Comparison of two adult mosquito sampling methods with human landing catches in south-central Ethiopia

The human landing catch (HLC) is the standard reference method for measuring human exposure to mosquito bites. However, HLC is labour-intensive, exposes collectors to infectious mosquito bites and is subjected to collector bias. These necessitate local calibration and application of alternative methods. This study was undertaken to determine the relative sampling efficiency (RSE) of light traps with or without yeast-produced carbon dioxide bait vs. HLC in south-central Ethiopia. The experiment was conducted for 39 nights in a 3 × 3 Latin square randomized design with Anopheles arabiensis as the target species in the period between July and November 2014 in Edo Kontola village, south-central Ethiopia. Center for Disease Control and Prevention light trap catches (LTC) and yeast-generated carbon dioxide-baited light trap catches (CB-LTC) were each evaluated against HLC. The total nightly mosquito catches for each Anopheles species in either method was compared with HLC by Pearson correlation and simple linear regression analysis on log-transformed [log10(x + 1)] values. To test if the RSE of each alternative method was affected by mosquito density, the ratio of the number of mosquitoes in each method to the number of mosquitoes in HLC was plotted against the average mosquito abundance. Overall, 7606 Anopheles females were collected by the three sampling methods. Among these 5228 (68.7%) were Anopheles ziemanni, 1153 (15.2%) An. arabiensis, 883 (11.6%) Anopheles funestus s.l., and 342 (4.5%) Anopheles pharoensis. HLC yielded 3392 (44.6%), CB-LTC 2150 (28.3%), and LTC 2064 (27.1%) Anopheles females. The RSEs of LTC and HLC for An. arabiensis were significantly correlated (p < 0.001) and density independent (p = 0.65). However, for outdoor collection of the same species, RSEs of LTC and CB-LTC were density dependent (p < 0.001). It was estimated that on average, indoor LTC and CB-LTC each caught 0.35 and 0.44 times that of indoor HLC for An. arabiensis respectively. Results showed that HLC was the most efficient method for sampling An. arabiensis. LTC can be used for large-scale indoor An. arabiensis surveillance and monitoring when it is difficult to use HLC. CB-LTC does not substantially improve sampling of this major vector compared to LTC in this setting. Trial registration PACTR201411000882128 (retrospectively registered 8 September, 2014)


