Next Article in Journal
Remote Sensing of Aerosols and Water-Leaving Radiance from Chinese FY-3/MERSI Based on a Simultaneous Method
Previous Article in Journal
Interannual Variability of Salinity in the Chukchi Sea and Its Relationships with the Dynamics of the East Siberian Current during 1993–2020
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

A Data Driven Approach for Analyzing the Effect of Climate Change on Mosquito Abundance in Europe

1
BEYOND Center for EO Research and Satellite Remote Sensing, Institute for Astronomy, Astrophysics Space Applications and Remote Sensing (IAASARS), National Observatory of Athens, 11523 Athens, Greece
2
Department of Meteorology and Climatology, School of Geology, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece
3
Atmospheric Chemistry Department, Max Planck Institute for Chemistry, 55128 Mainz, Germany
4
Istituto Zooprofilattico Sperimentale delle Venezie, Viale dell’Università, 35020 Legnaro, Italy
5
Kommunale Aktionsgemeinschaft zur Bekämpfung der Schnakenplage (KABS) e.V., 67346 Speyer, Germany
*
Author to whom correspondence should be addressed.
Remote Sens. 2023, 15(24), 5649; https://doi.org/10.3390/rs15245649
Submission received: 30 September 2023 / Revised: 29 November 2023 / Accepted: 30 November 2023 / Published: 6 December 2023
(This article belongs to the Section Biogeosciences Remote Sensing)

Abstract

:
Mosquito-borne diseases have been spreading across Europe over the past two decades, with climate change contributing to this spread. Temperature and precipitation are key factors in a mosquito’s life cycle, and are greatly affected by climate change. Using a machine learning framework, Earth Observation data, and future climate projections of temperature and precipitation, this work studies three different cases (Veneto region in Italy, Upper Rhine Valley in Germany and Pancevo, Serbia) and focuses on (i) evaluating the impact of climate factors on mosquito abundance and (ii) long-term forecasting of mosquito abundance based on EURO-CORDEX future climate projections under different Representative Concentration Pathways (RCPs) scenarios. The study shows that increases in precipitation and temperature are directly linked to increased mosquito abundance, with temperature being the main driving factor. Additionally, as the climatic conditions become more extreme, meaning higher variance, the mosquito abundance increases. Moreover, we show that in the upcoming decades mosquito abundance is expected to increase. In the worst-case scenario (RCP8.5) Serbia will face a 10% increase, Italy around a 40% increase, and Germany will reach almost a 200% increase by 2100, relative to the decade 2010–2020. However, in terms of absolute numbers both in Italy and Germany, the expected increase is similar. An interesting finding is that either strong (RCP2.6) or moderate mitigation actions (RCP4.5) against greenhouse gas concentration lead to similar levels of future mosquito abundance, as opposed to no mitigation action at all (RCP8.5), which is projected to lead to high mosquito abundance for all cases studied.

Graphical Abstract

1. Introduction

Vector-borne diseases (VBDs) are diseases transmitted by the bites of infectious vectors such as mosquitoes, and account for more than 17% of all infectious diseases, causing more than 700,000 deaths annually [1]. Although many of these diseases have been effectively controlled, in the last decades, an emergence of mosquito-borne diseases (MBDs) has been observed [2,3]. Mosquitoes are well known disease vectors, as diseases like Malaria, West Nile Virus (WNV), Chikungunya, Dengue and Zika are spread by the bite of infected mosquitoes. The changing climatic and ecological conditions, global travel and trade, human behavior, as well as rapid and unplanned urbanization, are key factors influencing the seasonal and geographic distribution of the vectors’ population and therefore the transmission of the related pathogens [4].
Large outbreaks of MBDs have recently struck the world [5]. Europe faced a WNV epidemic in 2010 with 1016 cases [6], while in 2018 the cases rose to 1516, with 166 of them resulting in death [7]. Furthermore, according to the European Center for Disease Prevention and Control (ECDC), the number of confirmed Malaria cases reported in the EU from 2008 to 2012 ranged between 5000 and 7000 [8], whereas, in 2018, it reached almost 8500 [9]. Also, over the past 50 years, a Dengue incidence increase of thirty times has been observed globally, following its spread into many new countries.
Machine learning techniques have been employed in the past to predict the short-term upcoming risk of MBDs in an area. Models such as Support Vector Machine have been used to estimate the potential number of human cases of Malaria and Dengue in China and India, respectively [10,11], while decision trees have been employed for predicting WVN incidence in the USA [12]. In [13] the authors used a K-Nearest Neighbors model in order to estimate the mosquito population, using it as a proxy for future MBD cases. Following a similar approach, the authors in [14] created a location and genus agnostic framework to estimate the mosquito population without any constraint regarding the area or the mosquito genus, with a reliable performance.
A lot of attention has been paid to investigating the factors that amplify MBD occurrence in the community. Temperature is one of the main drivers, as it may either increase the abundance of the vectors [4,15] or shorten the extrinsic incubation period of the pathogen in the vector [16]. Precipitation is also an important factor in the mosquito life cycle, as increased precipitation can positively affect the mosquito population by creating breeding sites [4], while on the other hand extensive precipitation can flush out mosquito larvae reducing their future population [17]. Besides temperature and precipitation, there are other factors influencing MBD incidence, such as vegetation, elevation, and wind [4], and there is evidence that rapidly changing climatic conditions over an area may drive the observed emergence of MBDs [15].
Climate change includes global warming which is accompanied by a shift in the Earth’s weather patterns. It is already affecting Europe in multiple ways. It mainly refers to the increase of temperature since pre-industrial times. Since then, the average temperature of our planet has risen by 1 °C and by 2100 it is expected to have risen by more than 2 °C [18]. However, besides temperature, climate change has affected other aspects of the environment. It is modifying the water cycle and affecting precipitation patterns. It includes changes in wetness and dryness, wind fields, snow and ice coverage, coastal areas, and oceans [18]. Furthermore, due to climate change, extreme weather events, such as heat waves [19], intensive precipitation [20] and floods [21] are likely to be observed more often locally. The cause of climate change lies in increasing greenhouse gases concentrations (particularly CO2), which are mainly induced by human activity [22]. Limiting the associated emissions is a key factor for the stabilization of temperature and climate [18].
However, besides the already known impacts of climate change on weather, there are less obvious impacts, such as those on mosquito abundance (MA). The impact of climate change in the Aedes genus mosquito has been studied in [23,24], which concluded that as the temperature rises due to greenhouse gas emissions, it will favor mosquito development in many regions, contributing to disease dissemination, while the authors in [25] mapped the potential spatial redistribution of Anopheles mosquitoes to new regions of Africa under the influence of temperature and precipitation increases. The same conclusion was drawn after studying the effects of Malaria incidence transmitted by Anopheles genus mosquitoes in coastal Kenya [26]. The authors in [27] concluded that global warming increases the length of the Malaria and Dengue transmission seasons. From a more general point of view, Ref. [28] studied the effects of climatic change and found that the spread of MBDs is directly linked to global warming, regardless of the area.
Using equations to link temperature with the mosquito life cycle, the authors in [29] found a link between increased temperature and seasonal population abundance, while the authors in [30] linked the altered frequency and intensity of rainfall due to climate change with increased peak of MA. Refs. [31,32] stated that temperature, available water bodies, precipitation, and vegetation can have significant effects on a mosquito’s life cycle and presence in an area, and therefore on the occurrence of a MBD in the community. These climatic features are directly affected by climate change and this dependence brings out the need for studying the effects of climate change on MA. However, the authors in [33], who studied the relation between climatic drivers such as temperature and precipitation, concluded that climatic change alone is not capable of increasing mosquito populations, rather only in combination with human activities.
All of the cases studied place primary emphasis on temperature and precipitation in an attempt to link these variables with MBD occurrence or MA, estimating the direction of change or the future spatial distribution. However, to the best of the authors’ knowledge, there are no studies quantifying this change for upcoming years. Combining climate change projections with the estimation of MA, which is often used as a proxy of MBs incidence [34], can give useful insights into the future mosquito population dynamics which, in turn can lead to the application of adequate measures that might positively impact mosquito population control and thus disease mitigation.
So, the contribution of this paper is to quantify the effects of the two basic climatic factors (temperature and precipitation) on MA. We use Machine Learning (ML) algorithms and heterogeneous spatial-temporal information, that includes historical entomological data, and satellite Earth Observation (EO) data, to build statistical models that capture the relationships between MA and the climatic factors. The first part of this study aims to understand the impact of each of the climatic factors on the MA separately. Further, we use future climate projections based on three RCPs (RCP2.6, a strong mitigation of greenhouse gas emissions scenario, RCP4.5 a medium mitigation, and RCP8.5, which is characterized by no mitigation actions at all) in order to estimate the impact of climate change on each factor (temperature and precipitation) as well as to quantify the effect of those changes on MA. We applied our analysis to three different areas in Europe, namely Italy, Germany, and Serbia. However, it is important to mention that the utilization of the EO data allows for the application of this study to any region with systematic entomological monitoring.

