Spatio-temporal variability of malaria infection in Chahbahar County, Iran: association with the ENSO and rainfall variability

Malaria is one of the most widespread communicable diseases in the southeast regions of Iran, particularly the Chabahar County. Although the outbreak of this disease is a climate-related phenomenon, a comprehensive analysis of the malaria-climate relationship has not yet been investigated in Iran. The aims of this study are as follows: a) analyzing the seasonal characteristics of the various species of the infection; b) differentiating between number of patients during El Niño and La Niña and also during the wet and dry years. The monthly malaria statistics collected from twelve health centers were firstly averaged into seasonal scale and then composited with the corresponding data of the ground-based meteorological records, Southern Oscillation Index (SOI), and the satellite-based rainfall data. The proper statistical tests were used to detect differences in the number of patients between El Niño and La Niña and also between the adopted wet and dry episodes. Infection rate from the highest to the lowest was associated with summer, autumn, spring, and winter, respectively. Plasmodium falciparum, P. vivax, and the other species were responsible for 22%, 75%, and 3% of the sickness, respectively. The outbreak of P. falciparum/P. vivax occurs during autumn/summer. Due to the malaria eradication programs in urban areas, infection statistics collected from the rural areas were found to be more climate-related than that of urban regions. For rural/urban areas, the infection statistics exhibited a significant decline/increase during El Niño episodes. In autumn, spring, and winter, the patient number has significantly increased/decreased during the dry/wet years, respectively. These relationships were, however, reversed in summer.


Introduction
Parasitic diseases, including incoming blood and tissue protozoa, have always originated from environmental pollution (Omidian et al. 2021;Rayani et al. 2020). Malaria is a leading life-threatening mosquito-borne parasitic disease across the globe. It causes a debilitating disease and sometimes terrible death in different parts of the world, especially in the tropics and subtropics, where Anopheles mosquitoes breed abundantly in mud and swamps (Hay et al. 2002;Maghsoodloorad et al. 2019). Researchers estimate that 241 million cases of malaria occurred worldwide in 2020, causing 627,000 deaths (World Health Organization 2020). In 1924, malaria infection was a serious barrier to Iran's economic growth since more than one-third of Iran's 13 million population suffered from this infection (Ziony 1950). Malaria control programs initiating from 1945, resulted in eradication/significant suppression of the disease in the northern/ southern parts of the country by 1977, respectively (Edrissian 2006; Raeisi et al. 2013). In 2010, 90% of the reported cases in Iran belonged to the southeastern parts of the country (Ghanbarnejad et al. 2021;Hatam 2010;Hemami et al. 2013;Kazemi et al. 2005;Mohammadzadeh et al. 2014;Rezanezhad et al. 2011). In spite of the fact that the climatemalaria connection is not yet completely understood, scientific communities know that there is an important balance of temperature, rainfall, and humidity that creates ideal conditions for mosquitoes to breed and transmit malaria parasites (Abiodun et al. 2016;Krefis et al. 2011;Parham et al. 2012;Sena et al. 2015;Soverow et al. 2009). In this regard, some studies have reported that the aggressive outbreaks are seen mainly during and after the "monsoon" season. A significant increase of malaria morbidity and mortality of patients of Isra University Hospital, Hyderabad, Pakistan, was reported after the monsoon flood of 2011 (Memon et al. 2014). Moreover, another study showed a significant effect of monsoon rainfall in the inter-annual variability of epidemic malaria in India (Laneri et al. 2010). Their non-linear model demonstrated the potential of forecasting malaria epidemics based on climate variability, in the desert and semi-arid regions of this country.
The El Niño-Southern Oscillation (ENSO) is a natural phenomenon described by quasi-periodic warming of the ocean surface between the eastern and western parts of the equatorial Pacific Ocean (Nazemosadat and Ghasemi 2004). The warmer water stream essentially oscillates back and forth across this part of the Pacific, much like water in a large container. The El Niño (or warm phase) and La Niña (or cold phase) are the extreme phases of ENSO, which contribute to the predictability of climate anomalies in the global scale. Comprehensive definition of ENSO and its effects on global climate are presented in the web pages of the Australian Bureau of Meteorology (BOM) and National Oceanic and atmospheric Administration (NOAA; http:// ocean servi ce. noaa. gov/ facts/ ninon ina. html; http:// www. bom. gov. au/ clima te/ enso/ histo ry/ ln-2010-12/ ENSO-what. shtml). Recent studies have shown that ENSO in conjunction with variations in the sea surface temperatures over tropical and extra-tropical water bodies induces a significant impact on precipitation variability and climate change in Iran and West Asia (Nazemosadat and Cordery 2000;Nazemosadat and Ghaedamini 2010;Nazemosadat and Ghasemi 2004;Nazemosadat et al. 2006;Nazemosadat and Shahgholian 2017).
Our study area, Chabahar, is a county located in Sistan-Baluchistan province, in the southeast of Iran with an area of 9739 km 2 and population of 291910 (Zubair et al. 2008). Southern and eastern borders of the county are confined by the warm waters of the Indian Ocean (115 Km), including a few small bays and gulfs, and the Iran-Pakistan border (131 Km), respectively (Fig. 1). These specific geographical attributes and also crossing the international borders by the infected locals, has made malaria a persistent disease in this region (Fekri et al. 2014;Halimi et al. 2016;Hanafi-Bojd et al. 2010).
In spite of the fact that monsoonal rainfall and floods are known as the influential parameters on the outbreaks of malaria in India and Pakistan (Laneri et al. 2010;Memon et al. 2014), Chabahar has a Mediterranean climate and precipitation generally occurs during boreal autumn and winter. Furthermore, our preliminary investigation did not show a significant relationship between the Central Indian Monsoon Precipitation and summertime air temperature in Chabahar.
In this study, we utilized our own methodology to understand the malaria-climate associations over an area with the highest rate of infection in Iran. In contrast to most of the previous studies, we analyzed this association on a seasonal scale. On this basis, the disease characteristics for each individual health center were evaluated and an infection map was provided for the county. The impact of the extreme ENSO phases on the station-based data of three variables, comprising rainfall, air temperature, and relative humidity, is investigated. We further categorized the patient statistics based on the malaria species. Later, the effects of the El Niño or La Niña events on the main species of the infection were also spatially mapped. To the best of our knowledge, for the first time, we have utilized the satellite-based rainfall data to specify the wet and dry episodes in each health center. The malaria statistics of these two episodes were compared and the results were demonstrated by spatiotemporal maps. To avoid misleading conclusions, the effects of the malaria-elimination policies on these examinations were also taken into account.

