Temperature is the key weather determinant of Aedes albopictus seasonal activity in southern France

9 The presence and the activity of Aedes albopictus are a growing concern for nuisance and public 10 health in Europe. Vector control operators, public health officers, and communities look for 11 weather-based decision support systems to inform mosquito management policies. Despite an 12 increasing number of entomological and modelling studies, our incomplete understanding of 13 mosquito population response to weather drivers in natural conditions restricts the development of 14 sound vector management policies. 15 Here, we aim to clarify the role of weather conditions on Ae. albopictus presence and abundance in 16 four sites in southwestern France. We rely on ovitrap longitudinal records collected on a 1 -2 weeks 17 basis and weather time series over 2023 and 2024 to model oviposition activity. Our analysis 18 combines a mechanistic model from literature and a new machine learning model fitted on cross -19 correlated lagged weather predictors. 20 Both models satisfactorily reproduce the observed oviposition dynamics, correctly predicting the 21 onset and the end of the activity – periods that existing models have often inadequately captured. 22 Temperature plays a major role in triggering the presence of Ae. albopictus , explaining the 23 interannual variation of oviposition in all sites, especially in spring and autumn. In fact, warm springs 24 and autumns extend the periods in which Ae. albopictus life-history traits (fertility, development, 25 survival) approach their thermal optima. In summer, a more prominent role of rain and humidity 26 emerges among secondary drivers of oviposition intensity. 27 This work contributes to the development of operational weather -driven forecasting tools for Ae. 28 albopictus activity to support vector control operations in different biogeographical contexts. 29 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint

30 The Asian tiger mosquito Aedes albopictus is an arthropod native to tropical rainforest of Southeast 31 Asia (Paupy et al., 2009). In the last century, facilitated by globalization of shipping, it invaded both 32 tropical and temperate areas worldwide. In mainland France, its first establishment occurred in 33 2004 in the southeastern regions (Roche et al., 2015) . Its biting nuisance and its competence to 34 transmit arbovirus, such as dengue and chikungunya, have raised widespread concerns about its 35 control. In mainland France, the first autochthonous transmission of these arboviruses occurred in 36 2010 in Nice (dengue) and Fréjus (chikungunya ; Franke et al., 2019) . Consistently with the 37 intensification of outbreaks in tropical areas, the number of autochthonous cases of Aedes-38 transmitted arboviral disease have gradually increased in Southern Europe, reaching more than 737 39 cases in mainland France and 353 cases in Italy up to the 8th of October 2025 (Istituto Superiore di 40 Sanità, 2025; Santé Publique France, 2025). 41 The need to anticipate both mitigate biting nuisance and vector -borne disease risk has prompted 42 vector control operators, public health officials, and citizens to try to interpret the seasonal 43 dynamics of Ae. albopictus under the light of possible drivers. These stakeholders adapt their 44 behavior following decision rules – generally supported by scientific evidence – based on a range of 45 models, from simple approache s, as calendar dates (Franke et al., 2019) , to complex model 46 ensembles (Da Re et al., 2025). Ultimately, most models assume – either implicitly or explicitly – that 47 the seasonal dynamics of Ae. albopictus depend on weather conditions. Eggs need water, usually 48 provided by precipitations , to hatch and develop into larvae and pupae (Paupy et al., 2009) . 49 Temperature affects Ae. albopictus development, survival and fertility (Mordecai et al., 2019) . 50 Humidity is also considered as important for adult survival, and wind patterns plays a minor role on 51 its dispersion (Waldock et al., 2013) . Beside these broad principles, there are evidences that Ae. 52 albopictus populations exhibit signatures of local adaptation to novel selective pressures, reflecting 53 evolutionary processes that facilitate the establishment in diverse environments. For instance, 54 populations in western Madagascar are adapted to long dry periods (Raharimalala et al., 2012) ; in 55 temperate areas, they survive to winter thanks to photoperiodically -induced egg diapause, which 56 evolved rapidly during invasion of poleward areas (Lacour et al., 2015; Urbanski et al., 2012) , with 57 eggs being resistant to colder temperature compared to their tropical counterparts (Kramer et al., 58 2021). These adaptive capabilities challenge our possibility to extrapolate current knowledge about 59 Ae. albopictus into newly colonized areas, making it more difficult to anticipate adapted measures 60 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint for the management, surveillance, and control of this species. For example, quantifying the risk that 61 – based on weather conditions – an introduction of an arbovirus -infected person could lead to an 62 outbreak helps operators to prioritize interventions according to available resources. Despite 63 growing research effort to clarify the role of weather determinants of Ae. albopictus activity (Brass et 64 al., 2024; Erguler et al., 2016; Kobayashi et al., 2002; Metelmann et al., 2019; Petrić et al., 2021; Tran 65 et al., 2013) , local authorities, public health services and mosquito control agencies often lack of 66 solid scientific evidence to support their policies. These are ultimately jeopardized by climate 67 changes, expected to affect the seasonality of this vector (Colón-González et al., 2021). 68 Our study aims to fill this gap and enable better anticipation of the nuisance and health risk 69 associated with Ae. albopictus. We focus on four different ovitrap longitudinal records collected on 70 a 1-2 weeks basis and weather time series over 2023 and 2024 in Occitanie and Nouvelle-Aquitaine, 71 southern France, which have been colonized after 2010. We model oviposition dynamics using both 72 weather-driven statistical and mechanistic approaches, and compare their predictions with ovitrap 73 observations. For these sites, we statistically infer the presence of an interannual trend in terms of 74 ovipositing activity. Therefore, we explore the importance of weather and weather -driven 75 demographical determinants of mosquito trends. Finally, we conclude on some remarks about the 76 possibility of forecasting ovipositing activity and its implications for optimizing mosquito control 77 operations in the context of the climate change. 78

