Climate Change 2001:
Working Group II: Impacts, Adaptation and Vulnerability
Other reports in this collection

9.7.1.1 Modeling the Impact of Climate Change on Malaria

Classical epidemiological models of infectious disease use the basic reproduction rate, R0. This measure is defined as the number of new cases of a disease that will arise from one current case when introduced into a nonimmune host population during a single transmission cycle (Anderson and May, 1992). The basic reproduction rate—or a related concept, "vectorial capacity"—can provide a relative index of the impact of different climate scenarios on the transmissibility of vector-borne diseases such as malaria. Vectorial capacity, however, is determined by complex interactions of many host, vector, pathogen, and environmental factors. Some of the variables are sensitive to temperature, including mosquito density, feeding frequency, mosquito survival, and the extrinsic incubation period (EIP) of the parasite (plasmodium) in the mosquito (Martens et al., 1999). The EIP is especially important, and, within the lower temperature range, it is very temperature-sensitive.

Figure 9-2: Potential impact of climate change on seasonal transmission of falciparum malaria. Output from MIASMA v2.0 malaria model (Martens et al., 1999) indicates the number of months per year when climate conditions are suitable for falciparum transmission and where there is competent mosquito vector: (a) months of potential transmission under current climate (1961-1990); (b) months of potential transmission under a GHG-only climate scenario (HadCM2 ensemble mean) in the 2080s. Future changes in mosquito distributions are not modeled. This model does not take into account control or eradication activities that have significantly limited the distribution of malaria.

Biological (or process-based) models have been used to estimate the potential transmission of malaria. This is a measure of the extent to which the natural world (the global environment-climate complex) would allow the transmission of malaria if there were no other human-imposed constraints on transmission. However, in some areas where human-imposed constraints have occurred as a result of economic growth, or were put in place purposely, malaria transmission has been successfully controlled, regardless of suitable local temperatures. There has been considerable evolution of models since the SAR (Martens et al., 1995, 1997, 1999; Martin and Lefebvre, 1995). One model (Martens et al., 1999) includes vector-specific information regarding the temperature-transmission relationship and mosquito distribution limits. Recent studies using that revised model applied to the HadCM2 climate scenarios project a global increase of 260-320 million people in 2080 living in the potential transmission zone (against a baseline expectation of about 8 billion—that is, a 2-4% increase in the number of people at risk) (Martens et al., 1999; McMichael et al., 2000a). This projection, by design, does not take into account the fact that much of this additional population at risk is in middle- or high-income countries where human-imposed constraints on transmission are greatest and where potential transmission therefore is unlikely to become actual transmission. The model also projects regional increases and a few decreases in the seasonal duration of transmission in current and prospective areas of malaria transmission. Constraining of GHG emissions to achieve CO2 stabilization within the range 550-750 ppm would reduce those projected increases by about one-third (Arnell et al., 2001).

On a global scale, all biological models show net increases in the potential transmission zone of malaria and changes in seasonal transmission under various climate scenarios (Martens et al., 1995, 1999; Martin and Lefebvre, 1995). Some local decreases in malaria transmission also are predicted to occur where declines in rainfall would limit mosquito survival. The outputs of these malaria models are very sensitive to assumptions about the minimum rainfall or humidity levels needed for malaria transmission.

Another global modeling study (Rogers and Randolph, 2000) used a statistical-empirical approach, in contrast to the aforementioned biological models. The outcome variable in this model is the number of people living in an actual transmission zone, as opposed to a potential transmission zone (as estimated by biological models). Using an IS92a (unmitigated) climate scenario, this study estimated no significant net change by 2080 in the portion of the world's population living in actual malaria transmission zones; modeled malaria transmission increased in some areas and decreased in others. This study made the assumption that the actual geographic distribution of malaria in today's world is a satisfactory approximation of its historical distribution prior to modern public health interventions. This assumption is likely to have biased the estimation of the underlying multivariate relationship between climatic variables and malaria occurrence because the sensitive climate-malaria relationship in the lower temperature range in temperate zones (especially Europe and the southern United States) would have been excluded from the empirically derived equation. Hence, the use of that derived equation to predict malaria risk in 2080 would have been relatively inert to marginal climatic changes at the fringes of the current geographic distribution.

Another type of modeling addresses changes in the distribution of mosquito vector species only. The CLIMEX model estimates changes in global and national (Australia) distribution of malaria vectors under a range of climate scenarios, based on the vectors' temperature and moisture requirements (Bryan et al., 1996; Sutherst, 1998). The distribution of Anopheles gambiae complex is projected to undergo a net increase in distribution in southern Africa under three climate change scenarios (Hulme, 1996). However, these models do not address complex ecological interactions, such as competition between species.

None of these models have been adequately validated at global or regional levels. Modeling to date has not satisfactorily addressed regional vulnerability to malaria or changes in risk in highland regions (Lindsay and Martens, 1998). This is principally because it is difficult to obtain sufficiently detailed geographic distribution maps of mosquitoes and malaria occurrence over time. An important criticism of biological models is that undue emphasis is placed on temperature changes, without consideration to other ecological complexities—including those influenced by rainfall, humidity, and host exposure—that influence transmission dynamics. Furthermore, the equations within a global model may be inappropriate for particular local conditions, and there is a need for cross-validation of large-scale and small-scale studies (Root and Schneider, 1995). Some attempts to apply these integrated modeling techniques to smaller scale regional settings have attempted to take account of local/regional conditions (Lindsay and Martens, 1998). None of the modeling to date has incorporated the modulating effect of public health strategies and other social adaptive responses to current or future malaria risk (Sutherst et al., 1998). Nevertheless, it remains a legitimate and important question to estimate, under scenarios of climate change, change in the extent to which the natural world (the global environment-climate complex) would allow transmission of malaria if there were no other human-imposed constraints on transmission.



Other reports in this collection