Relative contributions of climate factors and air pollution to childhood allergic diseases in Chongqing, China Authors: Rong Sun, Linlin Nie, Xinye Tang, Bo Pan, Jia Liang, Di Hu Published: 2026 ABSTRACT AbstractBackgroundClimate and air pollution are associated with pediatric allergic diseases; however, their relative impacts remain poorly understood, particularly in subtropical climates. This study quantified and compared the contributions of meteorological variables and air pollutants to childhood allergic rhinitis (AR), asthma and atopic dermatitis (AD).MethodsWe analyzed 1,516,996 pediatric hospital visits in Chongqing, China (2014–2019, 2023–2024), including AR (n = 456,807), asthma (n = 775,181), and AD (n = 285,008). Using distributed lag non-linear models, we examined associations with temperature, humidity, atmospheric pressure, wind speed, PM₂.₅, PM₁₀, NO₂, SO₂, CO, and O₃, comparing effect magnitudes through standardized coefficients and attributable disease burdens.ResultsTemperature demonstrated the strongest associations across all diseases, with opposing effects: positive associations with respiratory allergies (AR: RR = 1.25, 95% CI: 1.16–1.35 at 30 °C, attributing 84,640 cases [18.5% of total burden]; asthma: RR = 1.10, 95% CI: 1.05–1.16 at 32 °C, attributing 58,058 cases [7.5%]) versus inverse association with AD (RR = 0.75, 95% CI: 0.68–0.83 at 30 °C, associated with 58,270 fewer cases [20.4% reduction]). Seasonal analyses demonstrated marked heterogeneity: summer concentrated AR burden (temperature attributing 18,896 cases, 22.3% of annual burden), while winter concentrated asthma NO₂ effects (21,558 cases, 11.1% of winter burden). Individual air pollutants showed substantially smaller effects than temperature.ConclusionsTemperature demonstrated stronger short-term associations with childhood allergic disease burden than individual air pollutants in subtropical settings. Disease-specific environmental susceptibility patterns (opposing temperature effects for respiratory versus dermatological allergies) necessitate tailored prevention strategies. These results support integrating climate adaptation measures alongside air quality management in pediatric environmental health policies, with season-specific interventions during high-burden periods.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28203-5.Keywords: Allergic rhinitis, Asthma, Atopic dermatitis, Temperature, Air pollution, Climate factors Introduction In the past two decades, childhood allergic disease prevalence has dramatically increased globally, with atopic dermatitis, allergic rhinitis, and asthma each affecting 10–20% of the pediatric population [1–4]. These conditions are linked through the “allergic march”, where atopic dermatitis in infancy progresses to allergic rhinitis and later asthma [3]. In China, childhood allergic disease prevalence has increased substantially over recent decades, with allergic rhinitis and asthma prevalence rising from < 10% in the 1990s to 14–20% in major cities by 2011, affecting an estimated 240 million people nationwide [5]. These increases have occurred against the backdrop of considerable environmental changes, including climate warming and air quality deterioration, highlighting urgent needs for effective environmental health interventions. Children are uniquely vulnerable to these environmental changes because their immune systems and barrier functions are still developing, and early exposure creates long-term health impacts through developmental programming mechanisms [6–8]. Recent evidence highlights the critical importance of early-life environmental exposures in shaping lifelong allergic disease trajectories [9]. Early-life exposure to outdoor temperature and traffic-related air pollution (TRAP), including PM₁₀, NO₂ and SO₂, during pregnancy and early infancy demonstrates significant associations with childhood allergic diseases, with distinct roles for preconceptional, prenatal, and postnatal exposure windows [10, 11]. Recent studies have further demonstrated complex interactions between environmental factors and early-life exposures in childhood allergic diseases. For instance, nutritional supplements during pregnancy and infancy can modify the adverse effects of air pollution on food allergies, asthma, and eczema [12]. Similarly, parental physical activity interacts with particulate matter exposure to influence childhood respiratory allergic