79 Entomological longitudinal analysis 80 Our analysis basis on entomological surveillance observations provided by the Altopictus vector -81 control agency ovitrap collection in four different locations in mainland France, i.e. Pérols and 82 Murviel-les-Montpellier (Occitanie), Bayonne and Saint -Médard-en-Jalle (Nouvelle -Aquitaine; Fig. 83 1). According to the Corine Land Cover classification , the monitored areas correspond to 84 discontinuous urban fabric, whereas some ovitraps in Bayonne were located within continuous 85 urban fabric (European Union’s Copernicus Land Monitoring Service information, 2020) . Ovitraps 86 consist of artificial egg-laying containers made of 3-Liter black plastic buckets filled with 2 L of Bti -87 treated (Bacillus thuringiensis israelensis biolarvicide) tap water. A floating polystyrene square (5 x 88 5 x 2 cm) provided a substrate for oviposition. They are used to monitor the oviposition intensity, i.e. 89 the abundance of laid eggs, a proxy of the density of locally active females. These ovitraps have been 90 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint inspected every 1 -2 weeks between 2023 and 2024 (Tab. SI4). Surveillance operations were 91 interrupted from June 2024 onwards in Bayonne and Saint -Médard-en-Jalle. Records have been 92 aggregated using the mean value for each inspection date and site. 93 94 Fig. 1 - Spatial setup of the Altopictus ovitrap surveillance network in southwestern France: a) Pérols, b) Murviel -les-95 Montpellier, c) Bayonne, d) Saint-Médard-en-Jalle, over a OpenStreetMap land cover and land use layer. 96 To assess the impacts of different weather conditions over oviposition activity, we first analyzed the 97 ovitrap series site by site. Each time series was divided into three seasons, i.e. “spring” (May-June), 98 “summer” (July -August) and “autumn” (September -October). For each site and season, we 99 determined the presence of an interannual trend via the two -tailed two-samples Wilcoxon test by 100 comparing the statistical distribution of the ovitrap records of the two years of collection. 101 Input data 102 Both mechanistic and statistic models are driven by weather inputs. We retrieved daily precipitation 103 (mm), daily mean, maximum and minimum temperatures ( °C), relative humidity (%), average and 104 maximum wind speed (m/s) from MétéoFrance (details in Tab SI4). We retrieved human population 105 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint density raster at 30”, aggregated over a grid of 0.1° , based on the 2015 GPWv4 data for 2015 (Doxsey-106 Whitfield et al., 2015). 107 Statistical model 108 In statistical modelling, the response variables and their predictors have to be defined first. Here, we 109 considered the oviposition (measured as “ daily number of eggs per ovitrap ”), averaged by site and 110 collection date, as the response variable. To distinguish between those weather drivers that induce 111 the onset of mosquito activity and those governing oviposition intensity, we developed two separate 112 models. The first model, called “presence/absence model”, aims to identify the conditions that 113 allow mosquito activity with a binary response (1 = presence of eggs, 0 = absence). The second, 114 called “abundance model”, aims to predict the oviposition intensity as a continuous variable (mean 115 number of eggs per trap per day, restricted to positive trapping events). The modeling pipeline 116 described below was implemented for both model types. 117 To investigate the temporal relationships between mosquito activity and weather conditions, we first 118 computed cross-correlation maps (CCMs) between the response variable and weekly averaged 119 weather variables at lags ranging from 0 to 12 weeks prior to each trapping date (Curriero et al., 120 2005). CCMs enable to assess and visualize associations between two time series, such as 121 mosquito abundance and weather conditions, which may be lagged in time. Here, we measured the 122 strength of the association using the Distance correlation coefficient (Székely et al., 2007). Distance 123 correlation captures non-linear associations between variables, hence being more suited to analyze 124 complex ecological systems compared to linear correlation coefficients. These CCMs were 125 computed for each weather variable at each site to identify site -specific patterns, and were also 126 pooled across all sites to assess overall associations for each environmental variable. 127 For each pooled weather variable, the lag with the strongest correlation to the response variable was 128 retained as a predictor for multivariate modeling. Weather variables displaying weak associations 129 (distance correlation coefficient 0.7), by selecting those 131 predictors with the highest ecological relevance and interpretability. 132 The final set of selected predictors was used to train multivariate Random Forest (RF) models 133 (Breiman, 2001). Binary classification RFs were applied for presence modeling, while regression RFs 134 were used for abundance modeling. To avoid overfitting and inflated performance estimates, we 135 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint implemented a spatial leave-one-site-out cross-validation (LOSO-CV) framework. In this approach, 136 models were iteratively trained on data from three sites and tested on the fourth. This strategy not 137 only mitigates overfitting but also provides a robust assessment of the model’s generalizability to 138 previously unseen locations, for both presence/absence and abundance predictions (Meyer et al., 139 2018). Finally, the outputs of the presence and abundance models were combined to generate a 140 unified prediction. Specifically, if the predicted probability of presence exceeded the threshold value 141 of 0.5, the corresponding value from the abundance model was retained; otherwise, the prediction 142 was set to zero. 143 Mechanistic model 144 As a mechanistic model, we used the model by Metelmann et al. (2019), currently used and validated 145 in temperate areas in Europe (Barman et al., 2025; Radici et al., 2025), which describes the dynamics 146 of the mosquito population (mosquito/ha) via ordinary differential equations into five stages (eggs 𝐸, 147 diapausing eggs 𝐸𝑑, larvae and pupae as juveniles 𝐽, unfed immature female adults 𝐼, mature female 148 adults 𝐴). These equations specify mortality, development and fertility rates, which rely on 149 environmental (photoperiod 𝑃 and human density 𝐻) and weather (rainfall 𝑅 and temperature 𝑇) 150 drivers. 151 { 𝐸˙ =𝛽(𝑇)(1−𝜔(𝑃))𝐴−(ℎ(𝑅,𝐻)𝛿𝐸+𝜇𝐸(𝑇))𝐸 𝐽˙=ℎ(𝑅,𝐻)(𝛿𝐸𝐸+𝜎(𝑇,𝑃)𝛾(𝑇)𝐸𝑑)−(𝛿𝐽(𝑇)+𝜇𝐽(𝑇)+ 𝐽 𝐾(𝑅,𝐻))𝐽 𝐼˙=1 2𝛿𝐽(𝑇)𝐽−(𝛿𝐼(𝑇)+𝜇𝐴(𝑇))𝐼 𝐴˙=𝛿𝐼(𝑇)𝐼−𝜇𝐴(𝑇)𝐴 𝐸𝑑˙ =𝛽𝜔(𝑃)𝐴−ℎ(𝑅,𝐻)𝜎(𝑇,𝑃)𝐸𝑑 152 153 Where 𝛽 represent the fertility rate, 𝛿𝐸, 𝛿𝐽 and 𝛿𝐼the development rate of eggs, juveniles and adults, 154 𝜇𝐸, 𝜇𝐽, 𝜇𝐴, their mortality, 𝜔 the fraction of diapausing laid eggs, ℎ the hatching rate, 𝜎 the fraction of 155 diapausing eggs ready to hatch, 𝛾 the probability of winter survival, 𝐾 the carrying capacity of 156 juveniles (Table SI1 for details on weather dependencies ). This model allowed to estimate the 157 observed oviposition abundance via the indicator 𝐿𝐸𝑖𝑡 of eggs laid per day per hectare in site at time 158 in site 𝑖 at time 𝑡: 159 𝐿𝐸𝑖𝑡=𝛽̃𝑖𝑡𝐴𝑖𝑡 160 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Where 𝐴𝑖𝑡 is the simulated density of adult female mosquitoes, and 𝛽̃𝑖𝑡 is the temperature -161 dependent fertility rate. In order to compare our simulations with observed ovitrap data, we averaged 162 𝐿𝐸𝑖𝑡 with a moving window of two weeks and we normalized it with respect to the highest value. 