Does mosquito species diversity peak at intermediate levels of forest disturbance (Connell’s IDH model vs Fox’s criticism)?

In contrast with several previous studies that found patterns supporting the non-equilibrium maintenance of biological diversity3,4,5,6,7,8, Fox14 challenged Connell’s1 IDH model, citing a lack of theoretical support. Connell1 stated that at intermediate disturbance levels, competitive exclusion declines, allowing for stable coexistence in a non-equilibrium state. Huston18 simulated a Lotka-Volterra competition model with mortality events that mimicked the effects of disturbance and supported Connell’s ideas. However, Fox14 first cited the lack of evidence that supports competitive exclusion. Fox14 also argued that the Huston18 model has a subtle outcome. Over the long-term, the increase in average mortality rates cannot produce stable coexistence.

In agreement with the current scientific proposals challenging IDH11,12,13, we did not find a hump-shaped diversity-disturbance relationship using mosquito larvae diversity and forest cover percentage from the lowland tropical rainforest of central Panama. The present test therefore failed to empirically support IDH. Notwithstanding, the main claims of Fox14 pointed to logical flaws in the theoretical rationale of IDH: 1) lack of evidence of competitive displacement, 2) not all species have linear responses to disturbance, and 3) increase in average mortality rates in the long-term, are discussed herein, as follows.

The present study found Connell1 fractions - colonist and climax species. Colonist mosquito species were associated with low forest covers at one extreme of forest disturbance gradient, whereas climax species were found at the opposite end. This result resembles the mechanism of coexistence of closely related species at the landscape scale (although they can be spatially segregated at the habitat scale), proposed by Laporta and Sallum43, after Juliano44.

Similarly, it is shown here that colonist An. albimanus did not co-occur with two climax species, An. oswaldoi and An. triannulatus, at the habitat scale (i.e., larval habitat); although the species coexisted at the landscape scale. It is interesting to note further that both An. oswaldoi and An. triannulatus co-occurred at the larval habitat, but their abundances (n = 278 and n = 138, respectively) were lower than that of An. albimanus (n = 602). This may indicate that An. oswaldoi and An. triannulatus interact at the microhabitat scale43, so that they need to share resources at the surface of the water on those larval habitats where they co-exist (e.g., ponds and stream margins). The partitioning of resources between An. oswaldoi and An. triannulatus is likely due to favourable water pH and temperature conditions in shared larval habitats with high forest cover. However, when this high forest cover setting is disturbed, departing from 100% to 20%, it is plausible to assume that both An. oswaldoi and An. triannulatus (i.e., climax) get displaced45 by An. albimanus (i.e., colonist), an abundant species in colonising disturbed areas46, 47. An illustration of the possible competitive displacement of An. albimanus and An. oswaldoi/An. triannulatus mediated by disturbance was made herein.

Another group of closely related species is made up by the colonists Cx. coronator, Cx. declarator, and Cx. nigripalpus, with Cx. interrogator, another climax species of the same subgenus. These Culex species also showed co-occurrence at the landscape scale and, potentially, competitive displacement. The competitive displacement could occur when an assemblage dominated by Cx. interrogator in high forest cover sites is subjected to disturbance and gets replaced by other dominant Culex (Cux.) species in open areas.

Only 7 species (3 Anopheles, 4 Culex (Cux.)) out of 22 colonist and climax species in our dataset fit with a pattern of competitive displacement. This is partly because semi-permanent water habitats species (i.e., Coquillettidia, Mansonia, and Psorophora) mainly occurred in disturbed areas, whereas tree hole and fallen leaves species (i.e., Ae. terrens, An. eiseni, Li. asulleptus, Li. durhamii, Tr. digitatum) were mostly found in forested sites. For these latter species (e.g., Ae. terrens), a disturbance event cannot produce another pair of coexisting species, because tree holes were absent in low forest cover. In relation to the first claim of Fox14 regarding lack of evidence for competitive displacement, we state that we could find evidence of competitive displacement, but it does not seem to have a large effect.

