Thanks to the ubiquity of ggplot2, there are data and functions available for creating basic visualizations for spatial data analysis. We can use the map_data feature from ggplot2 to draw the regions as polygons without needing specialized packages.

The map_data enable us to access data for different levels of map data with coordinates to draw regions (and/or sub-regions) polygons.

the dataset (named globe here) is a data frame extracted from ggplot2 to enable to plot ‘world’ map, comprising coordinates to draw the world’s regions (countries) polygons.

globe <- map_data("world")

We prepare the Comtrade data to be compatible with the map data file. Fisrt, by making consolidations about different names used in thdatasets,ts

comtrade_countries<-dt_list[['2021']]$Reporter%>%unique()
mapdata_countriews<-globe$region%>%unique()
comtrade_countries[!comtrade_countries%in%mapdata_countriews]

## [1] "Brunei Darussalam"   NA                    "Czechia"            
## [4] "Republic of Moldova" "Trinidad and Tobago"

dt_list<-lapply(dt_list,function(dt){
  dt<-dt %>%
      mutate(Reporter = case_when((Reporter == 'Russian Federation') ~ 'Russia',
                             (Reporter == 'United Kingdom')~'UK',
                             (Reporter == "Republic of Korea") ~"South Korea",
                            (Reporter ==  "United States of America") ~'USA',
                            TRUE~Reporter),
             Partner=case_when((Partner=='Russian Federation') ~ 'Russia',
                             (Partner == 'United Kingdom')~'UK',
                            (Partner ==  "Republic of Korea") ~"South Korea",
                           ( Partner ==  "United States of America") ~'USA',
                           TRUE~Partner)
             )
  dt
})

And computing total trade per exporter:

dt_list<-lapply(dt_list,function(dt){
  dt<-dt%>%group_by(Reporter)%>%mutate(TotalTrade=sum(primaryValue))%>%ungroup()
  dt
})

We merge the globe data frame with the comtrade data to prepare a spatial data.

mapdata_list<-lapply(dt_list,function(dt){
  mapdata <- left_join(globe, dt, by=c("region"="Reporter"))
  mapdata
})

And also prepare another data frame in parallel, adding cities’ locations for bubble map, by acquiring the world.cities data from maps package, making consolidation in some country names; and merging them with the comtrade acquired data

# getting world.cities data ========== 
data(world.cities)
capital_cities<-world.cities[world.cities$capital==1,]
comtrade_countries<-dt_list[[1]]$Reporter%>%unique()
mapdata_countriews<-capital_cities$country.etc%>%unique()
comtrade_countries[!comtrade_countries%in%mapdata_countriews]

## [1] "Brunei Darussalam"   NA                    "Czechia"            
## [4] "Fiji"                "South Korea"         "Republic of Moldova"
## [7] "Montenegro"

capital_cities<-capital_cities%>%mutate(country.etc2=country.etc)
capital_cities[capital_cities$country.etc=='Korea South','country.etc']<-'South Korea'
capdata_list<-lapply(dt_list,function(dt){
  capdata<-left_join(dt,capital_cities, by=c("Reporter"="country.etc"))
  capdata<-capdata%>%filter(!is.na(lat),!is.na(long))
})

And treat the resultant data frame to have vectors of trade from reporter to partner to draw curves illustrating the flow of materials in the trade.Wee treat also to present the partner’s name.

capdata_vec_list<-lapply(capdata_list,function(capdata){
  # get geolocation information for partner ===========
  capdata_vec<-left_join(capdata,capital_cities, by=c("Partner"="country.etc"))
})

Visualization

We explore the data with three complementary visualization methods:

colored map;
bubble map;
simple bar plot;
and tree map;

colored map according to total exportation

Note how a colored map is interesting, but cannot illustrate well countries with small geographical areas.

TotalExpList<-lapply(c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990"),function(x){
  mapdata<-mapdata_list[[x]]
  g<-mapdata%>%ggplot() +
  geom_map(aes(map_id = region, fill = TotalTrade), map = globe) +
  geom_polygon(data = mapdata, aes(x=long, y = lat, group = group), colour = 'black', 
  fill='grey', alpha=0.2) +
  expand_limits(x = globe$long, y = globe$lat) +
  scale_size_continuous(breaks = 10^(1:10),trans='log10') +
    scale_color_viridis(trans="log10",
                      breaks = 10^(1:10),direction=-1,option='cividis') +
  theme_void() + coord_fixed() +ggtitle(paste0('Exports of ',prodClass, ' (U$): ',x,''))
  g
})
names(TotalExpList)<-c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990")