outcomes [13], while parental stress modulates the impact of indoor and outdoor pollution on childhood eczema [14]. These interaction patterns highlight the importance of considering multiple environmental and behavioral factors simultaneously when assessing pediatric allergic disease risk. The associations between exposure to air pollution and the development of asthma, allergic rhinitis, and atopic dermatitis in children have been documented in several meta-analyses; all demonstrated a significant positive relationship between exposure to particulate matter, nitrogen dioxide, and ozone [15, 16]. Children are physiologically more vulnerable to exposure to air pollution than adults; children have higher ventilation rates per body weight (6–8 vs. 4–6 L/min/kg for adults) and developing respiratory tracts with incomplete mucociliary clearance and immature antioxidant defense systems, rendering them more susceptible to oxidant injury from pollutants such as ozone and particulate matter [15, 17]. Furthermore, the developing immune system in children shows heightened Th2-skewed responses, potentially amplifying IgE-mediated allergic reactions to environmental triggers [18, 19]. Mechanistically, air pollutants induce oxidative stress, inflammatory responses, disrupt barrier function, and enhance allergen penetration [20, 21]. Longitudinal cohort studies provide evidence for associations [4, 22]. The mechanisms linking meteorological conditions to allergic disease burden operate through multiple interrelated pathways. Temperature directly modulates aeroallergen ecology, with studies documenting that temperatures exceeding 20 °C trigger peak pollen production in temperate regions, while subtropical climates demonstrate extended allergen seasons spanning 8–10 months annually [23–25]. Additionally, warm temperatures influence immune function through heat shock protein expression and may alter Th1/Th2 balance, while also affecting skin barrier lipid synthesis and transepidermal water loss in atopic dermatitis [19, 26]. Humidity and atmospheric pressure further modulate disease risks through effects on respiratory function and allergen dispersal patterns [27, 28]. Critical knowledge gaps persist. Few studies have systematically compared meteorological factors versus air pollutants using standardized frameworks within the same population, with most examining single diseases rather than multiple conditions simultaneously [29, 30]. Previous research has predominantly employed linear regression models that may fail to capture threshold effects and non-monotonic relationships, and the application of distributed lag non-linear models (DLNMs) remains limited in pediatric environmental health research [30, 31]. Most existing studies focus on temperate climates with limited evidence from subtropical regions representing 40% of the global population yet severely underrepresented in environmental health literature [21, 32]. Additionally, the temporal dynamics of environmental effects—including lag patterns and seasonal modifications—remain poorly characterized, and the lack of quantitative burden assessments hampers translation of epidemiological findings into actionable public health strategies [30, 33, 34]. While emerging evidence has identified critical exposure windows for air pollution effects on childhood allergic diseases [10, 35–37], these studies primarily examined individual pollutants or diseases separately. The relative importance of meteorological versus air pollution factors across multiple allergic conditions in subtropical climates, remains inadequately addressed. Whether findings from temperate climate studies are generalizable to subtropical monsoon climates with extended warm-humid seasons remains uncertain, necessitating systematic comparison using advanced non-linear models to elucidate primary environmental drivers [23]. We selected Chongqing, a subtropical megacity (> 30 million population) in southwestern China, characterized by humid subtropical monsoon climate with hot humid summers (> 35 °C, > 80% humidity) and mountainous basin topography restricting air circulation. This setting represents numerous mountainous cities globally (Bogotá, Quito, Kathmandu, Almaty) inhabited by > 500 million people, providing insights applicable to similar urban environments worldwide. Chongqing’s unique environmental characteristics create a distinct setting for studying climate-health