ARTIGO

ARTICLE
Variability in malaria cases and the
association with rainfall and rivers water
levels in Amazonas State, Brazil

Variabilidade dos casos de malária e sua relação

com a precipitação e nível d’água dos rios no
Estado do Amazonas, Brasil

Variabilidad de los casos de malaria y su relación

con las precipitaciones y nivel del agua de los
ríos en el estado del Amazonas, Brasil Bruna Wolfarth-Couto 1
Rosimeire Araújo da Silva 1
Naziano Filizola 2

doi: 10.1590/0102-311X00020218

Abstract Correspondence
B. Wolfarth-Couto
Instituto Nacional de Pesquisas da Amazônia.
Understanding the relations between rainfall and river water levels and ma- Av. André Araújo 2936, Manaus, AM 69067-375, Brasil.
laria cases can provide important clues on modulation of the disease in the brunaprojetoslba@gmail.com
context of local climatic variability. In order to demonstrate how these re-
lations can vary in the same endemic space, a coherence and wavelet phase 1 Instituto Nacional de Pesquisas da Amazônia, Manaus, Brasil.
2 Universidade Federal do Amazonas, Manaus, Brasil.
analysis was performed between environmental and epidemiological variables
from 2003 to 2010 for 8 municipalities (counties) in the state of Amazonas,
Brazil (Barcelos, Borba, Canutama, Carauari, Coari, Eirunepé, Humaitá,
and São Gabriel da Cachoeira). The results suggest significant coherences,
mainly on the scale of annual variability, but scales of less than 1 year and of
2 years were also found. The analyses show that malaria cases display a peak
at approximately 1 and a half months before or after peak rainfall and on
average 1-4 months after peak river water levels in most of the municipali-
ties studied. Each environmental variable displayed distinct local behavior in
time and in space, suggesting that other local variables (e.g. topography) may
control environmental conditions, favoring different patterns in each munici-
pality. However, when the analyses were performed jointly it was possible to
show a non-random order in these relations. Although environmental and
climatic factors indicate a certain influence on malaria dynamics, surveil-
lance, prevention, and control issues should not be overlooked, meaning that
government public health interventions can mask possible relations with local
hydrological and climatic conditions.

Malaria; Atmospheric Precipitation; Hydrology

This article is published in Open Access under the Creative Commons

Attribution license, which allows use, distribution, and reproduction in
any medium, without restrictions, as long as the original work is correctly
cited. Cad. Saúde Pública 2019; 35(2):e00020218
2 Wolfarth-Couto B et al.

Introduction

Malaria, a serious infectious disease, places a huge burden on the population’s health and social and
economic development. In Brazil, especially in the Legal Amazon, the disease is considered a serious
public health problem, with widespread incidence and debilitating effects due mainly to the favorable
environmental conditions for maintenance of the disease 1,2.
Although environmental and social conditions are important for malaria’s endemic levels, factors
such as health services access and quality can affect the disease dynamics 3,4. The interaction between
these factors favors the variations in case reporting, and many such factors are associated with weak-
nesses in epidemiological surveillance activities, responsible for delays in diagnosis and treatment of
the disease and conditions of population vulnerability 5,6,7.
Many factors (e.g. climatic, ecological, and environmental) can be responsible for the vector’s
seasonal characteristics 8. Research has constantly suggested the influence of climatic factors on the
occurrence of vector-borne diseases 9,10,11. Environmental variables such as temperature, humidity, and
land use and vegetation patterns affect the life cycle of various diseases, especially vector-borne ones 12.
Seasonality patterns in the malaria vector’s presence are closely related to the annual rainfall cycle
and meteorological and hydrological variations 13,14. The annual variability in rainfall contributes
to altering the vector density, in addition to providing an aquatic medium for the mosquitos’ life
cycle, increased humidity, and thus vector longevity 8,15. Although rainfall plays an important role
in malaria, its effect and intensity can vary with the circumstances in certain geographic regions 8.
Another relevant element is understanding the hydrological patterns in relation to malaria cases.
Seasonal variations in hydrological levels contribute to the formation of potential breeding sites, with
a significant impact on malaria fluctuation and incidence 14,16.
Since environmental and climatic characteristics ensure favorable environments for this endem-
ic’s perennial transmission 2, understanding the relations between rainfall and river water level and
their effects on malaria is important for understanding the heterogeneous epidemiological profile and
differences in variability in the Amazon Region.
Even in a region where malaria is considered endemic, the transmission dynamics can vary
depending on the interaction with environmental sociocultural, economic, and political factors 17.
According to Confalonieri 18, the environmental and social characteristics of the Brazilian Amazon
are relevant to the determination of epidemiological patterns. The region’s geographic and ecological
characteristics also substantially determine potential habitats for the vector’s reproduction.
Since malaria displays complex interaction between the parasite, vector, human population, and
environment, the disease also shows a complex spatial and temporal distribution. In addition to the
influence of public policies, various studies point to non-homogeneity as a function of different forms
of land occupation and distinct epidemiological conditions due to landscape characteristics and cli-
matic conditions 19,20,21.
Studies aimed at shedding light on the disease dynamics, with differences at the local level, and
identifying spatial differences based on the river basin’s hydrological variability and local rainfall
conditions provide relevant backing for the implementation of prevention and control strategies
based on distinct malaria patterns. The current study thus aimed to aimed to analyze the statistical
covariance in local rainfall and river water levels and malaria cases in order to demonstrate how these
relations vary in the same endemic space.

