Malaria Suitability Indicator

Earth Sciences department, Barcelona Supercomputing Center (BSC)

2022-08-29

Climate Suitability for Malaria Transmission

Introduction

Climate sensitive infectious diseases are a subject of major concern for public health around the world in the context of global change. Malaria is an infectious disease caused by parasites of the genus Plasmodium and transmitted by Anopheles mosquitoes (Boualam et al., 2021). Eradication efforts managed to achieve complete elimination of endemic malaria in Europe since 1974, although sporadic transmission events are frequently reported in travelers coming from other endemic areas. Studies evaluating the factors that keep malaria out of the continent found that high socio-economic conditions and increased life expectancy are key (Fischer et al., 2020). On the other hand, climatic conditions are becoming more suitable for malaria transmission (Zhao et al., 2016).

The Lancet Countdown in Europe collaboration was launched in 2021 with the purpose of using a variety of indicators to monitor trends in the impact of climate on the dynamic of different health processes. As part of thisinitiative, the Climate Suitability for Malaria Transmission indicator was adapted from the Global Lancet Countdown Collaboration (Romanello et al., 2021) for tracking the number of months suitable of malaria transmission per year in the continent between 1950 and 2021.This indicator is a threshold based model that overlaps the climatic and environmental requirements of Anopheles mosquitoes and Plasmodium vivax, which was the parasiteof endemic circulation in the continent until the mid 1970s (Boualam et al., 2021).

Although the CSIndicators was initially developed for providing a set of tools useful in the context of agricultural applications, it is also possible to expand its applications to the computation of other type of indicators. In this vignette, the Climate Suitability for Malaria indicator will be computed using the TotalTimeExceedingThreshold function. The output will be an multidimensional array containing the number of months per year suitable for malaria transmission.

Load libraries

In addition to the CSIndicators package, the CSTools and s2dv packages will be used for data extraction and visualization.

library(CSTools)
library(CSIndicators)
library(s2dv)

1. Data

1.1 Area of interest

The number of months suitable for malaria transmission between 2010 and 2020 will be computed for the Iberian Peninsula (40º14´24´´N 4º14´21´´W). This area comprises the territories of Portugal, Spain, France, Andorra and Gibraltar.

1.2 Data extraction

Malaria indicator is computed with climate and environmental data. Climate variables are extracted from the ERA5-Land reanalysis dataset (Muñoz-Sabater et al., 2021), which span between 1950 and present day, at a 9km resolution. Necessary climate variables are 2 meter temperature (tas), 2 meter dew point temperature (tdps) and total precipitation (prlr). Environmental information is extracted from the CORINE Land Cover inventory, which classifies land into 44 classes at a 100 m resolution. These data are produced by the Copernicus service and extracted and reprocessed by the Earth Sciences department from the Barcelona Supercomputing Center.

The function CST_Load from the CSTools package allows the user to access the processed data and delivers the data in the form of an s2dv_cube object. These multidimensional objects store data and metadata in several elements. For further information about these objects visit Data retrieval and storage.

CST_Load has its requirements for extracting data. First, the path to the files is introduced using whitecards delimited by dollar signs. Second, the temporal extent and resolution of the output con be provided with sdates, which takes a vector of all dates of interest. The spatial extent is inputted within lonmax, lonmin, latmax and latmin.

# Path to file locations using whitecards
path_ERA5_CDS <- list(path='/esarchive/recon/ecmwf/era5land/$STORE_FREQ$_mean/$VAR_NAME$_f1h/$VAR_NAME$_$YEAR$$MONTH$.nc')

# Temporal extent and resolution:
# We will work with all months from 2010 to 2020
year_in <- 2020 # initial year
year_fi <- 2021 # last year
month_in <- 1 # first month
month_fi <- 12 # last month

sdates <- paste0(year_in:year_fi, '0101')

# Extract data
vars <- c('tas', 'tdps', 'prlr')

out <- NULL
for(var in vars){
  out[[var]] <- CST_Load_s2dv(var = var,
                           exp = NULL, # We only require observed data
                           obs = list(path_ERA5_CDS),
                           sdates = sdates,
                           lonmax = 355, lonmin = 350,
                           latmax = 45, latmin = 43,
                           storefreq = 'monthly',
                           leadtimemin = month_in, 
                           leadtimemax = month_fi,
                           output = "lonlat")  
}

# library(zeallot)
# c(tas, tdps, prlr) %<-% CST_Load_s2dv(var = c('tas', 'tdps', 'prlr'),
#                                       exp = NULL, # We only require observed data
#                                       obs = list(path_ERA5_CDS),
#                                       sdates = sdates,
#                                       lonmax = 355, lonmin = 350,
#                                       latmax = 45, latmin = 43,
#                                       storefreq = 'monthly',
#                                       leadtimemin = month_in, 
#                                       leadtimemax = month_fi,
#                                       output = "lonlat")