relationships. The city’s environmental profile includes: (1) Basin topography: Chongqing sits in the Sichuan Basin surrounded by mountains reaching 1,000–2,000 m elevation, creating a natural bowl that restricts air circulation and traps pollutants, particularly during winter months when atmospheric temperature inversions occur frequently; (2) Subtropical monsoon climate: Hot humid summers (mean temperature > 28 °C, relative humidity > 80%) alternate with mild winters (mean ~ 8 °C), creating extended warm seasons (8–10 months annually) that substantially prolong aeroallergen exposure compared to temperate regions where warm seasons last only 4–6 months; (3) High ambient humidity: Year-round mean relative humidity of 78.8% (range: 48–100%) influences both aeroallergen dispersal patterns and skin barrier function through effects on transepidermal water loss; (4) Complex air pollution patterns: Winter pollution peaks result from residential heating emissions and atmospheric stagnation (mean PM₂.₅: 62.3 µg/m³, NO₂: 48.7 µg/m³), while summer features photochemical ozone formation (mean O₃: 55.8 µg/m³) due to high temperatures and solar radiation. These characteristics create a distinct environmental profile that may modify climate-health relationships compared to temperate settings, making Chongqing an ideal natural laboratory for examining subtropical climate impacts on pediatric allergic diseases. In this study, we sought to undertake the first systematic quantitative comparison of meteorological factors versus air pollutants as determinants of childhood allergic diseases, examining three distinct conditions (allergic rhinitis, asthma, atopic dermatitis) simultaneously within a large pediatric cohort from a subtropical urban environment. Our primary objective was to quantify and compare the relative contributions of meteorological factors (temperature, atmospheric pressure, humidity, wind speed) versus air pollutants (PM₂.₅, PM₁₀, NO₂, SO₂, CO, O₃) to disease-specific burdens. Secondary objectives included characterizing disease-specific exposure-response relationships using DLNMs to capture complex non-linear patterns and identify specific risk thresholds, examining seasonal effect modification to assess whether environmental-disease relationships vary across seasons with different climate-aeroallergen profiles, quantifying attributable disease burdens for individual environmental factors to inform intervention prioritization, and conducting comprehensive sensitivity analyses to assess robustness of findings to alternative model specifications. We employed DLNMs [31, 38] specifically to overcome limitations of conventional linear approaches, enabling identification of threshold effects, optimal exposure windows, and disease-specific susceptibility patterns. This methodological innovation allows characterization of complex temporal dynamics that inform targeted prevention strategies. We hypothesized that both meteorological factors and air pollutants would demonstrate significant associations with childhood allergic diseases, different allergic diseases would exhibit distinct environmental susceptibility patterns reflecting their divergent pathophysiological mechanisms, and environmental effects would vary across seasons due to changing climate conditions and aeroallergen profiles. Based on established biological mechanisms, we specifically hypothesized that temperature effects would demonstrate opposing directionality for respiratory versus dermatological allergies, reflecting distinct pathophysiological pathways (aeroallergen-mediated versus barrier dysfunction mechanisms) [23, 33]. We further hypothesized that subtropical climate characteristics might modify environmental-disease relationships compared to temperate climate findings. Methods Study design and population The objective of this retrospective time-series study was to determine the effects of environmental factors on daily pediatric hospital visits for three types of pediatric allergy (i.e., Asthma, Allergic Rhinitis and Atopic Dermatitis). Data on pediatric health was collected from the Children’s Hospital of Chongqing Medical University, China’s largest and most advanced Children’s Hospital in Southwest China and ranked the third in the country overall. The Children’s Hospital of Chongqing Medical University serves as the only specialized tertiary pediatric hospital in the entire municipality of Chongqing; therefore, all