Methodology

Study area

The state of Amazonas is located in the North of Brazil, with an area of 1,559,161.682km2. The state’s
climate is humid equatorial, with high temperatures and high mean rainfall. The region basically has
two well-defined seasons, from November to March (rainy season) and from May to September (dry
season), with the high river flooding season from May to August and the low river level season from
September to October 22.

Cad. Saúde Pública 2019; 35(2):e00020218

 MALARIA CASES AND ITS RELATIONSHIP WITH ENVIRONMENTAL FACTORS 3

Data

Data on total malaria cases and rainfall and water level anomalies for each municipality included
the historical series from 2003 to 2010 on a monthly scale. Since the data from the Global Historical
Climatology Network (GHCN) at the time of the data collection provided a historical series up to the
year 2010, we opted to standardize the size of the historical data series based on the availability of data
on total malaria cases from the Epidemiological Information System on Malaria (SIVEP-Malaria),
from 2003 to 2010, contained in the GHCN database. The water level data were obtained for this
same period.
The municipalities were selected randomly among those with complete hydrological data, pro-
viding a representative sample of the state of Amazonas, as well as municipalities located in different
river basins. The municipalities analyzed in the study were Barcelos, Borba, Canutama, Carauari,
Coari, Eirunepé, Humaitá and São Gabriel da Cachoeira (Figure 1).

• Malaria cases

Data on total malaria cases were obtained by processing crude data in dbf files, available in SIVEP-
Malaria, using only total new and autochthonous cases.

• Water levels

Water level data (hydrometric station measurements) were obtained from the database of the Brazil-
ian National Water Agency (ANA) and SO HYBAM (Observation Service “Geodynamical, Hydro-
logical and Biogeochemical Control of Erosion/Alteration and Material Transport in the Amazon,

Figure 1

Location of the study area in the state of Amazonas, Brazil. The eight municipalities analyzed were São Gabriel da Cachoeira, Barcelos, Eirunepé,
Carauari, Coari, Canutama, Humaitá, and Borba.

Cad. Saúde Pública 2019; 35(2):e00020218

4 Wolfarth-Couto B et al.

Orinoco and Congo Basins”). When choosing the target municipalities, the hydrometric stations were
selected on the basis of criteria including data comprehensiveness and consistency.

• Rainfall

Monthly rainfall data were used for global continental areas from 1948-2010, with a horizontal
resolution of 0.5º latitude and 0.5º longitude referring to latitudinal and longitudinal points in the
municipality, available through GHCN in the 2B and CAMS version. First, we calculated mean
rainfall and total mean for the series, and next, for the years 2003-2010, we calculated the monthly
rainfall anomalies. The rainfall anomalies allowed assessing the degree of rainfall variability and its
relationship to extreme events.

Data analysis

The study used the procedures and calculations proposed by Torrence & Compo 23. Coherence and
wavelet phase analysis allows determining the dominant modes of variability between the variables
and how these modes vary over time 24, i.e., breaking down and describing the function f(t) in the
frequency domain in order to analyze this function on different frequency and time scales. This tech-
nique is important for investigating non-stationary phenomena.
Coherence values can be viewed based on the chromatic scale from blue to red. Coherence values
of one (red) represent a strong relationship between the variables. Coherences from yellow to blue
represent weaker relations and without statistical significance. Statistically significant coherences
are demarcated by a black line, where the level of significance using the Monte Carlo method is 95%
confidence. Low and statistically non-significant coherence values indicate that the variables were
independent in the years. However, significant and high coherence values suggest that the series
presents a degree of interrelationship with variations on the same frequency.
Phase or lagged analysis between the series is characterized and illustrated by the vectors’ slope
angle. Horizontal arrows pointing to the right (0º) result in series in the same phase (joint relation-
ship); arrows pointing left (180º) reflect series in opposite phases (inverse relationship). Arrows point-
ing down (-45º, -90º, or -135º) suggest that the first series analyzed is lagged, occurring before the
second series; arrows pointing up (45º, 90º, or 135º) indicate that the first series is lagged, occurring
after the second phase.
For this study’s analyses and to understand the time lags (how much one variable antecedes or
precedes the other), the independent variables were rainfall and water level, and the dependent vari-
able was malaria cases. Calculation of the lag is based on the vector’s slope angle.
The abscissa represents the annual period, subdivided into fractions of a year, i.e.: levels 0.25, 0.5,
1.0, and 2.0, which appear standardized on the axes and correspond to periods of 3 months, 6 months,
1 year, and 2 years, respectively. The results for this type of analysis frequently vary in the intervals
at these levels.

Results

The results of coherence analysis generally indicate relations on an annual scale of variability (one
year), but smaller scales of 0.25-0.9 years (corresponding to 3-9 months) and on a biennial scale (two
years) were also observed. In relation to rainfall and malaria cases, the results vary in phase coherences
(relations between variables occurring at the same time) and lagged coherences (variables occurring at
different moments). For water level and malaria, the coherences were mainly lagged, with variability
of 1-4 months, according to the municipalities.
The variables rainfall and malaria cases in the municiplaty of São Gabriel da Cachoeria (Figure
2) feature phases coherences in the years 2004-2006 on a scale of variability of 0.6-0.8 years (cor-
responding to eight to 10 months). Lagged coherences on a scale of variability of 0.25-0.33 years
(3-4 months) for the year 2004 and 0.41-0.66 years (5-8 months) for 2006 and 2007 indicate that the
peaks in malaria cases anteceded the peaks in rainfall by 20-30 days. In the year 2007, on the scale of

Cad. Saúde Pública 2019; 35(2):e00020218

 MALARIA CASES AND ITS RELATIONSHIP WITH ENVIRONMENTAL FACTORS 5

Figure 2

Coherence (hatching from red to blue) and difference in wavelet phase (arrows) in the series “rainfall-malaria cases” and “water level-malaria cases” for
the municipalities of São Gabriel da Cachoeira and Barcelos, Amazonas State, Brasil, 2003-2010.

