ArticlePDF Available

Correction: Environmental Predictors of Seasonal Influenza Epidemics across Temperate and Tropical Climates

Authors:
  • Center for Welfare Metrics

Abstract and Figures

Human influenza infections exhibit a strong seasonal cycle in temperate regions. Recent laboratory and epidemiological evidence suggests that low specific humidity conditions facilitate the airborne survival and transmission of the influenza virus in temperate regions, resulting in annual winter epidemics. However, this relationship is unlikely to account for the epidemiology of influenza in tropical and subtropical regions where epidemics often occur during the rainy season or transmit year-round without a well-defined season. We assessed the role of specific humidity and other local climatic variables on influenza virus seasonality by modeling epidemiological and climatic information from 78 study sites sampled globally. We substantiated that there are two types of environmental conditions associated with seasonal influenza epidemics: "cold-dry" and "humid-rainy". For sites where monthly average specific humidity or temperature decreases below thresholds of approximately 11-12 g/kg and 18-21°C during the year, influenza activity peaks during the cold-dry season (i.e., winter) when specific humidity and temperature are at minimal levels. For sites where specific humidity and temperature do not decrease below these thresholds, seasonal influenza activity is more likely to peak in months when average precipitation totals are maximal and greater than 150 mm per month. These findings provide a simple climate-based model rooted in empirical data that accounts for the diversity of seasonal influenza patterns observed across temperate, subtropical and tropical climates.
Content may be subject to copyright.
Environmental Predictors of Seasonal Influenza
Epidemics across Temperate and Tropical Climates
James D. Tamerius
1,2
*, Jeffrey Shaman
1,2
, Wladmir J. Alonso
2
, Kimberly Bloom-Feshbach
3
,
Christopher K. Uejio
4
, Andrew Comrie
5
,Ce
´cile Viboud
2
1Environmental Health Sciences, Columbia University, New York, New York, United States of America, 2Fogarty International Center, National Institutes of Health,
Bethesda, Maryland, United States of America, 3Mount Sinai School of Medicine, New York, New York, United States of America, 4Department of Geography, Florida State
University, Tallahassee, Florida, United States of America, 5School of Geography and Development, University of Arizona, Tucson, Arizona, United States of America
Abstract
Human influenza infections exhibit a strong seasonal cycle in temperate regions. Recent laboratory and epidemiological
evidence suggests that low specific humidity conditions facilitate the airborne survival and transmission of the influenza
virus in temperate regions, resulting in annual winter epidemics. However, this relationship is unlikely to account for the
epidemiology of influenza in tropical and subtropical regions where epidemics often occur during the rainy season or
transmit year-round without a well-defined season. We assessed the role of specific humidity and other local climatic
variables on influenza virus seasonality by modeling epidemiological and climatic information from 78 study sites sampled
globally. We substantiated that there are two types of environmental conditions associated with seasonal influenza
epidemics: ‘‘cold-dry’’ and ‘‘humid-rainy’’. For sites where monthly average specific humidity or temperature decreases
below thresholds of approximately 11–12 g/kg and 18–21uC during the year, influenza activity peaks during the cold-dry
season (i.e., winter) when specific humidity and temperature are at minimal levels. For sites where specific humidity and
temperature do not decrease below these thresholds, seasonal influenza activity is more likely to peak in months when
average precipitation totals are maximal and greater than 150 mm per month. These findings provide a simple climate-
based model rooted in empirical data that accounts for the diversity of seasonal influenza patterns observed across
temperate, subtropical and tropical climates.
Citation: Tamerius JD, Shaman J, Alonso WJ, Bloom-Feshbach K, Uejio CK, et al. (2013) Environmental Predictors of Seasonal Influenza Epidemics across
Temperate and Tropical Climates. PLoS Pathog 9(3): e1003194. doi:10.1371/journal.ppat.1003194
Editor: Steven Riley, Imperial College London, United Kingdom
Received August 28, 2012; Accepted December 26, 2012; Published March 7, 2013
This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for
any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.
Funding: JDT acknowledges support from the Earth Institute Postdoctoral Fellowship program at Columbia University. Support was also provided by the in-
house Influenza Research Program of the Division of International Epidemiology and Population Studies, Fogarty International Center, National Institutes of
Health, which is funded by the International Influenza Unit, Office of Global Affairs, Department of Health and Human Services, the Research and Policy for
Infectious Disease Dynamics (RAPIDD) program of the Science and Technology Directorate, Department of Homeland Security, and the National Institutes of
Health Models of Infectious Disease Agent Study program (MIDAS) through cooperative agreement 1U54GM088558. The funders had no role in study design,
data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: jt2684@columbia.edu
Introduction
Influenza exerts a significant health burden on human
populations across temperate, subtropical and tropical regions
[1]. The striking seasonal pattern that characterizes influenza in
temperate populations has long suggested a causal link between
seasonal fluctuations in climatic and social factors and influenza
transmission [2–4]. Temperate regions of the northern and
southern hemispheres are characterized by highly synchronized
annual influenza epidemics during their respective winter months
[5,6]. In contrast, influenza seasonal characteristics are more
diverse in tropical and subtropical regions; some sites experience
annual epidemics coinciding with the local rainy season [7–11],
whereas others are characterized by semi-annual epidemics or
year-round influenza activity without well-defined influenza
seasons [7,12,13].
Recent epidemiological studies indicate that low levels of
specific humidity are associated with the onset of pandemic and
epidemic influenza in the US [14,15], consistent with laboratory
experiments and animal models suggesting that low specific
humidity favors virus survival and aerosol transmission [16–17].
There are several alternative explanations for the winter seasonal
transmission of influenza in temperate regions, including the
inhibition of host immune function due to decreased exposure to
solar radiation [18,19], and the inhibition of mucociliary clearance
by the inhalation of cold-dry air [20]. Person-to-person contact
rates may also strengthen in the winter due to increased indoor
crowding, modulated by school terms [21]. There are few
biological explanations for the association between precipitations
and influenza activity reported in some tropical and subtropical
regions, although rainy conditions may also favor indoor crowding
[4].
Although it is common for epidemiological studies to examine
relationships between seasonal influenza activity and climatic
factors for individual sites, few studies have assessed the
consistency of these relationships across a broad range of
temperate, subtropical and tropical sites. A recent study evidenced
a link between influenza and temperature based on aggregate
country-level data, but did not characterize the geographical and
climatic boundaries that define regions experiencing different
PLOS Pathogens | www.plospathogens.org 1 March 2013 | Volume 9 | Issue 3 | e1003194
influenza seasonality patterns [22]. Here we investigate both
relative and absolute associations between climatic factors and the
timing of seasonal influenza epidemics for 78 individual sites
sampled globally [23]. We develop models that predict the month
of peak influenza activity for each study site as a function of
climatic variables and identify climatic thresholds accounting for
the diversity of influenza seasonality patterns observed globally.
Methods
Data
Influenza epidemiological data. We used a recently
developed global database that provides information on the
month of maximum influenza activity (‘‘influenza peaks’’) for 78
sites worldwide, of which 39% were located in the tropics. A
detailed description of the database is provided in the Supplement
and [23], and is briefly summarized below.
The data were compiled based on a systematic literature review
of published influenza and respiratory virus surveillance studies,
reporting weekly or monthly laboratory-confirmed influenza cases
for a period of 12 consecutive months or more, augmented with
electronic data from regional or national influenza surveillance
schemes. Studies focused on the 2009 A/H1N1 pandemic virus
were excluded to restrict the analysis to seasonal patterns of inter-
pandemic influenza.
We identified 85 studies matching our inclusion criteria,
encompassing 78 sites in 40 countries sampled during the period
1975–2008, with median study duration of 2 years (Figure 1). A
majority of study sites (76%) represented a specific city rather than
a state, province, or region. The 24% of study sites representing
regional level data encompassed areas that were relatively small
and homogeneous with respect to climate, including countries
such as Italy and the Republic of Korea; states within large
countries such as Michigan, USA or Victoria, Australia; climato-
logically-homogeneous regions in Peru; provinces within Thai-
land; multiple cities in northern Argentina or in Taiwan; and a
subtropical island in Japan.
For sites with multiple years of data, the peak influenza month
of each year was determined and the average month of peak
influenza was calculated. Because some sites were characterized by
semi-annual influenza activity, influenza peaks separated by four
months or greater were considered distinct influenza seasons (see
Supplement for more information). Indeed, 17 sites (22% of the
dataset) concentrated in East and South-East Asia, and equatorial
regions of Central and South America, were characterized by two
distinct influenza peaks within the year (Figure 1). Of these 17 sites
with semi-annual influenza peaks, 15 had a primary peak (present
in all study years) and a secondary peak (present in a subset of
years), while primary and secondary influenza peaks were not
distinguishable in two sites. We report results from analyses
performed on the superset of sites consisting of all primary and
Author Summary
Human influenza infections have a pronounced seasonal
cycle in temperate regions. Recent laboratory and epide-
miological evidence suggests that low humidity conditions
in the winter may increase virus survival and enable the
virus to transmit efficiently between hosts. However,
seasonal influenza activity in some tropical locations
occurs during the rainy season, whereas other tropical
locations do not experience a well-defined influenza
season. The primary goal of this study was to identify
the relationship between the timing of seasonal influenza
epidemics and climate variability across the globe. We
show the importance of thresholds in humidity, temper-
ature and precipitation that affect the epidemiology, and
potentially the transmission route, of influenza. A better
understanding of the environmental, demographic and
behavioral drivers of influenza seasonality is important for
optimizing intervention strategies, especially in low and
middle-latitude regions.