Data
To characterize the main climatological characteristics of Chabahar ( Fig. 1), its station-based monthly data of air temperature, rainfall, and relative humidity were obtained from the Iranian meteorological organization over the period of 1986-2016. The corresponding monthly SOI data were also extracted from the BOM's website for 1986-2016 (http:// www. bom. gov. au/ clima te/ enso/ histo ry/ ln-2010-12/ ENSOwhat. shtml). It has already been documented that the SOIprecipitation relationships in Iran are marginally stronger than that of Niño-precipitation relationships (Nazemosadat 1999(Nazemosadat , 2001Nazemosadat and Shirvani 2004). Monthly statistics of the number of patients collected in fourteen health centers of Chabahar, were obtained from the regional office of the Ministry of Health and Medical Education in Iran for the period 2002 to 2011.
While the infection statistics were gathered from multiple health centers, the sole meteorological station was located near the city center on the coast of the Indian Ocean (southwest corner of Fig. 2). The difference between the number of health centers and meteorological stations was a big hinder for spatial assessment of the malaria-rainfall relationships. To resolve this problem, the satellite-based rainfall data were used on a calendar month temporal resolution and a 0.25° by 0.25° spatial resolution for the period of 2002-2011 (https:// pmm. nasa. gov/ data-access/ downl oads/ trmm; https:// trmm. gsfc. nasa. gov/). The data were extracted from the official website of the Tropical Rainfall Measuring Mission (TRMM) 3B43 (V7). The capabilities of the Geographical Information System (GIS) were utilized to convert the TRMM gridded data into the point data for the specified latitudes and longitudes of the health centers. Although the Goddard Earth Sciences Data and Information Services Center (GES-DISC) has recommended the application of  Hemami et al. (2013). B The study area in Chabahar county TRMM data for the Malaria-Modeling Approaches (https:// pmm. nasa. gov/ data-access/ downl oads/ trmm), the authors could not find any references for the application of this recommendation, as conducted in the present study.

Methods
Our preliminary examination revealed that the effects of climate indices on malaria are more obvious if their relationships are investigated in seasonal rather than monthly timescale. Therefore, all the collected monthly data were averaged into seasonal records. In this averaging process, January-March, April-June, July-September, and October-December periods were adopted as winter, spring, summer and autumn, respectively. The non-parametric Mann-Whitney and Fisher Exact tests were used to investigate the significance of the difference in the weather data or number of patients between various phases of ENSO or various phases of ENSO or between the adopted wet and dry epochs (Gagnon et al. 2002;Parham et al. 2012;Wardrop et al. 2013;Zubair et al. 2008).