Background
The current core malaria vector interventions are longlasting insecticidal nets (LLINs) and indoor residual spraying (IRS) [1]. In attempts to control malaria by attacking the vector, it is important to measure the impact of such interventions on mosquito populations. Measuring this requires an appropriate method of sampling mosquitoes biting humans [2,3]. The most direct way to do this is by the human landing catch (HLC) because mosquitoes are captured as they land and attempt to feed on collectors [4]. The HLC is the standard method for measuring exposure of humans to mosquito bites [5] and for estimating the human biting rate (HBR) which is a key determinant of the entomological inoculation rate (EIR), a measure of malaria transmission [6].
However, the HLC exposes collectors to potentially infectious mosquito bites, is labour-intensive, requires highly trained collectors and is difficult to supervise. Besides, results obtained by HLC can be biased due to natural human variations in attractiveness to mosquitoes [7,8]. These issues limit the application of HLC particularly for monitoring the effectiveness of vector control interventions and necessitate the search for alternative methods.
A recent review by Briet et al. [8] showed that LTC has been widely evaluated against HLC for collecting host-seeking vectors in several areas and is considered a safe and approximately equivalent alternative to HLC for measuring indoor exposure to mosquito bites and malaria transmission by African vectors. Light traps are affordable, easy to use and can be deployed in large-scale longitudinal, community-based trapping schemes using solar-powered battery chargers and provide valuable entomological data of the impact of vector control interventions [16,22].
Another promising alternative to HLC is carbon dioxide-baited traps. Many kairomones are released by the human body, but carbon dioxide (CO 2 ) is the most important for luring host-seeking mosquitoes [23]. Artificial sources of CO 2 , specifically from dry ice, industrial pressurized gas cylinders, or from propane are commonly used in mosquito traps [22,23]. However, adult mosquito surveillance in many rural areas is challenging due to lack of CO 2 , either in the form of dry ice or compressed gas. In resource-poor areas, like sub-Saharan Africa, it is hard to obtain CO 2 sources that are reliable, cheap, and durable [23].
In Japan, Saitoh et al. [24] developed an easy and cheap method to produce CO 2 by using a yeast-sugar solution in plastic bottles and evaluated the efficacy of yeast-generated CO 2 as an attractant for mosquitoes in field collections. They found that traps baited with yeast-generated CO 2 caught higher numbers of mosquitoes than unbaited traps. Results from trapping experiments conducted in the laboratory and semi-field systems in Kenya found that traps baited with yeast-produced CO 2 caught significantly more Anopheles gambiae sensu stricto than unbaited traps as well as traps baited with industrial CO 2 [23]. However, studies on other major African malaria vectors, such as An. arabiensis as a target species is lacking.
Anopheles arabiensis is the principal malaria vector in Ethiopia [25] and its control is primarily based on IRS and LLINs either in combination or separately. To generate evidence on the effect of IRS and LLIN combined interventions, a cluster randomized controlled trial was implemented in Adami Tullu district in south-central Ethiopia [26]. For this trial, an appropriate mosquito sampling method was required to monitor the impact of the interventions on An. arabiensis and other local Anopheles populations.
Although, CDC light traps have been used to estimate human biting rates and EIR equivalent to the ones obtained from HLC by locally calibrating against HLC and calculating conversion factors [2,27], no such evaluation and determination of conversion factors have been carried out in Ethiopia so far. For example, Animut et al. [28] estimated daily EIR for An. arabiensis in highland areas of south-central Ethiopia based on a conversion factor for LTC vs. HLC of 1.91, determined for this species in Zambia [27]. Likewise, Massebo et al. [29] estimated annual EIR for the same species in Chano, south-western Ethiopia, using 1.605 a factor determined in Tanzania [2]. The efficiency of a collection method can vary according to the composition of the mosquito species present, mosquito densities, availability of alternative hosts, and city lighting [8], therefore, it is difficult to extrapolate a conversion factor from one local epidemiological situation to another. Thus, as part of the trial, this study was undertaken to evaluate the collection efficacy of light traps, carbon dioxide baited light traps vs. HLC. The objective was to assess which method could be used in place of HLC for routine mosquito surveillance by determining conversion factors between LTC and HLC and between CB-LTC and HLC.

The study area
The study area has been described in detail elsewhere [25,26,30]. In brief, the study was conducted in Edo Kontola, in Adami Tullu district, central Ethiopia ( Fig. 1) during July-November 2014. This time coincides with the major malaria transmission which usually is from September to November. This village was selected based on results from preliminary mosquito collections showing high numbers of mosquitoes compared to other sites [31]. The capital of the district, Batu (formerly Zeway) is located at 7°56′N 38°42′E; 1640 meters above sea level. It is about 160 km south of Addis Ababa on the highway connecting Addis Ababa to Nairobi. The total annual rainfall was about 700 mm in 2013, with peaks in July (250 mm) and August (220 mm). The mean minimum and maximum annual temperatures were 14.5 and 27.7 °C, respectively. Most of the population in the district is living in rural areas in houses made of mud or cement walls and thatched or corrugated iron roofs. Local residents primarily depend on farming, livestock rearing, and fishing for subsistence from Lake Zeway [26].