Pediatric Residents in the area receive their health care from this hospital, regardless of whether they reside in a Urban or Rural Area. Thus, the hospital has an extensive Service Area for Health Care to include Urban, Suburban and Rural Regions. The time frame for this study was from January 1, 2014 to December 31, 2019, and from January 1, 2023 to December 31, 2024, with a total of 2922 days of observation included; this excludes the time period of January 1, 2020 to December 31, 2022 as it was impacted by the COVID-19 Pandemic, which disrupted health care-seeking behaviors in China, and caused an estimated 40–60% decrease in Pediatric Outpatient Volume, leading to an opportunity for Systematic Bias to be introduced into the Study. Outcome definitions Cases were identified using ICD-10 codes (allergic rhinitis: J30.0-J30.4; asthma: J45.0-J45.9; atopic dermatitis: L20.0-L20.9). Diagnoses were established by specialists following Chinese national guidelines aligned with international standards (ARIA, GINA, Hanifin-Rajka criteria). Daily visit counts were aggregated by disease. Environmental exposure assessment Daily weather and air quality records were collected using various monitoring systems managed by both the Chongqing Municipal Meteorological Bureau and the Chongqing Ecology and Environmental Bureau that utilize numerous stations throughout Chongqing’s central urban and outlying area (Fig. 1). Study area and environmental monitoring network in Chongqing, China. Map showing Chongqing location in southwestern China and environmental monitoring stations. Children’s Hospital of Chongqing Medical University (★), the sole specialized tertiary pediatric facility serving as the regional referral hub; environmental monitoring stations (▲) distributed across the study area. Base map source: National Geomatics Center of China Individual assessments of environmental exposure were performed for each patient based on the nearest monitoring station to where they lived (central urban area or outlying areas). This method of assessing individual exposure levels was developed to better understand how changes in the physical environment from the natural features of Chongqing (a mountainous region) influence how a person experiences environmental exposure. Daily averages of the following environmental conditions for each person were calculated: temperature measured in degrees Celsius (°C), air pressure measured in hectopascals (hPa), relative humidity expressed as a percentage (%), wind velocity (m/s), concentrations of PM₂.₅, PM₁₀, nitrogen dioxide (NO₂), sulfur dioxide (SO₂), carbon monoxide (CO), and ozone (O₃). These environmental factors were selected as they represent standard criteria air pollutants with established regulatory monitoring networks, documented associations with respiratory and allergic diseases in environmental epidemiology literature, and continuous data availability throughout the study period. Statistical analysis We employed quasi-Poisson generalized additive models with DLNMs [31, 38] to examine associations between environmental exposures and daily hospital visits. The model was specified as: Log[E(Yt)] = α + crossbasis(environmental factor, lag, df) + natural splines(confounders, df) + time trend + day-of-week + holidays, where E(Yt) represents expected daily visits, crossbasis functions captured non-linear exposure-response and lag-response relationships using natural cubic splines (4 df for exposure-response, 3 df for lag, maximum lag = 21 days), and confounders included meteorological variables not being examined as primary exposures. We selected a maximum lag of 21 days based on allergic response timelines and prior environmental epidemiology studies. Following environmental exposure, allergic responses evolve sequentially: allergen sensitization and immune priming initiate Th2 differentiation and IgE production within 3–7 days [18, 39], followed by tissue inflammation with mast cell degranulation and eosinophil recruitment over 7–14 days [20], culminating in clinical symptom manifestation and healthcare-seeking behavior within 10–21 days [40]. This window has been validated in previous time-series analyses of environmental exposures and allergic diseases, with distributed lag models using maximum lags of 14–28 days effectively capturing delayed health effects [28, 29, 33]. Shorter periods would miss delayed presentations, particularly for atopic