登革热是一种由伊蚊传播的病毒性疾病，广泛流行于全球热带、亚热带地区。近几十年来，全球登革热发病率急剧上升，已有超过120个国家受到登革热的侵袭，其中亚洲和美洲国家受到的影响最为严重。登革热疫情的大规模暴发给上述国家的人群健康和社会经济带来了沉重负担。由于目前仍缺乏相应的特效抗病毒治疗药物或疫苗，有针对地预防和控制是减少疫情规模和影响的主要手段。分析登革热流行趋势的变化规律，识别登革热流行的风险因素，构建登革热预测预警模型，评估登革热流行的高风险区域，对制定疫情应对方案、保护人群健康与减少经济损失非常重要。

对登革热疫情趋势特征和风险因素的系统研究，是优化登革热传播机制模型的基础，同时也为制定有效的防控策略提供了支持。登革热的暴发流行受到气候变化和人类活动的显著影响。一方面，气候变化能够促进或制约蚊虫的繁殖，影响其对登革病毒的传染性和易感性，从而改变登革热的流行规模。已有研究发现，厄尔尼诺-南方涛动现象通过调节特定地区的气候条件，影响登革热的流行动态，进而能够用于疫情风险的早期预测。然而，现有的登革热预测模型方法还存在对全球登革热长期流行趋势预测提前期不足的问题。另一方面，人类活动也会对登革热的传播产生影响。已有研究认为，能够改变人类活动行为的新冠疫情防控措施可能对登革热传播存在影响，但研究对影响方向的认识具有相悖性，具体影响尚不明确。

鉴于此，本研究以探清气候对登革热的影响机制、延长登革热的预测提前期、评估不同区域人类活动对登革热的影响程度为目标，基于登革热病例数据的可获取性，筛选了亚洲和美洲共46个国家为研究区域，选取了1990-2020年的登革热疫情变化为研究对象，并以2020年新冠疫情防控措施启动年份为时间节点，分三个层面逐步展开研究。首先，系统考察1990-2019年间亚洲和美洲地区登革热的长期流行趋势特征及气候影响机制；其次，基于气候影响机制，充分优化气候驱动的登革热传播机制模型方法；最后，基于模型预测方法，全面评估2020年亚洲和美洲地区登革热的异常流行趋势特征和风险因素。

本研究的主要结果总结如下：

1．热带印度洋海温指数，与亚洲、美洲地区登革热流行的季节和年际波动趋势关联最强，并通过遥相关作用长期影响登革热流行季的局地温度，从而推动登革热流行。

（1）南、北半球国家均表现出一致的登革热季节性流行趋势，即各自半球的夏季为流行季：北半球为7-10月，南半球为2-4月，二者之间保持稳定的6个月延迟。通过剔除发病率和气候数据的线性趋势发现，热带印度洋海温指数与登革热流行趋势密切相关。该指数代表了热带印度洋海温异常的区域平均值。登革热流行季的前一季度气候指数与南、北半球登革热月发病率均相关：当指数为正（负）值时，登革热月发病率偏高（低）。在30个全球气候指数中，热带印度洋海温指数与南、北半球登革热年发病率呈现最密切的相关性（北半球：r=0.53，P<0.01；南半球：r=0.57，P<0.01）。

（2）登革热流行季的前一季度热带印度洋海温指数与其后南、北半球登革热流行季的局地气温之间，呈现出显著正相关性（北半球：r=0.75，P<0.01；南半球：r=0.38，P<0.05）。用于探究遥相关的典型相关分析方法表明，登革热流行季的前一季度，热带印度洋全区海温异常偏暖（偏冷）事件，与随后发生的登革热流行季较暖（较冷）局地气温之间存在着显著的遥相关性（北半球：r=0.83，χ²=55.04，P<0.01；南半球：r=0.82，χ²=44.81，P<0.01）。

（3）小波分析发现，局地温度与登革热发病率的季节性变化特征呈现高度一致性：主周期都为一年，且高峰期都位于各自半球的夏季。经验动态建模的结果显示，局地温度对登革热的发病率具有显著的正中位数效应，即每升高1℃，平均每10万人中的感染病例数约增加7.35例（95% CI：6.50-8.21，Wilcox P<0.01）。