Study design and mosquito collections
This mosquito sampling method calibration study was conducted simultaneously with monitoring of humanbiting activities of local Anopheles species by HLC alone [30]. The experiment was conducted for 39 nights in a 3 × 3 Latin square randomized design replicated 13 times (cycles) in the period between July and November 2014 (Fig. 2). Three mosquito sampling methods were used:(1) Human landing catches (HLC), (2) CDC light trap beside a human-occupied bed net (LTC) and (3) CDC light trap baited with yeast-generated CO 2 (CB-LTC). Eight local volunteers were trained on how to collect mosquitoes. Three houses of approximately similar size and design were selected. The houses were of traditional style with thatched conical-shaped roofs, circular floors and plastered walls. It was also arranged in such a way that the selected houses were free of cattle, chicken and human occupants on all collection nights.
On each experimental night, four of the eight volunteers were allocated to one of the three houses to perform HLC, while the other four were assigned to sleep next to the traps individually under LLINs in indoor and outdoor venues of the other two houses. The sampling methods were rotated among the houses nightly. Mosquito collections were mainly done for three consecutive nights per week. The traps operated overnight from 19:00 to 06:00 h. Human landing catches were performed for 50 min each hour with 10 min break for the collectors. In a house assigned to HLC, four collectors conducted HLC in two rounds on each experimental night. During the first round, from 19:00 to 24:00 h, one person collected mosquitoes indoors and the other outdoors. From 24:00 to 06:00 h, the other collectors took over and performed the same activities. Collectors sat on chairs indoors and outdoors with their legs exposed; the outdoor collector was positioned at least 10 m from the house. Using flashlights, collectors caught landing mosquitoes with a hand-held mouth aspirator and each hour's collection was kept separately in labelled paper cups.
Indoor and outdoor collectors changed venues during the 10 min break whereas the two groups of collectors changed for pre-and post-midnight shifts alternately each night, i.e. the group that collected during pre-midnight hours worked during the post-midnight period the next night and vice versa. The collectors worked during different times and sites to reduce the effects of a particular site and compensate individual differences in attractiveness of the human baits. At a house assigned to CDC light trap alone, two light traps, one indoor and the other outdoor, were hung at the feet of sleeping volunteers, who were protected by a LLIN. Traps were generally positioned at 1.0 m from the floor or ground, where the outdoor light trap was set 10 m from the outer wall of the house and on the opposite wall from where the indoor light trap was placed.
Two plastic bottles of each 2.5 l volume were used to hold yeast-sugar solution for fermentation and production of CO 2 to be used in CB-LTC. This was made by mixing 17.5 g dry bakers' yeast and 250 g table sugar in 250 ml tap water [24] at ambient temperature 1 h before the set up and operation of the traps on each experimental night. Silicon tubes each with 0.70 mm diameter were fitted through a hole drilled in the screw cap to release CO 2 to the vicinity of the light bulb of the trap. At a house assigned to this collection method, a CB-LTC was set indoors at the feet of a sleeping volunteer who was protected by LLIN similar to LTC as described above. The same procedure was applied for the outdoor CB-LTC, which was placed at 10 m distance from the outer wall of the house.

Mosquito processing
Live mosquitoes captured by the above methods were killed with chloroform. At a field laboratory in Batu town, mosquitoes were first sorted out to Culicines and Anophelines. Adult female Anopheles mosquitoes were further identified by morphological criteria using identification keys [32]. An. arabiensis was confirmed to be the only member of An. gambiae species complex in the study area [25,31] and also in nearby areas [28,29]. In order to determine malaria sporozoite infection rate, the head and thorax of each mosquito (n = 6295) were carefully separated from the abdomen and tested for the presence of Plasmodium falciparum and Plasmodium vivax circumsporozoite proteins (CSP) by the direct enzyme-linked immunosorbent assay (ELISA) [33].

Ethical considerations
The ethical considerations for this study is described in more details in the published protocol [26] and is specifically outlined as follows:  [26]. The protocol for the trial was registered at the Pan African Clinical Trials Registry under the Registration Number PACTR201411000882128.

Information and informed consent
Verbal and written informed consent to take part in the study was obtained beforehand from volunteers for landing catches, who were all more than 18 years old and from house owners using the local Afan Oromo language. For the human landing catches, a separate written informed consent describing the potential risks and benefits of the study was obtained from the volunteers. These volunteers were selected from the study village. The participants were instructed that involvement in the study was voluntary, and that they had the right to withdraw at any time regardless of reason. Assurance was also given that a refusal to participate in this study would not affect their access to services at the health posts in the study villages in the community.