dermatitis where barrier dysfunction evolves gradually over 2–3 weeks [41], while longer periods (> 30 days) would introduce excessive complexity without biological justification as environmental effects beyond 3–4 weeks are likely confounded by subsequent exposures [22]. For atopic dermatitis specifically, the 21-day lag window reflects the disease’s distinct pathophysiology characterized by gradual barrier dysfunction rather than the immediate airway responses observed in allergic rhinitis and asthma [1]. Environmental exposures initiate progressive impairment of skin barrier integrity that evolves over days to weeks before clinical presentation and healthcare consultation [34]. This extended temporal window is consistent with established AD pathophysiology where environmental triggers do not cause immediate symptoms but rather initiate a cumulative barrier dysfunction process. Previous environmental epidemiology studies of allergic diseases have successfully employed lag windows ranging from 14 to 28 days [28, 29, 33], supporting the biological appropriateness of our 21-day specification for capturing the full spectrum of delayed environmental effects on AD. Cross-basis functions were constructed using bidimensional splines to model both the exposure-response dimension and the lag-response dimension. The exposure-response dimension employed natural cubic splines with 4 degrees of freedom, allowing flexible modeling of non-linear relationships including J-shaped, U-shaped, and monotonic patterns. The lag-response dimension used natural cubic splines with 3 degrees of freedom, with knots placed at equally spaced values on the log scale of lags to allow greater flexibility for short lags while maintaining parsimony for longer lags. Degrees of freedom for cross-basis functions were selected through systematic comparison of Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) values across multiple combinations. We tested exposure-response df of 3, 4, and 5, and lag-response df of 3 and 4. The selected specification (exposure df = 4, lag df = 3) minimized AIC across all disease outcomes while maintaining biological interpretability of exposure-response curves. For long-term trend control, we compared models with 4, 6, 8, 10, and 12 degrees of freedom per year. We selected df = 8/year based on: (1) minimization of AIC/BIC (2), visual inspection of residual autocorrelation functions showing no significant remaining temporal structure, and (3) sensitivity analyses demonstrating < 10% variation in primary effect estimates across df = 6–10 specifications. Detailed AIC and BIC values for all tested model specifications are presented in Supplementary Table S5. Minimum-risk levels were identified from exposure-response curves as the exposure level associated with lowest disease risk. For temperature, these varied by disease: AR (~ 10 °C), asthma (~ 18 °C), and AD (~ 28 °C), reflecting disease-specific physiological optima. For air pollutants, minimum-risk levels were set at lower bounds of observed ranges (NO₂: ~20 µg/m³, O₃: ~15 µg/m³), consistent with World Health Organization guidance and accumulating evidence that no safe threshold exists for air pollution health effects [42]. For other meteorological variables, minimum-risk levels were identified empirically from exposure-response curves: humidity (~ 78% for all three diseases), wind speed (~ 1.6 m/s), and atmospheric pressure (~ 970–971 hPa), representing the exposure levels associated with lowest disease burden across all three allergic conditions examined. Covariate selection followed a priori biological reasoning and collinearity assessment. For each disease, we constructed two alternative model specifications to address temperature-pressure collinearity (Spearman ρ = -0.857): temperature-based models (Models 1, 3, 5 for allergic rhinitis, asthma, and atopic dermatitis respectively) and pressure-based models (Models 2, 4, 6 for the three diseases respectively). Both model types included primary exposures modeled as cross-basis functions: temperature (or atmospheric pressure in Models 2, 4, 6), humidity, wind speed, nitrogen dioxide, and ozone. Confounders included natural splines for long-term trends (8 degrees of freedom per year), day-of-week indicators, and public holiday indicators. Air pollutants (nitrogen dioxide and ozone) were treated as primary exposures rather than confounders, as our research objective was to quantify their independent associations with allergic diseases. All covariates were assessed for multicollinearity using variance inflation factors (all values < 5). Long-term temporal trends were controlled using natural cubic splines (8 df/year), a flexible non-linear approach that captures seasonal patterns, multi-year trends, and potential non-linear time effects. This specification was selected based on Akaike Information Criterion and Bayesian Information Criterion optimization and visual inspection of residual autocorrelation functions, which showed no significant remaining temporal structure (detailed in Supplementary Table S5). Day-of-week effects were modeled as categorical indicators (6 dummy variables for Monday-Saturday, Sunday as reference) to capture weekly patterns in healthcare utilization. The temperature-based models were considered our primary models due to the stronger biological plausibility of direct temperature effects on allergic diseases, while pressure-based models served as sensitivity analyses to confirm result robustness. We quantified effects using standardized coefficients (regression coefficient × interquartile range) enabling cross-exposure comparison, and attributable burden comparing observed exposures to minimum-risk levels from exposure-response curves (95% confidence intervals via Monte Carlo simulation). Detailed formulas are in Supplementary Methods. Covariate selection ensured biological plausibility and low multicollinearity (VIF < 5). We quantified effects using: (1) Standardized coefficients (regression coefficient × IQR) enabling cross-exposure comparison; (2) Attributable burden comparing observed exposures to minimum-risk levels from exposure-response curves, with 95% CIs via Monte Carlo simulation). Detailed formulas are in Supplementary Methods. Comprehensive sensitivity analyses were performed to assess robustness of findings: (1) Degrees of freedom variation for time trends (df = 4, 6, 8, 10, 12 per year), with results presented in Supplementary Tables S3 and S5 and exposure-response curves in Supplementary Figure S1; (2) Season-specific models using meteorological season definitions (spring: March-May; summer: June-August; fall: September-November; winter: December-February), aligning with Chongqing’s subtropical monsoon climate patterns and local meteorological conventions, with seasonal exposure-response relationships in Supplementary Figure S2; (3) Two-pollutant models including pairs of pollutants to assess co-pollutant confounding. All sensitivity analyses confirmed the stability of primary findings, with key associations showing < 20% variation in effect estimates across alternative specifications. Statistical power calculations confirmed > 99% power to detect clinically meaningful associations (OR ≥ 1.03) for all disease-exposure combinations. Analyses were performed using R version 4.0.2 (dlnm, mgcv, ggplot2 packages). Statistical significance was defined as two-sided P < 0.05. This study followed STROBE guidelines [43]. Ethical considerations This study was approved by the Ethics Committee of Chongqing Children’s Hospital (approval number: 2025 − 530). As a retrospective analysis of de-identified administrative health data, the requirement for individual informed consent was waived in accordance with Chinese regulations on medical data research. All data were accessed and analyzed following hospital data governance policies. Data security was maintained through encrypted storage and restricted access protocols. Results Baseline characteristics and correlation patterns Baseline Characteristics are contained in Table 1.The mean daily hospital visits were: asthma 191.8, AR 108.5, and AD 74.2 over 2,922 observation days. Meteorological conditions for the study showed a mean temperature of 18.6 degrees Celsius (0.3 to 35.5 degrees), 78.8% relative humidity, and a typical variability of weather associated with subtropical monsoons. Mean air quality results of 42.1 µg/m³ PM₂.₅, 40.1 µg/m³ NO₂, and 44.2 µg/m³ O₃ confirm the level of PM₂.