0.08-0.41 years (1-5 months), lagged coherences suggest that the peak rainfall anteceded the peak in
malaria cases by 10-20 days. For the variables water level and malaria cases (Figure 2), no coherences
were observed on the annual scale (one year), except for late 2009, while on the biennial scale (two
years) significant coherences were observed for the year 2008. Minor coherences were observed at
smaller scales.
For the municipality of Barcelos, the variables rainfall and malaria cases showed quite important
coherences (Figure 2). On the annual scale, starting in the year 2007 until early 2009, lagged coher-
ences occurred in which rainfall anteceded malaria cases by approximately a month and a half. On
scales of variability of 0.08-0.83 years (1-10 months) and on a biennial scale, phase coherences were
observed in which rainfall and malaria cases displayed strong interdependence. As for the variables
water level and malaria cases (Figure 2) on the annual scale, in the first half of the series (the same
period in which no coherence was seen with rainfall), the relationship occurred in the opposite phase
to water level, but without statistical significance. Significant coherence was only observed in 2009 on
the scale of variability of 0.08-0.25 years (1-3 months), and in the opposite phase.

Cad. Saúde Pública 2019; 35(2):e00020218

6 Wolfarth-Couto B et al.

For the municipality of Eirunepé, analyzing rainfall and malaria cases (Figure 3), lagged coherenc-
es were observed on the annual scale for the years 2005 and 2006, in which the rainfall peaks occurred
a month and a half after the peaks in malaria cases. On the biennial scale, the variables showed phase
coherences between 2006 and 2008. On the annual scale, phase coherences were observed between
2007 and 2009. In relation to the variables water level and malaria cases (Figure 3) on the annual scale,
lagged coherences were observed in which the peak water levels anteceded the malaria peaks by 3-4
months. On the scale of variability of 0.08-0.41 years, there were opposite phase coherences between
the years 2004 and 2006.
For the municipality of Carauari, a strong relationship was observed for all the years in the series.
On the annual scale, lagged coherences were observed between 2007 and 2009 in which the peaks in
malaria cases anteceded the rainfall peaks by a month and a half (Figure 3). On the scale of 0.25 to 0.33
years for 2009, lagged coherence was seen in which the peak in malaria anteceded the peak rainfall by
10 to 15 days. In relation to water level and malaria cases (Figure 3) on the annual scale, relations of
moderate to strong dependence were observed for practically all the years with lagged coherences, in
which the water level series anteceded by 45 days in the initial years of the series (2004-2006) and by
three months for the rest of the years (2008-2010).

Figure 3

Coherence (hatching from red to blue) and difference in wavelet phase (arrows) in the series “rainfall-malaria cases” and “water level-malaria cases” for
the municipalities of Eirunepé and Carauari, Amazonas State, Brasil, 2003-2010.

Cad. Saúde Pública 2019; 35(2):e00020218

 MALARIA CASES AND ITS RELATIONSHIP WITH ENVIRONMENTAL FACTORS 7

For the municipality of Coari, phase coherences were observed between rainfall and malaria cases
(Figure 4). The phase coherences generally occurred on an annual scale, but lagged coherences were
observed between the years 2005 and 2007. For 2004 and 2008, lagged coherences were shown on the
scale of variability of 0.33-0.5 years (4-6 months), indicating that the peaks in rainfall anteceded the
peaks in malaria cases by 20-45 days. As for water level and malaria cases, the coherences were much
stronger. On an annual scale, lagged coherences were seen in which in the peak water levels anteceded
the peaks in malaria cases by three months (Figure 4). On the scale of variability of 0.33-0.66 years
(4-8 months), in the years 2004-2008, lagged coherences were observed in which the peak water levels
anteceded the peaks in malaria cases by 1-2 months. On a scale of variability of 0.08-0.25 years, phase
coherences were observed for the years 2007 and 2008.
The municipality of Canutama displayed near-phase coherences between rainfall and malaria cas-
es from 2006 to 2009 on the annual scale (Figure 4). Although there were phase coherences between
rainfall and malaria cases, inverse relations (opposite phase) were seen between water levels and
malaria cases in the same period during which the relationship with rainfall was significant (Figure 4).

Figure 4

Coherence (hatching from red to blue) and difference in wavelet phase (arrows) in the series “rainfall-malaria cases” and “water level-malaria cases” for
the municipalities of Coari and Canutama, Amazonas State, Brasil, 2003-2010.

Cad. Saúde Pública 2019; 35(2):e00020218

8 Wolfarth-Couto B et al.

For the municipality of Humaitá, the relationship between rainfall and malaria cases was only
observed on the scales of variability of 0.08-0.5 years (1-6 months) and in phase, from 2003 to 2007
(Figure 5). For the year 2009, on the scale of variability of 0.08-0.16 years (1-2 months), lagged coher-
ence was seen in which the peak in malaria cases anteceded the peak rainfall by seven to 15 days. As
for the variables water level and malaria cases on an annual scale, lagged coherences were seen in
practically the entire series, where peak water levels anteceded the peak in malaria cases by a month
and a half (Figure 5).
In the municipality of Borba, on the annual scale, lagged coherences were seen, where the peaks in
malaria cases anteceded the peak rainfall by a month and a half in practically the entire series (Figure
5). As for water level and malaria cases on the annual scale, there were significantly strong lagged
coherences in which the peak water levels anteceded the peaks in malaria cases by up to three months
in the entire series (Figure 5).

