Near-present and future distribution of Anopheles albimanus in Mesoamerica and the Caribbean Basin modeled with climate and topographic data
© Fuller et al.; licensee BioMed Central Ltd. 2012
Received: 5 March 2012
Accepted: 30 April 2012
Published: 30 April 2012
Anopheles albimanus is among the most important vectors of human malaria in Mesoamerica and the Caribbean Basin (M-C). Here, we use topographic data and 1950–2000 climate (near present), and future climate (2080) layers obtained from general circulation models (GCMs) to project the probability of the species’ presence, p(s), using the species distribution model MaxEnt.
The projected near-present distribution parameterized with 314 presence points related well to the known geographic distribution in the study region. Different model experiments suggest that the range of An. albimanus based on near-present climate surfaces covered at least 1.27 million km2 in the M-C, although 2080 range was projected to decrease to 1.19 million km2. Modeled p(s) was generally highest in Mesoamerica where many of the original specimens were collected. MaxEnt projected near-present maximum elevation at 1,937 m whereas 2080 maximum elevation was projected at 2,118 m. 2080 climate scenarios generally showed increased p(s) in Mesoamerica, although results varied for northern South America and no major range expansion into the mid-latitudes was projected by 2080.
MaxEnt experiments with near present and future climate data suggest that An. albimanus is likely to invade high-altitude (>2,000 m) areas by 2080 and therefore place many more people at risk of malaria in the M-C region even though latitudinal range expansion may be limited.
Malaria is an important vector-borne disease in Latin America and the Caribbean (LAC) with some 775,000 cases reported in 2007 in the region . The majority of cases occur in the Amazon basin, but cases in the non-Amazonian areas of Latin America account for some 40% of the total in LAC . Malaria remains problematic in some countries in Mesoamerica and the Caribbean (M-C) including Haiti (2.8%), Guatemala (3.8%), Honduras (1.5%) and Panama (0.4%) and parts of Mexico (0.3%) [1, 2]. Some 90 different anopheline species have been reported in LAC and nine, including Anopheles albimanus, are considered dominant malaria vectors . An. albimanus is also considered one of the major vectors of human plasmodia in Mesoamerica, northern South America and the Caribbean . Recent outbreaks of autochthonous malaria in the Cayman Islands, the Bahamas and Jamaica have been attributed to this vector . An. albimanus is widespread, and reports indicate that it exploits a diverse range of habitats and is both anthropophilic and zoophilic in its feeding behavior . Moreover, its larvae can tolerate a wide range of water quality conditions, including rice fields, flowing streams, river margins and brackish water associated with mangrove swamps . While the species has normally been found in low-lying areas, usually in locations less than 500 m above sea-level, several reports indicate its presence at higher elevations (up to 1,941 m) [4, 6]. Given its apparently broad tolerance limits, its importance in malarial transmission, its broad spatial extent and ample data related to its location, we selected this particular species for our analysis. Moreover, as a generalist species, it may be more likely than more specialist anophelines (e.g., An. bellator, which breeds in bromeliads) to invade suitable habitats in mid-latitude or high-altitude locations after decades of atmospheric warming and therefore accurate information on its potential distribution can inform current and future control efforts aimed at limiting its spread.
The distribution limits of many malaria vectors are clearly associated with climate conditions, particularly temperature and precipitation. Temperature ranges are important at each stage of the mosquito lifecycle and influence vectorial capacity [7, 8]. Water availability for breeding sites is also fundamental to the survival of malaria vectors and may include, permanent and temporary water bodies, which are generally associated with rainfall timing and amount; and in some anophelines temporary water bodies may provide critical breeding sites in dry environments . Therefore, while temperature can be considered to exert a direct effect on mosquito development, rainfall timing and amount function as indirect environmental factors that condition the environment in ways that affect the fitness of mosquitoes. Elevation functions analogously in that it controls climate conditions at relatively local scales such that elevational gradients in montane environments function much the same as latitudinal gradients, while slope and aspect may be associated with anopheline habitats at local scales within particular landscapes . Moreover, one recent study showed that temperature, rainfall and altitude can explain much of the variance in the distribution of An. albimanus in LAC .
The mapping of anopheline presence using species distribution models (SDMs) offers many potential advantages over point-based maps that indicate where species have been recorded. While dot maps provide a general indication of where a particular species may be found, SDMs also furnish a means to interpolate between collection points as well as extrapolate outside known bounds. Further, SDMs provide a way to improve disease risk mapping by revealing how environmental changes are likely to affect the future distribution potential of species such as An. albimanus. Such models typically use geo-referenced collection points in conjunction with environmental covariates that function as predictors to project a potential distribution or probability of presence . To date relatively few studies [12, 13] have employed SDMs to ascertain the distribution limits of malaria vectors. Moreover, we are unaware of any published studies that have used SDMs to project future anopheline distributions in the Western Hemisphere based on outputs from general circulation models (GCMs), which predict future temperature and rainfall conditions associated with a range of future socio-economic scenarios .