2. Data

This study combines various data from heterogeneous sources that need to be collected, harmonized, and augmented before being used as input for the machine learning models. The dataset is composed of EO data from various satellites, and entomological and climatic projection data. EO data consist of environmental and topographical data for each of the areas of interest, while entomological data consist of MA data gathered in situ. Last but not least, simulation-driven future projection climatic data are used.

2.1. EO Data

There are two types of satellite feature, the θ c for the environmental, climate-related features and the θ f for the topography and other features which are not affected by climate change.

2.1.1. Environmental Data

Land surface temperature (LST) measurements (in °C) from the MODIS sensors onboard the TERRA and AQUA satellites are used. LST is estimated using the top-of-the-atmosphere brightness temperatures from the infrared bands of the satellite sensors. The product incorporated into the model is V6.0, which provides day and night LST measurements from satellite overpasses with a spatial resolution of 1 kilometer (km). Additionally, cumulative temperature features are calculated, such as the mean temperature from January to April for each trap site in order to capture the seasonal trend of temperature, which, according to the literature, affects mosquito abundance [4].
The Integrated Multi-satellite Retrievals for GPM (IMERG) precipitation gridded dataset at a resolution of 0.1° × 0.1° was used to extract precipitation measurements. The accumulated precipitation values for one and two weeks before each trap’s date of placement, as well as the accumulated precipitation from the 1st of January of each year, were calculated.

2.1.2. Topographical Data

Topography has been indicated as a significant factor in the transmission of MBDs, while it also influences the biotic conditions of different mosquito species and indicates the most suitable breeding sites [4]. The Digital Elevation Model (DEM) product that was used to generate parameters such as elevation, slope, and aspect, was acquired from Copernicus LMS with a spatial resolution of 25 m. For each trap station, the mean elevation, slope, and aspect were calculated within a buffer zone of 1 km around the point. WWFHydroSHEDS (https://www.hydrosheds.org/ accessed on 12 March 2022) and Copernicus land cover products were used to calculate hydrological features such as distance from nearest water surface, as well as to create more complex features. The buffer radius was determined based on the flight range of the Culex spp. [35].

2.2. Entomological Data

In order to collect MA data, a systematic approach for entomological monitoring has been effective since 2010 for Europe, collecting data from stable station networks. The entomological surveillance of the Area of Interest (AOI) in this work made use of CDC-CO2 light traps and gravid traps, collecting mosquitoes each year on a roughly every other week basis, identifying the total number of mosquitoes and the number of mosquitoes that tested positive for the pathogen. As an example, Figure 1 depicts the entomological network in the three AOIs.

2.3. Climatic Data

In this study, we used an ensemble of Regional Climate Model (RCM) simulations driven by global climate models (GCMs) under three RCPs (RCP2.6, RCP4.5, and RCP8.5). RCPs are greenhouse gas concentration scenarios which refer to the radiative forcing caused by changes in atmospheric composition [36,37], and were adopted for the needs of the Fifth Assessment Report (AR5) of the IPCC [38]. RCP2.6 assumes a strict reduction plan of 70% for greenhouse gas emissions, with the main goal to limit the temperature increase to 2 °C by 2100 [39]. RCP4.5 is a mediocre scenario which assumes that emissions peak around 2040 and then decline [40], while RCP8.5 is a no-mitigation scenario where emissions continue to rise throughout the 21st century due to the lack of any environmental and climatic change policy [41] and it is commonly referred as a high-end scenario. It should be noted that high-end scenarios (like RCP8.5) can be very useful to explore high-end risks of climate change, but the rapid development of renewable energy technologies and emerging climate policy have made it considerably less likely that emissions could end up as high as RCP8.5 [42].
More specifically, here, daily near-surface temperature (in °C) and precipitation (in mm) data from 11 sets of high-resolution RCM simulations (Table A1) implemented within the framework of the EURO-CORDEX initiative (https://www.euro-cordex.net/ accessed on 5 March 2022) were used. The simulations cover the greater European area at a horizontal resolution of 0.11° (12.5 km) [43,44,45]. Each set covers the period 2010–2100 and incorporates simulations under the three RCPs described above (Table A1). Details about the RCMs that produced the specific simulations and the corresponding GCMs that drive them can be found in [46]. Although AR6 was recently published, based on a wealth of GCM simulations under the latest generation of climatic scenarios, the so-called Shared Socioeconomic Pathways (SSPs) [47], regionally down-scaled projections within the CORDEX program, are not yet available [46].
Although the above-mentioned RCMs provide a variety of variables, only temperature and precipitation were used for the purposes of this research as they were the most important factors identified in the literature. Given the coordinates of an area, the climatic variables of interest were extracted and statistically calibrated with respect to an observed reference climate of an area, due to the sensitivity of MBDs to climatological conditions. The calibration was based on the delta method—which is used for climate model bias correction [48]—using the EO and climatic data available for the decade 2010–2020, considered as the reference period.

2.4. Areas of Interest

All the aforementioned data were collected for the period from 2010 to 2020 for three different areas in Europe focusing on the Culex spp. mosquito collection samples. More specifically, the data consist of a network of 190 mosquito trap stations resulting in a dataset of more than 6000 observations for the Veneto region. In the case of Germany, the dataset consists of a network of 86 stations in the Upper Rhine Valley, with roughly 4000 observations, while in Pancevo, Serbia, the entomological network includes 20 different trap stations and 800 observations. Each AOI provides different characteristics (mean temperature, precipitation, and climatic zone) and a different amount of historical data. Table 1 summarizes the main climatic characteristics of each AOI for the reader to gain a basic understanding of the initial environmental conditions studied, as well as demographic information on population density.