Figure 5

Coherence (hatching from red to blue) and difference in wavelet phase (arrows) in the series “rainfall-malaria cases” and “water level-malaria cases” for
the municipalities of Humaitá and Borba, Amazonas State, Brasil, 2003-2010.

Cad. Saúde Pública 2019; 35(2):e00020218

 MALARIA CASES AND ITS RELATIONSHIP WITH ENVIRONMENTAL FACTORS 9

Discussion

In Amazonia, rainfall variability and its influence on the hydrological regime are frequently associated
with the dynamics of vector development and malaria transmission. Although these associations are
closely interrelated, factors pertaining to climatic and environmental conditions and epidemiological
surveillance activities establish distinct spatial scenarios for malaria.
The study showed that the relationship between rainfall and malaria cases resulted in phase
coherences, in which the frequency of peaks in the variables occurred at the same time, and lagged
coherences, in which cases of the disease occurred on average one and a half months before or after
the peaks in rainfall. According to Gurgel 6, peaks in malaria cases that antecede or precede the peaks
in rainfall characterize peaks in the disease at the start of the rainy season or in the dry season. This
variation is linked to the existence of ecological niches that allow the vector’s development and repro-
duction in these periods.
In many parts of Amazonia, peaks in malaria cases occur mainly during the dry season, following
the peak rains 6,25,26. In the post-rainy periods, high transmission rates are often reported, when the
environment becomes more favorable for the vector’s reproduction 27,28,29.
The municipalities that showed malaria peaks during the dry season (second half of the year) were
Coari and Canutama (mean yearly peak in August), São Gabriel da Cachoeira (mean yearly peak in
October), and Barcelos (meanly yearly peak in December). On the other hand, the municipalities of
Humaitá and Carauari (with mean yearly peaks in April) and Borba and Eirunepé (meanly yearly peaks
in May) showed peaks in malaria cases in the first half of the year. Wolfarth-Couto et al. 30 explain
that peaks in malaria cases can also occur before August in regions where the rainfall and hydrological
regime are anticipated naturally, thus leading to different seasonality patterns in malaria cases.
Although the highest reports of malaria cases in the literature are in the dry season, right after the
peak there is a decrease in the number of cases. Due to the low relative humidity and lack of tempo-
rary vector breeding sites caused by the low rainfall and high temperatures. the vector rates fall, thus
impacting disease transmission and case reporting. During this period, the mosquitos reproduce in
perennial bodies of water, mainly along the riverbanks 31, which play the role of year-round breeding
sites, even in the dry season.
The lags we observed between rainfall and malaria cases in the eight municipalities showed similar
characteristics to those reported by Grillet et al. 32. Lighter rainfall plays a critical role in the annual
variation in cases, directly impacting the peaks in malaria 6,32,33.
Local observations reveal the existence of different seasonal patterns for malaria in Amazonia,
which is not environmentally uniform, given the huge differences in topography, drainage patterns,
and economic and social development 25,34. Despite the importance of rainfall for malaria in the Ama-
zon Region and the fact that this relationship varies throughout the Amazon Basin 15,32, biological,
geographic, ecological, social, cultural, and economic factors can act synergistically in the produc-
tion, distribution, and especially control of vector-borne diseases 35. According to Barreto et al. 36,
geographic, economic, and social factors facilitate transmission and limit the application of standard
control measures.
Rainfall in Amazon does not display spatially and temporally uniform behavior 37. Analogous-
ly, malaria itself is not homogeneous. Although the results showed relations between rainfall and
malaria, phase coherences and lagged coherences express differences that may be related to the local
environmental characteristics and reflect the influence of epidemiological surveillance activities in
each location.
Another element that cannot be ruled out is that in the majority of the endemic areas, Plasmodium
vivax coexists with Plasmodium falciparum, and its recurrences follow a different temporal pattern 38.
Since P. vivax is more difficult to control and eliminate than P. falciparum due to its tendency to relapse
after the primary infection 38,39,40, this may affect the total number of malaria cases and the relations
between the variables, since the number of cases involving P. vivax (the principal etiological agent of
malaria in the region) is related more to relapses than to environmental determinants.
Coherence and wavelet phase analysis shows that local environmental conditions play an impor-
tant role in the vector’s seasonal cycle 41. Consistent with the findings by Hurtado et al. 42, the relations
between the variables indicated a strong annual pattern, in addition to significant coherences on the