MaxEnt results for different experiments that yielded relatively high mean probability values (~0.70 or greater) evaluated using 36 independent test points
MaxEnt Experiment or Crosstabulation (CT)
p(s) > 0.50 (eo, ec)
Elevation (LPT) E max (m)E mean (m)(sd)
Elevation p(s) > 0.50 E max (m)E mean (m)(sd)
Tmin, Tmax, PPT
1936 165 (797)
1886 147 (850)
Tmin, Tmax, SRTM
2208 197 (764)
1832 140 (778)
Tmin, Tmax, SRTM, PPT
2253 234 (322)
2253 239 (319)
Tmin, Tmax, SRTM, PPT Handey
2906 435 (541)
2258 267 (346)
Tmin, Tmax, SRTM, PPT CSIRO
2531 256 (352)
2253 248 (322)
Tmin, Tmax, SRTM, PPT CCCMA
2174 222 (302)
2174 221 (303)
CT Exp 56, 58, 61
1937 210 (277)
CT Exp 68, 69, 70
2118 231 (307)
Our model experiments with MaxEnt provide new insights into how the distribution of An. albimanus may change by 2080 as a result of climate change. The projected areas of range expansion appeared to be spatially coherent and proximate to projected areas of near-present range, which is to be expected. Although we expected that MaxEnt would project a northern expansion of An. albimanus range into parts of the Gulf Coastal region and South Florida by 2080 with warmer conditions, the model experiments for the most part did not reveal this. The lack of predicted range expansion is at odds with a study by Rogers and Randolph , which suggested an increased area of malaria potential in the southern United States by 2050 as a result of climate change. Of course, many other vectors may be implicated in future malaria outbreaks in the mid-latitude portions of our study area including dominant vectors such as An. aquasalis An. quadrimaculatus An. pseudopunctipennis An. darlingi An.vestitipennis, An. nuneztovari, which are all reported to be present in Central and northern South America [3, 8].
Consistent with general understanding of how climate relates to mosquito fitness, future climate changes that involve large increases in temperature (e.g., > 6°C) and much drier conditions tended to result in lower p(s). A small projected range contraction of approximately 80,000 km2 (or about 6.3%) suggests that overall climate conditions will become more unfavorable to An. albimanus in certain parts of the region. We interpret the potential decrease in p(s) as a function of drier and warmer conditions in parts of northern South America and Central America associated with El Niño events , which several GCM experiments suggest may become more frequent by the end of this century . The potential highland expansion of An. albimanus by 2080 was also expected and consistent with one other study that showed that populations of this species are encroaching into higher altitude regions in Ecuador, currently reaching 1,541 m there . The fact that MaxEnt projected a near-present maximum elevation of 1,937 m and that the Emax has been reported to reach 1,941 m in the M-C region  provides additional confidence in our MaxEnt results. Within the M-C region, approximately 354,000 km2 (or about 9.6% of the study area) is 2,000 m or greater in elevation, with most of this high-elevation terrain in Mexico, Guatemala, Costa Rica, and the northern Cordilleras of Andes in Colombia and Venezuela. It is worth noting that approximately 47 million people live in areas of 2,000 m or greater in these countries , and many more live in potentially high-risk areas in the Andean region to the south. Therefore, the approach we used here with MaxEnt may be applicable to many other areas where anophelines are likely to colonize high-elevation environments, such as East Africa .
