We thank Simon Hay and colleagues for acknowledging the uncontroversial evidence for warming trends in East African highlands, and for partially addressing the nature of the discrepancies in our earlier results on temperature trends in these regions along the lines indicated by Randolph (2009) and Omumbo et al. (2011).

It is unfortunate that almost a decade was lost with no clear resolution on an important empirical pattern that could have been clearly recognized by the end of the 1990s. Perhaps some lessons can be learned that can help malaria research avoid unnecessary protracted debates moving forward. One would hope in particular that the open discussion of scientific results can prevail in areas whose relevance goes well beyond science and academia, including the current prevalence maps and their implications for climate change (Bouma et al., 2011; Gething et al.,2010).

Our paper on temperature trends, Pascual et al. (2006), first questioned the validity of the argument put forward by Hay et al. (2002) in Nature; and subsequently repeatedly used by the authors in publications that suggested that climate change should not be considered as a factor behind the observed exacerbation of malaria in these regions in the 1990s as there was no evidence of statistically significant trends in the CRUTS1.0 temperature time series up to 1995 (seven years before the publication in Nature). We pointed out with more robust analyses that the time series did exhibit trends for the same sites when updated to 2002 (that is, in the CRU product known as CRUTS2.1 which followed and superseded CRUTS1.0). We were unable to identify at that time the reason for the discrepancy in the results when the data was considered only up to 1995, and could only offer discussion related to the methods themselves. The evidence for warming trends in the updated CRU products (including the more recent CRUTS3.) has now been confirmed and resolved, similar warming trends have now also been confirmed for local time series from meteorological stations in Kericho (Omumbo et al., 2011; Pascual et al., 2009; Alonso et al., 2010).

As pointed out by Omumbo and colleagues (2011) and Thomson et al. (2011), one central lesson of this whole debate is that the malaria community needs better and easier access to climate and disease data, especially quality-controlled data. A significantly more important lesson is that the whole debate could have been avoided and resolved years ago, in 2005, had there been the willingness to openly discuss the discrepancies in the analyses at the same NCEAS (National Center for Ecological Analysis and Synthesis) meeting on malaria and climate where the discussions were originally raised. This major impediment to scientific progress is an unfortunate reality of a field that has become heavily polarized into those ‘for’ and ‘against’ the influence of climate change on malaria and other infectious diseases (Chaves and Koenraadt, 2010). The conclusions put forward by these opposing sides tend to become influential and highly visible also in non-scientific circles, over different time periods they counter-react and feed on each other. In a complex disease such as malaria, the open discussion and consideration of different results and hypotheses is fundamental.

Had there been a better dialogue years ago we would have found that the discrepancies are not just data related, but also involve conceptual and methodological issues of general relevance to questions on trends and disease impacts; these will be developed in more detail elsewhere. One important issue is that a few apparent differences in the data regarding inter-annual variability (see Figure 2), can have an extraordinary influence in the outcome of statistical tests on trends. We should be concerned that the statistical analyses and the concepts we use for the detection of trends are often highly sensitive to the exact starting and ending times we choose for the time series analyzed, and to the associated differences in the variance of particular years. For example, had Hay et al. (2002) considered only 3 more years, from 1970 to 1998 (instead of 1995), even the Dicker-Fuller (DF) test applied in the original analyses would have detected significant trends (at the 0.05 significance level). Significance would have varied also depending on the starting date, increasing for example for the DF test applied from 1960 to 1995 (p-values<0.05), or 1960 to 1998 (p-values<0.005), and decreasing again if starting from 1950. Moreover, had the two data sets been compared earlier, pathological features in the time period between 1911 and 1940 would have been apparent, making questionable the application of the DF test starting in 1911 (Hay et al. 2002) , or any test applied to a data series starting in the first part of the last century. In particular, CRUTS1.0 exhibits time windows of large and anomalous excursions in the variance of mean temperature, that are no longer present in CRUTS2.1.

These points illustrate the difficulty of addressing trends in time series with many characteristic temporal scales of variation, in which ‘trends’ are intertwined with interannual variability at a number of different scales, in data that suffer from many sources of measurement error. This leads to important considerations on the robustness of the methods and on the concept of trends itself for biological considerations (Pascual et al 2006, Chaves and Koenraadt, 2010).

Methodological issues should not deflect attention however from the central question of whether the observed empirical patterns in temperature have consequences that are biologically significant ( a question we attempted to address in Pascual et al., 2006, and Alonso et al., 2011 with biological models).

Finally, we should not justify the decade-long delay in reaching a consensus on the occurrence of warming trends in E. African highlands, with the argument that this does no longer matter, because by the time we have agreed on what the data tells us about temperature trends, the burden of malaria is now decreasing. The recent decrease of malaria has been reported elsewhere for several highland regions together with its variations across different sites (e.g Chaves et al., 2011 ); and is consistent with the control efforts that followed the increase in incidence of the previous decade, however, it also coincides with prolonged droughts in the region that can also transiently reduce malaria prevalence. Importantly, this does not mean that we do not need to understand the role of climate change and other environmental drivers of transmission risk before control was implemented in the last decade, or to understand the effectiveness of control itself today in the context of climate variability and change. To conclude that we do not need to be concerned with a higher transmission risk because of current control measures would be short-sighted at best, and deflect from our understanding of sustainable control at a critical transition time when regional elimination has not yet been achieved. It is also a statement that ignores much of our ecological and evolutionary understanding on infectious diseases in dynamic environments.

Mercedes Pascual (Ecology and Evolutionary Biology, University of Michigan and Howard Hughes Medical Institute)
and
Xavier Rodó (ICREA Professor and Director, Institut Català de Ciències del Clima, IC3)

References:
Alonso, D. et al 2011. Epidemic malaria and warmer temperatures in recent decades in an East African highland 278:1661-1669. . (doi:: 10.1098/repb.2010.2020).

Bouma B.J. et al. 2011. Global malaria maps and climate change: a focus on East African highlands. Trends in Parasitology 27(10)

Chaves L.F. and C.J.M. Koenraadt. 2010. Climate change and highland malaria: fresh air for a hot debate. The Quarterly Review of Biology 85(1): 27

Gething P.W. et al, 2011. Climate change and the global malaria recession. Nature 465, 342–345

Hay S. et al. 2002. Climate change and the resurgence of malaria in the East African highlands. Nature 415: 905-909.

Omumbo J.A. et al. 2011. Raised temperatures over the Kericho tea estates: revisiting the climate in the East African highlands malaria debate. Malaria Journal 10:12 doi:10.1186/1475-2875-10-12

Pascual M. et al. 2006. Malaria resurgence in East African Highlands: temperature trends revisited. PNAS 103 (15): 5829-5834

Pascual, M. et al. . 2009. Underestimating malaria risk under variable temperatures. PNAS. 106(33): 13645-13646.

Randolph S.E. 2009. Perspectives on climate change impacts on infectious diseases. Ecology 90(4): 927-931.