Construction of the SOI-climate composites
Time-series of seasonal SOI  were firstly sorted in ascending order. All years with seasonal SOI equal or less than -5 or equal or greater than +5 were assigned as the El Niño or La Niña years, respectively. Table 1 depicts the years that are denoted as El Niño or La Niña for each season. Seasonal values of rainfall, temperature, and relative humidity in Chabahar station were matched with the El Niño and La Niña years to construct the ENSO composites for each variable (Table 2). Significant difference in the mean values of these variables between El Niño and La Niña was then investigated ( Table 2).

Construction of the SOI-malaria composites
Since the malaria data were available for the period 2002-2011, after obtaining the El Niño or La Niña years for this period from (Table 1), the ENSO-infection composites were constructed in a seasonal basis. The Number of Malaria Infected patients (NMI) during the El Niño and La Niña events (NMI El and NMI La , respectively) were then statistically compared for each individual season and health center. Furthermore, the values of the ration NMI El / NMI La , were used to assess the measure of the effects of ENSO extreme phases on the malaria infection. When this ratio is greater/ Fig. 2 The spatial distribution of the health centers around Chabahar. The position of the meteorological station is denoted by "chabahar" in red color

Rainfall-malaria composites
For each individual season and health center, the monthly TRMM rainfall data were obtained over the period of 2002 and 2011. After transferring monthly data into seasonal time-series, the series were sorted in ascending order. The sorted data series were then divided into three parts including three years with highest or lowest rainfall and remaining 4 years in the middle. Significant difference in the number of patients was then assessed between the years with highest and lowest rainfall (hereinafter wet and dry episodes). The ratio of number of patients during the wet years to the dry years (P w /P d ) was also examined for each individual health center to assess the departure from unity. This ratio was considered as an index for demonstrating the effects of rainfall variability on malaria statistics. When the ratio is greater/smaller than unity, patient numbers during wet episodes are more/less than that of during dry episodes. For strengthening the obtained results, the mentioned procedures for analyzing rainfall-malaria relationships were also re-examined when the considered wet or dry episodes contain 4 years.