The fact that MaxEnt projected near-present occurrence of the vector in Trinidad and other portions of the southern Caribbean islands where it is generally absent  suggests that the outputs from our experiments represent potential as well as actual distribution. In this sense, relatively high p(s) values obtained from climate-based MaxEnt experiments should be interpreted with some caution as the probability surfaces may be better construed as areas where the vector is likely to find suitable habitat if a population becomes established. Given the movement of people and goods within the region and the transportation connections between countries and islands, it is possible to imagine many potential An. albimanus invasions of insular locations from nearby mainland environments. However, such factors as competitive exclusion, existing vector control measures and variable socio-economic linkages between sources and sinks (i.e., potentially suitable areas for colonization) may limit the chance establishment of breeding populations. Indeed, socio-economic, land cover, and human population can be factored into future experiments. However, in this study, we constrained our analysis to climate and topographic covariates because these are relatively static over decadal time scales and we could design model experiments so that they were directly comparable between the near-present and 2080 conditions. Owing to the relatively coarse 2.5 arc-minute resolution of our data, we did not explicitly examine interactions between slope, aspect and rainfall variability, which can be readily discerned at finer resolutions and at more local scales of analysis. However, future work conducted at higher spatial resolutions and at national-to-subnational scales may benefit from use of higher resolution data as well as incorporation of topographic wetness indices, which have provided meaningful predictors of anopheline presence in other studies . Further, it may be possible to use MaxEnt and other SDMs with future human population and land cover projections in such a way as to reduce any “false alarms” or commission errors such as the one noted for Trinidad. In addition, among different SDMs, MaxEnt is potentially better at spatial interpolation than extrapolation, so other SDMs such as GARP may be better suited for projecting future latitudinal expansion of the vector’s range . For example, Townsend Peterson  used GARP to project future distribution of the An. gambiae complex and An. arabiensis in Africa and he demonstrated that model’s utility for projecting future distribution of these vectors. Interestingly, this study also suggested that future climate scenarios will result in net declines in the areas exposed to these important vector species.
Our climate-based investigation adds to the small group of published studies [12, 13, 22] that have used SDMs to project the ranges of anopheline mosquitoes and is among the first to use SDMs with different GCM outputs to project future ranges. We suggest that the results have the potential to inform current and future integrated control strategies, at broad scales, and are particularly applicable to further investigation of how elevation-climate linkages may control the spread of anopheline vectors of human malaria.
where f1(z) is the probability density of covariates across a landscape at known species locations and ƞ(z) = α + β·h(z). α is a normalizing constant that ensures f1(z) sums to 1 and β is an estimated parameter that weighs the contribution of each covariate using a log likelihood approach . For a complete explanation of MaxEnt, readers are referred to Elith et al. .
Mosquito presence data
We utilized three sources of presence data including points included in The Global Diversity Information Facility (http://www.gbif.org), MosquitoMap (http://www.mosquitomap.org) and published data from mosquito surveys in Colombia . The presence data included only collection records from 1950–2000 to coincide with the time period represented by the near-present climate layers. We gridded presence points to 2.5 arc-minute spatial resolution, which yielded 350 unique An. albimanus pixels throughout Caribbean Basin and northern South America, a sample size sufficient to model the species’ potential distribution based on environmental conditions at locations of known occurrence .
Model experiments and threshold selection
We conducted 70 different MaxEnt experiments using different combination of monthly minimum and maximum temperature, monthly precipitation and SRTM layers. An experiment was defined as a unique set of environmental covariates in which different combinations of temperature, rainfall, elevation, and bias layers were input to the model. Each set of twelve monthly Tmax, Tmin, and rainfall surfaces from the WorldClim data was treated as single covariate to include the full seasonal influence of these climate drivers (Table 1). Although MaxEnt can accept an unlimited number of covariates, including nominally scaled variables such as land cover, we restricted model experiments to covariates that are either expected to remain static, such as topography, or sets of climate covariates that are directly comparable between near-present and 2080 conditions. Of the total 350 unique presence points, we randomly selected 36 presence points to independently validate the results of each model experiment. These independent testing points (Figure 1) were used to calculate statistics such as the mean p(s) and standard deviation for each model experiment as well as commission and omission errors to evaluate the accuracy of different experiments. We developed several different bias layers intended to de-bias the data for unrepresentative (i.e., highly clustered) sampling of specimens. After a number of trials with different bias layers, we utilized a bias layer based on examination of the point distribution in Figure 1, which shows a clear tendency for collections near coastlines, as well as informed assumptions based on our own field survey experience; i.e., that the probability of sampling for An. albimanus declines in a negative exponential (i.e., following a decay function) fashion as a function of distance from major roads and coasts. To create this particular bias layer, we used GIS software to generate distance-weighted fuzzy membership layers, scaled from 0–255, with a negative J-shaped function fuzzy function and breakpoints at 5 km for both roads and coastlines. The two fuzzy layers were then combined using a weighted linear combination that assigned equal weights to each layer. We found that this approach consistently improved the mean p(s) values obtained from the validation points relative to other approaches (e.g., fuzzy layers based on distance from rivers, or distance from roads alone). Thus, each experiment was evaluated both qualitatively by comparing the MaxEnt outputs with literature reports from different countries and parts of the region and quantitatively by extracting probability values using the independent validation points selected randomly from the data set.