Malaria treatment
Mosquito collectors were trained to collect mosquitoes as soon mosquitoes land and before they bite. To help minimize risk, data collectors for the human landing catches were provided with an appropriate prophylactic drug (Malarone) before the collections. There were no reports on Malarone resistant Plasmodium parasites in Ethiopia. The project provided blood examination and treatment of malaria free of charge for any study participant or householder who fell ill or wished to check himself. The project follows the examination and treatment guidelines as described in the study protocol [26]. Fortunately, none of the mosquito collectors or householders was found parasite-positive during the study period.

Data analysis
Data entry and analyses were performed using the SPSS version 20.0. Data from the HLC were divided by 0.83, i.e. 50/60, to account for the fact that HLC was performed for only 50 min of each hour. The nightly number of mosquitoes (x) caught by each method was transformed to log10 (x + 1), to normalize the distribution. Differences among the sampling methods, collection venues (indoor/ outdoor), dates of collection and mosquito species were evaluated by analysis of variance (ANOVA) and Tukey's Post-hoc test. To determine whether each of the alternative sampling method were correlated with the reference method (HLC), the Pearson correlation coefficients for the relationship among log-transformed catches for each Anopheles species was analyzed. The nightly mosquito catches for each Anopheles species in each alternative method were compared with those of the HLC by a simple linear regression analysis on log-transformed values [34]. The relative sampling efficiency (RSE) was measured as the ratio of the number of mosquito species caught by each alternative method to the number caught by the reference method [34]. To test if the RSEs of LTC and CB-LTC were each affected by mosquito density, the ratios of the numbers of mosquitoes in each alternative method to the number of mosquitoes in HLC [log(HLC + 1)−log(LTC + 1)], was plotted against the average mosquito abundance, calculated as [log(HLC + 1) + log(LTC + 1)]/2 [34]. Results were considered significant at p < 0.05. Mean log ratio and its antilog (geometric mean ratio) was used to estimate conversion factors between each of the alternative traps (LTC and CB-LTC) and the reference method (HLC) for Anopheles species that showed consistent relative sampling efficiencies, i.e. that were not dependent on mosquito density [2].

Anopheles abundance and density
Overall, 7606 adult Anopheles females were collected by the three sampling methods over the 39 trap nights ( Anopheles arabiensis was most abundant in HLC (n = 833, 72.2%) and least in LTC (n = 140, 12.1%). Conversely, Anopheles funestus s.l. was most abundant in LTC (n = 416, 47.1%) and least in HLC (n = 60, 6.8%). All the Anopheles species were most frequent in HLC except Anopheles funestus s.l. All species obtained by HLC were also collected by the other methods. Out of 234 mosquito sampling occasions over 39 nights (indoor and outdoor LTC, CB-LTC and HLC combined), there were five (2.1%) occasions without any Anopheles mosquitoes collected. Of the five zero catches, two occurred in outdoor CB-LTC, and one each in indoor LTC, outdoor LTC, and indoor CB-LTC, respectively. No zero catches occurred in indoor and outdoor HLC.
The mean Anopheles mosquito catches per trap night for each species is given in Fig. 3 There were statistically significant differences among the average number of Anopheles species captured by HLC (F = 38.12, df = 3, p = 0.001), LTC (F = 11.17 df = 3, p = 0.001) and CB-LTC (F = 14.04, df = 3, p = 0.001). Post-hoc analyses showed that HLC yielded significantly There was a weak positive correlation between indoor LTC and HLC for this species (r = 0.31) and the regression slope was not significantly different from zero ( Table 2; Fig. 4a), which means that the RSE of the indoor light traps were not dependent on mosquito density. The correlation between LTC and HLC for outdoor catches was positive (r = 0.38) and significant, but the RSE of light traps was significantly dependent on outdoor abundance (Table 2; Fig. 4b). Significant positive correlation (r = 0.49) was found between indoor CB-LTC and HLC; the RSE was not significantly dependent on mosquito density ( Table 2; Fig. 4c). However, for the outdoor CB-LTC and HLC, the regression slope was significantly different from zero ( Table 2; Fig. 4d) meaning that RSE of outdoor CB-LTC was dependent on mosquito density.