3. Methodology

3.1. Problem Set-Up

To quantify the effect of climatic factors on MA, we assume an estimator F ( · ) that given its internal parameters μ and an observation x i θ , where θ indicates its features, is estimating the mosquito abundance m i ^ . This can be written as:
F ( x i θ | μ ) = m i ^ .
Assuming an input set consisting of N observations, x θ = { x 1 θ , x 2 θ , . . . , x N θ } , we note as μ * , the optimal internal parameters, having been selected in order to minimize the mean square error (MSE) optimization criterion between the estimated MA ( m i ^ ), and the actual MA (M), so the μ * can be expressed as
μ * = arg min μ i ( F ( x i θ | μ ) M i ) 2 .
Further, we separate the feature space ( θ ) on the variables that are affected by the climatic conditions θ c (e.g., temperature, precipitation, etc.) and the “fixed” variables that do not depend on climate θ f (e.g., distance from the coast, elevation, day of the year etc.).
θ = { θ c , θ f } .
Thus, the function that estimates MA can be written as:
F ( x θ | μ * ) = F ( x i { θ c , θ f } | μ * ) .
Once the function F ( · ) that estimates the MA is defined, and the optimal parameters μ * have been estimated by relying on the historical observations ( x { θ c , θ f } ) , we use F ( · ) to estimate the MA for different climatic conditions. In order to do that, we project our historical data to the expected conditions according to the climate change scenarios θ c = θ c + Δ θ c , where Δ θ c = { Δ T , Δ R } are the expected changes which the climate change studies indicate for the temperature and precipitation parameters. Thus,
F ( x i { θ c , θ f } | μ * ) = m i ,
where now m i is the estimation of MA on the conditions that are expected according to climate change.
By combining Equations (1)–(4) we can define the change ratio C R of mosquito abundance as:
C R = ( m i m i ^ ) / m i ^ ,
where m i is the F ( · ) estimation of MA on the conditions that are expected according to climate change, and m i ^ is the F ( · ) estimation of MA on the baseline conditions.
The change ratio, C R , indicates the percentage change of the mosquito abundance prediction based on the changed climatological conditions. The metric of C R reflects both the direction of change and the magnitude of change according to the initial state. However, the interpretation of the percentage change must take into consideration the initial distribution of the MA in each area in order to have a more comprehensive view of the increase or decrease of the C R .

3.2. Our Implementation

Our implementation consists of the following components:
  • Estimator F ( · ) : an XGboost regression model was used as the estimator [52]. For this selection we took into account the size of the available datasets (800–6K observations per AOI), and the fact that we had tabular data containing categorical variables extracted in the feature engineering process from the initial data (such as the province based on the coordinates of the observation) [53,54]. We also used recursive feature elimination in order to reduce the complexity of the model by eliminating the features that do not contribute to the model (a more detailed view on the estimator can be found in [14]).
  • Search of optimal parameters μ * : we trained our model by relying on historical data of the last 10 years. We selected the Mean Squared Error (MSE) as the optimization criterion,
    M S E = 1 N Σ i = 1 N ( M i m i ^ ) 2 ,
    where M i is the observed MA and m i ^ is estimation of MA by estimator F ( · ) .
  • The features θ : We created a wide range of features, reported in Table A2. To make our model more sensitive to the climatic factors of temperature and precipitation, we created more meta-features that rely on those two (e.g., accumulated precipitation over two weeks, etc.), increasing the sensitivity of the model to those two parameters. The exact number of features that the model selects for each case is not predefined and it is decided according to the recursive feature elimination process.
  • Calculation of Δ θ c : Future projections of climate from the EURO-CORDEX initiative framework for a bounding polygon of each area were used to calculate the difference { Δ T , Δ R } from the decade 2010–2020 which was used as the baseline decade. Thus, for each climatic factor, the difference of the mean decade value was calculated between every 2020 to 2100 decade and the baseline decade.
  • Calculation of C R : The impact of climate change is studied by injecting Δ θ c on the θ c variables to project x θ to each future decade. This future x θ set is given as input in F ( · ) so as to estimate m i and calculate C R compared with the initial estimations.

4. Results

4.1. Model Validation

As mentioned in Section 3, the estimation of MA is based on a regression model and so it is important to assess the performance of the model to verify the validity of the predictions. In other words, it is essential to quantify the error between the estimated MA and the observed one. For each case, a different model was created based on the total set of data. So for the rest of this subsection, the performance of each model will be assessed.
In the histograms of Figure 2, the blue hue describes the distribution of MA in the 20% of the original dataset that was used only for validation purposes. The orange hue corresponds to the distribution of the MA as predicted by the model in the same set of random observations. Besides the histogram, a very indicative metric is the Mean Absolute Error (MAE),
M A E = 1 N Σ i = 1 N | M i m i ^ | ,
where M i is the observed MA and m i ^ is estimation of MA by estimator F ( · ) .
Once the mosquitoes’ variability differs in each case of interest, another informative metric is to scale MAE according to the standard deviation of each case,
s c a l e d M A E = M A E ( 7 ) σ ( M ) ,
where σ ( M ) is the standard deviation of the observed mosquitoes, and penalizes the MAE based on the standard deviation of the MA. The smaller the scaled MAE is, the better, as the smaller the standard deviation is, the easier for the model to make predictions that are closer to reality.
As shown in Figure 2, the underlying distribution of MA in all of the AOI is left skewed. Although, in the case of Veneto, Italy and Upper Rhine Valley in Germany the model was able to capture and follow the pattern of the underlying distribution of MA (Figure 2a,b), in the Serbian case (Figure 2c) it failed to capture the full extent of the small values. One possible reason behind this right shift of the predicted distribution, compared with the actual one, is the small size of the dataset and the limited number of parameters, such as max depth of the trees, to avoid overfitting. However, this shift is acceptable considering that is a universal framework that can be applied to any region. Also worth mentioning is that in the German case the model performed better considering the lower scaled MAE. This is partly a result of the higher left skewed distribution, as the majority of the values are less than 50 mosquitoes, and the estimator can achieve higher performance due to the smaller variance of mosquito abundance.
Also worth mentioning is the negligible bias of the models against temperature and precipitation increase, as can be seen in Figure 3. This means that the model does not overestimate or underestimate the MA as the temperature or the precipitation rises, as the slope of the linear fit in all of the AOI is negligible and orders of magnitude less than the constant term. Since this model will be used to estimate the MA in different ranges of temperature and precipitation, it is crucial to be as little biased as possible towards those variables for the validity of our analysis. Overall, it can be concluded that the developed models perform with consistency throughout the three AOIs.
Table 2 summarizes the performance for each developed model. Despite the different MAE performance on each case which is inextricably linked with the underlined MA distribution, the performance of the scaled MAE (Equation (8)), is similar for all the AOIs, indicating similar behavior of the framework across the three cases, regardless of the different number of features selected. For the sake of completeness Table A3 presents the features selected for each case in order of importance.

4.2. Performance Analysis of Climatic Factors

We studied the influence of climatological conditions on MA in two ways; either by (i) changing the mean value of the θ c variables by adding a constant value to stimulate a constant change, or (ii) by keeping the mean value at the same level while changing the variance of θ c variables to stimulate the increase of extreme climatic conditions. Worth mentioning is that changing separately only one climatic factor is not feasible as there is a strong covariance among the climatic factors. Although, this analysis can provide useful information as a “’rule of thumb” for policy making.

