Malaria trends in Ethiopian highlands track the 2000 'slowdown' in global warming

Global warming is expected to promote malaria transmission in highlands. Highlands, at the edge of the disease's geographic distribution, experience seasonal and intermittent large epidemics, characterized by unstable transmission and low population immunity. This makes them ideal for observing the impact of climate conditions, control efforts, and other factors like immigration from lower endemic areas. While evidence exists linking warmer temperatures to increased malaria in East African highlands from the 1980s to 1990s, the subsequent decline in incidence challenges the importance of climate forcing versus control efforts. The complexity of malaria transmission necessitates a quantitative assessment of temperature's integrated effects, as lab findings on vector and parasite responses to temperature cannot be simply extrapolated. The shift from increasing to decreasing malaria incidence around 2000 provides an opportunity to examine the role of climate forcing based on epidemiological patterns, considering whether climate changes preceded and facilitated the impact of public health interventions. The turnaround in malaria incidence could be strongly linked to temperatures, potentially reflecting the transient "slowdown" in global warming. This slowdown, initially called the "global warming hiatus," involved a period of relatively stable global mean surface temperatures (GMST) from 1998 to 2005. Debate arose regarding the causes, with discussions focusing on Pacific Ocean internal variability (ENSO and PDO), land mass temperature trends, volcanic activity, and heat storage in deeper ocean layers. Ultimately, the slowdown was explained as a redistribution of heat within the Earth's system, with increased ocean heat uptake following a strong El Niño in 1998. This decadal variation in GMST resulted from regional differences and the dynamic contributions of major oceanic drivers. The timing of this global warming slowdown coincided with the observed decrease in seasonal malaria epidemics in East Africa, prompting investigation into their link. This study examines the connection between the reversal in malaria's decadal trend in Ethiopian highlands and the slowdown in global warming, considering its regional manifestation and the roles of ENSO and PDO. Using extensive malaria case records (1968-2007) for *Plasmodium falciparum* and *Plasmodium vivax* in Oromia, Ethiopia, where widespread public health interventions began only around 2004-2005, the study analyzes weather station data and explores connections between malaria cases, regional climate, and global climate variability. An atmospheric model prescribes ENSO and PDO variability to investigate their effects. A process-based transmission model for *P. falciparum* is employed to predict the impact of temperature changes on malaria cases in the absence of enhanced public health interventions. The findings would demonstrate strong coupling between malaria and climate, possibly preceding and facilitating public health efforts.

The study draws on previous research highlighting the complex relationship between climate change and malaria transmission in highland regions. Earlier work by Pascual et al. (2006) and Alonso et al. (2011) explored the association between temperature trends and malaria resurgence in East African highlands. Chaves and Koenraadt (2010) reviewed the debate surrounding climate change and highland malaria, while Stern et al. (2011) examined temperature and malaria trends in the region. Shanks et al. (2005) and Omumbo et al. (2011) focused on malaria in Kenya's Western Highlands. Siraj et al. (2014) investigated altitudinal changes in malaria incidence in Ethiopia and Colombia. The literature also includes discussions about the "global warming hiatus" and its potential causes, involving studies on the role of ocean heat uptake (Meehl et al., 2011), Pacific Ocean variability (Kosaka & Xie, 2013), and other factors. The study's methodology builds upon previous work on climate modulators of East African climate and malaria (Nicholson & Kim, 1997; Reason & Rouault, 2002; Anyamba et al., 2001). The use of a stochastic transmission model draws from established literature on malaria modeling (Caminade et al., 2014; Mordecai et al., 2019; Shapiro et al., 2017; Waite et al., 2019) and extends existing models to incorporate specific factors related to the study region and the observed data.