163 Model evaluation 164 We assessed the ability of both models to capture the temporal dynamics of mosquito oviposition 165 by computing the Spearman and Pearson correlation coefficients between predicted and observed 166 egg counts. The statistical significance of each correlation was evaluated using a two -sided 167 Student’s t-test. 168 Analysis of weather determinants of oviposition 169 To investigate the role of weather drivers in shaping oviposition dynamics, we interpreted the trained 170 machine learning (ML) models using Variable Importance Plots (VIPs) and Partial Dependence Plots 171 (PDPs). VIPs measure each variable’s contribution to the model performance, while PDPs visualize 172 its average effect on predictions, accounting for interactions with other predictors. These tools 173 enable to highlight which environmental variables mostly influence mosquito presence or 174 abundance. 175 To complement these global interpretations with fine-scale ecological insights, we applied the Local 176 Interpretable Model-agnostic Explanations (LIME) method (Ribeiro et al., 2016). LIME approximates 177 the complex ML model locally using a simpler and interpretable surrogate (e.g. a linear model). It 178 provides local contributions of each predictor to individual model predictions, indicating both the 179 direction (positive = increasing, negative = decreasing) and the magnitude of influence relative to 180 other predictors at that specific site and time point. 181 Analysis of weather-driven demographic determinants of oviposition 182 For the sites and seasons in which the Wilcoxon test of ovitrap abundance revealed a significant 183 difference between 2023 and 2024 (p -value < 0.05), we analyzed the statistical significance of the 184 weather-dependent demographic rates between those two years using the same Wilcoxon test. The 185 weather-dependent demographic rates considered here are the fertility (𝛽), juvenile development 186 (𝛿𝐽), immature adult development ( 𝛿𝐼), egg survival ( 𝑒−𝜇𝐸), juvenile survival ( 𝑒−𝜇𝐽), adult survival 187 (𝑒−𝜇𝐴), hatching (ℎ), carrying capacity (𝐾). The carrying capacity was considered as a demographic 188 trait as its role in intraspecific competition makes it proportional to survival rate. By testing the inter-189 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint annual variability of those rates , we identified th ose that varied the most “consistently” with 190 oviposition, and can be considered as determinants of mosquito activity. 191 Hereinafter, we will use “ inconsistent” as a synonym for “significantly sharing the opposite 192 interannual trend as the observed oviposition” and, “consistent” as a synonym for “significantly 193 sharing the same interannual trend as the observed oviposition”. 194 Software 195 All analyses were conducted using open -source software. The R programming language (R Core 196 Team, 2024) served as the primary platform. The Wilcoxon statistical analysis of time series was 197 performed using the ‘stat’ package (version 4.4.1). Correlation analyses were performed with the 198 ‘correlation’ package (0.8.8; Makowski et al., 2020) . Random forest models were fitted using the 199 ‘caret’ (7.0-1; Kuhn, 2008) and ‘ranger’ (0.17.0) packages, while spatial folds for cross -validation 200 were generated with the ‘CAST’ package (1.0.3). Model interpretability analyses relied on the ‘vip’ 201 (0.4.1; Greenwell & Boehmke, 2020) and ‘pdp’ (‘0.8.2’; Greenwell, 2017) packages to produce 202 variable importance and partial dependence plots, respectively, and the ‘lime’ package (0.5.3, 203 Hvitfeldt et al., 2022) to perform the LIME analysis. Among the inputs of the mechanistic model, we 204 retrieved the photoperiod (in hours) using the ‘suncalc’ (0.5.1), based on the latitude, longitude and 205 date. 206

207 Analysis of oviposition and weather time series 208 Compared to 2023, the oviposition onset of 2024 occurred later in the season for each site except 209 Bayonne (Fig. 2, Table SI4). For both Pérols and Murviel-les-Montpellier, the last positive ovitrap was 210 found later in the year in 2023 compared to 2024. Despite the variability of ovitraps records, the 211 Wilcoxon test revealed the presence of a significant difference in observed oviposition abundance 212 between seasons (Fig. 2, Fig. 6). In Pérols, oviposition followed a bimodal pattern in 2023 and was 213 higher in spring and autumn (but not significantly), contrarily to summer season, where oviposition 214 was higher in 2024 (Fig. 2a, Fig. 6a). In Murviel-les-Montpellier, oviposition was always higher in 2023, 215 but this difference was not significant in summer ( Fig. 2b, Fig. 6b). In Saint -Médard-en-Jalle and 216 Bayonne, spring oviposition was higher in 2023 than in 2024 ( Fig. 2c,d, Fig. 6c,d). Every season in 217 every site was warmer in 2023 compared to 2024 (with the exception of summer in Bayonne and 218 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Saint-Médard-en-Jalle, in which temperatures remained roughly stable). Except for Pérols, where 219 rainfall was higher in every season in 2024 compared to 2023, there was no clear interannual trend 220 in precipitation (Tab. SI4). 221 222 Fig. 2 - Ovitrap data (solid line; mean value for each reading) in 2023 - 2024 in the 4 sites, together with average daily 223 temperature (dashed line; averaged with a monthly moving window) and monthly rainfall (bars). 224 Lagged associations between weather and oviposition dynamics 225 Cross-correlation analyses between mosquito dynamics and weather variables revealed distinct 226 temporal patterns for eggs presence and abundance across the study sites (Fig.s SI2 and SI3). 227 Temperature-related variables – average, minimum, and maximum temperature – consistently 228 emerged as the most correlated predictors for both presence and abundance. For presence, the 229 strongest associations are generally observed at 1 -8 weeks lags, while for abundance, they are 230 observed at shorter time lags (1-4 weeks). Relative humidity and rainfall-related variables displayed 231 weaker and more heterogeneous correlations, with peak correlations generally occurring at longer 232 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint time lags (up to 12 weeks). Wind-related variables generally exhibited the weakest associations with 233 mosquito dynamics, with few exceptions. The temporal structure of associations displayed a 234 notable degree of consistency across sites, especially for temperature. 235 Among the weather variables, for the presence/absence RF model, we selected: (i) average 236 temperature between the 1st and 9th week (preceding oviposition; TM_0_8), and (ii) relative humidity 237 from the 6 th to the 12 th week (UM_5_11). For the abundance RF model, the retained predictors 238 included: (i) average temperature from the 1st to the 5th week (TM_0_4), (ii) relative humidity from the 239 1st to the 12th week (UM_0_11), and (iii) cumulative rainfall from the 2nd to the 6th week (RR_1_5). 240 Model performance assessment 241 Both statistic and mechanistic models globally captured the seasonal trend (Fig. 3) – except ML 242 models in Bayonne (Fig. 3g). The beginning and the end of the seasons were globally better captured 243 by ML models, especially in those locations where the whole time series was available (Pérols, 244 Murviel; Fig. 3e, f). In the same locations, ML models were particularly good at capturing the inter -245 seasonal trends (peaks and downs). The predictions of the models were always significantly close 246 to the observations (Tab. 1). 