4.2.1. Mean Value Change

The plots below show the impact of the mean value of temperature (Figure 4a–c), precipitation (Figure 4d–f), and the combination of those two factors (Figure 4g–i) on MA. The blue hue in each of the plots depicts the standard deviation of the change. To investigate the impact of a constant change in temperature and precipitation on MA, the temperature was increased in a [0, 4] range with steps of 0.2 °C, and precipitation value was increased in a [0%, 40%] range with steps of 5%. In Figure 4a–c it becomes clear that increasing temperature is linked to an increase of mosquito abundance. Germany faces the steeper increase of mosquito abundance, up to nearly 250% in the case of a 4 °C temperature rise, while Serbia follows a much more conservative pattern. The effect of precipitation on MA showed a mixed trend in the study areas. The maximum effect of increased precipitation on MA in Italy (Figure 4d) remained below 20%, while in Serbia increased precipitation had a negative impact on the MA, as it decreased by up to 10% (Figure 4f). In Germany precipitation has a markedly higher impact on MA (Figure 4e), with an even larger effect when considering the integration of simultaneous increase in temperature (Figure 4h) which led to a nearly 800% total increase on MA. Overall, it seems that the increase in the mean temperature is the driving factor behind the increase of predicted MA; however, mean precipitation increase acts as an amplifier on the MA prediction.

4.2.2. Variance Change

To control the variance of each factor (temperature, precipitation) a zero mean Gaussian noise is added to the data. The variance of the injected noise on the data will scale according to a factor β , N ( μ , σ β ) , where β [0, 2] with a step of 0.25, and σ is the variance of the data. The increase of variance represents the increase of extreme climatic conditions.
An increase in temperature or precipitation variance initially leads to increased MA in all of the AOIs (Figure 5a–f). In each region, temperature or precipitation variance increase is the driving factor behind MA increase, while the other factor acts as an amplifier of the MA increase. Once again, Upper Rhine Valley, Germany seems to be the most impacted AOI as both temperature and precipitation variance are increased individually where the MA climbed up to 300%, while the increase of those two factors simultaneously (Figure 5h) leads to a 600% increase of MA.

4.3. Climatic Analysis

For the climatic analysis, the decade 2010–2020 was used as the baseline. Thus, for each decade from 2020 to 2100 the change of each climatic factor relative to the 2010–2020 decade, Δ θ , was calculated. Figure 6 presents the mean difference of temperature and precipitation between the 2010–2020 decade and the decades from 2020 to 2100 for the three aforementioned RCPs for all the three AOIs. The upper plots focus on the mean temperature change from the baseline decade, while the lower ones focus on the mean change of the two weeks accumulated precipitation, with the hue showing the inter-modal variability of the 11 RCM simulations models.
As shown in Figure 6a–c the temperature change follows a distinctive pattern throughout the three cases. In the RCP2.6, the mean change is close to 0.5°, while the medium mitigation scenario follows an increasing pattern over the decades, which is further increased in the RCP8.5 scenario of no mitigation, with a mean change of more than 3.5 °C by 2100.
However, precipitation does not follow such a distinctive pattern. In the case of Italy (Figure 6d), precipitation in the RCP2.6 scenario seems to be increasing for all the years, except the last decade which reports a decline back to the 2010–2020 level. A very different future, mainly characterized by decreasing precipitation, is predicted by the RCP4.5 scenario. Based on the RCP8.5 scenario, unlike temperature, rainfall increases steadily until the 2060s and at some point it seems to decrease at the end of the century. Precipitation does not follow a clear pattern either in case of Germany (Figure 6e). In all three scenarios, precipitation is varying, with some decades facing a small increase and some others a small decrease, based on the mitigation actions taken. However, based on the RCP8.5 scenario, the temperature increase is linked with a slight increase in precipitation. Finally, in the case of Serbia, precipitation (Figure 6f) deviates from the above patterns, following a relative steady course throughout the years, with the exception on the RCP8.5 scenario where there is a constant increase.

Impact of Climate Change on Mosquito Abundance

Climatic projections can be combined with the MA estimator as described in the methodology, Section 3, to estimate the MA under the influence of climate change. Figure 7 shows the estimated mean change of MA based on 11 different RCM simulation models for each of the three climatic scenarios, while the hue shows the inter-modal variability of these 11 models.
Overall, the bigger picture describes that in Upper Rhine Valley in Germany we expect the highest increase of MA, that could reach up to 200% for the worst scenario RCP8.5, contrary to Veneto in Italy and Pancevo, Serbia, where it follows a moderate increase compared with the 2010–2020 decade. Veneto will face a 40% increase in the worst case scenario and Pancevo around a 10% increase in case of RCP8.5. In terms of absolute numbers, however, we can see that Upper Rhine Valley in Germany and Veneto, Italy expect similar increases of mosquitoes, and again Pancevo, Serbia is expecting the smallest (Table 2).
Scenario RCP8.5 leads to greater MA increase compared with the other two scenarios, as expected. An interesting finding is that in all of the three AOI scenarios, RCP2.6 and RCP4.5 have a similar impact on the MA. That means that the scenario of moderate effort to reduce greenhouse gas emissions performs similarly to the scenario characterized by strong mitigation actions.
It is also worth mentioning that the strong divergence of MA on the RCP8.5 compared with the other two scenarios in all of the AOIs is not observed sooner than the decade of 2040. Until then, the three RCP scenarios follow a similar pattern, indicating that any greenhouse gas mitigation strategy selected will have a similar impact on MA until 2040. More specifically, in Veneto, Italy and in Pancevo, Serbia this divergence in MA increase will take place in the 2050s, a decade later compared with Upper Rhine Valley, Germany, where a lack of any mitigation action will lead the MA to be increased sooner.

5. Discussion

In this paper, we proposed a method to estimate the impact of temperature and precipitation on MA, as well as an estimation of the future impact of climate change on MA. The ML estimator developed achieves accurate and reliable performance, as it captures the underlying distribution of the MA of each case, with negligible bias on its predictions regarding the temperature and precipitation, proving that it is feasible to quantify the future impact of climate change on MA.
Using this ML estimator and taking into consideration temperature and precipitation change, this paper is divided into two parts. The first part studies how temperature and precipitation affect MA, while the second part focuses on the impact of climate change on MA, by integrating future climate projections into the estimator. Both of these parts were implemented for three different AOIs in Europe, in Italy, Germany and Serbia, focusing on the Culex spp. genus of mosquitoes.
Regarding the effects of temperature and precipitation on MA in the first part, we can conclude that as the variance of both the climatic variables rises, so does MA. Furthermore, rise in temperature mean is directly linked to a MA increase in all of the AOIs, while an increment in precipitation mean value does not have a clear monotonic behavior.
In the second part of the study, the impact of climate change was further investigated to estimate the change of MA. Under the future climate projections, Germany will face the biggest increase of mosquitoes in the years to come, contrary to Serbia, where the smallest increase of MA is predicted. In all areas, the scenario RCP8.5 is linked to the greatest MA increase, while MA under RCP2.6 and RCP4.5 scenarios seems to range similarly. We need to highlight here that even with a moderate effort to reduce greenhouse gas emissions we can achieve as good results regarding MA as if strong mitigation actions were taken, contrary to the RCP8.5 scenario, where the increase level is greater compared with the other two, while any greenhouse gas mitigation strategy selected will have a similar impact on MA until 2040.
The study demonstrates the potential of using data from diverse sources to integrate into a synthesis aiming to control future mosquito populations and thus help monitor the spread of mosquito-borne disease. The advantage of this study compared with the existing literature is that it quantifies the effect of climate change on MA under three different RCP scenarios based on EO data, which also allow for application of the same methodology to any region.
However, it is important to mention that the current study did not take into consideration the changes in other environmental parameters which are directly or indirectly affected by climate change, such as the vegetation of an area. This study only considered the effects of temperature and precipitation to study the impact of climate change on MA. Thus, a next step of this research would be to consider more climate variables, as climate change affects many more parameters besides temperature and precipitation, such as the vegetation and humidity of an area. In addition, another direction to consider is studying the effects of climate change on the Aedes and Anopheles mosquito genera individually. This research focused only on the Culex spp. genus. Each genus responds differently to environmental factors, and they utilize different habitats, especially in their larval stages.