2．引入气候指数的登革热传播机制模型，能够在登革热流行季前有效捕捉亚洲、美洲各国登革热的季节和年际波动趋势，从而延长预测的提前期、提升预测的可靠性与稳定性。

（1）将与热带印度洋海温指数遥相关的局地温度因素纳入数学模型，不仅能够解释南、北半球一致的登革热季节性流行趋势（北半球：r=0.70，P<0.01；南半球：r=0.79，P<0.01），还能够在登革热流行季前有效预测各国在一年内的登革热规模（r =0.72）。已有全球气候模式能够提前6个月预测热带印度洋海温指数的变化，在此基础上本模型预计，能将登革热预测的准备时间提前9个月。

（2）将受到热带印度洋全区海温一致变化事件影响的局地温度数据，纳入登革热传播机制模型中，并进一步将全球气候模式生成的不受该事件影响的局地温度数据，也纳入到该模型中。比较两种模型模拟的和实际报告的登革热病例数间的相关性。发现纳入该事件的模型模拟结果（r =0.72）优于不纳入该事件的模型结果（r =0.42）。此外，未纳入该事件的模型未能捕捉各国登革热疫情的年际波动特征。

（3）在50种不同的初始条件、蚊子生命周期特征的热响应参数，以及25种不同的蚊子与人的初始比例下建立的模型，能够稳定模拟亚洲和美洲地区各国登革热的流行动态。

3．2020年亚洲和美洲地区登革热呈现异常下降的流行趋势，与气候突变无关，而与新冠疫情防控措施及其引发的人员流动变化密切相关，其中“关闭学校”和减少“非居住区”人员聚集与登革热发病率降低的关联性最强。

（1）在全面考虑与热带印度洋海温指数遥相关的局地温度、宿主免疫以及社会经济等登革热流行驱动因素的基础上，统计模型预测和观测结果综合表明，2020年3月后亚洲和美洲地区登革热发病率低于预期流行水平的35%。83%的国家发病水平低于前6年的均值。在不同地区，登革热疫情的缓和程度存在差异：流行季在3月之后的北半球国家，发病率下降幅度更大；而对于流行季在2-4月的南半球国家，尽管2020年初发病率高于历史同期平均水平，但在3月份之后的下降幅度也超过预期。

（2）在排除气候突变和登革热病例漏报对结果的影响后，对新冠疫情防控措施及其引发的人员流动变化进行研究，发现措施出台时间、“非居住区”人流量减少时间及登革热发病率下降时间均为2020年3-4月，三者相吻合。利用“关闭学校、关闭工作场所、取消公共活动、限制集会规模、关闭公共交通、居家令、限制国内流动和国际旅行管制”8种不同类型防控措施指数，以及“住宅、工作场所、交通站、公园、杂货店和药房，以及零售和娱乐场所”6类场所内人员流动强度指标，通过回归模型，筛选出显著风险因素，以解释预测和实际报告病例数之间的差异。结果表明，“关闭学校”措施的实施以及“非居住区”人流量的改变与登革热相对危险度（Relative Risk，RR）显著相关（“关闭学校”：RR=0.17，95% CI：0.09-0.31，P<0.05；“非居住区”：RR=0.28，95% CI：0.13-0.59，P<0.05）。

（3）模型预测表明，在2020年4-12月期间，新冠疫情防控措施的实施可能导致了亚洲和美洲地区登革热病例减少约72万例（95% CI：12-147万）。尤其在亚洲国家，登革热发病水平下降更加明显：除新加坡外，其余亚洲国家报告的病例数均呈现下降趋势。在16个美洲国家中，有12个国家报告的病例数低于模型预测值。

本研究在亚洲和美洲地区登革热流行趋势特征、风险因素和模型预测方面进行了深入研究，为登革热的提前预测和精准防控提供科学的理论依据，具有重要的实用价值。主要创新点如下：第一，揭示了热带印度洋海温异常通过影响局地气温与亚洲、美洲地区登革热流行间接关联的机制；第二，引入气候指数，从而延长登革热疫情趋势可靠预测周期；第三，明确高人流量和高混合度场所是登革热传播的高风险区域，强调学校等非居住区在未来的登革热防控工作和科学研究中的重要性。

Dengue is a viral disease transmitted by the Aedes aegypti mosquito and is widely prevalent in tropical and subtropical regions of the world. In recent decades, the incidence of dengue fever has risen sharply globally, and more than 120 countries have been affected by dengue fever, with countries in Asia and the Americas being the most severely affected. Large-scale outbreaks of dengue fever have imposed a heavy burden on the health of the population and the socio-economics of these countries. As there is still a lack of effective antiviral therapeutic drugs or vaccines, targeted prevention and control are the main means to reduce the scale and impact of the outbreaks. Analyzing the changing patterns of dengue epidemic trends, identifying risk factors for dengue epidemics, constructing dengue prediction and early warning models, and evaluating high-risk areas for dengue epidemics are very important for formulating outbreak response plans to protect the health of the population and reduce economic losses.