Anopheles pharoensis
For this species, there were no significant correlations between LTC and HLC indoors or outdoors nor between CB-LTC and HLC indoors and outdoors ( Table 2). The RSEs of indoor LTC and outdoor CB-LTC compared to HLC, respectively, were significantly dependent on mosquito density (Table 2; Fig. 5a, d). However, the RSEs of outdoor LTC and indoor CB-LTC compared to HLC, respectively did not depend on mosquito density ( Table 2; Fig. 5b, c).

Anopheles ziemanni
A positive significant correlation was found between LTC and HLC indoors (r = 0.43) as well as outdoors (r = 0.78). Similarly, there were positive and significant correlations between CB-LTC and HLC both indoors (r = 0.63) and outdoors (r = 0.80). None of the regression slopes were significantly different from zero (Table 2; Fig. 6a-d), meaning that the RSEs were not dependent on mosquito density.

Anopheles funestus s.l
For this species complex, all the regression slopes were significantly different from zero, indicating that both LTC: HLC ratio and CB-LTC:HLC ratio were dependent on the mosquito density indoors and outdoors. That means both of the alternative traps were not consistent for sampling An. funestus s.l. (Table 2; Fig. 7a-d).

Conversion factors
The mean log ratios of LTC and CB-LTC vs. HLC, respectively were negative (Table 3), meaning that the RSEs

Relative sampling efficiency of the light trap with versus without CO 2 bait
Relative sampling efficiency of LTC versus CB-LTC is given in Table 4. For An. arabiensis, there was a weak positive (r = 0.31) correlation between LTC and CB-LTC indoors but there was no significant correlation outdoors. The regression slopes were neither significantly different from zero indoors nor outdoors meaning that the RSEs were not dependent on indoor and outdoor An. arabiensis density in this setting. For An. pharoensis, the correlations between LTC and CB-LTC, both indoors and outdoors, were not significant. The RSEs of LTC and CB-LTC were significantly dependent on mosquito density

Malaria infection
A total of 4000 An. ziemanni, 1153 An. arabiensis, 342 An. pharoensis and 800 An. funestus s.l. were tested for the presence of CSP of P. falciparum and P. vivax. However, none was found positive. For this reason, the entomological inoculation rate (EIR) could not be determined and compared among the sampling methods.

Discussion
The ultimate aim of this study was to determine reliable conversion factors between either light trap (LTC) alone or light traps baited with yeast-generated carbon dioxide (CB-LTC) both set beside occupied long-lasting insecticidal nets against human landing (HLC) for entomological monitoring of the impact of malaria control interventions. The results showed that the HLC was the most efficient method compared to both LTC and CB-LTC for sampling the majority of Anopheles species including the major malaria vector, An. arabiensis. Despite lower relative Anopheles sampling efficiencies of both LTC and CB-LTC compared to HLC, they can be used as alternative to indoor HLC of An. arabiensis. It was estimated that on average, indoor LTC caught 0.35 times the number of An. arabiensis as compared Relationship between relative sampling efficiency of indoor (a) and outdoor (b) LTC (upper panels), indoor (c) and outdoor (d) CB-LTC (lower panels) and abundance of An. funestus s.l.. Relative sampling efficiency is the difference in the mosquito catches by either of the alternative methods and the HLC (y-axis). The mosquito abundance is the joint average of each alternative and the reference method (x-axis)