247 248 Fig. 3 - Observed oviposition vs simulated oviposition in each site. First row: normalized observations versus predictions of 249 the mechanistic model. Second row: absolute observations versus the machine-learning model predictions. 250 Table 1 - Performances of the abundance models. All the correlation values are significative at a p-value < 0.001. 251 Site Spearman’s correlation coefficient Pearson’s correlation coefficient Mechanistic ML abundance Mechanistic ML abundance Pérols 0.87 0.92 0.60 0.84 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Murviel-les-Montpellier 0.89 0.9 0.62 0.78 Bayonne 0.88 0.66 0.92 0.23 Saint-Médard-en-Jalle 0.92 0.95 0.91 0.86 Analysis of environmental determinants 252 Pooled-site effects of weather variables on oviposition activity – Variable importance plots (VIPs) 253 and partial dependence plots (PDPs) highlighted the dominant role of temperature in predicting both 254 mosquito presence and abundance (Fig. 4). For the presence model, average temperature between 255 the 1st and 9th weeks (before oviposition; TM_0_8) was by far the most important predictor across all 256 sites, while relative humidity during weeks 6 to 12 (UM_5_11) had only a minor impact ( Fig. 4a). The 257 PDPs showed a sharp increase in the probability of mosquito presence as TM_0_8 rises from 258 approximately 11 °C to 18 °C, after which the probability saturated ( Fig. 4b). Relative humidity 259 exhibited a weaker and more complex relationship – generally negative – with presence probability 260 (Fig. 4c). 261 In the abundance model , average temperature weeks 1 to 5 (TM_0_4) was the most influential 262 predictor, followed by relative humidity (UM_0_11) and cumulative rainfall (RR_1_5; Fig. 4d). The 263 PDPs revealed a strong, linear increase in predicted egg abundance with TM_0_4 ranging from 15 °C 264 to ~24 °C, after which the abundance saturated and then decreased ( Fig. 4e). In contrast, higher 265 relative humidity was generally associated with a decrease in predicted abundance ( Fig. 4f). The 266 effects of rainfall differed across sites, showing a slightly positive effect on oviposition intensity in 267 Pérols and Murviel -lès-Montpellier, while appearing negligible in Bayonne and Saint -Médard-en-268 Jalles ( Fig. 4g). The pooled bivariate PDP for TM_0_4 and RR_1_5 highlighted an optimal range of 269 conditions for oviposition activity , with maximum predicted egg abundance occurring near 25  °C 270 average temperature of the previous 5 weeks (TM_0_4) and approximately 150  mm cumulative 271 rainfall over the 2nd to 5th weeks before deposition (RR_1_5; Fig. 4h). 272 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint 273 Fig. 4 - VIPs and PDPs for the mosquito presence (top) and abundance (bottom) models. Bar plots (a,d; left) show variable 274 importance, indicating the relative contribution of each environmental predictor to the RF model. Line plots (b, c, e, f, g; 275 right) display partial dependance, representing the marginal effect of each predictor on the response variable while 276 averaging over the influence of other variables. The bivariate PDP in the bottom panel (h) illustrates the combined effect of 277 temperature (TM_0_4) and rainfall (RR_1_5). For the presence model, the response variable is the probability of oviposition 278 occurrence; for the abundance model, it is the predicted number of eggs per trap (restricted to presence-only data). These 279 plots are based on models trained on pooled data across all sites. 280 Local spatio-temporal effects of weather variables on oviposition activity – Across all sites and 281 weeks, temperature consistently emerged as the most influential predictor driver, with strong 282 positive contributions aligning closely with observed peaks in egg abundance, particularly during the 283 warmer months (Fig. 5, and SI4-7). 284 Rainfall and humidity, by contrast, showed more variable and context -dependent effects. Notably, 285 in Pérols and Murviel -lès-Montpellier, distinct declines in predicted abundance were observed 286 during summer ( Fig. 5a,b). In this season, temperature remained positively associated with 287 oviposition intensity, whereas rainfall made a negative contribution to the model predictions. 288 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint 289 Fig. 5 - Local contribution of environmental predictors to predicted mosquito abundance using LIME. 290 Each colored tile represents the contribution of a given environmental variable to the model’s predicted egg presence/ and 291 abundance for a specific site and time point (weekly resolution, 2023 –2024). Colors indicate whether the variable 292 contributed positively (blue) or negatively (red) to the prediction, with color intensity proportional to the magnitude of th e 293 effect. 294 Analysis of weather-driven mosquito demographic determinants 295 In the mechanistic model, temperature-related parameters, with the exclusion of egg survival, were 296 globally consistent with ovitrap observations ( Fig. 6). Fertility (𝛽), juvenile development ( 𝛿𝐽), 297 immature adult development (𝛿𝐼) and adult survival (𝑒−𝜇𝐴) were consistent 5 times out of 6, juvenile 298 survival (𝑒−𝜇𝐽) 3 out of 6. In all the other cases, the trend was neither significant nor inconsistent. By 299 contrast, observed oviposition was consistent with rainfall-dependent demographic traits – hatching 300 (ℎ) and carrying capacity ( 𝐾) – only in summer in Pérols. In all other cases, they were either not 301 significant or inconsistent. 302 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint 303 Fig. 6 - Consistency analysis per site, year and season between oviposition intensity (boxplots in the first row) against 304 mechanistic weather -driven parameters (boxplots from the second row). For each site and season, the boxplots 305 corresponding to the year of highest oviposition are colored in blue. Then, the boxplots of those combinations of site and 306 season in which the parameters varied consistently(/inconsistently) with oviposition are colored in green(/yellow). Gray 307 boxplots mean lack of significance. The stars indicate the p -value (*: < 0.05, ***: < 0.01, ***: < 0.001) according to the one -308 to-one Wilcoxon test. 309 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint

310 Average temperatures of the previous weeks are the main drivers of oviposition in Southern 311 France – In the study area, interannual trends of ovipositing activity – a widespread proxy used to 312 monitor the activity of adult females (Da Re et al., 2024) having taken a blood meal to develop their 313 eggs – can be easily explained by a limited number of weather drivers, which impact different life -314 history traits. Average temperature of the previous 9 weeks is by far the main determinant of the 315 presence of eggs, and of the previous 5 for their abundance. The importance of considering lagged 316 temperatures into the description of oviposition dynamics in southern Europe is confirmed by recent 317 research – despite the disagreement with these studies on the amplitude of the lag. For instance, 318 Torina et al. (2023), found that average temperature of the previous 5 weeks is the most correlated 319 driver to observed oviposition intensity in Palermo, Italy, while Da Re et al., (2025) found that average 320 temperature of the third-to-last week is the most important determinant of oviposition intensity over 321 the area covered by the VectAbundance survey in southeastern Europe (Da Re et al., 2024) . These 322 differences may reflect either local population dynamics or the results of different modelling 323 choices. Concerning diapause end, it is interesting to notice that our results allow to correctly 324 capture and satisfactorily explain the onset of mosquito activity also neglecting the photoperiod, the 325 main driver of egg diapause (i.e., the population overwintering through the production of eggs that 326 will hatch in spring; Lacour et al., 2015; Lounibos et al., 2003; Sturiale & Armbruster, 2023). 327 Warm temperatures facilitate longer and more intense activity seasons of Aedes albopictus in 328 a temperate climate – Wide research focuses on temperature as the main driver of the ecology of 329 ectotherm populations. Mordecai et al. (2019) formalized the performance of a corpus of mosquito 330 life-history traits as an (asymmetric) bell -shaped function of temperature (Briere et al., 1999; 331 Waldock et al., 2013). The optimal thermal response of Ae. albopictus on several key biological traits 332 varies between 24.2 °C (survival of aquatic stages) to 32.6 °C (adult development rate; de Souza & 333 Weaver, 2024). For temperatures lower than the optimal ones, as it regularly occurs in temperate 334 countries, a temperature increase corresponds to an increment of mosquito fitness (de Souza & 335 Weaver, 2024). Our analysis of demographic determinants in the mechanistic model corroborate s 336 this idea; temperature -dependent traits (i.e. survival, development or fertility rates) varied 337 consistently with oviposition intensity across all sites during cool seasons (spring and autumn ) 338 characterized by suboptimal temperatures. 339 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint It is worth noticing that climate change increases the likelihood of warmer springs and autumns, with 340 important implications on the epidemiological risk. In fact, in a temperate climate such as that of 341 southern France, the increase of temperatures could lead to the extension of the season at risk of 342 autochthonous transmission of arboviruses, as noted by other studies (Colón-González et al., 2021; 343 Radici et al., 2025). In fact, temperature affects the extrinsic incubation time of arboviruses, which 344 develop and are transmitted more easily at hot temperatures (Liu-Helmersson et al., 2014) . It is 345 therefore advisable that the calendar for enhanced surveillance, which currently begins in May in 346 mainland France (Franke et al., 2019), should be adjusted accordingly. 347 Rainfall has a minor role in sustaining mosquito activity in southern France – The analysis of 348 environmental determinants indicates that rainfall plays a secondary role in shaping population 349 abundance dynamics, while it does not affect the presence. Consistently, the analysis of 350 environmentally-driven demographics suggests that rainfall becomes determinant only during 351 summer. The primary effects of rainfall are to trigger egg hatching and to increase the availability of 352 larval breeding sites, the main limiting resource to mosquito population growth (Waldock et al., 353 2013). Yet, the importance of rainfall in sustaining larval survival – and therefore mosquito 354 population – in urban settings is increasingly debated (Boyer et al., 2014). Urbanization has led to an 355 increased capacity to retain water – whether from rainfall or irrigation – due to the expansion of 356 impervious surfaces (gutter silt traps, raised terraces, roadworks, flowers pots, rainfall collectors, 357 etc.). In many cases, human irrigation compensates for reduced precipitation, providing persistent 358 larval refugia that reduce mosquito populations dependence from natural water supply (Li et al., 359 2014). It is not surprising that rainfall -dependent parameters only become the limiting factors to 360 oviposition only during the driest season in the driest locations (~ 21 mm of cumulative rainfall in 361 June and July in Pérols, while Bayonne has ~110; Tab. SI4). Moreover, Fonseca et al. (2015) suggested 362 that in autumn, females laying diapausing eggs exhibit a preference for larger containers; this 363 behavioral shift reduces the dependance of the first generations of larvae on rainfall abundance (or 364 scarce evaporation). 365 The VIP analysis highlights a secondary role of relative humidity, which affects negatively both 366 presence and abundance of oviposition. Relative humidity has long been recognized as a challenging 367 variable to embed into a modelling scheme due to its complex interactions with other weather 368 variables (Waldock et al., 2013). For instance, high humidity is expected to enhance eggs survival at 369 26 °C, but the same does not hold consistently across other temperatures (Juliano et al., 2002). The 370 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint effect of humidity on adult mosquitoes is even less understood: European populations of Asian tiger 371 mosquito have been found also in regions with relative humidity levels below 35% (Waldock et al., 372 2013). In conclusion, the range of values we measured in the study area is probably not broad 373 enough – between 60 and 85% – to have an appreciable effect on mosquito dynamics. 374 The effects of extreme weather events on the mosquito activity in southern France – Summer 375 temperatures may exceed the optimal range of Ae. albopictus. This occurred in the warmer sites of 376 Pérols and Murviel-les-Montpellier, where the PDP reveal a slight decrease of abundance for average 377 monthly temperatures exceeding 24 °C. While on -field evidence of this phenomenon in temperate 378 climate is poor, other modelling study document the negative effect of summer heatwaves on 379 mosquito fitness (Garrido Zornoza et al., 2024) . As a matter of fact, it is difficult to disentangle the 380 effect of heatwaves from that of droughts, as in temperate climate they often occur simultaneously. 381 Notably, in Pérols and Murviel -lès-Montpellier, distinct declines in predicted abundance were 382 observed during summer 2023, despite suboptimal but still favorable temperature conditions. In 383 these dates, the LIME analysis provided an insight on the role of low rainfall, which makes a negative 384 contribution to the model predictions, suggesting that low or insufficient precipitation results in 385 limited oviposition. In contrast, in Bayonne and Saint -Médard-en-Jalles, rainfall and humidity show 386 minimal or inconsistent influence across the season, indicating a reduced predictive value of these 387 variables under the prevailing local oceanic conditions. 388 Across the available time series, we have not observed extreme rainfall events. Several studies 389 reported that excessive rain may flush out breeding sites, potentially disrupting mosquito population 390 dynamics (Waldock et al., 2013). However, the 2014 epidemic of chikungunya in the Montpellier area 391 – which includes Pérols and Murviel -les-Montpellier – following an exceptional rain event (299 mm 392 on the 29th of September) lead researchers to reconsider the effect of extreme precipitation. In such 393 cases, heavy rainfall may also create an unprecedented number of new breeding sites, ultimately 394 boosting the mosquito population (Roiz et al., 2015; White et al., 2025). While our analysis highlights 395 that rainfall plays a secondary role in driving oviposition intensity, the PDPs indicate that cumulative 396 rainfall exceeding 150–250 mm over the preceding five weeks may be associated with a decrease in 397 oviposition intensity. 398