plots_list2<-sapply(names(TotalExpList),function(x){
  p=TotalExpList[[x]]
  p+  theme(legend.position="none")+ggtitle(x)+
    theme(plot.title = element_text(size = 30, face = "bold")) 
},USE.NAMES = TRUE,simplify = FALSE)
legend = gtable_filter(
  ggplot_gtable(
    ggplot_build(
      TotalExpList[[1]]+
        theme(legend.position = "bottom",
              legend.text=element_text(size=25),
              legend.title = element_text(size=30)
      ))
    ), "guide-box")

p<-grid.arrange(arrangeGrob(grobs = plots_list2, ncol = 3),
             legend,
  nrow = 2,
             heights = c(9, 1)
)

TotalExport_countMap

bubble map with total exportation (values)

BubleTotal<-lapply(c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990"),function(x){
  capdata<-capdata_list[[x]]
  g<-ggplot() +
  geom_polygon(data = globe, aes(x=long, y = lat, group = group), colour = 'black', fill = NA)+
  geom_point(data=capdata,aes(x=long,y=lat,color=TotalTrade,size=TotalTrade))+
  scale_size_continuous(range=c(1,13),trans="log10",
                        breaks = 10^(1:10)) +
  scale_color_viridis(guide='legend',trans="log10",
                      breaks = 10^(1:10),direction=-1,option='cividis') +
  theme_void() + coord_map() +
    ggtitle(paste0('Exports of ',prodClass, ' (U$): ',x,''))
  g
})
names(BubleTotal)<-c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990")
plots_list2<-sapply(names(BubleTotal),function(x){
  p=BubleTotal[[x]]
  p+  theme(legend.position="none")+ggtitle(x)+
    theme(plot.title = element_text(size = 30, face = "bold")) 
},USE.NAMES = TRUE,simplify = FALSE)
legend = gtable_filter(
  ggplot_gtable(
    ggplot_build(
      BubleTotal[[1]]+
        theme(legend.position = "bottom",
              legend.text=element_text(size=25),
              legend.title = element_text(size=30)
      ))
    ), "guide-box")

p<-grid.arrange(arrangeGrob(grobs = plots_list2, ncol = 3),
             legend,
  nrow = 2,
             heights = c(9, 1)
)

TotalExport_bubbleMap

Simple bar plots

In fact, basic bar plots can already illustrate the above, although without the information about location being actually used.

Barplot_l<-lapply(c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990"),function(x){
  dt<-dt_list[[x]]
  M <- max(dt$TotalTrade)
  a<-dt%>%filter(!is.na(Reporter))%>%select(Reporter,TotalTrade)%>%unique()
  Ma<-sum(a$TotalTrade)
  g<-dt%>%filter(!is.na(Reporter))%>%select(Reporter,TotalTrade)%>%unique()%>%
  arrange(desc(TotalTrade))%>%head(20)%>%
    mutate(cum_TotalTrade = cumsum(TotalTrade)/M) %>%
  ggplot(aes(x=reorder(Reporter,TotalTrade),y=TotalTrade))+
    geom_bar(stat="identity")+
    geom_line(mapping=aes(y=cum_TotalTrade*M,group=1))+
    scale_y_continuous(limits = c(0,Ma),
      sec.axis = sec_axis(~ ./Ma, 
                          labels = scales::percent,
                          name = "Cummulative percentage"))+
    coord_flip()+
  labs(title=paste0('Top 10 Exporters in total traded (U$) - ',x),x='Country')
})
names(Barplot_l)<-c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990")
plots_list2<-sapply(names(Barplot_l),function(x){
  p=Barplot_l[[x]]
  p+  theme(legend.position="none")+ggtitle(x)+
    theme(plot.title = element_text(size = 14, face = "bold")) 
},USE.NAMES = TRUE,simplify = FALSE)
p<-grid.arrange(arrangeGrob(grobs = plots_list2, ncol = 3))

TotalExport_barPlot

And note how the spatial data visualization essentially intends to show the same information of the bar plot, but has the geographical information added, which brings is interesting in some sense, but yet harder to compare.

we are ignoring the 1990 data, as apparently, the amount of countries reporting the data was limited in that period.
we cannot make strict comparisons, as reporting methodologies can be different across countries, and it is even known such methodologies can even change in time in the same country.
the major exporters were almost stable until 2018
Looking at Pareto plots and tree maps, we see the major players in the global trade related to vaccines changed from 2018 to 2021 because Covid2019 crisis and what we can call as the entering of new players, but still the 5 bigger exporters handle 80% of all global trade.