Cad. Saúde Pública 2019; 35(2):e00020218

10 Wolfarth-Couto B et al.

scale of variability less than one year and on the biennial scale. According to Chowell et al. 43, a strong
annual pattern can be followed by biennial cycles, but differing by region and by type of malaria. Since
rainfall acts as a modulator of vector dynamics, interannual rainfall variability may have effects on
this relationship 31,43,44.
The scale of biennial oscillation, which appears heterogeneously between the municipalities, was
observed in the relationship between malaria cases and rainfall (Barcelos and Eirunepé) and water
level (São Gabriel da Cachoeira). Poveda & Mesa 45 suggest that biennial rainfall variability is associ-
ated with interference by ENOS (El Niño-Southern Oscillation), with responses in hydrological vari-
ability. Interannual oscillations in rainfall lead to responses in the Amazon Basin’s rivers, decreasing
(or increasing) the water discharges during El Niño/La Niña events 46.
According to Ferreira de Souza et al. 41, the biennial scale of ENOS is the principal phenomenon
that modulates climatic variability in Amazonia, as reflected mainly by the rainfall anomalies. In the
tropics, El Niño/La Niña heavily modulate the interannual changes in the predominant environmen-
tal variables. According to Rasmusson et al. 47, biennial oscillation is a fundamental element in the
variability of ENOS and evolves typically over the course of two years.
Studies on climatic influences on infectious diseases focus mainly on ENOS in public health and
malaria control. The malaria vector is susceptible to meteorological changes, which directly affect the
mosquito’s life cycle 10. Information on the potential impact of climatic factors on malaria incidence
can be useful for orienting disease prevention programs, aimed at the ultimate elimination of the
disease 42.
In relation to the scale of variability less than one year, the study showed an association between
malaria cases and water level. Hydrological phenomena on this scale may be associated with the
increase in the mosquito vector.
Consistent with Wolfarth 25 and Xavier 48, the data from rainfall series can display unusual hydro-
logical variability. Known as “repiquete”, characterized as an atypical hydrological phenomenon with
an oscillation outside the annual variability pattern, it consists of sudden variations in the water
level (in a mean period of one to three months) until the standard hydrological regime is established.
According to Wolfarth 25, the causes of repiquete are unknown, but the phenomenon may be associ-
ated with processes of climatic variability and have effects on the peaks in malaria cases.
Lagged coherences on the annual scale of variability in which malaria cases occur on average one
to four months after the peak river water levels are consistent with Wolfarth-Couto e al. 30, who state
that distinct hydrological regimes are important in the variable seasonality of malaria cases by shifting
the river water flooding.
Studies have shown that like rainfall, hydrological variability influences the fluctuation in vector
density 16,49,50. Rainfall not only provides larval habitats, but is also responsible for the rising river
levels that creates permanent breeding sites.
The findings comparing water levels and malaria cases suggest a better description of the depen-
dence between the study’s variables. According to Stefani et al. 14, malaria incidence receives a
relevant impact from water level variation. The rivers of the Amazon Basin and their hydrological
variability have a major influence on vector density 16,27,50, expressing year-round relations.
The independent variables and malaria cases may or may not present significant relations. Girod
et al. 21 explain the variation in these relations based on the characteristics of each location’s land-
scape. In addition, the climatic conditions in a given region are relevant for understanding the distinct
(or similar) dynamics of the disease 20.
Regardless of significant associations between the variables, the relationship between malaria and
climate is complex and indirect, especially when using data on malaria cases and not on the vector
(number of anopheline mosquitos). Despite this limitation, malaria cases serve as an excellent health
indicator and measure of epidemiological surveillance in the region, besides serving to back planning,
interventions, and control by health services 51. Grillet et al. 32 contend that better understanding of
temporal patterns in malaria would allow developing more effective surveillance and early warning
systems to prevent cases in response to climatic variations.
Environmental factors are not the only elements that determine malaria transmission. A better
understanding of the impacts of climate changes, hydrological variation, and ecological and epidemi-
ological factors are still necessary to assess the true local risk of malaria 26. Terrazas et al. 50 emphasize

Cad. Saúde Pública 2019; 35(2):e00020218

 MALARIA CASES AND ITS RELATIONSHIP WITH ENVIRONMENTAL FACTORS 11

the importance of socioeconomic policies to be implemented jointly with strategic environmental

protection and epidemiological surveillance measure for the population in the state of Amazonas.
Importantly, factors not assessed such as species-specific malaria cases, local temperatures, and
ENOS have distinct responses in association with malaria and can contribute to more robust conclu-
sions. Although environmental and climatic factors have a certain influence on malaria dynamics, sur-
veillance, prevention, and control issues should not be overlooked. Government measures in health
may (or may not) act effectively, masking possible relations with hydrological and climatic conditions.
Coherence and wavelet phase analysis has become an increasingly useful and significant tool for
interpreting natural phenomena in different fields of study 32,41,42,43. The technique provides a robust
mathematical base that encourages scientific research aimed at analyzing physical signs with com-
plex variabilities. It is thus possible to simply and quickly determine the covariance of variables and
their interrelations 52.

Conclusion

The statistical analyses show that malaria cases are strongly associated with rainfall and river water
levels as climatic factors. Significant coherences were thus demonstrated, mainly on the scale of
annual variability, in addition to displaying coherences on scales smaller than one year and on the
biennial scale. The analyses show that peaks in malaria cases are reported approximately one month
and a half before or after peaks in rainfall, and on average one to four months after the peak river
water levels in most of the municipalities studied here.
Importantly, each environmental variable displayed a distinct local behavior in time and space,
suggesting that other local variables (such as topography) may control the environmental conditions
and favor different behavior in each municipality. However, when the analyses are performed jointly
it is possible to see a non-random order in these relations.
The study suggests that intervention and control plans that contemplate each location’s environ-
mental and climatic reality, together with lagged conditions in the hydrological and rainfall regimes,
are fundamental elements for monitoring and helping to control the disease, with the potential to
assist in mitigating the burden caused by malaria in the state of Amazonas.

Contributors Acknowledgments

B. Wolfarth-Couto participated in the study con- The study was financed by the Amazonas State
ception and design, statistical analysis, data inter- Research Foundation (FAPEAM) and the Brazilian
pretation, and writing of the manuscript. R. A. Silva National Research Council (CNPq). The authors
participated in the statistical analysis and data inter- wish to acknowledge the support received from the
pretation. N. Filizola participated in the study’s ori- Institut de Recherche pour le Développmente (IRD;
entation, data interpretation, and critical revision of France), Maison de la Télédétection (Montpellier,
the manuscript. France), Potamology Laboratory of the Amazon
(LAPA), Federal University of Amazonas (UFAM),
State University of Amazonas (UEA) (especially
Additional informations MSc. Igor Oliveira for the technical support), and
the Graduate Studies Program in Climate and the
ORCID: Bruna Wolfarth-Couto (0000-0002-1445- Environment, National Institute of Amazonian
7840); Rosimeire Araújo da Silva (0000-0001-5828- Research (PPG-CLIAMB/INPA).
6193); Naziano Filizola (0000-0001-7285-7220).

Cad. Saúde Pública 2019; 35(2):e00020218

12 Wolfarth-Couto B et al.

References

1. Tadei WP, Thatcher DB. Malaria vectors in the 16. Barros FSM, Honório NA. Man biting rate
Brazilian Amazon: Anopheles of the subgenus seasonal variation of malaria vectors in Ror-
Nyssorhynchus. Rev Inst Med Trop São Paulo aima, Brazil. Mem Inst Oswaldo Cruz 2007;
2000 42:87-94. 102:299-302.
2. Tadei WP, Thatcher DB, Santos JMM, Scar- 17. Ministério da Saúde. Ações de controle da ma-
passa VM, Rodrigues IB, Rafael MS. Ecologic lária: manual para profissionais de saúde na
observations on anopheline vectors of malaria atenção básica. Brasília: Ministério da Saúde;
in the Brazilian Amazon. Am J Trop Med Hyg 2006.
1998; 59:325-35. 18. Confalonieri UEC. Saúde na Amazônia: um
3. Secretaria de Vigilância em Saúde, Ministério modelo conceitual para análise de paisagem e
da Saúde. Guia para gestão local do controle doenças. Estud Av 2005; 19:221-36.
da malária: diagnóstico e tratamento. Brasília: 19. Barata RCB. Malária no Brasil: panorama epi-
Ministério da Saúde; 2008. demiológico na última década. Cad Saúde Pú-
4. Ministério da Saúde. Relatório nacional de blica 1995; 11:128-36.
acompanhamento: combater o HIV/AIDS, a 20. Zhou G, Minakawa N, Githeko AK, Yan G. As-
malária e outras doenças. Brasília: Ministério sociation between climate variability and ma-
da Saúde; 2010. laria epidemics in the east African highlands.
5. Terrazas WCM. Desenvolvimento de SIG para Proc Natl Acad Sci U S A 2004; 101:2375-80.
análise epidemiológica da distribuição espacial 21. Girod R, Roux E, Berger F, Stefani A, Gaborit
da malaria no município de Manaus: um en- P, Carinci R, et al. Unraveling relationships
foque em nível local [Masters Thesis]. Rio de between Anopheles darlingi (Diptera: Culicidae)
Janeiro: Escola Nacional de Saúde Pública Ser- densities, environmental factors and malaria
gio Arouca, Fundação Oswaldo Cruz; 2005. incidences: understanding variable patterns of
6. Gurgel E. Paludisme et dynamiques environ- transmission in French Guiana (South Ameri-
nementalles dans l’état du Roraima au Brésil ca). Ann Trop Med Parasitol 2011; 105:107-22.
[Doctoral Dissertation]. Paris: Université de 22. Filizola N, Vicente AS, Santos AMC, Oliveira
Paris; 2006. MA. Cheias e secas na Amazônia: breve abor-
7. Barcellos C, Monteiro AMV, Corvalán C, Gur- dagem de um contraste na maior bacia hi-
gel HC, Carvalho MS, Artaxo P, et al. Mudan- drográfica do globo. T & C Amazônia 2006;
ças climáticas e ambientais e as doenças in- 9:42-9.
fecciosas: cenários e incertezas para o Brasil. 23. Torrence C, Compo GPA. Practical guide to
Epidemiol Serv Saúde 2009; 18:285-308. wavelet analysis. Bull Am Meteorol Soc 1998;
8. Alemu A, Abebe G, Tsegaye W, Golassa L. Cli- 79:61-78.
matic variables and malaria transmission dy- 24. Torrence C, Webster PJ. Interdecadal changes
namics in Jimma town, South West Ethiopia. in the ENSO-monsoon system. J Clim 1999;
Parasit Vectors 2011; 4:30. 12:2679-90.
9. Kovats RS, Bouma MJ, Hajat S, Worrall E, 25. Wolfarth BR. Análise epidemiológica espa-
Haines A. El Niño and health. Lancet 2003; cial, temporal e suas relações com as variáveis
362:1481-9. ambientais sobre a ocorrência da malária no
10. Thomson MC, Mason SJ, Phindela T, Connor período de 2003 a 2009 em 4 municípios do
SJ. Use of rainfall and sea surface temperature Estado do Amazonas, Brasil [Masters Thesis].
monitoring for malaria early warning in Bo- Manaus: Instituto Nacional de Pesquisas da
tswana. Am J Trop Med Hyg 2005; 7:214-21. Amazônia; 2011.
11. Patz JA, Campbell-Lendrum D, Holloway T, 26. Wolfarth BR, Filizola N, Tadei WP, Durieux
Foley JA. Impact of regional climate change on L. Epidemiological analysis of malaria and its
human health. Nature 2005; 438:310-7. relations with environmental variables in four
12. Hay SI, Guerra CA, Tatem AJ, Noor AM, Snow municipalities of the State of Amazonas – Bra-
RW. The global distribution and population at zil. Hydrol Sci J 2013; 58:1495-504.
risk of malaria: past, present, and future. Lan- 27. Magris M, Rubio-Palis Y, Menares C, Villegas
cet Infect Dis 2004; 4:327-36. L. Vector bionomics and malaria transmission
13. Hiwat H, Bretas G. Ecology of Anopheles dar- in Upper Orinoco river, southern Venezuela.
lingi Root with respect to vector importance: a Mem Inst Oswaldo Cruz 2007; 103:303-11.
review. Parasit Vectors 2011; 4:177. 28. Vittor AY, Pan W, Gilman RH, Tielsch J, Glass
14. Stefani A, Hanf M, Nacher M, Girod R, Carme G, Shields T, et al. Linking deforestation to
B. Environmental, entomological, socioeco- malaria in the Amazon: characterization of the
nomic and behavioural risk factors for malaria breeding habitat the principal malaria vector,
attacks in Amerindian children of Camopi, Anopheles darlingi. Am J Trop Med Hyg 2009;
French Guiana. Malar J 2011; 10:246. 81:5-12.
15. Olson SH, Gagnon R, Elguero E, Durieux L, 29. Himeidan YE, Elzaki MM, Kweka EJ, Ibrahim
Guégan J, Foley JA, et al. Links between cli- M, Elhassan MI. Pattern of malaria transmis-
mate, malaria, and wetlands in the Amazon sion along the Rahad river basin, eastern Su-
Basin. Emerg Infect Dis 2009; 15:659-62. dan. Parasit Vectors 2011; 4:109.

Cad. Saúde Pública 2019; 35(2):e00020218

 MALARIA CASES AND ITS RELATIONSHIP WITH ENVIRONMENTAL FACTORS 13

30. Wolfarth-Couto BR, Filizola N, Durieux L. Pa- 43. Chowell G, Munayco CV, Escalante AA, Mck-
drão sazonal dos casos de malária e a relação enzie E. The spatial and temporal patterns of
com a variabilidade hidrológica no Estado do falciparum and vivax malaria in Perú: 1994-
Amazonas, Brasil. Rev Bras Epidemiol; subme- 2006. Malar J 2009; 8:142.
tido. 44. Pascual M, Cazelles B, Bouma MJ, Chaves LF,
31. Confalonieri UEC. Variabilidade climática, Koelle K. Shifting patterns: malaria dynamics
vulnerabilidade social e saúde no Brasil. Terra and rainfall variability in an African highland.
Livre 2003; 1:193-204. Proc Biol Sci 2008; 275:123-32.
32. Grillet ME, Souki ME, Laguna F, León JR. The 45. Poveda G, Mesa OJ. Feedbacks between hy-
periodicity of Plasmodium vivax and Plasmodium drological processes in tropical South America
falciparum in Venezuela. Acta Trop 2014; and large-scale ocean-atmospheric phenom-
129:52-60. ena. J Clim 1997; 10:2690-702.
33. Aranha Camargo LM, Urbano M, Krieger H, 46. Ronchail J, Labat D, Callede J, Coconneau G,
Plessman E, Pereira L. Unstable hypoendemic Guyot JL, Filizola N, et al. Discharge variabili-
malaria in Rondônia (Western Amazon region, ty within the Amazon basin. In: Francks S, Wa-
Brazil): epidemic outbreaks and work-associ- gener T, Bogh E, Gupta HV, Bastidas L, Nobre
ated incidence in a agroindustrial rural settle- C, et al., editors. Regional hydrological impacts
ment. Am J Trop Med Hyg 1994; 51:16-25. of climatic change: hydroclimatic variability:
34. Chaves SS, Rodrigues LC. An initial examina- proceedings of symposium S6 held during the
tion of the epidemiology of malaria in the state seventh IAHS scientific assembly. Wallingford:
of Roraima, in the Brasilian Amazon Basin. International Association of Hydrological Sci-
Rev Inst Med Trop São Paulo 2000; 42:269-75. ences; 2005. p. 21-9. (IAHS Publication, 296).
35. Tauil PL. Controle de doenças transmitidas 47. Rasmusson EM, Wang X, Ropelwski CF. The
por vetores no Sistema Único de Saúde. Inf biennial component of ENSO variability. J
Epidemiol SUS 2002; 11:59-60. Mar Syst 1990; 1:71-96.
36. Barreto ML, Gloria Teixeira M, Bastos FI, 48. Xavier SDR. Variabilidade climática, vulne-
Ximenes RAA, Barata RB, Rodrigues LC. Suc- rabilidade ambiental e saúde: os níveis do rio
cesses and failures in the control of infectious Negro e as doenças relacionadas à água em
diseases in Brazil: social and environmental Manaus [Masters Thesis]. Rio de Janeiro: Es-
context, policies, interventions, and research cola Nacional de Saúde Pública Sergio Arouca,
needs. Lancet 2011; 377:1877-89. Fundação Oswaldo Cruz; 2014.
37. Espinoza Villar JC, Ronchail J, Guyot JL, 49. Sogoba N, Doumbia S, Vounatsou P, Baber I,
Cochonneau G, Filizola N, Lavado W, et al. Keita M, Maiga M, et al. Monitoring of larval
Spatio-temporal rainfall variability in the habitats and mosquito densities in the Sudan
Amazon basin countries (Brazil, Peru, Bolivia, Savanna of Mali: implications for malaria vec-
Colombia and Ecuador). Int J Climatol 2009; tor control. Am J Trop Med Hyg 2007; 77:
29:1574-94. 82-8.
38. White NJ. Determinants of relapse periodic- 50. Terrazas WCM, Sampaio VS, Castro DB, Pinto
ity in Plasmodium vivax malaria. Malar J 2011; RC, Albuquerque BC, Sadahiro M, et al. Defor-
10:297. estation, drainage network, indigenous status,
39. Tatem AJ, Smith DL, Gething PW, Kabaria and geographical differences of malaria in the
CW, Snow RW, Hay SI. Ranking of elimination State of Amazonas. Malar J 2015; 14:379.
feasibility between malaria-endemic countries. 51. Wieffels A, Wolfarth-Couto B, Filizola N,
Lancet 2010; 376:1579-91. Durieux L, Mangeas M. Accuracy of the ma-
40. Battle KE, Karhunen MS, Bhatt S, Gething laria epidemiological surveillance system data
PW, Howes RE, Golding N, et al. Geographical in the state of Amazonas. Acta Amaz 2016;
variation in Plasmodium vivax relapse. Malar J 46:383-90.
2014; 13:144. 52. Holdefer AE, Severo DL. Análise por ondaletas
41. Ferreira de Souza RA, Andreoli RV, Kayano sobre níveis de rios submetidos à influência de
MT, Carvalho A. American cutaneous leish- maré. Revista Brasileira de Recursos Hídricos
maniasis cases in the Metropolitan Region of 2015; 20:192-201.
Manaus, Brazil: association with climate vari-
ables over time. Geospat Health 2015; 10:40-7.
42. Hurtado LA, Calzada JE, Rigg CA, Castillo M,
Chaves LF. Climatic fluctuations and malaria
transmission dynamics, prior to elimination,
in Guna Yala, República de Panamá. Malar J
2018; 17:85.