Fox14 also claimed that not all species have linear responses to disturbance. This claim was supported by our outcomes. Highly abundant species in colonist and climax fractions were found. These species had a linear response to the disturbance gradient. For instance, An. albimanus, Cx. coronator, Ps. cingulata, and Cq. nigricans had peak abundance in the most disturbed areas, and this peak declined towards zero abundance in high forest cover sites. Similarly, Ad. squamipennis, Ae. terrens, An. eiseni, An. trianullatus, and Ur. geometrica had very sharp responses on larval habitats with high forest cover, while their abundances were zero at the other extreme. Notwithstanding, there were other species that did not respond linearly to the disturbance gradient, as claimed by Fox14. The colonist Ma. titillans had one peak of high abundance at low forest cover, which is supported by the study of Alfonzo et al.42, and another with mid-abundance in the middle of the gradient. The climax species Li. durhamii had two peaks as well, a very high one at high forest cover, supported by Suaza-Vasco41, and another with low abundance at intermediate/low forest cover. Additionally, we found five species that can be considered disturbance-generalists, because they had at least two peaks of equal abundances in different forest disturbance categories. This was noticed before for Cx. erraticus by Alfonzo et al.42, where they collected 75% of specimens in open areas and 25%, in tall undisturbed forest habitats. Finally, we also found a mid-disturbance species, Wy. simsi, which had a peak of abundance at the intermediate portion of forest disturbance gradient. The finding of a mid-disturbance species in our dataset is interesting because it means that a community dominated by mid-disturbance species would naturally lead to the expected unimodal diversity-disturbance relationship. Considering specific attributes that enabled species to persist at any given disturbance frequency (e.g., refs 48, 49), our data could support the second claim of Fox14.

The third claim of Fox14 was that disturbance could increase the average mortality rates in the long-term for rare species. The author simulated from a series of population-based differential equations that species with low abundance (rare) were very likely to become extinct in places exposed to frequent disturbances in the long-term. This claim was supported by our data in full. Twenty-six (48%) of 54 species in our dataset were classified as rare species. Each of these species persists in disturbed areas in a source-sink population dynamics kind of fashion, with Barro Colorado Island as the source of their populations and surrounding areas such as PVAS and ACH as the sink, according to MacArthur and Wilson50. Hypothetically, if Barro Colorado Island is exposed to frequent disturbance events, these rare species would probably become locally extinct in the long-term, unless another source for their populations appears.

The fact that we could not empirically support Connell’s1 IDH pattern with our dataset is related to the claims of Fox14. Competitive displacement did not seem to have a large effect. Species had too many peculiar responses to the disturbance gradient, including a group of disturbance-generalists and one mid-disturbance species. In addition, the high number of rare species indicates that they would not coexist in the long-term in a scenario exposed to constant disturbances, because they would probably become extinct.

Mosquito species diversity increased towards relatively undisturbed forest environments

We rejected IDH because mosquito species diversity increased with increasing values of forest cover, rather than peaking in the middle of the forest disturbance gradient. Our results depicted an effect of forest cover on mosquito species diversity, meaning that the number of species increased towards healthier and relatively undisturbed forest sites. Although positive and statistically significant, this effect was small. One analogous example of such a small effect is: an increase of 0 to 50% forest cover would increase species diversity by 0.2. However, all 26-rare species (48% of the whole community) were more frequently found in high forest cover.

Underlying mechanisms that could support higher species diversity at high forest cover settings might be habitat complexity and habitat constraints. The former relates to vegetation complexity, which determines the availability and spectrum of different types of larval habitats, whereas the latter is related to chemical composition of the water (i.e., temperature and pH) in those habitats. Vegetation complexity tends to decrease from climax to colonising forest scenarios51, thus supporting fewer mosquito species along the disturbance axis due to fewer opportunities to develop and coexist. In contrast, habitat constraints tend to increase in the same direction making it harder for climax and rare species to survive in harsh environments51. Fluctuations of habitat complexity and constraints lead to variation in larval habitat properties that are more evident at the opposite extremes of the disturbance gradient.