For further analysis, we selected three model experiments for the both near-present and 2080 projections based on their relatively high mean values (i.e., ~0.70) from the independent test points and to illustrate model sensitivity with inclusion/exclusion of different covariates. We then ran the model with the 2080 climate layers and the SRTM layer to produce future probability surfaces using the same combination of covariates used in experiment 61. We used the lowest presence threshold (LPT) method for setting thresholds to evaluate presence/absence from the MaxEnt experiments. This method uses the lowest predicted value associated with any one of the observed presence points and it can be interpreted ecologically as pixels predicted as being at least as suitable as those where a species’ presence has been recorded . It is thus considered a highly conservative way to map the minimum predicted distribution. In addition, the LPT approach reduces omission error to zero in the training data set . LPT-based maps were then cross-tabulated for near-present and 2080 conditions to indicate areas of model agreement and disagreement. We also applied a logical threshold of p(s) > 0.50 after examining mean values for validation points. We combined different presence/absence maps by cross-tabulating the maps thresholded with the LPT criterion and retaining the area where all three models agreed. This approach of combining different maps provides a conservative estimate of near-present and 2080 model experiments and reduces further the likelihood of false alarms (commission errors).
This investigation received financial support from the NIH (grant 5U19AI089702-02 to SH) with subcontract to DOF and JCB. MLQ was supported by COLCIENCIAS (grant 110134319196). We gratefully acknowledge Paula Pareja for providing georeferenced points of some Colombian records of An. albimanus from the Instituto Nacional de Salud de Colombia. JCB was partially supported by the Abess Center for Ecosystem Science and Policy, University of Miami.
- WHO/PAHO (World Health Organization/Pan American Health Organization): Malaria in the Americas: time series epidemiological data from 2000 to 2007. 2008, Pan American Health Organization, Regional Office for the Americas, Washington DCGoogle Scholar
- Arevalo-Herrera M, Quiñones ML, Guerra C, Céspedes N, Giron S, Ahumada M, Piñeros JG, Padilla N, Terrientes Z, Rosas Á, Padilla JC, Escalante AA, Herrera S: Malaria in selected non-Amazonian countries of Latin America. Acta Trop. 2011, 121: 303-314.PubMed CentralView ArticlePubMedGoogle Scholar
- Sinka ME, Rubio-Palis Y, Manguin S, Patil AP, Temperley WH, Gething PW, van Boeckel T, Kabaria CW, Harbach RE, Hay SI: The dominant Anopheles vectors of human malaria in the Americas: occurrence data, distribution maps and bionomic précis. Parasites Vectors. 2010, 3: 72-10.1186/1756-3305-3-72.PubMed CentralView ArticlePubMedGoogle Scholar
- Montoya-Lerma J, Solarte YA, Quiñones ML, López-Ruiz F, Wilkerson RC, González R, Giraldo-Calderón: Malaria vector species in Colombia – A review. Mem Inst Oswaldo Cruz. 2011, 106 (Suppl. I): 223-238.View ArticlePubMedGoogle Scholar
- Rawlins SC, Hinds A, Rawlins JM: Malaria and its vectors in the Caribbean: The continuing challenge of the disease forty-five years after eradication from the islands. West Indian Med J. 2008, 57: 462-PubMedGoogle Scholar
- Pinault LL, Hunter FF: New highland distribution records of multiple Anopheles species in the Ecuadorian Andes. Malar J. 2011, 10: 236-10.1186/1475-2875-10-236.PubMed CentralView ArticlePubMedGoogle Scholar
- Stresman GH: Beyond temperature and precipitation: Ecological risk factors that modify malaria transmission. Acta Trop. 2010, 116: 167-172. 10.1016/j.actatropica.2010.08.005.View ArticlePubMedGoogle Scholar
- Rúa GL, Quiñones ML, Vélez ID, Zuluaga JS, Rojas W, Poveda G, Ruiz D: Laboratory estimation of the effects of increasing temperatures on the duration of gonotrophic cycle of Anopheles albimanus (Diptera: Culicidae). Mem Inst Oswaldo Cruz. 2005, 100 (5): 515-520. 10.1590/S0074-02762005000500011.View ArticlePubMedGoogle Scholar