The study used malaria case data from Debre Zeit, Oromia, Ethiopia (1968-2007) for *Plasmodium falciparum* and *Plasmodium vivax*, confirmed through microscopy. Regional climate data (temperature and rainfall) were obtained by averaging data from eleven meteorological stations closest to the study site, ensuring robustness and accounting for inconsistencies. The data were validated by comparing it with a composite from all 24 stations in Oromia. Global climate data, including El Niño (Niño 3.4 index) and Pacific Decadal Oscillation (PDO) indices, were obtained from NOAA and CRU. Singular Spectrum Analysis (SSA) decomposed the time series into seasonal, interannual, and interdecadal components, enabling the examination of covariation at different timescales. Scale-Dependent Correlation Analysis (SDC) identified spatial correlations between malaria cases and global sea surface temperature anomalies (SSTa) for specific periods. A stochastic transmission model for *P. falciparum* was developed, incorporating seasonality, temperature effects in two crucial windows preceding transmission seasons, environmental noise, and a measurement model (negative binomial distribution) to account for under-reporting and error. The model, based on a compartmental SEIR structure with an additional asymptomatic class, was fitted to data from 1980-1999 using a sequential Monte Carlo method (Iterated Filtering) to obtain Maximum Likelihood Estimates (MLEs) for the parameters. Post-2000 predictions were generated through simulation, initializing with estimates from the filtering algorithm. The model's robustness was assessed by fitting an alternative model with a fixed incubation period. Finally, Community Atmosphere Model version 5 (CAM5) simulations were used to investigate the influence of ENSO and PDO on regional climate before and after 2000, comparing their correlation with Niño3.4 and PDO indices.

The study revealed a strong concordance between long-term variations in minimum temperature and malaria incidence (r > 0.95, p < 0.01), suggesting a significant role of temperature in the observed turnaround of epidemic size around 2000. A process-based transmission model, driven solely by temperature changes and assuming no enhanced public health intervention post-2000, accurately predicted the decline in malaria cases until 2004. This suggests that changes in temperature facilitated, and possibly partially drove, the observed decline in malaria epidemics before the implementation of substantial control measures. SSA analysis confirmed the covariation between malaria cases and minimum temperature at multiple temporal scales (seasonal, interannual, and interdecadal). The interannual components of malaria cases and temperature covaried with the PDO index, indicating the influence of large-scale climate variability. SDC analysis identified significant correlations between malaria cases during the 1997/1998 El Niño event and Pacific Ocean SSTa, indicative of the effects of ENSO and PDO. The atmospheric model (CAM5) simulations supported the role of ENSO and PDO in modulating East African climate, with the influence of ENSO weakening and that of the PDO strengthening after 2000. This aligns with the findings from the time-series analyses. The analyses of *P. vivax* yielded similar results despite its different epidemiology. The model accurately predicted the observed decline in malaria incidence even though it didn't account for the weak interventions in place before 2005. The study ruled out drug resistance as the primary driver of the observed decline, due to the similar patterns in *P. falciparum* and *P. vivax* incidence.

The study's findings demonstrate a strong coupling between malaria dynamics and climate variability in the Ethiopian highlands. The close tracking of malaria trends with regional temperatures, particularly the accurate prediction of the post-2000 decline by a temperature-driven model, underscores the importance of climate as a driver of malaria transmission. This suggests that the observed reduction in malaria cases around 2000 was at least partially driven by the slowdown in global warming and associated changes in regional climate variability, acting synergistically with public health interventions. The similar responses of *P. falciparum* and *P. vivax*, which have different epidemiological characteristics, further supports this finding and rules out other possible factors like drug resistance as the primary driver. While the study isolates the effect of temperature, other factors like land-use, migration, and socioeconomic conditions undoubtedly play a role in malaria transmission dynamics. The study does not claim control measures were completely absent before 2005 but asks what the impact of temperature would have been if they remained constant. The results highlight the need to consider climate variability when evaluating the effectiveness of malaria control interventions and planning future strategies. The observed impact of climate conditions even after strong control measures were introduced underscores the persistent role of climate in influencing malaria risk.

This study provides strong evidence that the slowdown in global warming around 2000 significantly impacted malaria transmission in the Ethiopian highlands, preceding and facilitating the effects of enhanced public health interventions. The close coupling between climate variability and malaria dynamics underscores the critical need for incorporating climate projections into malaria control strategies. Future research could extend these analyses to other highland regions in East Africa to assess the generalizability of the findings and explore the interactions between climate change, interventions, and malaria transmission in more detail. Further investigation of the evolving roles of ENSO and the PDO is warranted, particularly concerning the weakening influence of ENSO and strengthening influence of PDO after 2000. The results emphasize the crucial interplay between environmental factors and public health interventions in the fight against malaria.

The study focused on a single region in Ethiopia and its findings might not be universally applicable. The transmission model was parameterized for *P. falciparum*, potentially overlooking some complexities of *P. vivax* transmission dynamics. While the study accounted for potential biases in malaria reporting, the possibility of residual confounding factors remains. The analysis primarily focuses on the impact of temperature, with less focus on the nuanced roles of rainfall and other factors that may influence malaria transmission. The study's timeframe might limit the assessment of long-term trends and the long-term impact of the PDO.