of the study – Some caveats must be considered before generalizing the outcomes of 399 this study. The results we obtained need to be contextualized within a specific environment in terms 400 of landscape, climate and social features – specifically, a temperate climate within a western 401 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint European urban fabric. These highly anthropized sites are located approximately at the same 402 latitude, close to the coasts. We believe that our results can be generalized to other contexts that 403 share similar environmental features in terms of climate and urbanization. However, extending 404 entomological observations both in time and space would help capturing a wider range of 405 environmental conditions and associated mosquito responses. This would help solving site-specific 406 discrepancies (such as underestimating extreme abundance peaks), typical of ML modelling, 407 ultimately strengthening its operational value for vector control. 408 Some obstacles to the generalizability of any study on Ae. albopictus are its high phenotypical 409 plasticity and adaptation capability, for example of diapause, whose rapid evolution due to the 410 latitudinal expansion of this species requires constant surveillance. Although photoperiod has a 411 minor role in our study , it remains the primary regulator of winter egg diapause, and its evolution 412 would alter this species’ seasonality. Some studies suggested that year-round activity may become 413 possible in the future due to climate change in Southern Europe (Del Lesto et al., 2022). 414 Weather-driven models as decision support systems in mosquito management – Our models 415 accurately predicted the onset of mosquito activity after diapause, a hard benchmark for many 416 entomological models, as well as variations of abundances during the mosquito activity season. 417 These results reinforce the complementary of statistical and mechanistic tools to analyze vector 418 dynamics – as previously noted by Tran et al. (2020) in La Réunion – but also their potential to support 419 operational interventions. The outcome of these models can assist vector control operators in 420 prioritizing periods or geographical areas for vector control interventions, or enable health 421 authorities to activate targeted surveillance activities. For instance, the early prediction of the spring 422 onset of mosquito activity would allow for the planning of timing anti -larval interventions. Similarly, 423 health services and mosquito control operators may be mobilized as average temperatures exceed 424 ~18°C over five weeks, at which we obtained the steepest abundance increase, which has 425 consequences over arbovirus transmission . The integration of such modelling outcomes into the 426 routines of mosquito control operators, health services, and local authorities would allow quick 427 actions to abate the health risk. 428 While examples of similar operational tools exist – such as the ARBOCARTO framework (Marti et al., 429 2022), which is already used in some regions of France and abroad , or the ea rly warning system 430 described by Díaz et al. (2024) and used in the Caribbean – there remains substantial room to 431 improve their societal uptake. The development of ARBOCARTO highlights that creating effective 432 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint tools to predict mosquito activity requires strong coordination among a wide range of stakeholders, 433 which is not limited to the collaboration between field entomologists and ecological modelers. 434 Beyond the scientific community, such an effort calls for collaboration between public health 435 authorities, vector -control operators, local political officials, urban management services, and 436 representatives of civil society. Only through this integrative, multi-actor approach can surveillance 437 tools be effectively translated into actionable strategies for public health and vector control. Lastly, 438 to enhance the effectiveness of these models and tools, they should be designed to forecast 439 mosquito activity (in the future) – and not only predict it (in the present). 440

441 The authors acknowledge the support of RIVOC funded by the Occitanie region in France for the 442 VECTOCLIM project. We thanks Colombine Bartholomée, Annelise Tran, Renaud Marti, Florence 443 Fournet and Cyril Caminade for fruitful exchanges. 444

445 Barman, S., Semenza, J. C., Singh, P., Sjödin, H., Rocklöv, J., & Wallin, J. (2025). A climate and 446 population dependent diffusion model forecasts the spread of Aedes Albopictus 447 mosquitoes in Europe. Communications Earth & Environment, 6(1), 276. 448 https://doi.org/10.1038/s43247-025-02199-z 449 Boyer, S., Foray, C., & Dehecq, J.-S. (2014). Spatial and Temporal Heterogeneities of Aedes 450 albopictus Density in La Reunion Island: Rise and Weakness of Entomological Indices. 451 PLOS ONE, 9(3), e91170. https://doi.org/10.1371/journal.pone.0091170 452 Brass, D. P., Cobbold, C. A., Purse, B. V., Ewing, D. A., Callaghan, A., & White, S. M. (2024). Role of 453 vector phenotypic plasticity in disease transmission as illustrated by the spread of dengue 454 virus by Aedes albopictus. Nature Communications, 15(1), 7823. 455 https://doi.org/10.1038/s41467-024-52144-5 456 Breiman, L. (2001). Random Forests. Machine Learning, 45(1), 5–32. 457 https://doi.org/10.1023/A:1010933404324 458 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Briere, J.-F., Pracros, P., Le Roux, A.-Y., & Pierre, J.-S. (1999). A Novel Rate Model of Temperature-459 Dependent Development for Arthropods. Environmental Entomology, 28(1), 22–29. 460 https://doi.org/10.1093/ee/28.1.22 461 Colón-González, F. J., Sewe, M. O., Tompkins, A. M., Sjödin, H., Casallas, A., Rocklöv, J., 462 Caminade, C., & Lowe, R. (2021). Projecting the risk of mosquito-borne diseases in a 463 warmer and more populated world: A multi-model, multi-scenario intercomparison 464 modelling study. The Lancet Planetary Health, 5(7), e404–e414. 465 https://doi.org/10.1016/S2542-5196(21)00132-7 466 Curriero, F., Shone, S., & Glass, G. (2005). Cross Correlation Maps: A Tool for Visualizing and 467 Modeling Time Lagged Associations. Counselling Psychology Review, 5(3), 267–275. 