The call above creates a list with the three datasets. Each s2dv_cube object has the following dimensions:

dim(out$tas$data)
## dataset  member   sdate   ftime     lat     lon 
##       1       1       2      12      21      51

And it is possible to have a look at the data:

# A summary of the entire dataset
summary(out$tas$data)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##   272.5   281.8   284.7   284.9   288.7   292.9   19680
# Sneak peek of a sample of the temperature data in Jan 2010
out$tas$data[1,1,1,1,,][15:20, 15:20]
##        [,1]   [,2]   [,3]   [,4]   [,5]   [,6]
## [1,]     NA     NA     NA     NA 283.70 283.39
## [2,]     NA     NA     NA     NA 283.36 282.70
## [3,]     NA     NA     NA 283.27 283.00 282.03
## [4,] 282.94 282.79 282.59 282.57 282.37 281.64
## [5,] 281.89 281.58 281.53 281.56 281.39 280.84
## [6,] 281.18 280.96 280.88 280.82 280.56 280.11

2. Data transformation

2.1 Temperature and dew point temperature

This variables will be used to compute relative humidity, which requires them to be in Degrees Celsius (C):

for(var in c('tas', 'tdps')){
  out[[var]]$data <- out[[var]]$data - 273.15
}

summary(out$tas$data)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##   -0.67    8.61   11.56   11.77   15.56   19.78   19680

2.3 Relative humidity

After transforming temperature (\(T\)) and dew point temperature (\(T_d\)) units from Kelvin (K) to Degrees Celsius (C), it is possible to compute the relative humidity using August-Roche-Magnus equation (Alduchov 1996):

\[ RH=\frac{100*exp(\frac{aT_d}{b+T_d})}{exp(\frac{aT}{c+T})} \] Where a and b are the coefficients 17.625 and 243.04, respectively.

# Create a template with the same dimensions as the datasets
out$rh <- out$tas

# Compute relative humidity (RH)
out$rh$data <- 100*(exp((17.625*out$tdps$data)/(243.04+out$tdps$data)) / 
                      exp((17.625*out$tas$data) / (243.04+out$tas$data)))

summary(out$rh$data)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##   67.55   79.30   81.90   81.88   83.97  116.08   19680

2.2 Precipitation

Precipitation is extracted from the hourly reanalaysis dataset and processes as monthly meters per second. For the computation of the malaria indicator, it is necessary to transform it into accumulated mm per month:

# To compute the accumulated precipitation per month it is necessary to multiply by:
# - total number of seconds in one hour: 3600
# - total number of hours in a day: 24
# - average number of days in a month: 30.41
# - change factor from meter to mm: 1000
out$prlr$data <- out$prlr$data*3600*24*30.41*1000

summary(out$prlr$data)
##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
##    2.22   61.84   99.72  107.85  142.38  395.53   19680

2.4 Land use

Add chunk

3. Compute indicator

The suitability for malaria transmission indicator estimated the number of months suitable for transmission of P. vivax, calculated from the following thresholds:

Additionally, Anopheles suitability is influenced by the surrounding conditions, so that the following land cover classes are considered highly suitable (Benali et al., 2014):

# Create a template with the same dimensions as the datasets
malariaSuit <- out$tas

# Determine suitability
malariaSuit$data <- ifelse(out$tas$data >= 14.5 & 
                             out$tas$data <= 33 & 
                             out$rh$data >= 60 & 
                             out$prlr$data >= 80, 1, 0)

# Compute number of number of months per year that are suitable
# (i.e. suitability = 1)
malariaInd <- CSIndicators::CST_TotalTimeExceedingThreshold(malariaSuit, 
                                                            threshold=0.5)

# Sneak peek of the data
malariaInd$data[1,1,1,,][15:20, 15:20]
##      [,1] [,2] [,3] [,4] [,5] [,6]
## [1,]   NA   NA   NA   NA    2    2
## [2,]   NA   NA   NA   NA    2    2
## [3,]   NA   NA   NA    2    1    2
## [4,]    1    1    1    1    1    2
## [5,]    1    1    1    1    1    1
## [6,]    0    1    1    1    1    1

4. Visualization

4.1. Map

Using the function PlotEquiMap() from the s2dv package, it is possible to have a visual inspection of the number of suitable months in January 2010.

PlotLayout(PlotEquiMap, c('lat', 'lon'),
           var=malariaInd$data[1,1,,,],
           nrow=1, ncol=2,
           lon=malariaInd$lon,
           lat=malariaInd$lat,
           filled.continents=FALSE,
           brks=seq(0, 12, by=1),
           toptitle="Number of months per year suitable for malaria in the Iberian Peninsula",
           titles=c("2010", "2011"),
           title_scale=0.5,
           coast_width=2,
           filled.oceans=TRUE,
           country.borders=TRUE,
           intylat=1,
           intxlon=1)

4.2 Time series

Add chunk

5. Summary statistics

Add chunk

6. Conclusions

Add chunk

References