The systematic study of dengue epidemic trend characteristics and risk factors is the basis for optimizing the model of dengue fever transmission mechanism, and also provides support for the development of effective prevention and control strategies. Dengue fever outbreaks and epidemics are significantly affected by climate change and human activities. On the one hand, climate change can promote or constrain mosquito reproduction and affect their infectivity and susceptibility to dengue viruses, thus altering the scale of dengue epidemics. The El Niño-Southern Oscillation (ENSO) phenomenon has been found to influence the dynamics of dengue epidemics by modulating climatic conditions in specific regions, which in turn can be used for early prediction of outbreak risk. However, existing dengue forecasting modeling methods also suffer from insufficient lead time for predicting long-term global dengue epidemic trends. On the other hand, human activities can also have an impact on the spread of dengue fever. It has been suggested that new crown epidemic prevention and control measures that can change the behavior of human activities may have an impact on the spread of dengue fever, but the studies have contradictory knowledge about the direction of the impact, and the specific impact is not clear.

In view of this, this study aims to explore the mechanism of climate influence on dengue fever, extend the prediction lead time of dengue fever, and assess the degree of influence of human activities on dengue fever in different regions, and based on the accessibility of dengue fever case data, we screened a total of 46 countries in Asia and the Americas as the study area, and selected the change of dengue fever outbreaks from 1990 to 2020 as the study object, and took the year of launching preventive and control measures of the new crown epidemic in 2020 as the time node. The year of the start of the prevention and control measures was taken as the time node, and the study was gradually carried out at three levels. Firstly, the long-term epidemiological trend characteristics of dengue fever in Asia and the Americas during 1990-2019 and the climate influence mechanism are systematically examined; secondly, based on the climate influence mechanism, the climate-driven modeling method of dengue fever transmission mechanism is fully optimized; and lastly, based on the model prediction method, the trend characteristics of the abnormal dengue fever epidemiology in 2020 in Asia and the Americas and the risk factors are comprehensively assessed.

The study's principal findings are summarized as follows:

1. The Tropical Indian Ocean Sea Surface Temperature (SST) Index exhibits the strongest association with the seasonal and interannual variations in dengue epidemic across Asia and the Americas. This index, through teleconnections, exerts prolonged influence on local temperatures during dengue season, thereby propelling its epidemic.

(1) Dengue epidemics coincide with summer in each hemisphere: July to October in the Northern Hemisphere and February to April in the Southern Hemisphere, maintaining a steady six-month lag between the two. After removing the linear trends from incidence and climate data, a close association emerges between the Tropical Indian Ocean SST Index and dengue epidemic trends. This index represents the regional average anomaly of SSTs in the Tropical Indian Ocean. The climate index of the preceding quarter of the dengue season correlates with monthly dengue incidence in both hemispheres. Positive (negative) index values correspond to higher (lower) dengue incidence. Among 30 global climate indices, the Tropical Indian Ocean SST Index displays the highest correlation with annual dengue incidence in both hemispheres (Northern Hemisphere: r=0.53, P<0.01; Southern Hemisphere: r=0.57, P<0.01).

(2) Investigation of the teleconnection between climate index and local temperatures reveals a significant positive correlation between the Tropical Indian Ocean SST Index of the preceding quarter of dengue season and local temperatures during subsequent dengue seasons in both hemispheres (Northern Hemisphere: r=0.75, P<0.01; Southern Hemisphere: r=0.38, P<0.05). Utilizing a Canonical Correlation Analysis method for teleconnections, it's found that the occurrence of anomalously warm (cool) events in the Tropical Indian Ocean during the preceding quarter of dengue season correlates significantly with warmer (cooler) local temperatures during the subsequent dengue season (Northern Hemisphere: r=0.83, χ2=55.04, P<0.01; Southern Hemisphere: r=0.82, χ2=44.81, P<0.01).

(3) Research on the driving effect of local temperatures on dengue transmission demonstrates a strong consistency between the seasonal variations of local temperatures and dengue incidence. Both exhibit a primary annual cycle, with peaks during the respective summers of each hemisphere. Empirical Dynamic Modeling results illustrate a significant positive median effect of local temperatures on dengue incidence, suggesting that a 1°C increase corresponds to an average increase of 7.35 cases per 100,000 individuals (95% CI: 0.13-0.59 cases, Wilcox P<0.01).