Author Contributions

Conceptualization, A.T. and G.A.; Methodology, A.T., G.A., A.K.G. and D.A.; Writing—original draft, A.T. and G.A.; Writing—review & editing, A.K.G., D.A., P.Z., D.F., D.W. and C.K. All authors have read and agreed to the published version of the manuscript.

Funding

This study was conducted within the scope of the EYWA project, an EarlY WArning System for Mosquito-Borne Diseases, an initiative under the flag of the EuroGEO Action Group “EO4EViDence—Earth Observation for Epidemics of Vector-Borne Diseases”.

Data Availability Statement

The data for this study are not available due to privacy constraints. However, we acknowledge the importance of data transparency and we are committed to providing information and insights to the extent permissible by relevant policies and regulations.

Acknowledgments

The authors would like to express their gratitude to their colleagues from Kommunale Aktionsgemeinschaft zur Bekämpfung der Schnakenplage (KABS) and EKO-SAN PLUS d.o.o. for their entomological data contribution and their valuable expert entomological comments. We acknowledge the World Climate Research Programme’s Working Group on Regional Climate, and the Working Group on Coupled Modelling, former coordinating body of CORDEX and responsible panel for CMIP5. We also thank the climate modelling groups (listed in Table A1) for producing and making available their model output. We also acknowledge the Earth System Grid Federation infrastructure, an international effort led by the U.S. Department of Energy’s Program for Climate Model Diagnosis and Intercomparison, the European Network for Earth System Modelling and other partners in the Global Organization for Earth System Science Portals (GO-ESSP). Finally, we would like to acknowledge the Veneto Region of Italy for sharing their entomological data for our research purposes.

Conflicts of Interest

The authors declare no conflict of interest.

Abbreviations

The following abbreviations are used in this manuscript:
RCPRepresentative Concentration Pathway
VBDVector Borne Disease
MBDMosquito-Borne Disease
WNVWest Nile Virus
ECDCEuropean Center for Disease Prevention and Control
MAMosquito Abundance
MLMachine Learning
EOEarth Observation
LSTLand Surface Temperature
IMERGIntegrated Multi-satellite Retrievals for GPM
DEMDigital Elevation Model
AOIArea of Interest
RCMRegional Climate Model
GCMGlobal Climate Model
MSEMean Squared Error
MAEMean Absolute Error

Appendix A

Table A1. List of the 11 EURO-CORDEX sets of simulations used in the present study.
Table A1. List of the 11 EURO-CORDEX sets of simulations used in the present study.
RCMDriving GCMRealization
1ALADIN63.v2CNRM.CNRM-CERFACS-CNRM-CM5r1i1p1
2CCLM4-8-17.v1CLMcom.ICHEC-EC-EARTHr12i1p1
3HIRHAM5.v2DMI.ICHEC-EC-EARTHr3i1p1
4RACMO22E.v1KNMI.ICHEC-EC-EARTHr12i1p1
5RACMO22E.v2KNMI.MOHC-HadGEM2-ESr1i1p1
6RACMO22E.v2KNMI.CNRM-CERFACS-CNRM-CM5r1i1p1
7RCA4.v1SMHI.MOHC-HadGEM2-ESr1i1p1
8RCA4.v1SMHI.MPI-M-MPI-ESM-LRr1i1p1
9RCA4.v1SMHI.ICHEC-EC-EARTHr12i1p1
10REMO2009.v1MPI-CSC.MPI-M-MPI-ESM-LRr1i1p1
11REMO2009.v1MPI-CSC.MPI-M-MPI-ESM-LRr2i1p1
Table A2. List of all the features available to the models.
Table A2. List of all the features available to the models.
FeatureExplanation
dt_placementDate of the observation
station_idStation ID
xLongitude
yLatitude
lst_dayLand surface temperature at day
lst_nightLand surface temperature at night
lstMean Land surface temperature of day and night
lst_Jan_meanMean land surface temperature in January
lst_Feb_meanMean land surface temperature in February
lst_Mar_meanMean land surface temperature in March
lst_Apr_meanMean land surface temperature in April
acc_precipitation_1weekAccumulated precipitation counting towards
one week before the date of placement
acc_precipitation_2weekAccumulated precipitation counting towards
two weeks before the date of placement
acc_precipitation_janAccumulated precipitation counting from the 1st of January of each year
wc_l_1000Water course length of national hydrological data within
a buffer zone of 1000 m around each sampling/trapping site
dem_1000Mean elevation (resolution = 12.5 m),
within a buffer of 1000 m around trapping sites
aspect_1000Mean aspect (12.5 m), within a buffer of 1000 m around trapping sites
slope_1000Mean slope (12.5 m), within a buffer of 1000 m around trapping sites
coast_dist_1000Mean Distance of sampling/trapping site
within a buffer of 1000 m from coastline
wc_dist_1000Nearest distance of watercourses of national hydrological data
within a buffer zone of 1000 m around each sampling/trapping site
flow_acc_1000Mean flow accumulation within a buffer of
1000 m around trapping sites
waw_mean_1000Distance of sampling/trapping sites from nearest surface
water polygon area within a buffer of 1000 m around trapping sites
days_distanceTime difference in days between the date of
placement and 1 January of the corresponding year
provinceProvince in which trap is located based on the coordinates
mo_cosCosine transformation of the month of date of placement
mo_sinSine transformation of the month of date of placement
summer_days_yearNumber of days with over 30 °C within the year
summer_days_monthNumber of days with over 30 °C within the month
distanceEuclidean distance of coordinates between
a specific point and the trap site
PCA3 PCA components constructed out of all the above mentioned features
climatic_zoneClimatic zone of x,y point based on Köppen climatic classification
Table A3. Features selected per case in order of descending significance.
Table A3. Features selected per case in order of descending significance.
ItalyGermanySerbia
dem_1000acc_precipitation_2weekdays_distance
days_distanceacc_precipitation_janPCA_1
xdays_distanceacc_precipitation_2week
yflow_acc_1000y
lst_mar_meanPCA_2x
PCA_2waw_mean_1000acc_precipitation_jan
flow_acc_1000slope_mean_1kmacc_precipitation_1week
slope_1000PCA_3lst
wc_l_1000lstPCA_3
lst_feb_meandem_1000lst_night
acc_precipitation_janlst_feb_meanPCA_2
PCA_3lst_mar_meanlst_feb_mean
PCA_1acc_precipitation_1week
waw_mean_1000lst_day
lst_jan_meanaspect_1000
aspect_1000lst_night
lstPCA_1
acc_precipitation_2week
wc_dist_1000
coast_dist_1000
lst_apr_mean
acc_precipitation_1week
lst_night
lst_day