Table 3 Mean log ratios and corresponding geometric mean ratios of alternative mosquito collection methods (light traps, LTC and CO 2 -baited light traps, CB-LTC) against the reference method (human landing catches, HLC) for sampling Anopheles species in Edo Kontola, Ethiopia
a Negative mean log ratio indicates that the efficiency of LTC or CB_LTC were lower than HLC  [7,13,15,18,35] which support that HLC is the most efficient sampling method for anthropophilic Anopheles mosquitoes and for routine monitoring of malaria vectors. In contrast to these results, several studies [2,4,11,35,36] showed higher sampling efficiency of LTC compared with HLC for different Anopheles species including An. arabiensis. This was particularly so in Ahero, Kenya [13] and in Macha, Zambia [27] where An. arabiensis was the sole An. gambiae sibling species as is the case in the present study. These observed differences in sampling efficiencies could be explained by local variations in host-seeking behaviours of An. arabiensis across Africa [37]. Further possible reasons could be attributed to the crude nature of both sampling methods due to lack of operational standard procedures regarding trap placement, operation time, in real world settings [8]. To make more valid comparisons these procedures should be standardized.
Moreover, the high efficiency of HLC versus LTC for sampling host-seeking An. arabiensis reflects the basic differences between the two methods. In the case of HLC, a cocktail of stimuli that attract host-seeking mosquitoes such as olfactory, visual cues, volatiles, body heat and humidity are present [38]. Mosquitoes respond to such stimuli and can target the appropriate site for taking a blood meal. By contrast, LTC use mainly visual stimuli. Further, although the presence of human-occupied LLINs besides LTC is expected to augment the trap catches, the excito-repellent effect of the net might have decreased the efficiency of the LTC [39]. However, some studies have shown that using LLINs have little or no impact on the efficiency of LTC [14,40].
The present results also indicate that the correlation between LTC and HLC for outdoor An. arabiensis was statistically significant, but the RSE of LTC was significantly dependent on mosquito density. Such results were expected because An. arabiensis can have more diverse alternative hosts outdoors than indoors which might have diverted more host-seeking individuals from outdoor light traps to human and other animal hosts. An. arabiensis is known to be flexible in host-preferences and indoor/outdoor feeding based on availability of domestic animals [41,42]. Furthermore, feeding behaviour of An. arabiensis can be influenced by availability of cattle in the homestead [43,44]. There were plenty of cattle in Edo Kontola and this might be a potential cause for poor performance of the LTC in outdoor situations. Similar to the present results, some previous reports indicate that light traps were less efficient outdoors [45].The present findings, therefore, suggest that light traps may not be a reliable alternative to HLC for sampling An. arabiensis outdoors in this setting.
Results also revealed that CB-LTC was less efficient than HLC. It was estimated that on average the RSE of indoor CB-LTC was 0.44 times that of indoor HLC for An. arabiensis. The RSE of the trap was not significantly dependent on indoor density of An. arabiensis. This finding is in line with a study on Anopheles aquasalis in Suriname that showed high efficiency of HLC compared to carbon dioxide baited traps [46]. Yeast-produced CO 2 was originally developed to be compared with the standard and industrial mosquito attractants specifically CO 2