Figure 1. Map of 78 study sites included in this study. The site symbols indicate whether a location has annual or semi-annual influenza
activity, and symbol size is proportional to the duration of the epidemiological studies used to determine the month of peak activity for each
location.
doi:10.1371/journal.ppat.1003194.g001
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 2 March 2013 | Volume 9 | Issue 3 | e1003194
secondary influenza peaks (n = 96), as well as the subset restricted
to primary influenza peaks (n = 76).
In addition to the epidemiological influenza database gleaned
from the literature and encompassing 78 sites, we collected
laboratory-confirmed influenza epidemiological data for 9 coun-
tries from FluNet, the WHO global influenza surveillance effort
[24], to ensure proper model validation with an independent
disease dataset. The group of 9 countries (Spain, Tunisia, Senegal,
Philippines, Vietnam, Colombia, Paraguay, South Africa, and
Argentina) was selected because they were latitudinally diverse,
with a heavy focus on subtropical and tropical regions, each
country was relatively small geographically, and provided several
years of data. We also favored countries that were not represented
in the original 78-location database.
Climate data. For the 78 influenza sites, we compiled
average monthly temperature (uC), relative humidity (%) and
precipitation (mm) data from the CRU/Oxford/IWMI 109
latitude/longitude gridded dataset (CRU CL 2.0) [25]. This
dataset was selected because of its global coverage, high spatial
resolution, and monthly temporal resolution (equal to the temporal
resolution of the epidemiological dataset). We then calculated the
average monthly specific humidity (g/kg)—a measure of absolute
humidity—from relative humidity, temperature and surface
pressure (estimated from elevation) [14].
Because the CRU climate dataset consists of monthly averages
for the arbitrary multiyear period of 1960–1991 and does not
necessarily represent the local monthly conditions in the years
sampled for influenza activity, we also considered a more recent
meteorological dataset from the NCEP/NCAR Global Reanalysis
(GR) project [26]. Unlike the CRU dataset, the GR dataset
provides comprehensive time series data for 1948-present, which
enables calculation of average monthly meteorological conditions
for the appropriate time periods at all sites. The major drawback
of the GR dataset is its coarser spatial scale (2u62ulatitude/
longitude), which can obscure local variability in weather and
climate.
To evaluate potential biases associated with a mismatched time
period (CRU dataset) and a coarse spatial resolution (GR dataset),
we performed a comparative analysis of both datasets (Figure S1 in
Text S1). We established that the root-mean-square (RMS) error
introduced by the mismatched temporal period was 3–5 times
smaller than the RMS error introduced by a coarse spatial
resolution. Thus, herein we report the results of analyses based on
temperature, specific humidity, relative humidity and precipitation
data from the more spatially-resolved CRU dataset. We also use
the solar radiation variable from the GR dataset as the CRU
dataset does not have appropriate solar radiation information
(Supplement). Finally, we obtained climate data from the CRU
and GR datasets for the most populous city in each of the 9
countries selected from FluNet for additional model validation.
Statistical Analyses of Influenza and Climate
Exploratory rank analyses. We conducted exploratory
analyses based on a non-parametric rank order approach to assess
the relative association between influenza peaks and seasonal
climate variation. Specifically, we ranked the monthly values of
temperature, solar radiation, specific humidity, and precipitation
for each site in ascending order. We calculated the mean of the
ranks corresponding to the peak influenza month(s) for each site.
Assuming that there is no relative relationship between climatic
factors and timing of influenza peaks, we would expect mean ranks
of 6.5 (‘‘null value’’). We also performed the rank order analysis
across latitude by employing a window spanning 10uof latitude
and sliding it across 5–50uN/S latitude at 2.5uintervals. We
calculated the mean rank for each climate variable corresponding
to influenza peaks for all the sites within each interval. To test for
significance we generated a null distribution (p = 0.05) by boot-
strapping randomly generated distributions for each latitudinal
interval. We evaluated lag relationships of up to 4 months between
climate factors and influenza peaks.
Univariate and multivariate regression. To assess the
absolute relationship between influenza seasonality and climate,
we developed univariate and multivariate logistic regression
models. Because exploratory analyses revealed a bimodal
relationship between influenza peaks and some of the environ-
mental predictors we employed second-degree polynomial func-
tions for climate predictors. We also added interaction terms
describing the deviation of each monthly predictor from its annual
average. The dependent variable was a vector of months
indicating the presence or absence of an influenza peak for each
site and month.
As a sensitivity analysis, we considered mixed effects logistic
models with independent intercepts (exchangeable covariance
structure) to control for the repeated measurements in each site
and different disease periodicities (i.e., annual versus semi-annual
influenza activity). We found this did not have a significant effect
on the modeled relationships and therefore report the results from
classical logistic regression for simplicity.
We used the ‘‘all possible subsets’’ multivariate regression
approach and retained models in which all predictors were
significant (p = 0.05). In conjunction with a jackknife leave-one-out
method we assessed model fit with a ‘‘peak prediction metric’’ that
was defined by calculating the difference between the observed
month of peak influenza activity versus the month with the highest
predicted probability of a peak for each site. The cumulative
proportion of peaks predicted within +/21 month of the observed
peaks were compared to the upper 97.5%, 99% and 99.9%
thresholds of the cumulative proportion of influenza peaks
randomly distributed within +/21 months of the observed peaks
(10,000 runs). In addition to using the peak prediction metric to
quantify model performances across all sites, we compared model
performances between sites in high latitudes (poleward of 25uN/S),
middle latitudes (12.5–25uN/S) and low latitudes (equatorward of
12.5uN/S). We evaluated lag relationships of up to 4 months
between the predictors and influenza peaks.
For additional model validation, we confronted the predictions
of the selected multivariate climate model against the seasonal
distribution of influenza viruses in independent sites selected from
the FluNet database.
Defining influenza geographical and climatic boun-
daries. The rank order and logistic regression analyses indicat-
ed that the relationship between seasonal climate and influenza
peaks was not consistent globally. Low temperatures, solar
radiation and specific humidity corresponded to epidemics in
high latitudes, whereas high levels of precipitation, specific
humidity and relative humidity corresponded to epidemics in
low latitudes. To try to synthesize these results, we explored
whether the seasonal range of climatic factors in a site was
predictive of the environmental conditions during the local
influenza season. Specifically, we constructed a binary dependent
variable by classifying each influenza peak as either ‘cold-dry’ or
‘humid-rainy’ based on whether the influenza peak corresponded
to a month with a specific humidity rank less than or equal to 6 or
greater than or equal to 7, respectively. Specific humidity was
chosen to classify the peaks because the rank-relationship between
influenza peaks in the low latitudes was the opposite of the
relationship observed in high latitudes, and because specific
humidity was significantly correlated with other relevant climate
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 3 March 2013 | Volume 9 | Issue 3 | e1003194
predictors, including temperature (Pearson rho = 0.87, p,0.0001)
and precipitation (Pearson rho = 0.62, p,0.0001). We then used
the site-specific annual minimum and maximum of each
environmental predictor to generate a conditional probability
function through logistic regression. We defined the value at which
the function equaled 0.50 to be the threshold between cold-dry
and humid-rainy influenza locations. We used a jackknife leave-
one-out method to assess the accuracy of the logistic model in
predicting the climate conditions corresponding to each influenza
peak. Finally, for additional model validation, we confronted the
predicted probability function against observed influenza activity
patterns in the independent study sites selected from FluNet.
Results
Influenza Peaks and Climate: Rank Analysis
Across the 78 sites, influenza peaks generally coincided with
months of lower temperature, lower solar radiation and lower
specific humidity than expected under the null hypothesis (mean
rank = 4.3 [95% CI: 3.7, 5.0] for temperature, 4.5 [95% CI: 3.9,
5.1] for solar radiation and 4.8 [95% CI: 4.0, 5.6] for specific
humidity). In contrast, relative humidity and precipitation did not
significantly deviate from the null value (mean rank not
significantly different than 6.5). The association between influenza
peaks, temperature, solar radiation, relative humidity and specific
humidity was most significant when the influenza peaks lagged 1-
month behind the environmental predictors. We obtained similar
results when the analysis was restricted to primary influenza peaks.
A similar analysis performed with a sliding geographical window
revealed that the association between influenza peaks and climatic
variables varied with latitude (Figure 2). The strongest association
was found at high latitudes poleward of 25uN/S, with influenza
peaks preferentially occurring in months with the lowest temper-
ature, solar radiation and specific humidity. Influenza peaks
occurred in the months with the highest levels of relative humidity
and lowest levels of precipitation poleward of approximately
40uN/S. Primary influenza peaks equatorward of 10uN/S
corresponded to the months with the highest annual levels of
specific humidity and precipitation (p,0.05); whereas there was
no association with temperature, solar radiation and relative
humidity. In middle latitudes ranging between 12.5–25uN/S,
there was no significant association between influenza peaks and
climatic variables.