Climatic features and associations with ENSO
The mean annual rainfall in Chabahar meteorological station is 120 mm for the period 1986-2016 (Table 2). Winter, spring, summer and autumn rainfalls constitute 53.5%, 8%, 4.5%, and 34% of this rainfall, respectively. As indicated, the corresponding seasonal values of temperature or relative humidity were 22. °C, 29.3°C, 29.2°C, and 25.1°C or 69%, 79%, 82%, and 70 %, respectively. These statistics infer that, in contrast to rainfall, relative humidity is generally greater in the warm rather than cool months of the year. Autumn/winter rainfall is significantly/remarkably greater for the El Niño as compared with La Niña years (Table 2). This relationship, however, reverses for spring and summer, when the mean seasonal rainfall is much less than that of winter or autumn (from 9.6 to 5.4 mm / season, respectively). Since rainfall time-series in Chabahar are affected by frequent near zero and sporadic torrential rain, the presented ENSO-rainfall relationships could change over time for spring and summer; when these relationships are not strongly significant. Compared to El Niño, air temperature tends to be warmer during La Niña episodes from 0.2°C to 0.4°C. Relative humidity, however, is consistently greater during El Niño episodes from about 0.4% in spring to 7.3% in winter.
The impact of ENSO on malaria potentially increases when the difference in the climatic features is significant between El Niño and La Niña. This influence is, hence, more remarkable during autumn, winter compared to summer and spring. Zubair et al. concluded that, for the areas that are less affected by the disease control program, the long-term records of malaria data are mostly influenced by the ENSO extreme conditions (Zubair et al. 2008). Table 3 illustrates the summer records of the number of patients, in fourteen health centers, during 2002-2011, when seasonal infection is in its highest status. Since the dataset of two of the stations contained some missing data, their statistics are not presented. Similar statistics were also obtained for other seasons, but due to result briefing, these data are presented in Appendix 1. As indicated, mean values of the number of patients for winter, spring, summer and autumn are correspondingly 267, 1501, 5159, and 2885. While the highest records are related to summer and autumn, the lowest values are linked to winter, when seasonal rainfall (air temperature) is remarkably higher (lower) than the other seasons. The difference in the statistics of winter and summer could be mostly attributed to the air temperature and relative humidity rather than low values of seasonal rainfall. As indicated in Tables 2 and 3 and Appendix 1, air temperature, relative humidity and patient numbers are maximized/ minimized during summertime/ wintertime.  2  3  4  5  6  7  8  9  10  11  12   Summer  2002  4  25  25  6  28  37  3  13  -185  92  8  2003  88  215  420  19  54  171  254  127  -254  64  161  2004  15  14  44  8  6  44  40  17  2  104  61  20  2005  22  55  65  12  14  78  59  20  7  70  33  49  2006  2  15  53  3  5  23  9  3  3  36  17  2  2007  20  108  132  11  4  108  153  66  0  47  11  30  2008  14  49  312  5  1  83  50  19  6  26  4  21  2009  6  18  210  3  3  17  2  5  1  presented statistics the outbreaks of P. falciparum and P. vivax are generally related to autumn and summer, respectively. Our findings suggest that P. falciparum, P. vivax and the other types of malaria were responsible for 22%, 75%, and 3% of infections, respectively. Figure 5 illustrates the spatial distribution of autumnal P. falciparum-infection; when the frequency of this type of infection is greater than the other seasons. Similar illustrations related to other seasons are presented in Appendix 2. In spite of the fact that the infection statistics are different from season to season, the spatial patterns of the epidemic are in general agreement, suggesting that the frequency of seasonal infection is firstly highest over the northeastern and secondly over the southwestern parts of the county (Fig. 5 and Figs in Appendix 2). According to Fig. 3, after autumn, statistics of summer, spring, and winter are in the second to fourth ranks, respectively.  In spite of this coherence, patient number is more than threefold higher for P. vivax in Fig. 6 as compared with P. falciparum in Fig. 5. After summer, the risk of P. vivax is, respectively, more severe during autumn, spring and winter (Fig. 4). The highest frequencies of P. falciparum and P. vivax are related to the year 2003; one of the driest years during the study period (Figs. 3 and 4). However, our examination did not detect any exceptional condition in the temperature and relative humidity for this year. It is noteworthy to mention that the unusually high incidence of malaria is not necessarily driven by unusual weather patterns, and rather, is mediated by other factors such as human behavior, the mosquitos movement patterns, migration and agricultural practices (Wardrop et al. 2013). Table 4 illustrates the autumnal values of the number of patients for the warm and cold phases of the ENSO at Dagres health center as an example. The mean values of patients' number and their associated NMIEl/NMILa are 73.0, 96.0, and 0.76. These statistics infer that the patient's number is smaller during El Niño as compared to La Niña. It is worthwhile to note that Dagres health center is located in the rural areas near the Iran-Pakistan border. In contrast, the stations located in the southwestern corner of the county (namely Shahri 1 and Shahri 2 in Fig 2) represent the urban region. Figure 7 depicts the spatial distribution of the NMIEl/ NMILa for autumn, when the difference between NMIEl and NMILa is significant for most of the health centers. According to this figure, for urban/rural areas, the ratio is predominantly less/greater than unity. Similar figures were also obtained for the other seasons but are not shown here. Gagnon et al. found a statistically significant relationship between ENSO and malaria epidemics in Colombia, Guyana, Peru, and Venezuela (Gagnon et al. 2002) As indicated in Fig. 7, opposite to the northern and particularly northeastern parts of the county, for the urban areas in the southwestern corner of the study area, the ratios of NMIEl/NMILa derived from 1.5 to 2.36 which are higher On the other hand, for the areas denoted by blue or yellow color for which the ratio deviates from 0.26 to 1.00, the occurrence of El Niño decreases the patient number up to 400%. Therefore, the effects of the ENSO extreme phases on the number of patients are different between the rural areas that are mostly located in the eastern side of the county and urban regions in the western and southern parts of the study area. Further research is recommended to consolidate the given results.