Tree map

Tree map is another alternative to show the same information, also without considering spatial/location related information.

# Tree map =============
TreeMapList<-lapply(c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990"),function(x){
  dt<-dt_list[[x]]
  g<-dt%>%filter(!is.na(Reporter))%>%select(Reporter,TotalTrade)%>%unique()%>%
  arrange(desc(TotalTrade))%>%
  ggplot(aes(area = TotalTrade, fill = Reporter,label = paste(Reporter, round(TotalTrade/1000000), sep = "\n"))) +
  geom_treemap()+
  geom_treemap_text(grow = TRUE)+
  theme(legend.position = "none")+
    labs(title=paste0('Total Export (10⁶U$) - ',x))
  g
})
names(TreeMapList)<-c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990")
plots_list2<-sapply(names(TreeMapList),function(x){
  p=TreeMapList[[x]]
  p+  theme(legend.position="none")+ggtitle(x)+
    theme(plot.title = element_text(size = 30, face = "bold")) 
},USE.NAMES = TRUE,simplify = FALSE)
legend = gtable_filter(
  ggplot_gtable(
    ggplot_build(
      TreeMapList[[1]]+
        theme(legend.position = "right",
              legend.text=element_text(size=18),
              legend.title = element_text(size=20),
              legend.key.size = unit(0.2, "cm")
      )+guides(fill=guide_legend(ncol =1))
      )
    ), "guide-box")

p<-grid.arrange(arrangeGrob(grobs = plots_list2, ncol = 3),
             legend,
  ncol = 2,
             widths = c(9, 1)
)

TotalExport_TreeMap

Before the Covid Outbreak, the vaccine market was led by companies like Merck (US), GlaxoSmithCline GSK (mostly manufacturing at Belgium) and Sanofi (France, US, Canada, Mexico, China, India). Now, Pfizer escalated their participation in the vaccines world (Sagonowky 2022).

Note the counts above do not reflect the manufacturing capability of each country, as the attendance of local demand is not included in the data, just exports. So, it highlights small countries with small (or at least not big) population which are set as strategical for global manufacturers, than those targeting local demand.

The Ireland participation is detached as it set itself as strategic for implantation of manufacturing plants of foreign-owned enterprises comprising global top 10 biopharmaceutical companies like Pfizer, Eli Lilly, WuXi, Janssen, MSD, Sanofi, BMS, Alexion, and Allergan.

The major change is the increase of China’s participation, which is based mainly because Sinovac (China) outranking Moderna and several other previous important players.

The USA participation is mostly stable on time. Pre-covid, such contribution were because major players like Merk and Sanofi. Currently, they have been outranked by Pfizer, which also has manufacturing plants in the US, thought also in Germany, Belgium, Ireland, and Croatia. And also Moderna, another relevant player in providing solutions for Covid Outbreak, is a US company, and manufactures vaccines in the US, thought partnered with Lonza with sites in US and Switzerland; Recipharm, from France, and Rovi from Spain to offer vaccines. The Germany’s participation is because Pfizer and Moderna’s partner BioNTech, the mRNA technology innovator, is in Germany. AstraZeneca used manufacturing facilities across Europe, North, Central and South America, Asia and Australia

Plotting trade connections

Trade is not just about the total amount, but also, or more, about the interactions. Here, beyond the information on the exported amount/value, it is also about connections, about who exports to who.

The below figure confirms how the world vaccine trade is really interconnected, but yet does not bring clarity on critical supply sources and bottlenecks which can make, for example, about how susceptible is the supply chain for disruption.

library(ggnewscale)
plots_list<-(lapply(c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990"),function(x){
  capdata_vec<-capdata_vec_list[[x]]%>%filter(!is.na(long.x),!is.na(lat.x))
  g<-ggplot() +
  geom_polygon(data = globe, aes(x=long, y = lat, group = group), colour = 'black', fill = NA)+
  geom_curve(data = capdata_vec, aes(x = long.x,    y = lat.x, 
                               xend = long.y, yend = lat.y,
                               linewidth=primaryValue,color=primaryValue#,color=NetWgt_cat
                               ),alpha=0.3,
                              #color="blue",
             show.legend=c(linewidth=TRUE,color=TRUE))+
    scale_size_continuous(range=c(0.01,1),trans="log10",limits=c(1,10^9),
                        breaks = 10^(c(1,3,5,7,9))) +
  scale_color_viridis(guide='legend',trans="log10",limits=c(1,10^9),
                      breaks = 10^(c(1,3,5,7,9)),direction=-1,option="mako") +
    new_scale("size") +new_scale_color()+
  geom_point(data=capdata_vec,aes(x=long.x,y=lat.x,
                                  color=TotalTrade,
                                  size=TotalTrade),show.legend=c(size=TRUE,color=TRUE))+
    scale_size_continuous(range=c(1,13),trans="log10",limits=c(1,10^10),
                        breaks = 10^(1:10))+
  scale_color_viridis(guide='legend',trans="log10",limits=c(1,10^10),
                      breaks = 10^(1:10),direction=-1,option='cividis') +
  #scale_size_continuous(range=c(1,9),trans="log10") +
  #scale_color_viridis(trans="log10") +
  coord_equal()+
    ggtitle(paste0('Exports of ',prodClass, '(U$): ',x,''))
  g
}))
names(plots_list)<-c("2021","2020","2019","2018","2017","2016","2015",
         "2010","2005","2000","1995", "1990")