References

  1. World Health Organization. Global Vector Control Response 2017–2030. 2017. Available online: https://www.who.int/publications/i/item/9789241512978 (accessed on 3 March 2022).
  2. Gubler, D. Dengue, Urbanization and Globalization: The Unholy Trinity of the 21 Century. Trop. Med. Health 2011, 39, 3–11. [Google Scholar] [CrossRef]
  3. Reisen, W.K. Ecology of West Nile Virus in North America. Viruses 2013, 5, 2079–2105. [Google Scholar] [CrossRef] [PubMed]
  4. Parselia, E.; Kontoes, C.; Tsouni, A.; Hadjichristodoulou, C.; Kioutsioukis, I.; Magiorkinis, G.; Stilianakis, N.I. Satellite Earth Observation Data in Epidemiological Modeling of Malaria, Dengue and West Nile Virus: A Scoping Review. Remote Sens. 2019, 11, 1862. [Google Scholar] [CrossRef]
  5. Wilson, A.L.; Courtenay, O.; Kelly-Hope, L.A.; Scott, T.W.; Takken, W.; Torr, S.J.; Lindsay, S.W. The importance of vector control for the control and elimination of vector-borne diseases. PLoS Neglected Trop. Dis. 2020, 14, e0007831. [Google Scholar] [CrossRef] [PubMed]
  6. Paz, S.; Semenza, J.C. Environmental Drivers of West Nile Fever Epidemiology in Europe and Western Asia—A Review. Int. J. Environ. Res. Public Health 2013, 10, 3543–3562. [Google Scholar] [CrossRef] [PubMed]
  7. European Centre for Disease Prevention and Control. West Nile Virus Infection. Annual Epidemiological Report for 2018. 2019. Available online: https://www.ecdc.europa.eu/en/publications-data/west-nile-virus-infection-annual-epidemiological-report-2018 (accessed on 3 March 2022).
  8. European Centre for Disease Prevention and Control. Number and Rates of Confirmed Malaria Reported Cases, EU/EEA 2008–2012. 2012. Available online: https://www.ecdc.europa.eu/en/publications-data/number-and-rates-confirmed-malaria-reported-cases-eueea-2008-2012 (accessed on 3 March 2022).
  9. European Centre for Disease Prevention and Control. Malaria—Annual Epidemiological Report for 2018. 2020. Available online: https://www.ecdc.europa.eu/en/publications-data/malaria-annual-epidemiological-report-2018 (accessed on 3 March 2022).
  10. Chintalapati, D.S.; Sohani, S.; Kumar, D.; Malik, A.; Chahar, B.; Nema, A.; Panigrahi, B.; Dhiman, R. A Support Vector Machine-Firefly Algorithm based forecasting model to determine malaria transmission. Neurocomputing 2014, 129, 279–288. [Google Scholar] [CrossRef]
  11. Guo, P.; Liu, T.; Zhang, Q.; Wang, L.; Xiao, J.; Zhang, Q.; Luo, G.; Li, Z.H.; He, J.; Zhang, Y.; et al. Developing a dengue forecast model using machine learning: A case study in China. PLoS Neglected Trop. Dis. 2017, 11, e0005973. [Google Scholar] [CrossRef]
  12. Young, S.; Tullis, J.; Cothren, J. A remote sensing and GIS-assisted landscape epidemiology approach to West Nile virus. Appl. Geogr. 2013, 45, 241–249. [Google Scholar] [CrossRef]
  13. Scavuzzo, J.; Trucco, F.; Espinosa, M.; Tauro, C.; Abril, M.; Scavuzzo, C.; Frery, A. Modeling Dengue Vector Population Using Remotely Sensed Data and Machine Learning. Acta Trop. 2018, 185, 167–175. [Google Scholar] [CrossRef]
  14. Tsantalidou, A.; Parselia, E.; Arvanitakis, G.; Kyratzi, K.; Gewehr, S.; Vakali, A.; Kontoes, C. MAMOTH: An Earth Observational Data-Driven Model for Mosquitoes Abundance Prediction. Remote Sens. 2021, 13, 2557. [Google Scholar] [CrossRef]
  15. Franklinos, L.; Jones, K.; Redding, D.; Abubakar, I. The effect of global change on mosquito-borne disease. Lancet Infect. Dis. 2019, 19, e302–e312. [Google Scholar] [CrossRef] [PubMed]
  16. Dohm, D.J.; O’Guinn, M.L.; Turell, M.J. Effect of environmental temperature on the ability of Culex pipiens (Diptera: Culicidae) to transmit West Nile virus. J. Med. Entomol. 2002, 39, 221–225. [Google Scholar] [CrossRef]
  17. Benedum, C.; Seidahmed, O.; Eltahir, E.; Markuzon, N. Statistical modeling of the effect of rainfall flushing on dengue transmission in Singapore. PLoS Neglected Trop. Dis. 2018, 12, e0006935. [Google Scholar] [CrossRef] [PubMed]
  18. Masson-Delmotte, V.; Zhai, P.; Pirani, A.; Connors, S.; Péan, C.; Berger, S.; Caud, N.; Chen, Y.; Goldfarb, L.; Gomis, M.; et al. IPCC, 2021: Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change; Cambridge University Press: Cambridge, MA, USA, 2021. [Google Scholar] [CrossRef]
  19. Perkins-Kirkpatrick, S.E.; Lewis, S.C. Increasing trends in regional heatwaves. Nat. Commun. 2020, 11, 3357. [Google Scholar] [CrossRef]
  20. Zittis, G.; Bruggeman, A.; Lelieveld, J. Revisiting future extreme precipitation trends in the Mediterranean. Weather Clim. Extrem. 2021, 34, 100380. [Google Scholar] [CrossRef]
  21. Hirabayashi, Y.; Roobavannan, M.; Koirala, S.; Konoshima, L.; Yamazaki, D.; Watanabe, S.; Kim, H.; Kanae, S. Global flood risk under climate change. Nat. Clim. Chang. 2013, 3, 816–821. [Google Scholar] [CrossRef]
  22. Hansen, J.; Sato, M.; Kharecha, P.; Russell, G.; Lea, D.; Siddall, M. Climate change and trace gases. Philos. Trans. R. Soc. A 2007, 365, 1925–1954. [Google Scholar] [CrossRef]
  23. Iwamura, T.; Guzman-Holst, A.; Murray, K. Accelerating invasion potential of disease vector Aedes aegypti under climate change. Nat. Commun. 2020, 11, 2130. [Google Scholar] [CrossRef]
  24. Ryan, S.; Carlson, C.; Mordecai, E.; Johnson, L. Global expansion and redistribution of Aedes-borne virus transmission risk with climate change. bioRxiv 2019. [Google Scholar] [CrossRef]
  25. Tonnang, H.; Kangalawe, R.; Yanda, P. Predicting and mapping malaria under climate change scenarios: The potential redistribution of malaria vectors in Africa. Malar. J. 2010, 9, 111. [Google Scholar] [CrossRef]
  26. Le, P.; Kumar, P.; Ruiz, M.; Mbogo, C.; Muturi, J. Predicting the direct and indirect impacts of climate change on malaria in coastal Kenya. PLoS ONE 2019, 14, e0211258. [Google Scholar] [CrossRef]
  27. Colón-González, F.J.; Sewe, M.; Tompkins, A.; Sjödin, H.; Casallas García, A.; Rocklöv, J.; Caminade, C.; Lowe, R. Projecting the risk of mosquito-borne diseases in a warmer and more populated world—A multi-model, multi-scenario intercomparison modelling study. Lancet Planet. Health 2021, 5, e404–e414. [Google Scholar] [CrossRef] [PubMed]