Table 4 Correlation and regression analysis of log-transformed indoor and outdoor light trap catch (LTC) with yeast-generated Co 2 -baited light trap catches (CB-LTC) of Anopheles species in Edo Kontola, Ethiopia 2014
The correlation coefficients show the relationship between log(LTC + 1) and log(CB-LTC + 1). The regression slopes are from regressing relative sampling efficiencies (log(CB-LTC + 1)−log(LTC + 1)) on average abundance ([log(CB-LTC + 1) + log(LTC + 1)]/2) n sample size, r Pearson's correlation coefficient, b regression slope, CI confidence interval, t t-test value, p probability value from dry ice, pressurized gas cylinders or propane as cheaper and more accessible alternative in remote localities [24]. As a result, most existing evidence show the efficacy of traps baited with yeast generated CO 2 versus traps baited with the standard attractants [23,24]. Smallegange et al. [23] reported that traps baited with yeast-produced CO 2 caught similar number of An. arabiensis as traps baited with the standard industrial CO 2 and addition of human odour increased the trap catches. Based on this, it can be recommended that yeast generated CO 2 is a promising alternative to HLC and standard mosquito attractants for indoor collection of An. arabiensis. Further studies are required to optimize the efficacy of CB-LTC, industrial CO 2 baited traps and the HLC in Ethiopia and elsewhere.
For the outdoor collection of the same species, there was consistent correlation between CB-LTC and HLC. However, the RSE of CB-LTC was significantly dependent on An. arabiensis density. This could be attributed to the diverse alternative hosts that are available in outdoor settings. Further, outdoor environmental factors such as temperature, humidity and wind speed might affect fermentation of yeast-sugar solution and hence the trap efficacy. In addition, persistence, flow rate and impact radius (attractive range) of the CO 2 volatile have not been optimized and warrant further study.
For An. pharoensis, although outdoor LTC captured 0.24 times that of outdoor HLC and indoor CB-LTC caught 0.86 times that of indoor HLC, the RSE of this species in indoor LTC and outdoor CB-LTC were affected by the mosquito density. These collection methods may not be appropriate for estimating reliable An. pharoensis human biting rates. This might be attributed to exophilic and zoophilic behaviour of this species [25]. The presence of cattle in the surrounding area might have affected comparison of the sampling methods for this species. Therefore, further studies should consider animal baited traps to assess the impact of cattle on the efficacy of the sampling methods.
However, for indoor and outdoor An. ziemanni catches there were significantly consistent relationships between HLC and either of the two methods, respectively. It was estimated that on average, the efficiency of indoor LTC was 0.73 times that of indoor HLC and the corresponding outdoor LTC was 0.39 times that of outdoor HLC. The RSE of LTC was not significantly dependent on either indoor or outdoor An. ziemanni density. Based on these results it can be suggested that despite relatively low efficiency of LTC for collecting An. ziemanni indoors and outdoors, LTC can be used to determine reliable conversion factors for estimating human biting rates for this species.
Likewise, the relationship between CB-LTC and HLC for indoor and outdoor An. ziemanni catches was statistically significant regardless of its density. On average, indoor CB-LTC yielded 0.58 times the number of An. ziemanni captured by indoor HLC, whereas outdoor CB-LTC caught 0.36 times the number of the species collected by outdoor HLC. Though An. ziemanni is known to feed predominantly on cattle in Ethiopia [31,47,48], the present results clearly showed its anthropophilic tendencies as captured by HLC. These contrast some studies [48] which support that CO 2 attract more zoophilic and opportunistic anopheline species than anthropophilic ones.
For An. funestus s.l., there was no consistency between either LTC or CB-LTC compared to HLC indoors and outdoors. The RSE of both methods were significantly dependent on the mosquito density. This means that both methods may not be suitable for collection of hostseeking An. funestus s.l. as an alternative to HLC in this area. This is in contrast to some studies [13] that found consistent proportionality between LTC and HLC for An. funestus s.l. and the recent analyses [8] that showed that LTC were able to collect similar or proportional number of An. funestus s.l. with HLC in Africa. The differences might be attributed to geographical and ecological variations [27] coupled with variations in behaviour of the subspecies in the An. funestus group.
Finally, although the main objective of this study was to estimate the RSE of either LTC or CB-LTC against HLC and hence determine conversion factors for effective monitoring of the impact of IRS and LLINs interventions, the RSE of LTC against CB-LTC were also compared. The correlation between LTC and CB-LTC indoors for An. arabiensis was weakly positive regardless of mosquito density. In outdoor venues there was no significant correlation for An. arabiensis catches using these two methods regardless of mosquito density. Based on these results it is suggested that CB-LTC does not substantially improve sampling of this major vector compared to LTC in this setting for both indoor and outdoor venues.

Conclusions
Results showed that mosquito collection efficiency of the sampling methods varied by Anopheles species. The HLC was more efficient than either of the alternative methods (LTC and CB-LTC) for sampling An. arabiensis, the major malaria vector in the study area. However, the RSE of either of the two alternative methods were consistent and comparable with HLC for monitoring An. arabiensis indoors, but not outdoors. Therefore, CDC light traps with or without yeast-produced CO 2 represents an alternative to HLC for large scale indoor An. arabiensis surveillance and monitoring because of the various problems associated with using HLC. Adding yeast-produced CO 2 to light traps does not seem to improve the sampling effectiveness of these traps.