Impacts of the malaria elimination policies
Since ENSO is a global climate phenomenon, it is difficult to accept that the prevalence of the El Niño or La Niña induces an opposite effect on the malaria statistics of the study area, as presented in the previous section. Our close examination revealed that the implementation of malaria eradication policies by Iranian government is probably the main cause of the observed contradiction. While the number of infected people in urban regions has been consistently reduced from 2002 to 2011, such decline in the number of patients is not evident for rural regions Table 5 delineates the ratio of the number of patients during 5 years period of 2002-2006 to the corresponding statistics during 2007-2011. As indicated, except for two cases (summer in Dargas and winter in Bahookalat), the ratio is consistently greater than unity. This indicates a considerable reduction in the infected people during the 2007-2011, as compared to its preceding 5 years. The reduction rate is, however, different between health centers and from one season to another. The biggest/smallest ratio values are generally associated to the urban (i.e., Shahri 2 and Shahri 3)/rural areas (i.e., Dagres and Bahookalat). Statistics in Table 5 prove the effectiveness of the malaria elimination policies in urban areas as compared to the rural parts of the county. The given results by (Fekri et al. 2014) and (Hanafi-Bojd et al. 2010) generally support these findings.
According to our analysis and as partially presented in Table 1, among the 10 years of the study period, the autumnal SOI was negative for the first 4 years and positive for the remaining 6 years. In other words, El Niño/La Niña events had been mostly occurred during the first/second half of the study period. Therefore, the malaria eradication programs were more effective during the La Niña rather than El Niño years. Comparing the information presented in Tables 4 and 5 implies that if the effects of the eradication programs on the malaria statistic were removed, El Niño or La Niña events are generally associated with the decrease or increase in the malaria epidemic as indicated for Sangan, Arabzehi, Dagres, and Bahookalat in Fig. 7. In other words, the ENSO-malaria relationships are more accurate if these relations are assessed for the rural rather than urban regions. Table 6 depicts the mean values of autumnal rainfall for the assigned dry and wet periods (2005, 2007, and 2003 or 2011, 2006, and 2004) in Dagres health center (Fig. 2), which is presented as an example. The mean value of rainfall data for each episode, the corresponding patient number and the ratio of Pw(wet)/ Pd(dry) are also presented. Additional statistics are also provided for the case that each episode contains 4 years. According to the given information the upsurge or decline in the patient number is statistically associated with the adopted dry or wet events, respectively. The results of the statistical tests have remained stationary when Pw and Pd were computed for 4 years. Table 7 is the seasonal values of Pw/Pd for all twelve stations. As indicated, ratios are consistently and significantly less than unity for autumn, winter and spring and greater than 1.0 for summer. This means that the occurrence of wet or dry episodes harmonizes with the reduction or increase in the number of patients during autumn, winter, and spring, respectively. This relationship, however, reverses for summer; the season that water availability and hot weather can increase the infection. Some direct or indirect malaria-rainfall associations were also reported by other investigators (Odongo-Aginya et al. 2005;Wardrop et al. 2013). In spite of the fact that summertime precipitation contributes to only 4.5% of annual rainfall, this small and sporadic rainfall event is an essential component for increasing the infection rate. Since these rainfalls events are generally heavy, they can supply enough water for local ponds and wetlands. Although precipitation provides essential habitat for larvae during the aquatic stages of some infectious diseases, drought can indirectly expand the vector's range (Paz and Semenza 2016). Briët et al. have also reported some negative relationships between rainfall amount and malaria during persistent regional drought (Briët et al. 2008).

Conclusion
The mean annual rainfall in Chabahar is about 120 mm, out of which 53.5%, 8%, 4.5%, and 34% are related to winter, spring, summer, and autumn, respectively. Autumnal and winter rainfalls are mostly greater/smaller than the climatological mean during the El Niño/La Niña years. Compared to La Niña, air temperature is slightly 0.2 to 0.4 o C colder during El Niño. Seasonal values of relative humidity are consistently greater during El Niño from 0.4% in spring to 7.3 % in winter. The highest/lowest records of patient numbers are related to summer/winter indicating the importance of air temperature on the infection rates. P. falciparum, P. vivax, and the other types of malaria are responsible for 22%, 75%, and 3% of the disease, respectively. While P. falciparum is the prevailing infection during autumn, P. vivax is dominant in summer. For both types, the seasonal infection is greater/lesser over the northeastern and southwestern/central parts of the county. For urban/rural areas, the ratio of the patient number during El Niño to that of La Niña (NMIEl / NMIL) is found to be predominantly greater/less than unity.
While the ratio was about 0.25 in rural areas, it escalated to about 2.4 in the stations located in the urban regions. This difference was mostly attributed to the effectiveness of the malaria eradication programs in Iran. The seasonal values of the ratio of number of patients during the wet period to that of the dry period (Pw/Pd), were found to be consistently and significantly less than unity for autumn, winter and spring and greater than 1.0 for summer. This suggests that, while the occurrence of wet episodes reduces the patient number during autumn, winter, and spring, these episodes increase the patient statistics during the summertime. In other words, although summertime precipitation is very low, the sporadic heavy rainfall events can significantly increase the infection. For the other seasons, the patient number reduces during the wet years.