Univariate and Multivariate Models for Influenza Peaks
Temperature and specific humidity were the best individual
predictors of influenza peaks. The model fits improved slightly
when influenza peaks lagged 1-month behind these predictors, and
accurately predicted 56–66% of the peaks in the global datasets,
with highest accuracy at latitudes poleward of 25uN/S (Tables 1
and 2). The modeled relationship between specific humidity and
all influenza peaks was U-shaped, with lowest probability of an
influenza peak at 12 g/kg of specific humidity and increasing
probabilities at lower and higher values (Figure 3). The analysis
restricted to primary peaks revealed a similar relationship, with a
minimum influenza probability at 11 g/kg. Unlike specific
humidity the modeled relationship between temperature and
influenza peaks was monotonic, with the greatest probability of a
peak corresponding to low temperatures. Although the specific
humidity and temperature models were the best predictors of the
timing of influenza peaks across all sites, they were not significant
predictors of influenza peaks equatorward of 25uN/S.
There was a strong inverse relationship between solar radiation
and the probability of an influenza peak, especially when influenza
peaks were lagged by 1-month. The solar radiation model
outperformed the temperature and specific humidity models
based on AIC, but it was not as strong a predictor of the timing
of the influenza peaks (Tables 1 and 2).
Precipitation was a weak predictor of influenza peaks overall,
but it was a strong predictor of influenza peaks equatorward of
12.5uN/S, particularly for primary influenza peaks (p,0.01)
(Tables 1 and 2). Unlike the other climate variables, precipita-
tion-based models performed slightly better when no lag was
considered between precipitation and influenza activity.
Relative humidity was a strong predictor of influenza peaks
globally, particularly when a 1-month lag was applied to the
influenza peaks. There was a positive association between relative
humidity and influenza peaks in high and low latitudes, but this
model was a poor predictor of influenza peaks in middle latitudes.
Further, the relative humidity model was not as strong as the
specific humidity, solar radiation and temperature models
(Tables 1 and 2).
Overall, the multivariate models most predictive of influenza
peak timing included combinations of temperature and precipi-
tation (all peaks, Table 1), and temperature and specific humidity
(primary peaks, Table 2). These models accurately predicted peak
influenza months in 78% and 89% of the 9 independent sites
selected from the FluNet database, respectively (Figure 4). Further,
the models predicted a nearly uniform probability of influenza
peaks every month of the year in equatorial Colombia, a location
that experiences minimal seasonal climatic fluctuation (Figure 4).
Taken together, this analysis exploring the shape of the
relationship between climatic variables and influenza highlights
the covariability between specific humidity and temperature, and
the significant predictive power of these variables at high latitudes.
In contrast, precipitation and relative humidity were predictive of
influenza peaks at low latitudes. Interaction terms describing
monthly deviations from the annual average of each environmen-
tal predictor marginally improved some models but did not affect
the main conclusions (Tables S1 and S2 in Text S1).
Geographic and Climatic Boundaries Predictive of
Influenza Peaks
To further characterize the distribution of influenza peaks
globally and identify the geographical and climatic boundaries
defining influenza seasonality patterns, we categorized sites based
on whether influenza epidemics occurred in months with low
(cold-dry season) or high (humid-rainy season) levels of specific
humidity relative to the local climatology (Figure 5). We found that
the annual minimum level of specific humidity in a site was
predictive of the seasonal characteristics of influenza activity
locally. Sites characterized by cold-dry influenza peaks generally
experienced annual minimum specific humidity values less than
12 g/kg when all influenza peaks were considered, and approx-
imately 11 g/kg when the analysis was restricted to primary
influenza peaks (Figure 5). The minimum specific humidity models
were statistically significant, and classified 75% of the 96 total
influenza peaks correctly (p,0.001, Figure 5C), and 82% of the 76
primary peaks correctly (p,0.001). Annual minimum temperature
was a slightly better predictor of the type of influenza peak
characterizing a site, correctly classifying 77% and 87% of all
peaks and primary peaks, respectively (p,0.001). Sites character-
ized by cold-dry influenza peaks generally had annual minimum
temperature values less than 21uC when all influenza peaks were
considered, and approximately 18uC when the analysis was
restricted to primary influenza peaks. The temperature and
specific humidity models differentiated between the 6 cold-dry and
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 4 March 2013 | Volume 9 | Issue 3 | e1003194
Figure 2. Influenza peaks and climate by latitude. The mean monthly rank of each climate variable corresponding to the month of peak
influenza for each 10ulatitudinal band. Solar radiation, temperature and specific humidity are lagged by 1 month. The background interval
corresponds to the 95% null distribution. (A) displays the results for both primary and secondary influenza peaks; whereas (B) shows the results for
primary influenza peaks only. Influenza peaks corresponded to months characterized by low ranks of temperature, solar radiation, and specific
humidity in high latitudes. Primary influenza peaks corresponded to months with high ranks of humidity (both relative and specific) and precipitation
in low latitudes.
doi:10.1371/journal.ppat.1003194.g002
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 5 March 2013 | Volume 9 | Issue 3 | e1003194
3 humid-rainy influenza peaks available in the independent FluNet
sites with 100% and 78% accuracy, respectively.
Annual minimum solar radiation was a significant predictor of
the type of influenza peaks, comparable to temperature and
specific humidity, correctly classifying 71% and 79%of all peaks
and primary peaks, respectively (p,0.001). Annual minimum
relative humidity and annual monthly maximum precipitation
were also significant predictors, but the models were significantly
weaker than the other models. It should be noted that 12 sites had
both a cold-dry and humid-rainy influenza peak, assuring that one
peak would be classified incorrectly when all influenza peaks were
considered.
Taken together, this analysis indicates that thresholds in specific
humidity and temperature, and perhaps solar radiation, are
associated with the timing of the influenza season and the
occurrence of influenza activity in the dry-cold or humid-rainy
months of the year. The specific humidity and temperature models
were then used to predict the expected seasonal characteristics of
influenza globally (Figure 5C–D). Both models suggest that
seasonal influenza activity coincides with the humid-rainy season
Table 1. Environmental models for all influenza peaks, by latitudinal interval.
Proportion of peaks accurately predicted by each model
Climatologies Coefficients (SE) AIC All Peaks n = 96
a
High Latit.
n=50
b
Middle Latit.
n=31
c
Low Latit.
n=15
Temperature (Precipitation)
2
20.05 (4.66e-6)
1.10e-5 (7.78e-5)
590 0.61*** 0.82*** 0.39 0.40
Temperature 20.03 (0.01) 611 0.59*** 0.84*** 0.39 0.20
Specific Humidity (Specific
Humidity)
2
20.54 (20.09)
0.02 (3.82e-3)
593 0.56*** 0.78*** 0.32 0.33
(Solar Radiation)
2
22.06e-5 (4.18e-6) 578 0.52*** 0.70*** 0.39 0.20
Relative Humidity 0.03 (0.01) 613 0.41* 0.52*** 0.19 0.47
(Precipitation)
2
7.28e-06 (1.95e-6) 615 0.31 0.22 0.39 0.47
Expected Values i.e. null
distribution (95% CI)
0.25 (0.16 0.34) 0.25 (0.14 0.38) 0.24 (0.09 0.42) 0.25 (0.00 0.53)
*p,0.05,
** p,0.01,
***p,0.001.
a
high latitudes are regions poleward of 25uN/S.
b
middle latitudes are regions between 12.5uN/S and 25uN/S.
c
low latitudes are regions equatorward of 12.5uN/S.
The results of selected logistic regression models, based on Aikake Information Criterion (AIC) and the proportion of peaks accurately predicted by each model using a
jackknife leave-one-out method. These values can be compared against the expected values and corresponding confidence intervals under the null distribution in the
bottom row. The models are in descending order based on the proportion of peaks accurately predicted. Influenza peaks were lagged by 1-month with respect to each
environmental variable with the exception of precipitation.
doi:10.1371/journal.ppat.1003194.t001
Table 2. Environmental models for primary influenza peaks.
Proportion of peaks accurately predicted by each model
Climatologies Coefficients AIC All Peaks n = 76
a
High Latit.
n=47
b
Middle Latit.
n=20
c
Low Latit. n = 9
Temperature Specific
Humidity
20.10 (2.56e-3)
0.10 (1.76e-3)
501 0.70*** 0.87*** 0.30 0.67**
Temperature 20.05 (0.01) 504 0.66*** 0.85*** 0.35 0.33
Specific Humidity (Specific
Humidity)
2
20.55 (0.09)
0.02 (4.18e-3)
493 0.62*** 0.81*** 0.35 0.22
(Solar Radiation)
2
23.50e-5 (5.14e-6) 471 0.59*** 0.72*** 0.40 0.33
Relative Humidity 0.02 (0.01) 521 0.43*** 0.53*** 0.10 0.67**
(Precipitation)
2
5.24e-06 (2.57e-6) 524 0.32 0.21 0.40 0.67**
Expected Values i.e. null
distribution (95% CI)
0.24 (0.13 0.38) 0.24 (0.16 0.36) 0.20 (0.05 0.45) 0.20 (0.00 0.55)
*p,0.05,
**p,0.01,
***p,0.001.
a
high latitudes are regions poleward of 25uN/S.
b
middle latitudes are regions between 12.5uN/S and 25uN/S.
c
low latitudes are regions equatorward of 12.5uN/S.