2. Introduction of climate index into the dengue transmission model effectively captures the seasonal and interannual variations in dengue epidemic across Asia and the Americas before the onset of the disease season. This improves the reliability and stability of prediction.

(1) Mathematical model including locally correlated temperature factors linked to the Tropical Indian Ocean SST Index can explain the synchronized seasonal epidemic trend in both Hemispheres (Northern Hemisphere: r=0.70, P<0.01; Southern Hemisphere: r=0.79, P<0.01), and effectively predict the magnitude of dengue in each country within a year (r =0.72).

(2) Local temperatures influenced by Tropical Indian Ocean region-wide consistent SST variation events are incorporated into the dengue transmission mechanism model. Additionally, generated local temperatures unaffected by these events, derived from global climate models, are included in the model. A comparison is drawn between the correlation of simulated dengue prevalence trends in these two models and actual reported trends. It is observed that the model results incorporating these events (r =0.72) outperform those without such considerations (r =0.42). Furthermore, models that exclude these events fail to capture the interannual fluctuations in dengue epidemics across different countries.

(3) A model built upon 50 different initial conditions and thermal responses of mosquitoes’ life history traits, and 25 initial ratio of mosquito-to-human maintains stable simulations of dengue dynamics across Asia and the Americas.

3. An anomalous reduction in dengue epidemic occurred in Asia and the Americas in 2020. This phenomenon is unrelated to climate change but closely correlates with COVID-19 pandemic control measures and resulting changes in population movement. Notably, the association between "school closures" and reduced gatherings in "non-residential areas" exhibits the strongest correlation with decreased dengue incidence.

(1) Comprehensive consideration of climatic factors, host immunity, socioeconomic elements, and other dengue driving factors, combined with statistical modeling and observational result, indicates that dengue incidence in Asia and the Americas were 35% lower than the expected epidemic levels after March 2020. 82% of countries reported incidence lower than the six-year historical mean. Variability in the mitigation of dengue differed across regions: Northern Hemisphere countries with outbreak seasons after March experienced a greater reduction in incidence, while countries in the Southern Hemisphere with outbreak seasons between February and April saw incidence decline after March, exceeding expectations.

(2) This research establishes no significant correlation between climate change and COVID-19 pandemic control measures, including changes in population movement. After accounting for the effects of climate change and underreporting of dengue cases, investigating COVID-19 pandemic control measures and population movement alterations reveals that the times of implementing these measures, reducing non-residential area foot traffic, and the reduction in dengue incidence are all aligned around March-April 2020. Employing indices for various control measures (school closing, workplace closing, cancelling of public events, restrictions on gathering sizes, closing public transport, stay at home requirements, restrictions on internal movement and international travel controls), as well as indicators for the intensity of people's movements within categories such as (residential, workplace, transit stations, parks, grocery and pharmacy, retail and recreation) this study uses regression models to identify significant risk factors that explain differences between predicted and reported case numbers. Results demonstrate a significant correlation between the implementation of "school closures" and changes in non-residential area foot traffic with the Relative Risk (RR) of dengue ("school closures": RR=0.17, 95% CI: 0.09-0.31, P<0.05; "non-residential areas": RR=0.28, 95% CI: 0.13-0.59, P<0.05).

(3) Model predictions indicate that the implementation of COVID-19 pandemic control measures may have led to a reduction of approximately 720,000 dengue cases (95% CI: 120,000 to 1,470,000) across Asia and the Americas from April to December 2020. Particularly in Asian countries, the decline in dengue cases is more pronounced. Except for Singapore, all other Asian countries experienced a downward trend in reported cases. Among 16 American countries, 12 reported cases lower than model predictions.

This study conducted in-depth research on the characteristics of dengue epidemic trend, risk factors and model prediction in Asia and America, providing scientific theoretical basis for advance prediction and precise prevention and control of dengue, which is of great practical value. The main innovations are as follows: first, revealing the mechanism by which SST anomalies in the tropical Indian Ocean are indirectly associated with dengue epidemics in Asia and the Americas through influencing local air temperature; second, introducing climate indices, thus extending the cycle of reliable prediction of dengue epidemic trends; and third, making it clear that high-volume and high-mixing venues are high-risk areas for the spread of dengue, and stressing the importance of non-residential areas, such as schools, in the future dengue prevention and control work and scientific research.