library(gtable)
library(gridExtra)
plots_list2<-sapply(names(plots_list),function(x){
  p=plots_list[[x]]
  p+  theme(legend.position="none")+ggtitle(x)+
    theme(plot.title = element_text(size = 30, face = "bold")) 
},USE.NAMES = TRUE,simplify = FALSE)
legend = gtable_filter(
  ggplot_gtable(
    ggplot_build(
      plots_list[[1]]+
        theme(legend.position = "bottom",
              legend.text=element_text(size=25),
              legend.title = element_text(size=30)
      ))
    ), "guide-box")
p<-grid.arrange(arrangeGrob(grobs = plots_list2, ncol = 3),
             legend,
  nrow = 2,
             heights = c(9, 1)
)

TotalExport_Paths

Conclusion remarks

This markdown has illustrated the value of spatial data, displaying both the count information along with the localization and geographical details, then connecting them to show trade routes. This illustration gave an example of the capability of this type of visualization, which goes beyond bar graphs or treemaps, and how it can be assembled using R’s ggplot2 package.

But when trying to describe the evolution in time, year-by-year visualizations of any kind are limited. And abridging measures are more beneficial. Network analysis is a useful tool to measure the intricacy of the global trade network and its progression over time, a matter we will evaluate more deeply in the upcoming markdown.

References

D. Kahle and H. Wickham. ggmap: Spatial Visualization with ggplot2. The R Journal, 5(1), 144-161. URL http://journal.r-project.org/archive/2013-1/kahle-wickham.pdf

Laker, Benjamin. Global Supply Chain Crisis: Lessons For Leaders. Forbes, jan 11 2023. available at: https://www.forbes.com/sites/benjaminlaker/2023/01/11/global-supply-chain-crisis-lessons-for-leaders/?sh=20f96e534a8d

Shiffling, Sarah; Kanellos N.V. 5 challenges facing global supply chains. World Economic Forum. Sep 7, 2022. available at: https://www.weforum.org/agenda/2022/09/5-challenges-global-supply-chains-trade

Sagonowky, E.; Liu, A.; Kansteiner, F.; Becker, Z.; Dunleavy, K. The top 10 vaccine companies worldwide. Fierce Pharma Special Report- Vaccine Players. October 17, 2022. Available at: https://www.fiercepharma.comSanofi/pharma/top-10-vaccine-companies-worldwide

Hannah Finch. Where the Moderna vaccine is being made: The Moderna vaccine is the third Covid-19 vaccine available in the UK and has been instrumental in the booster programme. BusinessLive. 29 MAR 2021. available at: https://www.business-live.co.uk/technology/moderna-vaccine-being-made-20278092

At a glance. GSK. https://be.gsk.com/en-be/company/at-a-glance/#:~:text=Belgium%20is%20the%20heart%20of,industrial%20network%20of%20vaccines%20worldwide.

Vaccines. Merk. https://www.merck.com/research/vaccines/

Making vaccines for public health. Sanofi. https://www.sanofi.com/en/your-health/vaccines/production

Pushing boundaries to deliver COVID-19 vaccine across the Globe: Delivering COVID-19 vaccine part 2. AstraZeneca. https://www.astrazeneca.com/what-science-can-do/topics/technologies/pushing-boundaries-to-deliver-covid-19-vaccine-accross-the-globe.html

Department of Business, Enterprise and Innovation. Government of Ireland. Focus on Biopharmachem. August 2020 https://enterprise.gov.ie/en/publications/publication-files/focus-on-biopharmachem-2020.pdf

Assessing the global trade of vaccines using spatial data visualization and UN’s Comtrade data

Tomoe Gusberti, Dr. Eng.