Moreover, higher habitat complexity in undisrupted forest environments might promote niche differentiation due to a higher variety of larval habitats51, while lower habitat constraints in these settings may also favour optimum values of temperatures and pH to yield a larger number of interacting species as larvae in shared aquatic habitats52. Our findings further suggest that habitat complexity and constraints are shaped by a successional gradient of forest physiognomic stages, translated into different habitat conditions under three discrete categories of forest cover. Differences in mosquito diversity and abundances across a gradient of forest disturbance could be explained by shifting ecological conditions that affect larval breeding site availability and quality.

Limitations of the study

The limitations of the present study can be divided into three categories. Ideally, the best study design would match characteristics to cope with these three limitations: (1) The gradient of forest disturbance in this study was not continuous. We had three categories of forest disturbance: a) low forest cover (PVAS), b) medium forest cover (ACH), and c) high forest cover (BCI). In each category, there was a gradient of forest cover, but data were lacking in-between categories. Having more data to fulfil these gaps could help to have a more accurate picture of the effect of forest disturbance on mosquito diversity. (2) Our study design was not a follow-up of successional stages in a single landscape. As our study was a cross-sectional one, the major assumption is that BCI is an undisturbed scenario for ACH (a mid-disturbance scenario) and PVAS (a high-disturbance scenario). During the study design, these areas were selected in the same ecosystem (lowland tropical rainforest), so that they could hypothetically represent the same temporal process of disturbance. Using a cross-sectional study to test a temporal-dependent phenomenon such as the successional stages of Connell’s1 IDH is not only challenging, but this was the main critique of Sheil and Burslem15, when they replied to the “IDH should be abandoned” by Fox14. According to these authors, replacing space-for-time in such a case would not give the same evidence that Connell1 found in successional stages of forest in Uganda. This critique made Fox16 lower the tone of his statements and agree with the importance of the narrow Connell-based definition of IDH in order to avoid confounding factors. (3) Our study did not experimentally test the effects of forest disturbance on mosquito diversity. It is not trivial to propose an experiment with an intervention such as ‘forest disturbance’. Despite the importance of such an experiment, the opportunities to accomplish this can be very scant. Nevertheless, in places where forest disturbance is out of control, such as in the Brazilian Amazon53, one type of intervention could be to compensate landowners for keeping their forest in good shape, and track changes in the mosquito community over time with nearby territories in which landowners would not receive any bonus.

The ecological process that make risky scenarios for vector-borne disease emergence

While it is true that our results did not provide empirical support for IDH in the field of ecology, using this hypothesis and the criticism around it (i.e., by Fox14), as a framework, allowed us to better understand the role of forest disturbance in mosquito-borne disease transmission. Our considerations here also have implications for disease prevention and control. For instance, Lounibos45 discussed the applications of competitive displacement as practice for targeting vector population reduction. Our outcomes suggest that competitive displacement could be applied in the Republic of Panama as well as more broadly (i.e., Colombia, Peru, and Central America)20, 21, 46, 47 to prevent malaria epidemics. This can be achieved by increasing forest cover recovery in highly disturbed areas (i.e., PVAS and ACH), favouring the likelihood of auxiliary malarial vectors (i.e., An. oswaldoi/An. triannulatus)36, 37 over primary ones (i.e., An. albimanus). This control scheme might also be applied to prevent epidemics of arboviral pathogens involving colonists (i.e., Cx. coronator, Cx. declarator, and Cx. nigripalpus) and climax species (i.e., Cx. interrogator) within the subgenus Culex of Culex.

In terms of disease transmission54,55,56, focusing on Connell’s fractions (e.g., climax and colonist mosquito species) gave us the opportunity to investigate the role of these assemblages on disease emergence in relation to forest disturbance. Changes in biodiversity caused by recent landscape disturbance in Panama, or in any other place of the tropical rainforest domain57, are likely to impact vector species spectra and status over time58, 59. We found that the number of mosquito species being incriminated as vectors of pathogens was significantly higher in the colonist fraction than in the climax fraction. Not only was the association colonist–vector species significant, but also it was the most prominent at highly disturbed forest sites, while its likelihood decreased notably towards relatively undisturbed forest settings.