- Ageep TB, Cox J, Hassan MM, Knols BG, Benedict MQ, Malcolm CA, Babiker A, El Sayed BB: Spatial and temporal distribution of the malaria mosquito Anopheles arabiensis in northern Sudan: influence of environmental factors and implications for vector control. Malar J. 2009, 7: 123-136.View ArticleGoogle Scholar
- Moss WJ, Hamapumbu H, Kobayashi T, Shields T, Kamanga A, Clennon J, Mharakurwa S, Thurma PE, Glass G: Use of remote sensing to identify spatial risk factors for malaria in a region of declining transmission: a cross-sectional and longitudinal community survey. Malar J. 2011, 10: 163-10.1186/1475-2875-10-163.PubMed CentralView ArticlePubMedGoogle Scholar
- Elith J, Leathwick JR: Species distribution models: Ecological explanation and prediction across space and time. Annu Rev Ecol Evol Syst. 2009, 40: 677-697. 10.1146/annurev.ecolsys.110308.120159.View ArticleGoogle Scholar
- Foley DH, Rueda LM, Peterson AT, Wilkerson RC: Potential distribution of two species in the medically important Anopheles minimus complex (Diptera: Culicidae). J Med Entomol. 2008, 45: 852-860. 10.1603/0022-2585(2008)45[852:PDOTSI]2.0.CO;2.View ArticlePubMedGoogle Scholar
- Foley DH, Klein TA, Kim HC, Brown T, Wilkerson RC, Rueda LM: Validation of ecological niche models for potential malaria vectors in the Republic of Korea. J Am Mosq Control Assoc. 2010, 26: 210-213. 10.2987/09-5939.1.View ArticlePubMedGoogle Scholar
- Intergovernmental Panel on Climate Change (IPCC): Fourth Assessment Report. 2007, Cambridge University Press, Cambridge, UKGoogle Scholar
- Pope KO, Rejmankova E, Savage HM, Arredondojimenez JI, Rodriguez MH, Roberts DR: Remote sensing of tropical wetlands for malaria control in Chiapas. Mex Ecol Appl. 1994, 4: 81-90.View ArticleGoogle Scholar
- Grieco JP, Johnson S, Achee NL, Masuoka P, Pope K, Rejmankova E, Vanzie E, Andre R, Roberts D: Distribution of Anopheles albimanus, Anopheles vestitipennis, and Anopheles crucians associated with land use in northern Belize. J Med Entomol. 2006, 43: 614-622. 10.1603/0022-2585(2006)43[614:DOAAAV]2.0.CO;2.View ArticlePubMedGoogle Scholar
- Darsie RF, Vlach JJ, Fussell EM: New addition to the mosquito fauna of United States, Anopheles grabhamii (Diptera: Culicidae). J Med Entomol. 2002, 39: 430-431. 10.1603/0022-2585-39.3.430.View ArticlePubMedGoogle Scholar
- Rogers DJ, Randolph SE: The global spread of malaria in a future, warmer world. Science. 2000, 289: 1763-1766.View ArticlePubMedGoogle Scholar
- Proveda G, Waylen PR, Pulwarty RS: Annual and inter-annual variability of the present climate in northern South America and southern Mesoamerica. Paleogeogr Paleoclimatol Paleoecol. 2006, 234: 3-27. 10.1016/j.palaeo.2005.10.031.View ArticleGoogle Scholar
- http://sedac.ciesin.columbia.edu/gpw/, accessed 6 February 2012.
- Chaves LF, Koenraadt CJM: Climate change and highland malaria: Fresh air for a hot debate. Q Rev Biol. 2010, 85: 27-55. 10.1086/650284.View ArticlePubMedGoogle Scholar
- Townshend Peterson A: Shifting suitability for malaria vectors across Africa with warming climates. BMC Infect Dis. 2009, 9: 59-10.1186/1471-2334-9-59.View ArticleGoogle Scholar
- Phillips SJ, Anderson RP, Schapire RE: Maximum entropy modeling of species geographic distributions. Ecol Model. 2006, 190: 231-259. 10.1016/j.ecolmodel.2005.03.026.View ArticleGoogle Scholar
- Elith J, Phillips SJ, Hastie T, Dudίk M, Chee YE, Yates CJ: A statistical explanation of MaxEnt for ecologists. Divers Distrib. 2011, 17: 43-57. 10.1111/j.1472-4642.2010.00725.x.View ArticleGoogle Scholar
- Elith J, Graham C: Do They? How do they? Why do they differ? – on finding reasons for differing performances of species distribution models. Ecography. 2009, 30: 129-151.Google Scholar
- Feeley KJ, Silman MR: Keep collecting: accurate species distribution modelling requires more collections than previously thought. Divers Distrib. 2011, 17: 1132-1140. 10.1111/j.1472-4642.2011.00813.x.View ArticleGoogle Scholar
- Pearson RG, Raxworthy CJ, Nakamura M: Peterson Townshend A: Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar. J Biogeogr. 2007, 34: 102-117.View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.