Same as Table 1 but these are the results for primary influenza peaks only.
doi:10.1371/journal.ppat.1003194.t002
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 6 March 2013 | Volume 9 | Issue 3 | e1003194
in large areas of Central and South America, and Southern Asia;
while predictions were more uncertain in middle latitudes and
there were inconsistencies between the two models for parts of
Central Africa. In particular, the model driven by minimum
temperature predicted the occurrence of humid-rainy influenza
peaks in most of Central Africa, while the model driven by
minimum specific humidity predicted a more restricted zone of
humid-rainy peaks concentrated on the Western coast of this
region. These discrepancies can be explained by a combination of
warm year-round temperatures in this area, with low specific
humidity values in parts of the year.
Discussion
We explored the association between influenza seasonality and
climate in a representative sample of 78 global sites, spanning an
Figure 3. Influenza peaks, specific humidity and precipitation. (A) Estimated U-shaped relationship between the likelihood of an influenza
peak and average monthly specific humidity across all sites, based on logistic regression (Table 1). The left side of the curve is strongly correlated with
the relationship between specific humidity and influenza survival and transmission observed in laboratory studies [17,32,33]. However, the
mechanism that causes the pattern on the right side of the curve is not readily explained. (B) The relationship between average monthly specific
humidity and precipitation across all sites. Influenza peaks clustered in months associated with low specific humidity and high precipitation
conditions. This suggests that precipitation may explain the occurrence of humid-rainy influenza peaks and may be responsible for the right hand
side of the U-shaped curve between specific humidity and influenza (A).
doi:10.1371/journal.ppat.1003194.g003
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 7 March 2013 | Volume 9 | Issue 3 | e1003194
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 8 March 2013 | Volume 9 | Issue 3 | e1003194
absolute latitudinal range between 1uand 60u. Our analyses
revealed two distinct types of climatic conditions associated with
influenza seasons observed globally: ‘‘cold-dry’’ and ‘‘humid-
rainy’’. In general, sites that experienced low levels of specific
humidity and temperature (less than 11–12 g/kg and 18–21uC) for
at least one month during the year were characterized by seasonal
influenza activity during the months with minimal levels of specific
humidity and temperature. In contrast, sites that maintained high
levels of specific humidity and temperature were generally
characterized by influenza epidemics during the most humid
and rainy months of the year. The predictions of our climate-
based models compared favorably to influenza epidemiological
information collected independently of the dataset used for the
model-building exercise.
The bimodal nature of the relationship, in both relative and
absolute terms, between specific humidity and influenza peaks,
and its strong relationship to other climate variables such as
temperature and precipitation, makes specific humidity a useful
gauge of the environmental favorability of influenza activity across
all latitudes (Figure 5). However, although the specific humidity
models were significant predictors of influenza peaks globally, this
was primarily due to their performance in high latitudes. In low
latitudes, precipitation was a stronger predictor of the timing of
influenza activity, with peaks typically occurring in months with
average precipitation greater than 150 mm (Figure 3).
Overall, although precipitation was strongly associated with
influenza peaks in low latitudes, the timing of influenza peaks in
this region was more difficult to predict than in high-latitude sites.
Several sites in this region were not characterized by well-defined
influenza season; rather, influenza activity was present year-round
likely due to the limited seasonal environmental variation that
characterizes much of the region. For example, equatorial sites
such as Iquitos, Peru, and Singapore—where influenza seasonality
is weak [27,28]— experience limited fluctuations in precipitation,
with monthly averages constrained to a narrow range of 150–
300 mm year-round. In contrast, middle and low-latitude sites
such as Fortaleza, Brazil and Yangon, Myanmar —which are
noted for their well-defined influenza seasons [9,11]—are charac-
terized by large amplitude range in average monthly precipitation
from 25 mm in the dry season to over 300 mm and 600 mm in the
rainy season, respectively.
Model performances were particularly poor in a number of
middle latitudes sites. Predicting influenza peaks in these sites may
be complicated by large seasonal swings in climate that
characterize the region, generating both cold-dry and humid-
rainy seasons that are equally favorable for seasonal influenza
activity, such as in Senegal (Figure 4). For these sites other factors
might play a critical role in determining the timing of influenza
activity, including population mixing (i.e., travel) with regions that
do experience well-defined influenza seasons [29,30], or the
seasonal phasing with school cycles [21]. Moreover, the presence
of both cold-dry and humid-rainy seasons could explain the
occurrence of semi-annual influenza epidemics in some of these
middle-latitude sites. For example, Hong Kong has a primary
influenza peak in winter when average monthly specific humidity
and temperature are less than 8 g/kg and 17uC, and a secondary
influenza peak in summer when average monthly precipitation is
near 400 mm.
Temperature was a strong predictor of influenza seasonality in
high latitudes, suggesting that cold temperatures may drive
seasonal epidemics in these regions. However, previous analyses
of laboratory experiments have indicated that specific humidity is
a more parsimonious predictor of virus survival and transmission
than temperature [17]. Furthermore, individuals in temperate
regions spend a vast majority of their time indoors where
temperature is managed and does not correlate well with outdoor
temperatures. Yet, temperature may affect the timing of influenza
epidemics through mechanisms independent of virus survival; for
example, low outdoor temperatures may promote indoor crowd-
ing, thereby increasing person-to-person contact rates [2–4]. It is
also possible that even limited exposure to cold outdoor
temperatures can have long-lasting physiological effects on hosts
that make them more susceptible to infection or affect viral
shedding [16]. Additional experimental and observational work is
needed to disentangle the contribution of specific humidity and
temperature on influenza seasonality; epidemiological information
from Central Africa would be particularly useful in this respect as
our climate-based predictive models disagreed in this region.
The findings that both cold-dry and humid-rainy conditions are
associated with influenza peaks could be used to support the
hypothesis that two distinct mechanisms account for influenza
seasonality in temperate and tropical climates, perhaps due to
changes in the dominant mode of transmission [31]. For example,
specific humidity may drive the timing of influenza epidemics in
high latitudes by increasing virus survival and enabling aerosol
transmission; whereas direct transmission or transmission by
fomites may dominate in low-latitude sites where rainy conditions
favor indoor crowding. Middle latitudes may be a transition zone
where influenza seasons are driven by low specific humidity or
high levels of precipitation depending on local climate. Another
intriguing possibility is that the relationship between specific
humidity and virus survival underlies influenza transmission across
all latitudes. For example, a few experimental studies have
indicated a U-shaped relationship between relative humidity and
influenza virus survival, suggesting a similar relationship for
specific humidity given that experiments were held at constant
temperature [32–34]. Other laboratory studies, however, have
indicated that virus survival and transmission increase monoton-
ically as specific humidity decreases [17,35,36]. Further, the
hypothesis that specific humidity drives influenza transmission
globally is inconsistent with the low predictive power of this
climatic variable in middle and low-latitude sites in our study.
Relative humidity was a strong predictor of influenza peaks in
high and low latitudes, but a poor predictor in middle-latitude. In
high-latitude regions, relative humidity can vary significantly
between indoor and outdoor environments, and it is typically
minimal indoors during the winter when building air is heated.
Our analysis relied on outdoor humidity and hence we cannot rule
out that winter influenza epidemics in high latitudes could be
related to low indoor relative humidity and associated changes to
Figure 4. Influenza seasonal distribution for 9 sites selected from an independent epidemiological dataset and climate model
outputs. (A,C,E,G,I,K,M,O,Q) Box plots indicate the proportion of influenza cases occurring in each month of the year for 9 countries with multiyear
data selected from FluNet. Results of the best-fit climate models for all and primary peaks (Tables 1 and 2) are displayed for comparison. Specific
humidity and temperature were advanced one month to account for the one month lag between influenza peaks and these variables. Although the
models were designed to estimate the timing of peak influenza activity, they also provide estimates of the seasonal distribution of influenza virus
circulation. (B,D,F,H,J,L,N,P,R) The right column displays the monthly precipitation, temperature and specific humidity for each location. Dotted lines
indicate the climatic thresholds for each variable. In general, when temperature or specific humidity drops below their respective thresholds, or
precipitation surpasses its threshold, there is an increase in influenza activity.
doi:10.1371/journal.ppat.1003194.g004
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 9 March 2013 | Volume 9 | Issue 3 | e1003194
Figure 5. Climatic thresholds predictive of influenza seasonal characteristics. (A) density plot showing the specific humidity in absolute
terms (x-axis) and relative terms (y-axis) during influenza peaks across all sites. The plot shows that a vast majority of influenza peaks occurred in
‘‘cold-dry’’ conditions when specific humidity was lower than 8 g/kg and ranks were less than 4, or during ‘‘humid-rainy’’ conditions when specific
humidity was greater than 14 g/kg and ranks were greater than 9. (B) a line plot showing the average annual range of specific humidity (y-axis) for
each location (x-axis). Sites are ordered based on minimum specific humidity. The black dots indicate the specific humidity during the month of the
primary peak and circles indicate specific humidity during secondary peaks. Together, the plots suggest that sites with the lowest annual minimum
specific humidity have influenza peaks when specific humidity is at locally-minimal levels. (C) a map displaying the predictions of a logistic regression
indicating the probability of an influenza peak during the cold-dry season, versus the humid-rainy season, based on annual minimum specific
humidity. The markers indicate the 78 study sites with influenza peaks classified as cold-dry (circles) and humid-rainy (squares). (D) same as (C) but the
model is based on annual minimum temperature.
doi:10.1371/journal.ppat.1003194.g005
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 10 March 2013 | Volume 9 | Issue 3 | e1003194
host physiology, such as reduced mucociliary clearance [20]. In
low latitudes it is possible that relative humidity is confounding
precipitation in our analysis. Disentangling these two factors will
require more highly-resolved epidemiological data from equatorial
regions, and further experimental and observational studies.