  28. El-Sayed, A.; Kamel, M. Climatic changes and their role in emergence and re-emergence of diseases. Environ. Sci. Pollut. Res. 2020, 27, 22336–22352. [Google Scholar] [CrossRef] [PubMed]
  29. Ewing, D.; Cobbold, C.; Purse, B.; Nunn, M.; White, S. Modelling the Effect of Temperature on the Seasonal Population Dynamics of Temperate Mosquitoes. J. Theor. Biol. 2016, 400, 65–79. [Google Scholar] [CrossRef] [PubMed]
  30. Valdez, L.; Sibona, G.; Diaz, L.; Contigiani, M.; Condat, C. Effects of rainfall on Culex mosquito population dynamics. J. Theor. Biol. 2017, 421, 28–38. [Google Scholar] [CrossRef] [PubMed]
  31. Bartlow, A.; Manore, C.; Xu, C.; Kaufeld, K.; Del Valle, S.; Ziemann, A.; Fairchild, G.; Fair, J. Forecasting Zoonotic Infectious Disease Response to Climate Change: Mosquito Vectors and a Changing Environment. Vet. Sci. 2019, 6, 40. [Google Scholar] [CrossRef]
  32. Brugueras, S.; Fernandez-Martinez, B.; Martínez-de la Puente, J.; Figuerola, J.; Tomas, M.; Rius, C.; Larrauri, A.; gomez barroso, D. Environmental drivers, climate change and emergent diseases transmitted by mosquitoes and their vectors in southern Europe: A systematic review. Environ. Res. 2020, 191, 110038. [Google Scholar] [CrossRef]
  33. Roiz, D.; Ruiz, S.; Soriguer, R.; Figuerola, J. Climatic effects on mosquito abundance in Mediterranean wetlands. Parasites Vectors 2014, 7, 333. [Google Scholar] [CrossRef]
  34. Cailly, P.; Tran, A.; Balenghien, T.; L’Ambert, G.; Toty, C.; Ezanno, P. A climate-driven abundance model to assess mosquito control strategies. Ecol. Model. 2012, 227, 7–17. [Google Scholar] [CrossRef]
  35. Chuang, T.W.; Hildreth, M.; Vanroekel, D.; Wimberly, M. Weather and land cover influences on mosquito populations in Sioux Falls, South Dakota. J. Med. Entomol. 2011, 48, 669–679. [Google Scholar] [CrossRef]
  36. Moss, R.; Edmonds, J.; Hibbard, K.; Manning, M.; Rose, S.; Vuuren, D.; Carter, T.; Emori, S.; Kainuma, M.; Kram, T.; et al. The Next Generation of Scenarios for Climate Change Research and Assessment. Nature 2010, 463, 747–756. [Google Scholar] [CrossRef]
  37. Vuuren, D.; Edmonds, J.; Kainuma, M.; Riahi, K.; Thomson, A.; Hibbard, K.; Hurtt, G.; Kram, T.; Krey, V.; Lamarque, J.F.; et al. The representative concentration pathways: An overview. Climatic Change. This issue. Clim. Chang. 2011, 109, 5–31. [Google Scholar] [CrossRef]
  38. Core Writing Team; Pachauri, R. (Eds.) Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change; IPCC: Geneva, Switzerland, 2014. [Google Scholar]
  39. Vuuren, D.; Stehfest, E.; Elzen, M.; Kram, T.; Vliet, J.; Deetman, S.; Isaac, M.; Klein Goldewijk, K.; Hof, A.; Mendoza Beltran, A.; et al. RCP2.6: Exploring the possibility to keep global mean temperature increase below 2 °C. Clim. Chang. 2011, 109, 95. [Google Scholar] [CrossRef]
  40. Thomson, A.; Calvin, K.; Smith, S.; Kyle, P.; Volke, A.; Patel, P.; Delgado Arias, S.; Bond-Lamberty, B.; Wise, M.; Clarke, L.; et al. RCP4.5: A pathway for stabilization of radiative forcing by 2100. Clim. Chang. 2011, 109, 77–94. [Google Scholar] [CrossRef]
  41. Riahi, K.; Rao, S.; Krey, V.; Cho, C.; Chirkov, V.; Fischer, G.; Kindermann, G.; Nakicenovic, N.; Rafaj, P. RCP 8.5—A scenario of comparatively high greenhouse gas emissions. Clim. Chang. 2011, 109, 33–57. [Google Scholar] [CrossRef]
  42. Hausfather, Z.; Peters, G. Emissions—The `business as usual’ story is misleading. Nature 2020, 577, 618–620. [Google Scholar] [CrossRef] [PubMed]
  43. Jacob, D.; Teichmann, C.; Sobolowski, S.; Katragkou, E.; Anders, I.; Belda, M.; Benestad, R.; Boberg, F.; Buonomo, E.; Cardoso, R.; et al. Regional climate downscaling over Europe: Perspectives from the EURO-CORDEX community. Reg. Environ. Chang. 2020, 20, 51. [Google Scholar] [CrossRef]
  44. Jacob, D.; Petersen, J.; Eggert, B.; Alias, A.; Christensen, O.; Bouwer, L.; Braun, A.; Colette, A.; Déqué, M.; Georgievski, G.; et al. EURO-CORDEX: New high-resolution climate change projections for European impact research. Reg. Environ. Chang. 2014, 14, 563–578. [Google Scholar] [CrossRef]
  45. Vautard, R.; Gobiet, A.; Jacob, D.; Belda, M.; Colette, A.; Déqué, M.; Fernández, J.; Garcia-Diez, M.; Goergen, K.; Güttler, I.; et al. The simulation of European heat waves from an ensemble of regional climate models within the EURO-CORDEX project. Clim. Dyn. 2013, 41, 2555–2575. [Google Scholar] [CrossRef]
  46. Georgoulias, A.; Akritidis, D.; Kalisoras, A.; Kapsomenakis, J.; Melas, D.; Zerefos, C.; Zanis, P. Climate change projections for Greece in the 21st century from high-resolution EURO-CORDEX RCM simulations. Atmos. Res. 2022, 271, 106049. [Google Scholar] [CrossRef]
  47. Riahi, K.; van Vuuren, D.P.; Kriegler, E.; Edmonds, J.; O’Neill, B.C.; Fujimori, S.; Bauer, N.; Calvin, K.; Dellink, R.; Fricko, O.; et al. The Shared Socioeconomic Pathways and their energy, land use, and greenhouse gas emissions implications: An overview. Glob. Environ. Chang. 2017, 42, 153–168. [Google Scholar] [CrossRef]
  48. Navarro-Racines, C.; Tarapues, J.; Thornton, P.; Jarvis, A.; Ramirez-Villegas, J. High-resolution and bias-corrected CMIP5 projections for climate change impact assessments. Sci. Data 2020, 7, 7. [Google Scholar] [CrossRef] [PubMed]
  49. Available online: https://tcktcktck.org/europe (accessed on 14 September 2023).
  50. Beck, H.; Zimmermann, N.; McVicar, T.; Vergopolan, N.; Berg, A.; Wood, E. Present and future Köppen-Geiger climate classification maps at 1-km resolution. Sci. Data 2018, 5, 180214. [Google Scholar] [CrossRef] [PubMed]
  51. Available online: https://www.citypopulation.de/Europe.html (accessed on 14 September 2023).
  52. Chen, T.; Guestrin, C. XGBoost: A Scalable Tree Boosting System; Association for Computing Machinery: New York, NY, USA, 2016; pp. 785–794. [Google Scholar] [CrossRef]
  53. Grinsztajn, L.; Oyallon, E.; Varoquaux, G. Why do tree-based models still outperform deep learning on tabular data? arXiv 2022, arXiv:2207.08815. [Google Scholar] [CrossRef]
  54. Borisov, V.; Leemann, T.; Sessler, K.; Haug, J.; Pawelczyk, M.; Kasneci, G. Deep Neural Networks and Tabular Data: A Survey. IEEE Trans. Neural Networks Learn. Syst. 2022, 153–168. [Google Scholar] [CrossRef]