468 https://doi.org/10.53841/bpscpr.2016.31.1.67 469 Da Re, D., Marini, G., Bonannella, C., Laurini, F., Manica, M., Anicic, N., Albieri, A., Angelini, P., 470 Arnoldi, D., Bertola, F., Caputo, B., De Liberato, C., della Torre, A., Flacio, E., Franceschini, 471 A., Gradoni, F., Kadriaj, P., Lencioni, V., Del Lesto, I., … Rosà, R. (2025). Modelling the 472 seasonal dynamics of Aedes albopictus populations using a spatio-temporal stacked 473 machine learning model. Scientific Reports, 15(1), 3750. https://doi.org/10.1038/s41598-474 025-87554-y 475 Da Re, D., Marini, G., Bonannella, C., Laurini, F., Manica, M., Anicic, N., Albieri, A., Angelini, P., 476 Arnoldi, D., Blaha, M., Bertola, F., Caputo, B., De Liberato, C., della Torre, A., Flacio, E., 477 Franceschini, A., Gradoni, F., Kadriaj, P., Lencioni, V., … Rosà, R. (2024). VectAbundance: A 478 spatio-temporal database of Aedes mosquitoes observations. Scientific Data, 11(1), 636. 479 https://doi.org/10.1038/s41597-024-03482-y 480 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint de Souza, W. M., & Weaver, S. C. (2024). Effects of climate change and human activities on vector-481 borne diseases. Nature Reviews Microbiology, 1–16. https://doi.org/10.1038/s41579-024-482 01026-0 483 Del Lesto, I., De Liberato, C., Casini, R., Magliano, A., Ermenegildi, A., & Romiti, F. (2022). Is Asian 484 tiger mosquito (Aedes albopictus) going to become homodynamic in Southern Europe in 485 the next decades due to climate change? Royal Society Open Science, 9(12), 220967. 486 https://doi.org/10.1098/rsos.220967 487 Díaz, A. R., Rollock, L., Boodram, L.-L. G., Mahon, R., Best, S., Trotman, A., Meerbeeck, C. J. V., 488 Fletcher, C., Dunbar, W., Lippi, C. A., Lührsen, D., Sorensen, C., Muñoz, Á. G., Ryan, S. J., 489 Stewart-Ibarra, A. M., & Lowe, R. (2024). A demand-driven climate services for health 490 implementation framework: A case study for climate-sensitive diseases in Caribbean Small 491 Island Developing States. PLOS Climate, 3(10), e0000282. 492 https://doi.org/10.1371/journal.pclm.0000282 493 Doxsey-Whitfield, E., MacManus, K., Adamo, S. B., Pistolesi, L., Squires, J., Borkovska, O., & 494 Baptista, S. R. (2015). Taking Advantage of the Improved Availability of Census Data: A First 495 Look at the Gridded Population of the World, Version 4. 226–234. 496 Erguler, K., Smith-Unna, S. E., Waldock, J., Proestos, Y., Christophides, G. K., Lelieveld, J., & 497 Parham, P. E. (2016). Large-scale modelling of the environmentally-driven population 498 dynamics of temperate aedes albopictus (Skuse). PLoS ONE, 11(2), 1–28. 499 https://doi.org/10.1371/journal.pone.0149282 500 European Union’s Copernicus Land Monitoring Service information. (2020). CORINE Land Cover 501 2018 (vector/raster 100 m), Europe, 6-yearlye [Dataset]. https://doi.org/10.2909/960998c1-502 1870-4e82-8051-6485205ebbac 503 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Fonseca, D. M., Kaplan, L. R., Heiry, R. A., & Strickman, D. (2015). Density-Dependent Oviposition 504 by Female Aedes albopictus (Diptera: Culicidae) Spreads Eggs Among Containers During 505 the Summer but Accumulates Them in the Fall. Journal of Medical Entomology, 52(4), 705–506 712. https://doi.org/10.1093/jme/tjv060 507 Franke, F., Giron, S., Cochet, A., Jeannin, C., Leparc-Goffart, I., De Valk, H., & others. (2019). 508 Émergences de dengue et de chikungunya en France métropolitaine, 2010–2018. Bull. 509 Epidémiol. Hebd, 19(20), 374–382. 510 Garrido Zornoza, M., Caminade, C., & Tompkins, A. M. (2024). The effect of climate change and 511 temperature extremes on Aedes albopictus populations: A regional case study for Italy. 512 Journal of The Royal Society Interface, 21(220), 20240319. 513 https://doi.org/10.1098/rsif.2024.0319 514 Greenwell, B. M. (2017). pdp: An R Package for Constructing Partial Dependence Plots. The R 515 Journal, 9(1), 421–436. 516 Greenwell, B. M., & Boehmke, B. C. (2020). Variable Importance Plots—An Introduction to the vip 517 Package. The R Journal, 12(1), 343–366. 518 Hvitfeldt, E., Lin Pedersen, T., & Benesty, M. (2022). lime: Local Interpretable Model-Agnostic 519 Explanations — lime-package. https://lime.data-imaginist.com/reference/lime-520 package.html 521 Istituto Superiore di Sanità. (2025). Arbovirosi • bollettini periodici arbovirosi. 522 https://www.epicentro.iss.it/arbovirosi/dashboard 523 Juliano, S. A., O’Meara, G. F., Morrill, J. R., & Cutwa, M. M. (2002). Desiccation and thermal 524 tolerance of eggs and the coexistence of competing mosquitoes. Oecologia, 130(3), 458–525 469. https://doi.org/10.1007/s004420100811 526 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Kobayashi, M., Nihei, N., & Kurihaha, T. (2002). Analysis of northern distribution of Aedes 527 albopictus (Diptera: Culicidae) in Japan by geographical information system. Journal of 528 Medical Entomology, 39(1), 4–11. https://doi.org/10.1603/0022-2585-39.1.4 529 Kramer, I. M., Pfeiffer, M., Steffens, O., Schneider, F., Gerger, V., Phuyal, P., Braun, M., Magdeburg, 530 A., Ahrens, B., Groneberg, D. A., Kuch, U., Dhimal, M., & Müller, R. (2021). The 531 ecophysiological plasticity of Aedes aegypti and Aedes albopictus concerning 532 overwintering in cooler ecoregions is driven by local climate and acclimation capacity. 533 Science of The Total Environment, 778, 146128. 534 https://doi.org/10.1016/j.scitotenv.2021.146128 535 Kuhn, M. (2008). Building Predictive Models in R Using the caret Package. Journal of Statistical 536 Software, 28, 1–26. https://doi.org/10.18637/jss.v028.i05 537 Lacour, G., Chanaud, L., L’Ambert, G., & Hance, T. (2015). Seasonal Synchronization of Diapause 538 Phases in Aedes albopictus (Diptera: Culicidae). PLoS ONE, 10(12), 1–16. 539 https://doi.org/10.1371/journal.pone.0145311 540 Li, Y., Kamara, F., Zhou, G., Puthiyakunnon, S., Li, C., Liu, Y., Zhou, Y., Yao, L., Yan, G., & Chen, X.-541 G. (2014). Urbanization Increases Aedes albopictus Larval Habitats and Accelerates 542 Mosquito Development and Survivorship. PLOS Neglected Tropical Diseases, 8(11), e3301. 543 https://doi.org/10.1371/journal.pntd.0003301 544 Liu-Helmersson, J., Stenlund, H., Wilder-Smith, A., & Rocklöv, J. (2014). Vectorial capacity of Aedes 545 aegypti: Effects of temperature and implications for global dengue epidemic potential. PloS 546 One, 9(3), e89783. https://doi.org/10.1371/journal.pone.0089783 547 Lounibos, L. P., Escher, R. L., & Lourenço-De-Oliveira, R. (2003). Asymmetric Evolution of 548 Photoperiodic Diapause in Temperate and Tropical Invasive Populations of Aedes 549 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint albopictus (Diptera: Culicidae). Annals of the Entomological Society of America, 96(4), 512–550 518. https://doi.org/10.1603/0013-8746(2003)096%255B0512:AEOPDI%255D2.0.CO;2 551 Makowski, D., Ben-Shachar, M. S., Patil, I., & Lüdecke, D. (2020). Methods and Algorithms for 552 Correlation Analysis in R. Journal of Open Source Software, 5(51), 2306. 553 https://doi.org/10.21105/joss.02306 554 Marti, R., Castets, M., Demarchi, M., Catry, T., Besnard, G., Chouin, S., Clément, C., Esteve-555 Moussion, I., Etienne, M., Foussadier, R., Habchi-Hanriot, N., Jouanthoua, F., L’Ambert, G., 556 Godal, A., & Tran, A. (2022). ARBOCARTO: An operational spatial modeling tool to predict 557 the dynamics of Aedes mosquito species from weather and environmental variables 558 [Conference_item]. ESA. https://agritrop.cirad.fr/602160/ 559 Mcmaster, G. S., & Wilhelm, W. W. (1997). Growing degree-days: One equation, two 560 interpretations. Agricultural and Forest Meteorology, 87(1). 