Solar radiation was also a significant predictor of influenza
peaks in high latitudes suggesting that it may also underlie
influenza seasonality in these regions, perhaps through variation in
vitamin D intake [18]. However, solar radiation was not as strong
a predictor of influenza peaks as were specific humidity and
temperature. This corroborates recent studies indicating that
specific humidity is a stronger predictor of seasonal influenza
activity than solar radiation and vitamin D variability in the U.S.
[14,37]. Still, the potential seasonal forcing of solar radiation on
influenza transmission warrants further experimental and obser-
vational investigation.
The power of this study was rooted in the large number of
spatially diverse sites used to develop the epidemiological and
climatic databases and associated models. However, the challenge
of describing seasonal influenza activity consistently across a
variety of data sources required a crude epidemiological measure,
such as the average month of peak influenza activity. This measure
of influenza activity has several key drawbacks. Foremost, all
months with the exception of the peak influenza months were
considered equal, whether they had substantial influenza activity
or not. Second, the month of peak influenza activity may not be
contemporaneous with the month in which transmission is under
the most environmentally favorable conditions, since non-
environmental factors such as viral seeding, population suscepti-
bility, and person-to-person contact rates likely play a role in the
timing of influenza epidemics [15,21]. In this respect, it is
reassuring that a 1-month lag maximized the association between
influenza peak and most of the climatic variables, which is broadly
consistent with the time scale of the ascending phase of a local
epidemic. Third, we could not assess putative geographical
variation in the transmission potential or intensity of influenza
epidemics. For example, we may expect locations that have the
most favorable environmental conditions to experience the
greatest influenza annual attack rates and reproduction numbers,
holding all other relevant variables equal. A further limitation
relates to between-year variability in influenza timing and the
limited temporal sampling of our dataset, which may have resulted
in imprecise estimates of the average influenza peak in some sites,
especially sites that had only one year of influenza data. However,
sensitivity analyses limited to multi-year studies revealed similar
relationships between climate predictors and influenza peaks,
confirming the robustness of our results. Finally, we were unable to
check whether between-year fluctuations in climatic variables may
result in departures from average influenza seasonal characteristics
in specific years. This question could be an interesting area for
future research with more temporally refined epidemiological
datasets.
A number of follow-up studies could help refine our
understanding of the small and large-scale processes underlying
influenza seasonality. For example, experimental infections in
humans under controlled temperature and humidity conditions
could determine which environmentally-mediated mechanisms are
most important for human-to-human transmission. However,
there are several ethical and methodological hurdles to overcome
in such studies [38]. Seasonal fluctuation in contact rates could be
monitored by wireless sensor technology, which has recently
proved successful in estimating dynamic contact patterns in
schools and at conferences [39–40]. On a broader spatial scale,
determining regional differences in influenza transmission dynam-
ics and attack rates would be most informative. A recent study has
suggested that the reproduction number of seasonal epidemics was
lower on average in Brazil than in temperate countries, which
could be mediated by environmental factors [41]. Modeling of
long-term influenza time series data could help assess the
transmission impact of seasonal fluctuations in population mixing
in different regions, such as those associated with school cycles [21]
and transportation networks [29,30,42]. For example, epidemio-
logical evidence indicates that influenza circulation was weakly
seasonal in Iceland prior to the 1930s, presumably because of low
connectivity with other populations, and epidemics only became
fully synchronized with those in Europe and the USA following a
dramatic increase in international travel in the 1990s [43]. Hence,
efforts to collate multiyear influenza epidemiological information
retrospectively and prospectively in various regions of the globe,
especially from middle and low latitude regions, will be of
tremendous help to further elucidate the environmental and
population drivers of seasonality.
In conclusion, our study broadens our understanding of the
relationships between seasonal influenza epidemics and environ-
mental factors and provides a synthesis of epidemiological and
climatic characteristics across temperate, subtropical and tropical
regions. We have highlighted the importance of thresholds in
specific humidity, temperature and precipitation that are associ-
ated with the epidemiology (and potentially the modes of
transmission) of influenza. The results of this study could help
improve existing influenza transmission models by providing a
more accurate estimate of the environmental forcing on transmis-
sion processes, particularly in low and middle latitudes [14,44].
Further, our models could be used to predict the seasonal timing of
influenza activity in locations with little or no observational data
on influenza activity, and help target surveillance efforts and
optimize the timing of seasonal vaccine delivery, [45]. More
broadly, we hope that our work will generate interest in testing the
association between climatic patterns and infectious disease across
a wide range of diseases and latitudes, particularly for respiratory
and enteric pathogens that display marked seasonality [2,46]. A
better understanding of the environmental, demographic and
social drivers of infectious disease seasonality is key for improving
transmission models and optimizing interventions [47].
Supporting Information
Text S1 Additional information regarding data and
results. Included is a more detailed description of the influenza
database and the algorithm developed to define each ‘‘peak’’. We
also provide a comparison of CRU and GR climate datasets, and
provide results from additional predictive models of peak influenza
timing.
(DOC)
Acknowledgments
This research was conducted in the context of the MISMS Study, an
ongoing international collaborative effort to understand influenza epide-
miological and evolutionary patterns, led by the Fogarty International
Center, National Institutes of Health (http://www.origem.info/misms/
index.php).
Author Contributions
Conceived and designed the experiments: JDT CV JS WJA AC KBF.
Performed the experiments: JDT. Analyzed the data: JDT CV JS.
Contributed reagents/materials/analysis tools: JDT KBF CV JS CKU.
Wrote the paper: JDT CV JS WJA AC.
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 11 March 2013 | Volume 9 | Issue 3 | e1003194
References
1. Simonsen L (1999) The global impact of influenza on morbidity and mortality.
Vaccine 17:S3–10.
2. Lipsitch M, Viboud C (2009) Influenza seasonality: lifting the fog. Proc Nat
Acad Sci 106:3645–3646.
3. Lofgren E, Fefferman NH, Naumov YN, Gorski J, Naumova EN (2007)
Influenza seasonality: underlying causes and modeling theories. J Virol 81:5429–
5436.
4. Tamerius JD, Nelson MI, Zhou SZ, Viboud C, Miller MA, et al. (2011) Global
influenza seasonality: reconciling patterns across temperate and tropical regions.
Environ Health Persp 119:439–445.
5. Finkelman BS, Viboud C, Koelle K, Ferrari MJ, Bharti N, et al. (2007) Global
patterns in seasonal activity of influenza A/H3N2, A/H1N1, and B from 1997
to 2005: viral coexistence and latitudinal gradients. PLoS ONE 2:e1296.
6. Viboud C, Boe¨lle P, Pakdaman K, Carrat F, Valleron A, et al. (2004) Influenza
epidemics in the United States, France, and Australia, 1972–1997. Emerg Infect
Dis 10:32–39.
7. Moura FEA (2010) Influenza in the tropics. Curr Opin Infect Dis 23:415–420.
8. Rao BL, Banerjee K (1993) Influenza surveillance in Pune, India, 1978–90.
B World Health Organ 71:177–181.
9. Dapat C, Saito R, Kyaw Y, Naito M, Hasegawa G, et al. (2009) Epidemiology of
human influenza A and B viruses in Myanmar from 2005 to 2007. Intervirology
52:310–320.
10. Dosseh A, Ndiaye K, Spiegel A, Sagna M, Mathiot C (2000) Epidemiological
and virological influenza survey in Dakar, Senegal: 1996–1998. Am J Trop Med
Hyg 62:639–643.
11. Moura FE, Perdiga˜ o, Siqueira MM (2009) Seasonality of influenza in the tropics:
a distinct pattern in northeastern Brazil. Am J Trop Med Hyg 81:180–183.
12. Chiu SS, Chan KH, Chen H, Young BW, Lim W, et al. (2009). Virologically
confirmed population-based burden of hospitalization caused by influenza A and
B among children in Hong Kong. Clin Infect Dis 49(7):1016–1021.
doi:10.1086/605570
13. Lee VJ, Yap J, Ong JBS, Chan KP, Lin RTP, et al. (2009). Influenza excess
mortality from 1950–2000 in tropical Singapore. PLoS ONE 4(12):e8096.
doi:10.1371/journal.pone.0008096
14. Shaman J, Pitzer VE, Viboud C, Grenfell BT, Lipsitch M (2010) Absolute
humidity and the seasonal onset of influenza in the continental United States.
PLoS Biol 8:e1000316.
15. Shaman J, Goldstein E, Lipsitch M. (2011). Absolute humidity and pandemic
versus epidemic influenza. Am J Epidemiol 173(2):127–135. doi:10.1093/aje/
kwq347
16. Lowen AC, Mubareka S, Steel J, Palese P (2007) Influenza virus transmission is
dependent on relative humidity and temperature. PLoS Pathog 3:1470–1476.