Figure 1. Network of entomological traps in (a) Veneto, Italy, (b) Upper Rhine Valley, Germany, and (c) Pancevo, Serbia. Red dots represent the locations of the mosquito traps.
Figure 1. Network of entomological traps in (a) Veneto, Italy, (b) Upper Rhine Valley, Germany, and (c) Pancevo, Serbia. Red dots represent the locations of the mosquito traps.
Remotesensing 15 05649 g001
Figure 2. Histogram of actual vs predicted MA values. The blue hue shows the distribution of the actual values of the MA in the validation set. The orange hue shows the distribution of the predicted values of the MA for the same validation set.
Figure 2. Histogram of actual vs predicted MA values. The blue hue shows the distribution of the actual values of the MA in the validation set. The orange hue shows the distribution of the predicted values of the MA for the same validation set.
Remotesensing 15 05649 g002
Figure 3. Scatter plot of Error vs. Temperature (ac) and Error vs. Precipitation (df). The slope of the fitted red line shows that the error between the actual and predicted MA values is independent of the rise in temperature and precipitation and the three models are not biased towards those two variables.
Figure 3. Scatter plot of Error vs. Temperature (ac) and Error vs. Precipitation (df). The slope of the fitted red line shows that the error between the actual and predicted MA values is independent of the rise in temperature and precipitation and the three models are not biased towards those two variables.
Remotesensing 15 05649 g003
Figure 4. MA prediction in relation to temperature (ac), precipitation (df) and temperature and precipitation (gi) mean increase.
Figure 4. MA prediction in relation to temperature (ac), precipitation (df) and temperature and precipitation (gi) mean increase.
Remotesensing 15 05649 g004
Figure 5. MA prediction in relation to temperature (ac), precipitation (df) and temperature and precipitation (gi) variation increase.
Figure 5. MA prediction in relation to temperature (ac), precipitation (df) and temperature and precipitation (gi) variation increase.
Remotesensing 15 05649 g005
Figure 6. Mean change of daily temperature (ac) and 2 weeks accumulated precipitation (df) based on the three RCP scenarios. The hue of each line shows the inter-model variability of that change based on the 11 EURO-CORDEX sets of simulations.
Figure 6. Mean change of daily temperature (ac) and 2 weeks accumulated precipitation (df) based on the three RCP scenarios. The hue of each line shows the inter-model variability of that change based on the 11 EURO-CORDEX sets of simulations.
Remotesensing 15 05649 g006
Figure 7. MA prediction based on future projections of climate change based on the three RCP scenarios. The hue of each line shows the inter-model variability of that change based on the 11 EURO-CORDEX sets of simulations.
Figure 7. MA prediction based on future projections of climate change based on the three RCP scenarios. The hue of each line shows the inter-model variability of that change based on the 11 EURO-CORDEX sets of simulations.
Remotesensing 15 05649 g007
Table 1. Mean Temperature, precipitation [49] and Köppen climatic zone classification for each AOI [50]. The Köppen climate classification scheme divides climates into five main climate groups: A (tropical), B (arid), C (temperate), D (continental), and E (polar). The second letter indicates the seasonal precipitation type, while the third letter indicates the temperature. Population density is given per km2 [51].
Table 1. Mean Temperature, precipitation [49] and Köppen climatic zone classification for each AOI [50]. The Köppen climate classification scheme divides climates into five main climate groups: A (tropical), B (arid), C (temperate), D (continental), and E (polar). The second letter indicates the seasonal precipitation type, while the third letter indicates the temperature. Population density is given per km2 [51].
CaseMean
Temperature (°C)
Mean
Precipitation (mm)
Climatic
Zone
Population
Density/km2
Veneto, Italy15.9152.54CSA
(Temperate
Dry Summer
Hot Summer)
263.7
Upper Rhine Valley,
Germany
10.7943.85CFB
(Temperate
No dry season
Warm Summer)
311.2
Pancevo, Serbia15.1284.86CFB
(Temperate
No dry season
Warm Summer)
80.51
Table 2. Performance of each AOI model. The table shows the number of features for each model, the mean and standard deviation (std) values of the MA per mosquito trap in the validation set, the MAE between the actual and the predicted MA, and the MAE with respect to the distribution of the actual values.
Table 2. Performance of each AOI model. The table shows the number of features for each model, the mean and standard deviation (std) values of the MA per mosquito trap in the validation set, the MAE between the actual and the predicted MA, and the MAE with respect to the distribution of the actual values.
Case# FeaturesMeanStdMAEScaled MAE
Veneto, Italy2487112570.51
Upper Rhine Valley, Germany171651160.31
Pancevo, Serbia121892021180.58
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Tsantalidou, A.; Arvanitakis, G.; Georgoulias, A.K.; Akritidis, D.; Zanis, P.; Fornasiero, D.; Wohlgemuth, D.; Kontoes, C. A Data Driven Approach for Analyzing the Effect of Climate Change on Mosquito Abundance in Europe. Remote Sens. 2023, 15, 5649. https://doi.org/10.3390/rs15245649

AMA Style

Tsantalidou A, Arvanitakis G, Georgoulias AK, Akritidis D, Zanis P, Fornasiero D, Wohlgemuth D, Kontoes C. A Data Driven Approach for Analyzing the Effect of Climate Change on Mosquito Abundance in Europe. Remote Sensing. 2023; 15(24):5649. https://doi.org/10.3390/rs15245649

Chicago/Turabian Style

Tsantalidou, Argyro, George Arvanitakis, Aristeidis K. Georgoulias, Dimitris Akritidis, Prodromos Zanis, Diletta Fornasiero, Daniel Wohlgemuth, and Charalampos Kontoes. 2023. "A Data Driven Approach for Analyzing the Effect of Climate Change on Mosquito Abundance in Europe" Remote Sensing 15, no. 24: 5649. https://doi.org/10.3390/rs15245649

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop