It is possible to differentiate three seasons: one warm and dry, which corresponds to Spring September-December , one warm and rainy, which corresponds to Summer January-April , and one that is cold and humid which corresponds to Autumn and Winter May-August [ 17 ].
The city is crossed by the Tartagal River, and its urban area covers approximately 15 km 2 [ 43 ]. Daily recorded climate data sets for Tartagal city for minimum and maximum relative humidity, rainfall, and minimum and maximum temperature were provided by the National Meteorological Service of Argentina. Rainfall data had too many missing values to perform the analysis, since the applied methodology requires short-time variability that cannot be recovered using seasonal means or data from other localities.
Convergent cross-mapping CCM is a method for detecting causality in nonlinear dynamic systems [ 26 ]. For example, consider two coupled variables, x and y , in a dynamical system, such that x affects y. The shadow manifold M y based on y must map the corresponding contemporaneous value of the shadow manifold M x based on x. CCM determines how well local neighborhoods on a given shadow manifold correspond to local neighborhoods on the other. From the reconstructed shadow manifolds, we can use M y to predict states x p in M x.
Convergence means that the estimates from the cross-mapping improve as we increase the length of the time-series L , because the larger the sample of the dynamics the better the shadow manifold portrays the true attractor. If there is causation, we expect to see convergence, since the correlation coefficient between predicted and observed increases as L increases. Note that the causal variable leaves a signature on the affected one, but not vice-versa.
This means that, if x causes y as above, M y maps onto M x , but not the other way around, because x time-series contains no information about y. In the dynamics of disease transmission, some particularities must be taken into account. For example, if temperature is a causal factor in malaria cases, then it is reasonable to expect that this relationship may not be instantaneous, since mosquito population dynamics and parasite cycle occur on the scale of several weeks.
In this case, temperature may affect malaria cases with a delay, which we address by performing the CCM with increasing time lags between number of cases and temperature, and finding the delay with maximum prediction skill [ 31 ].
Of course, the same holds for other variables. Another relevant aspect in this context is seasonality. Variables such as temperature and humidity are seasonal and so very predictable. Therefore, it is necessary to distinguish between the influence of the season and of the climatic variables themselves. This can be resolved by constructing a null hypothesis with time series surrogates [ 45 ].
For a given variable x t e. These residuals are then shuffled and added back to the seasonal pattern, generating a surrogate series which has the same seasonality as the observed data but with shuffled residuals. Thus, if x t is really causally linked to another variable e. In other words, y t will be sensitive not only to the seasonal pattern but also to the anomalies residuals of the variable x.
We overcome the first issue by reestimating the number of malaria cases using a state space model, described in the next section. The problem with synchrony is dealt with in two parts: first we use a surrogate method, explained above, to make sure there is a causal link. Secondly, we analyze the time lag between the variables—a negative time lag is a reliable indication that the causal direction is correct [ 46 — 48 ].
It consists essentially of the same method of state space reconstruction, but including the causal variables into the state space axes. From that reconstruction, we recover a coefficient that is a proxy for the interaction strength between each driver and the target variable [ 41 ]. The S-map method produces a local linear model C that predicts the value y t from the multivariate reconstructed state-space vector x t as follows: 1. The model C contains the coefficients C 0 , C 1 … C E for each time point t and is obtained from the singular value decomposition solution according to the method presented in [ 41 ].
The coefficients C t are related to changes in the magnitude of causal factors over time and are used here to infer how malaria cases are affected by climate drivers. Epidemiological time series usually show epidemic bursts separated by intervals with no cases recorded. Nevertheless the number of sequential zeros in a time series should be smaller than the dimension of the state-space reconstruction intended [ 45 ].
Moreover, an unknown number of cases are not detected, which introduces observational noise into the series. Observational and process noise can blur the causality links to be detected by CCM [ 44 , 46 ].
To circumvent these problems we applied the CCM analyses to the expected number of recorded cases, which we estimated from the observed time series of number of cases, with minimal assumptions. To do that we fit to the time series a state-space model of count data dynamics in discrete time [ 49 ]. In this model the expected size of the population of infected people is the single state variable, and has a growth rate that can change at each time step. The number of infected people at each time step follow Poisson distributions, which have the expected number of cases as their only parameters.
Finally, the expected number of recorded cases is a fixed proportion of the number of infected people that are detected. This model thus describes the number of recorded cases at each time step as a zero-inflated Poisson variable that evolves freely in time. Under this model, the observed time series of recorded cases is a realization of a stochastic process, from which we estimate the expected trajectory in time.
Therefore, the time series of expected number of cases estimated by this model sorts out observation noise caused by detection failures and also averages out process noise.
We ran four MCMC chains of 3. Chain convergence was checked by the statistic, which had a maximum of 1. Effective sample size of posteriors were Each variable was tested for lagged effects up to 30 weeks, since we intended to capture the causality relationships in a single epidemic and mosquito cycle.
As previously shown [ 31 ], several secondary time-lags may also be causally related as a by-product of the main lag. Within the studied interval 02 Jan — 11 Dec , a total of cases of malaria caused by P. Malaria cases in Tartagal are typically observed from November to April summer season , while during winter July to September they drop to zero over several months Fig 1. The estimated mean weekly number of cases were compatible with the observed counts, ranging from 4.
The number of observed cases were in general larger than the estimated averages, which is expected for count variates with low mean values. All climate variables showed a clear seasonal trend, that express the alternation of a rainy, warm and wet summer December — February to a colder and not so wet winter June — August, Fig 2. We found causal relationships between four climate variables and malaria cases in Tartagal. Maximum temperature was causally linked to expected number of malaria cases 5 weeks later, while minimum temperature, around both 0 and 22 weeks of time-lag, was also important.
Maximum and minimum humidity showed causal links to expected number of cases as well, with a lag of ca. We note that correlation was weak between the climatic variables and number of malaria cases: the correlation coefficients are, at most, of 0.
This means that, in our study, correlation would not be enough to detect the patterns we found. Here we look at each causal variable found above and analyze the effect it has on the number of malaria cases Fig 4. As expected for non-linear dynamics, both the strength and direction of the causalities that we detected changed markedly along the time series, showing that the causal effects can emerge, wane, and even reverse as this complex dynamics unfolds.
Minimum humidity lagged by 13 weeks has a positive effect on number of malaria cases almost all over the time series, but the strength of the interaction decreases with increasing values of minimum humidity Fig 4a , that is, an increase in this variable has a stronger positive effect when minimum humidity is low. In general a rise in temperature causes an increase in the number of cases, although the strength of this effect is contingent on current and lagged temperatures.
The effect of the minimum temperature lagged by 22 weeks, on the other hand, gets weaker with increasing temperature, even though its effect was also usually positive. The thresholds for the interaction strength were checked by comparing the mean effect below and above the assumed threshold using a Mann-Whitney Wilcoxon rank sum test.
We used a causality criterion from dynamic systems theory to identify causal relationships between climatic factors and number of malaria cases in Northwestern Argentina. With this approach we were able not only to quantify causal effects summarized in Table 1 but also to highlight key features of the disease dynamics without the need of an explicit model.
In a nutshell, we show that causal links to climatic variables occur at different time lags with respect to number of cases, and can be interpreted as pertaining to two distinct scales: a longer one relevant for mosquito population dynamics, and a shorter one compatible with pathogen cycle.
These processes can bring on high vector abundance and conditions favoring parasite development, which are sine qua non factors for the transmission of the disease. Number of cases was causally linked to minimum and maximum temperatures lagged by 0 and 5 weeks respectively, with increasing temperature leading to higher number of cases. These short time lags are probably linked to the Extrinsic Incubation Period EIP of the malaria parasite inside the vector, which strongly depends on temperature [ 22 , 53 , 54 ].
Malaria symptoms, and therefore cases, appear approximately one week after EIP completion: some works report a range from 10 to 50 days for EIP [ 9 , 22 ], taking longer for lower temperatures.
In a paper published online yesterday in the journal Science , Pascual and her collaborators looked at how malaria moved up in elevation with temperature in Ethiopia and Colombia.
Tracking year-to-year temperature variations from to , researchers observed how malaria's range shifted. Infection rates tend to increase as temperatures go up, since the Plasmodium parasite that causes the disease reproduces faster inside vector mosquitoes when it's warmer, increasing the infection likelihood when the mosquito bites someone, Pascual explained. The Anopheles mosquitoes that spread the disease also thrive in the heat.
The results confirmed for the first time that malaria creeps uphill during warmer years and recedes as temperatures cool, a dangerous effect as the climate warms. The findings hold promise for better forecasting. In previous work, Pascual found she could predict malaria up to four months in advance in parts of India by monitoring monsoons ClimateWire , March 4, Highland cities could face severe outbreaks He noted that some previous studies suggested that as malaria risk expanded in some areas, it would contract in others, so the net burden of disease would stay the same.
The theory was that some areas would get too hot for the vectors to survive, so the infections would taper off ClimateWire , Oct. They then dissected them and examined their guts and salivary glands, looking for malaria parasites.
Open an infected mosquito and you will find bags filled with replicating parasites. Infected mosquitoes then pass on these parasites to humans when they draw blood. It takes time for the parasites to grow and then make their way to the salivary glands — but not as much time as was previously thought. Thus, if climate change continues to cause cooler regions to get warmer — even just a little bit — mosquitoes will get a jump start toward becoming fully infectious before they die.
Waite is counting on that not happening. Marlene Cimons writes for Nexus Media , a syndicated newswire covering climate, energy, policy, art and culture.
About Films Series News. Credit: Pixabay. Snow, "Climate change and the resurgence of malaria in the East African highlands", Nature 21 : : p. Kuhn, D. Campbell-Lendrum, C. Davies, "A continental risk map for malaria mosquito Diptera: Culicidae vectors in Europe", Journal of Medical Entomology, 39 4 : p.
Kuhn, "Malaria. Menne, K. Ebi Eds. Jetten, W. Martens, W. Takken, "Model stimulations to estimate malaria risk under climate change", Journal of Medical Entomology, 33 3 : p. Rogers, "Changes in disease vector distributions. Hulme Ed. Sutherst, "Implications of global change and climate variability for vector-borne diseases: generic approaches to impact assessments", International Journal for Parasitology 28 : p. Krishnamurthi, A. Chakraborty, V. Mehta, A.
Mehta, "Experimental prediction of climate related malaria incidence", Monsoon and impacts workshop, ; Ahmadabad, India. The development of conflict-sensitive approaches highlights how sustainable development can be made more effective through a consideration of peace and security. Despite numerous challenges and obstacles, which are far greater now than prior to the events of August , there are still many ways for the international community to help education move forward in Afghanistan.
0コメント