17. Shaman J, Kohn M (2009) Absolute humidity modulates influenza survival,
transmission, and seasonality. Proc Nat Acad Sci 106:3243–3248.
18. Cannell JJ, Vieth R, Umhau JC, Holick MF, Grant WB, et al. (2006) Epidemic
influenza and vitamin D. Epidemiol Infect 134:1129–1140.
19. Dowell SF (2001) Seasonal variation in host susceptibility and cycles of certain
infectious diseases. Emerg Infect Dis 7:369–374.
20. Salah B, Dinh Xuan AT, Fouilladieu JL, Lockhart A, Regnard J (1988) Nasal
mucociliary transport in healthy subjects is slower when breathing dry air. Eur
Respir J 1:852–855.
21. Cauchemez S, Valleron AJ, Boe¨lle PY, Flahault A, Ferguson NM (2008).
Estimating the impact of school closure on influenza transmission from Sentinel
data. Nature 452:750–754.
22. Azziz-Baumgartner E, Dao C, Nasreen S, Bhuiyan MU, Mah-E-Muneer S, et
al. (2012). Seasona lity, timing, and climate drivers of inf luenza activity
worldwide. J Infect Dis 206: 838–46. doi:10.1093/infdis/jis467
23. Bloom-Feshbach K, Alonso WJ, Charu V, Tamerius J, Simonsen L (2012)
Latitudinal variation in seasonal activity of influenza and respiratory syncytial
virus (RSV): a global comparative review. Plos One. In Press.
24. World Health Organization. (2012) FluNet. Available: http://www.who.int/
flunet. Accessed Oct 12, 2011.
25. New M, Lister D, Hulme M, Makin I (2002) A high-resolution data set of surface
climate over global land areas. Clim Res 21:1–25.
26. Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, et al. (1996) The
NCEP/NCAR 40-year reanalysis project. B Am Meteorol Soc 77:437–472.
27. Laguna-Torres AV, Go´mez J, Ocan˜ a V, Aguilar P, Saldarriaga T, et al. (2009)
Influenza-like illness sentinel surveillance in Peru. PLoS ONE 4:e6118.
28. Chow A, Ma S, Ling AE, Chew SK (2006) Influenza-associated deaths in
tropical Singapore. Emerg Infect Dis 12:114–121.
29. Colizza V, Barrat A, Barthelemy M, Valleron A. (2007). Modeling the
Worldwide Spread of Pandemic Influenza: Baseline Case and Containment
Interventions. PLoS Medicine 4(1):e13. doi:10.1371/journal.pmed.0040013
30. Kenah E, Chao DL, Matrajt L, Halloran ME, Longini IM Jr (2011) The Global
Transmission and Control of Influenza. PLoS ONE 6(5): e19515. doi:10.1371/
journal.pone.0019515
31. Lowen AC, Steel J, Mubareka S, Palese P (2008) High temperature (30uC)
blocks aerosol but not contact transmission of influenza virus. J Virol 78: 5650–
5652.
32. Shechmeister I (1950) Studies on the experimental epidemiology of respiratory
infections: III. certain aspects of the behavior of type A influenza virus as an air-
borne cloud. J Infect Dis 87: 128–132.
33. Schaffer F, Soergel M, Straube D (1976) Survival of airborne influenza virus:
effects of propagating host, relative humidity, and composition of spray fluids.
Arch Virol 51: 263–278.
34. Yang W, Elankumaran S, Marr LC (2012) Relationship between humidity and
influenza A viability in droplets and implications for influenza’s seasonality.
PLoS ONE 7(10): e46789.
35. McDevitt J, Rudnick S, First M, Spengler J (2010) Role of absolute humidity in
the inactivation of influenza viruses on stainless steel surfaces at elevated
temperatures. Appl Environ Microb 76: 3943–3947.
36. Harper G (1961) Airborne micro-organisms: survival tests wit h four viruses.
J Hyg - Cambridge 59:479–486.
37. Shaman J, Jeon CY, Giovannucci E, Lipsitch M (2011) Shortcomings of vitamin
D-based model simulations of seasonal influenza. PLoS ONE 6:e20743.
38. Killingley B, Enstone J, Booy R, Hayward A (2011) Potential role of human
challenge studies for investigation of influenza transmission. Lancet Infect Dis
11: 70142–6
39. Salathe´ M, Kazandjieva M, Lee JW, Levis P, Feldman MW, et al. (2010). A
high-resolution human contact network for infectious disease transmission. Proc
Natl Acad Sci U S A 107:22020–5.
40. Stehle´ J, Voirin N, Barrat A, Cattuto C, Colizza V, et al. (2011).Simulation of an
SEIR infectious disease model on the dynamic contact network of conference
attendees. BMC Med. 9:87.
41. Chowell G, Vibou d C, Simonsen L, Miller M, Alonso W (2010) The
reproduction number of seasonal influenza epidemics in Brazil, 1996–20 06.
Proc R Soc B 277(1689): 1857–1866.
42. Viboud C, Bjørnstad ON, Smith DL, Simonsen L, Miller MA, et al. (2006)
Synchrony, waves, and spatial hierarchies in the spread of influenza. Science
312(5772):447–451.
43. Weinberger DM, Krause TG, Mølbak K, Cliff A, Briem H, et al. (2012).
Influenza epidemics in iceland over 9 decades: changes in timing and synchrony
with the United States and Europe. Am J Epidemiol 176(7): 649–655.
44. Truscott J, Fraser C, Cauchemez S, Meeyai A, Hinsley W, et al. (2012). Essential
epidemiological mechanisms underpinning the transmission dynamics of
seasonal influenza. J R Soc Interface 9(67): 304–312.
45. de Mello WA, de Paiva TM, Ishida MA, Benega MA, Santos MC, et al. (2009)
The dilemma of influenza vaccine recommendations when applied to the
tropics: the Brazilian case examined under alternative scenarios. PLoS ONE
4(4):e5095.
46. Stensballe LG Devasundaram, JK Simoes EA (2003). Respiratory syncytial virus
epidemics: the ups and downs of a seasonal virus. Pediatr Infect Dis J 22:S21–32.
47. Pitzer VE, Viboud C, Lopman BA, Patel MM, Parashar UD, et al. (2011)
Influence of birth rates and transmission rates on the global seasonality of
rotavirus incidence. J R Soc Interface 8(64):1584–1593.
Environmental Predictors of Influenza
PLOS Pathogens | www.plospathogens.org 12 March 2013 | Volume 9 | Issue 3 | e1003194
... Globally, inuenza virus infects about three to ve million individuals annually, leading to [9] approximately 650,000 deaths . The rst inuenza pandemic of the st 21 century broke in Mexico in 2009, as a result of re-assortment between avian, human, and swine inuenza viruses to develop a new [6], [10] strain of Inuenza A (H1N1) . ...
... A study by Lin et al. from Guangdong, China reported two peaks for inuenza infection; one in July caused by seasonal inuenza and [16] another in November by pandemic 2009 strain . The temperate regions reported higher incidences of the human inuenza viral infection during the winter season (December -March), while tropical and subtropical areas displayed a year round circulation and biannual [9], [10] peaks . ...
Article
BACKGROUND & OBJECTIVE: To study region-wise monthly incidence, alteration in seasonal trend and geographical distribution of Swine Flu (H1N1) cases in district of western India for two years. After Ethics Committee approval, a cross-sectionalMATERIAL AND METHOD: study was carried out with analysis of two years' demographic prole, geographical distribution and seasonal variation of real time Reverse Transcription–Polymerase Chain Reaction (RT-PCR) - H1N1 positive cases. Annual change in pattern of positive cases was analyzed by independent sample student's t - test (P < 0.05). Out of 3,524 Seasonal Flu cases investigated by Inuenza A (H1N1) RT-PCR, total 1,600RESULT: Seasonal Flu–positive cases were analyzed. On looking at Age-wise distribution of patients, during rst year, maximum number of Seasonal Flu positive cases was from age group of 1 to 10 years, while during next year, maximum number of cases was from age group of 51 to 60 years. During rst year of epidemic, incidence of 836.12 cases per 1,00,000 population per year was detected with peak during winter as compared to 121.49 cases per 1,00,000 population per year with peak during monsoon in second year of epidemic. DuringINTERPRETATION & CONCLUSION: Epidemic of Seasonal Flu, Swine Flu (H1N1) positive cases got reduced annually, and seasonal pattern of H1N1 positive cases changed with peak during monsoon as compared to conventional winter peak. Overall change in seasonal trend, pediatric age distribution suggests need of in depth analysis of development of herd immunity, and risk factors involved for monsoon peak with alarming precaution by epidemic preparedness.
... Modeling approaches for epidemic forecasting can be broadly categorized into mechanistic (Zhang et al. 2017;Shaman and Karspeck 2012) and statistical (Tizzoni et al. 2012;Osthus et al. 2019;Brooks et al. 2018), the latter closely related to time series analysis (Box et al. 2015;Jha et al. 2015). Past work, especially in the context of influenza, has explored leveraging multiple sources of data, from search engine (Ginsberg et al. 2009;Yuan et al. 2013) to weather data (Shaman, Goldstein, and Lipsitch 2010;Tamerius et al. 2013;Volkova et al. 2017). Deep learning for epidemic forecasting is gaining more research interest lately. ...
Article
How do we forecast an emerging pandemic in real time in a purely data-driven manner? How to leverage rich heterogeneous data based on various signals such as mobility, testing, and/or disease exposure for forecasting? How to handle noisy data and generate uncertainties in the forecast? In this paper, we present DeepCOVID, an operational deep learning framework designed for real-time COVID-19 forecasting. DeepCOVID works well with sparse data and can handle noisy heterogeneous data signals by propagating the uncertainty from the data in a principled manner resulting in meaningful uncertainties in the forecast. The deployed framework also consists of modules for both real-time and retrospective exploratory analysis to enable interpretation of the forecasts. Results from real-time predictions (featured on the CDC website and FiveThirtyEight.com) since April 2020 indicates that our approach is competitive among the methods in the COVID-19 Forecast Hub, especially for short-term predictions.
... Epidemic Forecasting: Several approaches for epidemic forecasting have been proposed including statistical (Tizzoni et al. 2012;Adhikari et al. 2019;Osthus et al. 2019;Brooks et al. 2018), mechanistic (Shaman and Karspeck 2012;Zhang et al. 2017), and ensemble (Reich et al. 2019b) approaches. Several approaches rely on external signals such as environmental conditions and weather reports (Shaman, Goldstein, and Lipsitch 2010;Tamerius et al. 2013;Volkova et al. 2017), social media (Chen et al. 2016;Lee, Agrawal, and Choudhary 2013), search engine data (Ginsberg et al. 2009;Yuan et al. 2013), and a combination of multiple sources (Chakraborty et al. 2014). Recently, there has been increasing interest in deep learning for epidemic forecasting (Adhikari et al. 2019;Wang, Chen, and Marathe 2019;Rodríguez et al. 2020). ...
Article
Forecasting influenza in a timely manner aids health organizations and policymakers in adequate preparation and decision making. However, effective influenza forecasting still remains a challenge despite increasing research interest. It is even more challenging amidst the COVID pandemic, when the influenza-like illness (ILI) counts are affected by various factors such as symptomatic similarities with COVID-19 and shift in healthcare seeking patterns of the general population. Under the current pandemic, historical influenza models carry valuable expertise about the disease dynamics but face difficulties adapting. Therefore, we propose CALI-Net, a neural transfer learning architecture which allows us to 'steer' a historical disease forecasting model to new scenarios where flu and COVID co-exist. Our framework enables this adaptation by automatically learning when it should emphasize learning from COVID-related signals and when it should learn from the historical model. Thus, we exploit representations learned from historical ILI data as well as the limited COVID-related signals. Our experiments demonstrate that our approach is successful in adapting a historical forecasting model to the current pandemic. In addition, we show that success in our primary goal, adaptation, does not sacrifice overall performance as compared with state-of-the-art influenza forecasting approaches.
... The transmission of respiratory viruses, such as severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and influenza, is much more common indoors than outdoors [11]. According to Zhao et al., temperature and RH are the best predictors of predicting influenza transmission [12], and low temperature and low RH will promote influenza transmission [13,14]. With the use of heating systems, on one side, it could rise the indoor temperature, and reduce the potential influenza and SARS-CoV-2 transmission, but on the other hand, it also reduces the indoor RH, and then enhances the potential indoor transmission of respiratory infections. ...
Article
Full-text available
Since people spend much more time indoors, indoor conditions may be better indicators of personal exposure than outdoor. Several studies have explored the associations between indoor and outdoor temperature, humidity in USA or in tropical regions. Few studies examined these associations in China. In this study, we collected the daily indoor temperature and relative humidity (RH) in 1577 household in four megacities across temperate and subtropical regions in China, from March 2017 to July 2018. We use Pearson correlation analysis to examine their correlations with outdoor. We found that the correlation between indoor and outdoor RH was stronger in Northern China (r=0.76 V.S. r=0.58), while the correlation between indoor and outdoor temperature was stronger in Southern China (r=0.92 V.S. r=0.80). In the winter-spring months, there was no statistically significant linear relationship between indoor and outdoor temperature. However, the correlation between indoor and outdoor RH in high urbanized regions in winter-spring was stronger than that in low urbanized regions (r=0.85 V.S. r=0.70). This may be due to the comment use of air conditioning and ground heating in the high urbanized regions. This study provides insight for assessing the indoor-outdoor environment correlations in China.
... 3 Even though seasonal influenza epidemics are most prevalent in the winter months in temperate zones, they can occur any time of year in tropical regions. 2,4 The annual epidemic is predicted to affect up to 5 million cases of severe illness, with 650,000 deaths worldwide. 5 Patients with underlying chronic conditions (such as cardiovascular disease, diabetes mellitus, or asthma), extreme age, pregnant females, and children less than five years old. ...
Article
Full-text available
Introduction: Influenza vaccination is the primary control measure for severe complications caused by influenza viruses. Moreover, in the face of the COVID-19 pandemic, Saudi Arabia recommends vaccinating people at risk against influenza to minimise co-infection risk with SARS-CoV2. Therefore, this study aims to assess the Saudi population's knowledge, attitude, and practice toward influenza vaccination during the COVID-19 pandemic. Moreover, we evaluate the impact of the COVID-19 pandemic on seasonal influenza vaccination. Methods: This cross-sectional study was conducted using an online survey in Saudi Arabia between July to August 2021. Participants were invited to complete the questionnaire through a survey link sent to social media platforms. Results: A total of 2410 participants were included in this study. Our data demonstrate a lack of practice, attitude, and knowledge, especially on the influenza virus's symptoms, viral transmission, and vaccine efficacy. Moreover, this study showed that the COVID-19 pandemic significantly impacted seasonal influenza vaccination in the Saudi population by 1.5-times compared to the previous years. Conclusion: COVID-19 pandemic has increased the hesitancy of Saudi participants in influenza vaccination due to the lack of knowledge. As the pandemic of COVID-19 is fading, awareness campaigns are needed to encourage the public about the importance of receiving the influenza vaccine, especially for those at high risk each year.
... Seasonality is a hallmark of influenza epidemiology in the inter-pandemic period, a complex phenomenon shaped by the interplay of population contact patterns, virus survival and host immunity (Tamerius et al., 2011). Environmental and climatic influences have been shown to play a role in influenza seasonality but do not fully explain spatio-temporal variability in the occurrence of seasonal outbreaks (Shaman et al. 2010;Tamerius et al., 2013). During the interpandemic period, influenza seasons are well-synchronized in temperate climate zones of Northern and Southern Hemispheres and organized around their respective winters (Wenger et al., 2010;Lam et al., 2019;Morris et al., 2018). ...
Article
Seasonal influenza causes vast public health and economic impact globally. The prevention and control of the annual epidemics remain a challenge due to the antigenic evolution of the viruses. Here, we presented a novel modeling framework based on changes in amino acid sequences and relevant epidemiological data to retrospectively investigate the competitive evolution and transmission of H1N1 and H3N2 influenza viruses in the United States during October 2002 and April 2019. To do so, we estimated the time-varying disease transmission rate from the reported influenza cases and the time-varying antigenic change rate of the viruses from the changes in amino acid sequences. By incorporating the time-varying antigenic change rate into the transmission models, we found that the models could capture the evolutionary transmission dynamics of influenza viruses in the United States. Our modeling results also showed that the antigenic change of the virus plays an essential role in seasonal influenza dynamics.
Article
Understanding the role of meteorological factors in the transmission dynamics of respiratory infectious diseases remains challenging. Our study was to comprehensively investigate the nonlinear effects of environmental factors on influenza transmission, based on multi-region surveillance data from mainland China. An approach related to time-varying reproduction number (Rt) was proposed, which extracts the environment-related components from Rt to estimate the relationship between environmental factors and influenza transmission based on a mixed-effects regression model. Nonlinear relationships for absolute humidity (the lowest transmission was observed at absolute humidity of 12 g/m³) and mean temperature (the lowest transmission was observed at the mean temperature of 18 °C) with influenza transmission were observed. Influenza transmission holds almost constant with the average precipitation below 1 mm or sunshine hour below 9 h/day, but increases for the precipitation and decreases for the sunshine hour afterward. The environmental dependence varies across subtypes: A(H3N2) maintains relatively higher transmission in high temperature and humidity conditions, compared with other influenza subtypes. Overall, the subtypes specified environmental dependence of influenza transmission could explain 23.1 %, 29.2 % and 27.1 % of the variations for A(H1N1)pdm09, A(H3N2) and B-lineage in China. The projected seasonal transmission rates based on our approach could be used as a valuable seasonal proxy to model the influenza dynamics under various meteorological spaces. Finally, our approach is also applicable to obtain novel insights into the impact of environmental factors on other respiratory infectious diseases.
Chapter
Full-text available
The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which causes COVID-19, emerged in late 2019, halfway through the preparation of the IPCC WGII Sixth Assessment Report. This Cross-Chapter Box assesses how the massive shock of the pandemic and response measures interact with climate-related impacts and risks as well as its significant implications for risk management and climate resilient development.
