General patterns and variability
Previous reports from the Amazon region showed that An. darlingi had two main biting periods (crepuscular and nocturnal), with the pattern varying depending on the geographical location of the collection
[3, 14, 26]. In our study area, An. darlingi was generally active from 18:30 until 06:30 (Figure
4, Additional files
3), and its nightly biting numbers per unit time showed the lowest variability (V) among the five studied species (Table
4). Our collections throughout the night also call into question studies from other areas in Amazonian Brazil that ranked importance of malaria vector species based solely on crepuscular collections
Anopheles nuneztovari and An. triannulatus were the two species that showed consistent crepuscular activity, with the vast majority of mosquitoes biting soon after dusk and tapering off rapidly around 20:00 to 21:00. Both species, however, exhibited relatively high co-efficients of variation (V) in numbers per unit time (Table
3 and Table
Very little is known about An. intermedius in the Amazon Region. We have previously reported on its local potential as a vector of malaria
, host preference
 and seasonality
. Human landing catches showed that An. intermedius nightly biting patterns varied considerably (Figures
12, Additional file
4), the two villages with sufficient numbers for nested ANOVA analyses showing significant variation in V between years and among months and nights (Table
Skewness and kurtosis patterns
Skewness and kurtosis patterns of An. darlingi were markedly different from those of the other four species. The g1 and g2 moments for An. darlingi were typically non-significant, with few exceptions, which are associated with the predominantly nocturnal (17:30–06:30) activity pattern of this species, rather than the crepuscular pattern typical of An. marajoara, An. nuneztovari and An. triannulatus. The human landing catch profiles of An. darlingi did vary among sites and villages. The skewness frequency analysis demonstrated that significant differences occurred between sites in two villages. In São João, Site 1 had significantly lower positive skewness frequencies than Sites 2–4 (Table
7). This difference could have been due to the fact that Site 1 was at the end of the village that ran parallel to the Matapi River, where fewer people lived (one residence) compared to the other sites in São João. The isolation of this site may have caused a delay in biting activity until later in the evening. Site 4 was most similar to Site 1 (11% positive skewness) and was at the other end of the village with two houses near the collection site. Site 3 (20% positive skewness) was closer to the centre of the village and near a corral that had many pigs. Site 2 (43.8% positive skewness) was at the midpoint of the village where many people gathered at dusk. These results suggest that the availability of hosts influences early evening attraction, which probably accounts for the increases in significantly positive skewness. All the sites were equally distant from potential larval habitats of An. darlingi.
In contrast, in Santo Antônio the houses were located uphill, perpendicular to the river. Site 1 was along the Matapì River and next to a small stream (igarapé), while Site 2 was the first site up the hill from the river, Site 3 was at the back side of the village but very close to a larval habitat, and Site 4 was to the north of the village near a buffalo corral. The position of Sites 1 and 3 near larval habitats may have influenced the skewness patterns, suggesting a relation to the proximity of potential aquatic habitats and resting places. However, the main difference in skewness frequencies was that Site 2 and Site 4 had one and two collecting periods, respectively, with significantly negative skewness (Table
7). Little difference was apparent in percent of positively skewed activity patterns (7.4% to 11.5%). More detailed studies would be needed to examine intrapopulation skewness differences within a village.
Difference in kurtosis frequencies was observed for An. darlingi and An. marajoara in São Raimundo; all the kurtosis moments for Site 2 were non-significant, while the other sites had some positive kurtoses. As stated before for intra-site skewness comparisons, the differences between Site 2 and the other sites were that more people lived near this site and the community gathered here late into the night. Prolonged human activity around this site may have extended the biting activity, producing a higher percentage of non-significant kurtosis frequencies compared to the other three sites. In São João, all the kurtosis moments for Site 1 were non-significant for An. darlingi, while the other sites had some positive kurtosis. Site 1 was at the end of the village with only one residence. The lack of human activity around this site may have shortened its biting activity.
 reviewed the factors that possibly influence the form of the biting cycle of An. darlingi. He suggested that the age of the population, moonphase and distance from the oviposition or mating site may influence the biting pattern of this species. In our study, we observed that the distance from larval habitats may have influenced anopheline activity patterns within a village. Therefore, characteristics of the site of collection need to be considered when examining host-seeking activity patterns of these species. Studies conducted over longer time periods and at various sites would provide better inferences about the factors that influence An. darlingi’s activity patterns. This would further clarify the biting behaviour plasticity proposed by several authors
[3, 16]. Further research on these factors is warranted for all the vector species studied.