A positive linear response by colonist–vector species to forest disturbance seems to be a tangible process that might increase disease transmission risk in forest-altered tropical areas58. ‘The colonist–vector fraction’ including the malaria vector (i.e., Anopheles albimanus) and zoonotic and bridge vectors of several arboviruses (i.e., Culex pedroi, Culex nigripalpus, Psorophora cingulata, Psorophora ferox, Mansonia titillans, and Coquillettidia venezuelensis) increased in abundance as a function of forest disturbance. According to this view, the so-called spill-over effect might not be a random process, but rather a consequence of forest degradation leading to a higher probability of contact between colonist–vector mosquito species and humans at forest-altered sites. The proportion of different reservoir/hosts existing in forest degraded areas of Panama, will then determine the ultimate disease outcome there, as most mosquito species in forest-altered sites tend to feed opportunistically and upon what is most available nearby. All together, these results suggest a likely role of forest disturbance into vector-borne disease emergence in recently disturbed tropical regions58, 59, including Panama.

Nevertheless, considering the well-known suite of entomological studies by Dr. Pedro Galindo in the Republic of Panama54,55,56, just to mention a few, and more recently the theoretical insights by Randolph and Dobson60, finding vector species in the colonist and climax fractions should be equally expected. This is theoretically anticipated because female mosquitoes of any particular species have the same opportunities to blood-feed upon vertebrate animals that harbour pathogens. This argument might sound compelling, but it has not been supported by prior studies on vector competence. For instance, Turell by himself26 and with colleagues24, 28, 30 experimentally tested many mosquito species, thought to be competent transmitters, for infection with different arboviral pathogens, but only partial results could be translated into information on vector incrimination. Cohuet et al.61 further stated that pathogens could shape vectors.

If it were true that all mosquito species in the community could transmit pathogens equally efficiently, why are those in the colonist fraction of our study more likely to do so? On the one hand, we could rely upon the work of Myers and Patz58, who stated that forest disturbance selects the most abundant species, which could lead to higher contact rates with domestic animals and humans, thus increasing the odds of pathogen transmission on these hosts. On the other hand, we should also consider random genetic processes (e.g., neutral effects62) as a mechanism that could enhance vector potential at any given time-space. For example, a single mutation event made Ae. albopictus a superior vector of Chikungunya virus during several outbreaks across Asia and Europe, ten years ago63.

Considering the hypothetical synthesis of our work (Fig. 3), a logical statement is that forest disturbance could cause the overall decline of mosquito species diversity, yet at the same time, it could increase disease transmission risk. Although we did not measure disease risk directly, high abundance of colonist–vector species at immature stages was applied here as a proxy for female mosquito population size in forest altered sites. According to this, the deduction would be that one possible mechanism of interconnection of diversity-disease might be related to species interactions in the larval habitat. The authors are conscious of the limitations of the present work, and so the synthetic diagram in Fig. 3 is only intended to guide future research efforts in this direction.

Figure 3 Theoretical synthesis of the results obtained in the present work. The main factor, forest disturbance, decreases habitat complexity and increases habitat constraints. These underlying mechanisms decrease diversity and increase vectors of important pathogens. Full size image

Randolph and Dobson60 made a strong critique about the likely effects of high biodiversity as the buffering element to prevent the emergence of vector-borne disease in relatively undisturbed forest habitats. In our study, colonist–vector species such as An. albimanus (main vector of malaria in Central America20, 21, 46, 47), Cx. nigripalpus (main vector of St. Louis Encephalitis Virus in the US26), and Cx. pedroi (main vector of the Eastern Equine Encephalitis Virus in Peru27) were conspicuous in highly altered forest settings, while their abundances were almost zero at relatively undisturbed forest sites (i.e., old-growth forest). The fundamental conclusion derived from the field of disease ecology is therefore that old-growth forest conservation in tropical rainforests is a reasonable and highly recommended solution for preventing new outbreaks of arboviral and parasitic diseases in anthropic environments.