Cad. Saúde Pública 2019; 35(2):e00020218

14 Wolfarth-Couto B et al.

Resumo Resumen

O entendimento das relações entre as variáveis de La comprensión de las relaciones entre las varia-
precipitação e nível d’água dos rios com os casos bles de precipitaciones y el nivel de agua de los ríos
de malária podem fornecer indícios importantes con los casos de malaria pueden proporcionar in-
da modulação da doença no contexto da variabi- dicios importantes sobre la modulación de la en-
lidade climática local. No intuito de demonstrar fermedad en el contexto de la variabilidad climá-
como essas relações variam no mesmo espaço en- tica local. Con el fin de demonstrar cómo varían
dêmico, realizou-se a análise de coerência e fase esas relaciones en el mismo espacio endémico, se
de ondeletas entre as variáveis ambientais e epide- realizó un análisis de coherencia y fase de ondele-
miológica no período de 2003 a 2010 para 8 mu- tas entre las variables ambientales y epidemiológi-
nicípios do Estado do Amazonas (Barcelos, Borba, cas, durante el período de 2003 a 2010, en 8 mu-
Canutama, Carauari, Coari, Eirunepé, Humaitá e nicipios del estado de Amazonas (Barcelos, Borba,
São Gabriel da Cachoeira). Os resultados indicam Canutama, Carauari, Coari, Eirunepé, Humaitá
coerências significativas principalmente na escala y São Gabriel da Cachoeira). Los resultados indi-
de variabilidade anual, contudo, escalas menores can coherencias significativas, principalmente en
que 1 ano e bienal também foram encontradas. la escala de variabilidad anual, sin embargo, tam-
As análises mostram que casos de malária apre- bién se detectaron escalas menores de 1 año y bie-
sentam pico com aproximadamente 1 mês e meio nal. Los análisis muestran que los casos de mala-
antes ou depois dos picos de chuva, e em média 1-4 ria presentan un pico con aproximadamente 1 mes
meses após o pico dos rios para grande parte dos y medio antes o después de la pluviosidad más alta,
municípios estudados. Foi notado que cada variá- y de media 1-4 meses tras el pico de los ríos para
vel ambiental apresentou atuação local distinta no gran parte de los municipios estudiados. Se obser-
tempo e no espaço, sugerindo que outras variáveis vó que cada variable ambiental presentó una ac-
locais (a topografia é um exemplo) possam con- tuación local distinta en el tiempo y en el espacio,
trolar as condições ambientais favorecendo uma sugiriendo que otras variables locales (la topogra-
atuação diferenciada em cada município, porém, fía es un ejemplo) puedan controlar las condiciones
quando as análises são feitas em conjunto é pos- ambientales, favoreciendo una actuación diferen-
sível ver uma ordem não aleatória destas relações ciada en cada municipio, no obstante, cuando los
acontecerem. Embora os fatores ambientais e cli- análisis se realizan en conjunto es posible ver un
máticos denotem certa influência sobre a dinâmi- orden no aleatorio de estas relaciones para que se
ca da malária, questões de vigilância, prevenção e produzcan. A pesar de que los factores ambientales
controle não devem ser desprezadas, significando y climáticos denoten una cierta influencia sobre la
que as atuações governamentais de saúde podem dinámica de la malaria, cuestiones de vigilancia,
mascarar possíveis relações com as condições hi- prevención y control no se deben despreciar, lo que
drológicas e climáticas locais. significa que las actuaciones gubernamentales de
salud pueden enmascarar posibles relaciones con
Malária; Precipitação Atmosférica; Hidrologia las condiciones hidrológicas y climáticas locales.

Malaria; Precipitación Atmosférica; Hidrología

Submitted on 01/Feb/2018
Final version resubmitted on 04/Jul/2018
Approved on 05/Sep/2018

Cad. Saúde Pública 2019; 35(2):e00020218