561 Metelmann, S., Caminade, C., Jones, A. E., Medlock, J. M., Baylis, M., & Morse, A. P. (2019). The 562 UK’s suitability for Aedes albopictus in current and future climates. Journal of the Royal 563 Society Interface, 16(152). https://doi.org/10.1098/rsif.2018.0761 564 Meyer, H., Reudenbach, C., Hengl, T., Katurji, M., & Nauss, T. (2018). Improving performance of 565 spatio-temporal machine learning models using forward feature selection and target-566 oriented validation. Environmental Modelling & Software, 101, 1–9. 567 https://doi.org/10.1016/j.envsoft.2017.12.001 568 Mordecai, E. A., Caldwell, J. M., Grossman, M. K., Lippi, C. A., Johnson, L. R., Neira, M., Rohr, J. R., 569 Ryan, S. J., Savage, V., Shocket, M. S., Sippy, R., Stewart Ibarra, A. M., Thomas, M. B., & 570 Villena, O. (2019). Thermal biology of mosquito-borne disease. Ecology Letters, 22(10), 571 1690–1708. https://doi.org/10.1111/ele.13335 572 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Paupy, C., Delatte, H., Bagny, L., Corbel, V., & Fontenille, D. (2009). Aedes albopictus, an arbovirus 573 vector: From the darkness to the light. Microbes and Infection, 11(14), 1177–1185. 574 https://doi.org/10.1016/j.micinf.2009.05.005 575 Petrić, M., Ducheyne, E., Gossner, C. M., Marsboom, C., Nicolas, G., Venail, R., Hendrickx, G., & 576 Schaffner, F. (2021). Seasonality and timing of peak abundance of Aedes 577 albopictus in Europe: Implications to public and animal health. Geospatial Health, 578 16(1), Article 1. https://doi.org/10.4081/gh.2021.996 579 R Core Team. (2024). R: A Language and Environment for Statistical Computing. R Foundation for 580 Statistical Computing. https://www.R-project.org/ 581 Radici, A., Hammami, P., Cannet, A., L’Ambert, G., Lacour, G., Fournet, F., Garros, C., Guis, H., 582 Fontenille, D., & Caminade, C. (2025). Aedes albopictus is rapidly invading its climatic 583 niche in France: Wider implications for biting nuisance and arbovirus control in Western 584 Europe (p. 2025.02.14.638223). bioRxiv. https://doi.org/10.1101/2025.02.14.638223 585 Raharimalala, F. N., Ravaomanarivo, L. H., Ravelonandro, P., Rafarasoa, L. S., Zouache, K., Tran-586 Van, V., Mousson, L., Failloux, A.-B., Hellard, E., Moro, C. V., Ralisoa, B. O., & Mavingui, P. 587 (2012). Biogeography of the two major arbovirus mosquito vectors, Aedes aegypti and 588 Aedes albopictus (Diptera, Culicidae), in Madagascar. Parasites & Vectors, 5(1), 56. 589 https://doi.org/10.1186/1756-3305-5-56 590 Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). “Why Should I Trust You?”: Explaining the 591 Predictions of Any Classifier. Proceedings of the 22nd ACM SIGKDD International 592 Conference on Knowledge Discovery and Data Mining, 1135–1144. 593 https://doi.org/10.1145/2939672.2939778 594 Roche, B., Léger, L., L’Ambert, G., Lacour, G., Foussadier, R., Besnard, G., Barré-Cardi, H., Simard, 595 F., & Fontenille, D. (2015). The Spread of Aedes albopictus in Metropolitan France: 596 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Contribution of Environmental Drivers and Human Activities and Predictions for a Near 597 Future. PLOS ONE, 10(5), e0125600. https://doi.org/10.1371/journal.pone.0125600 598 Roiz, D., Boussès, P., Simard, F., Paupy, C., & Fontenille, D. (2015). Autochthonous Chikungunya 599 transmission and extreme climate events in Southern France. PLoS Neglected Tropical 600 Diseases, 9(6), 1–8. https://doi.org/10.1371/journal.pntd.0003854 601 Roques, L., & Bonnefon, O. (2016). Modelling Population Dynamics in Realistic Landscapes with 602 Linear Elements: A Mechanistic-Statistical Reaction-Diffusion Approach. PLOS ONE, 11(3), 603 e0151217. https://doi.org/10.1371/journal.pone.0151217 604 Santé Publique France. (2025). Chikungunya, dengue, Zika et West Nile en France hexagonale. 605 Bulletin de la surveillance renforcée du 17 septembre 2025. 606 https://www.santepubliquefrance.fr/maladies-et-traumatismes/maladies-a-transmission-607 vectorielle/chikungunya/documents/bulletin-national/chikungunya-dengue-zika-et-west-608 nile-en-france-hexagonale.-bulletin-de-la-surveillance-renforcee-du-17-septembre-2025 609 Sherpa, S., Blum, M. G. B., Capblancq, T., Cumer, T., Rioux, D., & Després, L. (2019). Unravelling 610 the invasion history of the Asian tiger mosquito in Europe. Molecular Ecology, 28(9), 2360–611 2377. https://doi.org/10.1111/mec.15071 612 Sturiale, S. L., & Armbruster, P. A. (2023). Contrasting effects of an extended fall period and winter 613 heatwaves on the overwintering fitness of diapausing disease vector, Aedes albopictus. 614 Current Research in Insect Science, 4, 100067. https://doi.org/10.1016/j.cris.2023.100067 615 Székely, G. J., Rizzo, M. L., & Bakirov, N. K. (2007). Measuring and testing dependence by 616 correlation of distances. The Annals of Statistics, 35(6), 2769–2794. 617 https://doi.org/10.1214/009053607000000505 618 Torina, A., La Russa, F., Blanda, V., Peralbo-Moreno, A., Casades-Martí, L., Di Pasquale, L., 619 Bongiorno, C., Vitale Badaco, V., Toma, L., & Ruiz-Fons, F. (2023). Modelling time-series 620 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint Aedes albopictus abundance as a forecasting tool in urban environments. Ecological 621 Indicators, 150(April), 110232. https://doi.org/10.1016/j.ecolind.2023.110232 622 Tran, A., L’Ambert, G., Lacour, G., Benoît, R., Demarchi, M., Cros, M., Cailly, P., Aubry-Kientz, M., 623 Balenghien, T., & Ezanno, P. (2013). A rainfall- and temperature-driven abundance model 624 for Aedes albopictus populations. International Journal of Environmental Research and 625 Public Health, 10(5), 1698–1719. https://doi.org/10.3390/ijerph10051698 626 Tran, A., Mangeas, M., Demarchi, M., Roux, E., Degenne, P., Haramboure, M., Goff, G. L., Damiens, 627 D., Gouagna, L. C., Herbreteau, V., & Dehecq, J. S. (2020). Complementarity of empirical 628 and processbased approaches to modelling mosquito population dynamics with Aedes 629 albopictus as an example-Application to the development of an operational mapping tool 630 of vector populations. PLoS ONE, 15(1), 1–21. 631 https://doi.org/10.1371/journal.pone.0227407 632 Urbanski, J., Mogi, M., O’Donnell, D., DeCotiis, M., Toma, T., & Armbruster, P. (2012). Rapid 633 Adaptive Evolution of Photoperiodic Response during Invasion and Range Expansion across 634 a Climatic Gradient. The American Naturalist, 179(4), 490–500. 635 https://doi.org/10.1086/664709 636 Vanalli, C., Casagrandi, R., Gatto, M., & Bevacqua, D. (2021). Shifts in the thermal niche of fruit 637 trees under climate change: The case of peach cultivation in France. Agricultural and Forest 638 Meteorology, 300(September 2020). https://doi.org/10.1016/j.agrformet.2021.108327 639 Waldock, J., Chandra, N. L., Lelieveld, J., Proestos, Y., Michael, E., Christophides, G., & Parham, P. 640 E. (2013). The role of environmental variables on Aedes albopictus biology and chikungunya 641 epidemiology. Pathogens and Global Health, 107(5), 224–241. 642 https://doi.org/10.1179/2047773213Y.0000000100 643 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint White, S. M., Tegar, S., Purse, B. V., Cobbold, C. A., & Brass, D. P. (2025). Modelling the Lodi, 2023 644 and Fano 2024, Italy Dengue Outbreaks: The Effects of Control Strategies and 645 Environmental Extremes. Transboundary and Emerging Diseases, 2025(1), 5542740. 646 https://doi.org/10.1155/tbed/5542740 647 648 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 10, 2025. ; https://doi.org/10.1101/2025.10.10.681559doi: bioRxiv preprint