Article
The transmission of most respiratory pathogens, including SARS-CoV-2, occurs via virus-containing respiratory droplets, and thus, factors that affect virus viability in droplet residues on surfaces are of critical medical and public health importance. Relative humidity (RH) is known to play a role in virus survival, with a U-shaped relationship between RH and virus viability. The mechanisms affecting virus viability in droplet residues, however, are unclear. This study examines the structure and evaporation dynamics of virus-containing saliva droplets on fomites and their impact on virus viability using four model viruses: vesicular stomatitis virus, herpes simplex virus 1, Newcastle disease virus, and coronavirus HCoV-OC43. The results support the hypothesis that the direct contact of antiviral proteins and virions within the "coffee ring" region of the droplet residue gives rise to the observed U-shaped relationship between virus viability and RH. Viruses survive much better at low and high RH, and their viability is substantially reduced at intermediate RH. A phenomenological theory explaining this phenomenon and a quantitative model analyzing and correlating the experimentally measured virus survivability are developed on the basis of the observations. The mechanisms by which RH affects virus viability are explored. At intermediate RH, antiviral proteins have optimal influence on virions because of their largest contact time and overlap area, which leads to the lowest level of virus activity.
Article
Full-text available
Background: There is limited information on influenza and respiratory syncytial virus (RSV) seasonal patterns in tropical areas, although there is renewed interest in understanding the seasonal drivers of respiratory viruses. Methods: We review geographic variations in seasonality of laboratory-confirmed influenza and RSV epidemics in 137 global locations based on literature review and electronic sources. We assessed peak timing and epidemic duration and explored their association with geography and study settings. We fitted time series model to weekly national data available from the WHO influenza surveillance system (FluNet) to further characterize seasonal parameters. Results: Influenza and RSV activity consistently peaked during winter months in temperate locales, while there was greater diversity in the tropics. Several temperate locations experienced semi-annual influenza activity with peaks occurring in winter and summer. Semi-annual activity was relatively common in tropical areas of Southeast Asia for both viruses. Biennial cycles of RSV activity were identified in Northern Europe. Both viruses exhibited weak latitudinal gradients in the timing of epidemics by hemisphere, with peak timing occurring later in the calendar year with increasing latitude (P<0.03). Time series model applied to influenza data from 85 countries confirmed the presence of latitudinal gradients in timing, duration, seasonal amplitude, and between-year variability of epidemics. Overall, 80% of tropical locations experienced distinct RSV seasons lasting 6 months or less, while the percentage was 50% for influenza. Conclusion: Our review combining literature and electronic data sources suggests that a large fraction of tropical locations experience focused seasons of respiratory virus activity in individual years. Information on seasonal patterns remains limited in large undersampled regions, included Africa and Central America. Future studies should attempt to link the observed latitudinal gradients in seasonality of viral epidemics with climatic and population factors, and explore regional differences in disease transmission dynamics and attack rates.
Article
Full-text available
Humidity has been associated with influenza's seasonality, but the mechanisms underlying the relationship remain unclear. There is no consistent explanation for influenza's transmission patterns that applies to both temperate and tropical regions. This study aimed to determine the relationship between ambient humidity and viability of the influenza A virus (IAV) during transmission between hosts and to explain the mechanisms underlying it. We measured the viability of IAV in droplets consisting of various model media, chosen to isolate effects of salts and proteins found in respiratory fluid, and in human mucus, at relative humidities (RH) ranging from 17% to 100%. In all media and mucus, viability was highest when RH was either close to 100% or below ∼50%. When RH decreased from 84% to 50%, the relationship between viability and RH depended on droplet composition: viability decreased in saline solutions, did not change significantly in solutions supplemented with proteins, and increased dramatically in mucus. Additionally, viral decay increased linearly with salt concentration in saline solutions but not when they were supplemented with proteins. There appear to be three regimes of IAV viability in droplets, defined by humidity: physiological conditions (∼100% RH) with high viability, concentrated conditions (50% to near 100% RH) with lower viability depending on the composition of media, and dry conditions (<50% RH) with high viability. This paradigm could help resolve conflicting findings in the literature on the relationship between IAV viability in aerosols and humidity, and results in human mucus could help explain influenza's seasonality in different regions.
Article
Full-text available
Influenza epidemics exhibit a strongly seasonal pattern, with winter peaks that occur with similar timing across temperate areas of the Northern Hemisphere. This synchrony could be influenced by population movements, environmental factors, host immunity, and viral characteristics. The historical isolation of Iceland and subsequent increase in international contacts make it an ideal setting to study epidemic timing. The authors evaluated changes in the timing and regional synchrony of influenza epidemics using mortality and morbidity data from Iceland, North America, and Europe during the period from 1915 to 2007. Cross-correlations and wavelet analyses highlighted 2 major changes in influenza epidemic patterns in Iceland: first was a shift from nonseasonal epidemics prior to the 1930s to a regular winter-seasonal pattern, and second was a change in the early 1990s when a 1-month lag between Iceland and the United States and Europe was no longer detectable with monthly data. There was a moderate association between increased synchrony and the number of foreign visitors to Iceland, providing a plausible explanation for the second shift in epidemic timing. This suggests that transportation might have a minor effect on epidemic timing, but efforts to restrict air travel during influenza epidemics would likely have a limited impact, even for island populations.
Article
Full-text available
Background: Although influenza is a vaccine-preventable disease that annually causes substantial disease burden, data on virus activity in tropical countries are limited. We analyzed publicly available influenza data to better understand the global circulation of influenza viruses. Method: We reviewed open-source, laboratory-confirmed influenza surveillance data. For each country, we abstracted data on the percentage of samples testing positive for influenza each epidemiologic week from the annual number of samples testing positive for influenza. The start of influenza season was defined as the first week when the proportion of samples that tested positive remained above the annual mean. We assessed the relationship between percentage of samples testing positive and mean monthly temperature with use of regression models. Findings: We identified data on laboratory-confirmed influenza virus infection from 85 countries. More than one influenza epidemic period per year was more common in tropical countries (41%) than in temperate countries (15%). Year-round activity (ie, influenza virus identified each week having ≥ 10 specimens submitted) occurred in 3 (7%) of 43 temperate, 1 (17%) of 6 subtropical, and 11 (37%) of 30 tropical countries with available data (P = .006). Percentage positivity was associated with low temperature (P = .001). Interpretation: Annual influenza epidemics occur in consistent temporal patterns depending on climate.
Article
Full-text available
The NCEP and NCAR are cooperating in a project (denoted "reanalysis") to produce a 40-year record of global analyses of atmospheric fields in support of the needs of the research and climate monitoring communities. This effort involves the recovery of land surface, ship, rawinsonde, pibal, aircraft, satellite, and other data; quality controlling and assimilating these data with a data assimilation system that is kept unchanged over the reanalysis period 1957-96. This eliminates perceived climate jumps associated with changes in the data assimilation system. The NCEP/NCAR 40-yr reanalysis uses a frozen state-of-the-art global data assimilation system and a database as complete as possible. The data assimilation and the model used are identical to the global system implemented opera- tionally at the NCEP on 11 January 1995, except that the horizontal resolution is T62 (about 210 km). The database has been enhanced with many sources of observations not available in real time for operations, provided by different coun- tries and organizations. The system has been designed with advanced quality control and monitoring components, and can produce 1 mon of reanalysis per day on a Cray YMPI8 supercomputer. Different types of output archives are being created to satisfy different user needs, including a "quick look" CD-ROM (one per year) with six tropospheric and stratospheric fields available twice daily, as well as surface, top-of-the-atmosphere, and isentropic fields. Reanalysis information and selected output is also available on-line via the Internet (http//:nic.fb4.noaa.gov:8000). A special CD- ROM, containing 13 years of selected observed, daily, monthly, and climatological data from the NCEP/NCAR Re- analysis, is included with this issue. Output variables are classified into four classes, depending on the degree to which they are influenced by the observations andlor the model. For example, "C" variables (such as precipitation and surface fluxes) are completely determined by the model during the data assimilation and should be used with caution. Never- theless, a comparison of these variables with observations and with several climatologies shows that they generally contain considerable useful information. Eight-day forecasts, produced every 5 days, should be useful for predictabil- ity studies and for monitoring the quality of the observing systems. The 40 years of reanalysis (1957-96) should be completed in early 1997. A continuation into the future through an identical Climate Data Assimilation System will allow researchers to reliably compare recent anomalies with those in earlier decades. Since changes in the observing systems will inevitably produce perceived changes in the climate, par- allel reanalyses (at least 1 year long) will be generated for the periods immediately after the introduction of new ob- serving systems, such as new types of satellite data. NCEP plans currently call for an updated reanalysis using a state-of-the-art system every five years or so. The suc- cessive reanalyses will be greatly facilitated by the generation of the comprehensive database in the present reanalysis.
Article
Full-text available
We describe the construction of a 10′ latitude/longitude data set of mean monthly surface climate over global land areas, excluding Antarctica. The climatology includes 8 climate elements-precipitation, wet-day frequency, temperature, temperature, diurnal temperature range, relative humidity, sunshine duration, ground frost frequency and windspeed-and was interpolated from a data set of station means for the period centred on 1961 to 1990. Precipitation was first defined in terms of the parameters of the Gamma distribution, enabling the calculation of monthly precipitation at any given return period. The data are compared to an earlier data set at 0.5° latitude/longitude resolution and show added value over most regions. The data will have many applications in applied climatology, biogeochemical modelling, hydrology and agricultural meteorology and are available through the International Water Management Institute World Water and Climate Atlas (http://www.iwmi.org) and the Climatic Research Unit (http://www.cru.uea.ac.uk).