The percentage of sites that had significantly positive co-efficients of skewness and kurtosis were high for An. nuneztovari and An. triannulatus, the positive skewness typically derived from ‘tails’ in nocturnal biting activity after the crepuscular peak, and the leptokurtosis showing that the biting peak is characteristically large in relation to tails (Figures
3). The crepuscular activity peaks associated with significantly positive skewness were typically captures made in the second hour of nightly collections (Figures
8), suggesting that these deviations from a normal distribution are not caused by inadequate sampling (e g, in pre-dusk hours) of potential biting times.
ML ANOVA comparisons showed no significant difference in the frequencies of skewness or kurtosis categories between An. nuneztovari and An. marajoara (Table
9). However, An. marajoara had a higher percentage of mesokurtic distributions (41.9%) (i e, not significantly different from normal) compared to An. nuneztovari (24.6%) (Table
8 and Table
9). This difference is evident by visual inspections of their activity patterns: An. marajoara host-seeking activity was more dispersed, peaking between 18:30 and 19:30 with additional activity throughout the night, while An. nuneztovari activity occurred mainly between 17:30 and 21:30 (Figures
4, Additional files
3). Differences among sites within villages in skewness and kurtosis frequency categories (Table
7 and Table
10) suggest that An. marajoara may be similar to An. darlingi in its capacity to alter activity patterns in response to environmental cues, such as host availability.
The skewness and kurtosis co-efficients for An. intermedius were usually significantly positive or not significantly different from normal, with one exception for kurtosis in São João (Table
5 and Table
8). Insufficient numbers of this species and An. triannulatus precluded some analyses conducted on the more abundant species.
Several species showed pre-dawn increases in host seeking during the final hour of nightly collections (e g, of An. darlingi) (Figure
9, Additional file
1), An. marajoara (Figure
4, Additional file
1) and An. triannulatus (Figure
8, Additional file
4). Possibly these same species might also have been captured after dawn, if collections had been made, which would further alter skewness and kurtosis moments and interpretations.
Charlwood and others
 hypothesized that biting patterns may have evolved in relation to host choice; with anthropophilic species being nocturnal and zoophilic species being crepuscular. Bloodmeal host identifications in the study area
 indicated that the most crepuscular species, An. nuneztovari and An. triannulatus were more zoophilic compared to the more nocturnally active An. darlingi and An. marajoara. These results seem to support a host choice hypothesis. However, the authors formerly showed that host availability was the most important factor in determining host blood meal choice in these same three villages
Inferences for malaria control
The degree of kurtosis indicates whether the risk of being bitten occurs over a short time period (leptokurtic), is normally distributed (mesokurtic), or occurs over a longer period (platykurtic; Figure
1). In this study An. nuneztovari and An. triannulatus, were mainly crepuscular being most active from 17:30 to 21:30. They both frequently showed positive skewness and high leptokurtosis. Therefore, vector control programmes would need to consider developing a control component that includes the time period when people are still active outside. Anopheles darlingi was active at night (17:30–06:30), had mainly non-significant skewness and non-significant kurtosis (mesokurtic). In this case, assuming that host-seeking activity patterns are similar for exophagic and endophagic An. darlingi, an indoor vector control strategy, such as impregnated materials
[28–30] may work the best. It is more difficult to determine a risk activity period for species like An. marajoara and An. intermedius. Anopheles marajoara had positive skewness but a lower frequency of positive kurtosis than did An. nuneztovari and An. triannulatus. Its activity pattern was more dispersed with additional activity throughout the night. Anopheles intermedius had no striking difference in the percentage of significantly positive skewness and kurtosis compared to non-significant distributions (Table
5 and Table
8). One would have to take into account other risk factors related to malaria transmission before recommending vector control for these two species.
The appropriate methods of malaria and vector control are highly dependent on the epidemiology of malaria in a particular location, including Plasmodium species, vector competency, health service availability and control methods
[5, 7]. In the study area the presence of five potential vectors
 with different biting cycles further complicates any vector management programme. Previous research in these villages showed that An. darlingi and An. marajoara are the most abundant and anthropophilic species
[22, 23] and have the highest Plasmodium spp. infection rates
. Therefore, one would most likely focus on these two species when developing an integrated malaria control programme for this region of the Brazilian Amazon.