₅ exceeded the WHO guideline for the majority of study days. Table 1Baseline characteristics of daily hospital visits, meteorological factors, and air pollutants in Chongqing, China (2014–2019, 2023–2024)VariablesMeanMinP25P50P75MaxIQRAllergic rhinitis108.50.080.0105.0131.0281.051.0Asthma191.80.0118.0203.0259.0479.0141.0AD74.20.048.067.097.0209.049.0Meteorological factors Mean temperature (°C)18.60.311.518.724.535.513.0 Daily temperature difference (°C)6.50.53.66.09.317.95.7 Mean air pressure (hPa)970.9951.0963.8970.5977.4998.913.6 Daily air pressure difference (hPa)5.51.43.95.06.420.02.5 Sunshine(h)3.00.00.00.85.812.75.8 Relative humidity (%)78.840.071.080.588.298.817.2 Precipitation (mm)3.40.00.00.02.5104.12.5 Wind speed (m/s)1.60.31.31.51.83.80.6Air pollution factors AQI64.615.342.757.677.2223.634.6 NO₂ (μg/m³)40.112.729.938.849.197.819.2 O₃ (μg/m³)44.23.920.037.864.4159.144.4 PM₁₀ (μg/m³)65.613.139.458.083.1251.843.7 PM₂.₅ (μg/m³)42.18.323.234.353.0175.029.8 SO₂ (μg/m³)9.44.06.78.211.033.44.3 CO (mg/m³)0.90.40.70.81.02.20.3Data are presented as mean ± SD with range and IQR (25th-75th percentile). The study period spanned 2,922 observation days, excluding 2020–2022 due to COVID-19 disruptionsAR allergic rhinitis, AD atopic dermatitis, SD standard deviation, IQR interquartile range, PM₂.₅ particulate matter ≤ 2.5 μm, PM₁₀ particulate matter ≤ 10 μm, NO₂ nitrogen dioxide, SO₂ sulfur dioxide, CO carbon monoxide, O₃ ozone Baseline characteristics of daily hospital visits, meteorological factors, and air pollutants in Chongqing, China (2014–2019, 2023–2024) Data are presented as mean ± SD with range and IQR (25th-75th percentile). The study period spanned 2,922 observation days, excluding 2020–2022 due to COVID-19 disruptions AR allergic rhinitis, AD atopic dermatitis, SD standard deviation, IQR interquartile range, PM₂.₅ particulate matter ≤ 2.5 μm, PM₁₀ particulate matter ≤ 10 μm, NO₂ nitrogen dioxide, SO₂ sulfur dioxide, CO carbon monoxide, O₃ ozone The temporal pattern(s) of allergic disease visits showed distinct characteristcs of day-to-day variability that support different environmental susceptibilities. The AR visit patterns had the highest CV (61.3%), indicating considerable day-to-day fluctuations consistent with strong seasonal and weather influences. The asthma visits averaged 235.8 visits/d (0-687, CV = 58.9%); the AD visits averaged 86.7 visits/d (0-289, CV = 63.2%). The high coefficients of variation across all three diseases indicate considerable temporal variability, supporting investigation of environmental drivers. Environmental exposures demonstrated characteristic subtropical monsoon patterns. Temperature demonstrated significant seasonal variation, with summer mean of 28.5 °C (range 25.0–35.5 °C) and winter mean of 8.2 °C (range: 0.3–15.0 °C), reflecting subtropical climate characteristics, while there is also significant consistency in the relative humidity throughout the year (average 78.8% [between 48 and 100%]), which is representative of monsoon climate. There were significant seasonal differences in air quality, with peaks of PM₂.₅ in winter (62.3 µg/m³) attributed to heating emissions from burning fuels, along with atmospheric inversion effects, and lowest levels of PM₂.₅ in summer (28.4 µg/m³) attributed to increased precipitation and atmospheric mixing. Seasonal trends exist for NO₂ but they are very similar to those of PM₂.₅ (NO₂ for winter: 48.7 µg/m³, and summer: 32.4 µg/m³). However, the trends for ozone (O₃) are reversed from PM₂.₅ and NO₂ in that O₃’s peak (photochemical) concentrations occurred in the summer, while PM₂.₅’s and NO₂’s peak concentrations occurred in the winter months. Inter-variable correlations (Table 2) revealed disease-specific patterns. AR showed strongest correlations with temperature (ρ = 0.376) and inverse NO₂ correlation (ρ = -0.341), suggesting warm-season predominance. AD demonstrated opposite patterns with inverse temperature (ρ = -0.613) and positive NO₂ (ρ = 0.468) correlations, revealing fundamentally distinct environmental susceptibility profiles between respiratory and dermatological allergic diseases. Table 2Spearman correlation coefficients between environmental variables and daily hospital visits for childhood allergic rhinitis, asthma, and atopic dermatitisADMean tempDTDMean APDAPDSunshineRHPrecipitationWind speedAQINO₂O₃PM₁₀PM₂.₅SO₂-0.613-0.179*0.540*0.536*-0.857*-0.437*0.318*-0.3470.202*0.296*-0.215*0.540*0.868*-0.385*0.119*0.006-0.369*-0.715*0.284*-0.175*-0.722*-0.203*0.024-0.38-0.072-0.207-0.429*0.636*-0.1450.284*0.310*-0.3240.156*0.265*-0.330.020.361*-0.3370.084*0.289*0.226*0.060*-0.214*-0.467-0.302*0.468*-0.3310.0220.311*0.141*-0.013-0.075-0.351-0.382*0.738*-0.4440.763*0.729*-0.67-0.0680.709*-0.63-0.1280.406*-0.233-0.376*0.395*-0.3610.064*0.310*0.240*0.027-0.191-0.451-0.2930.980*0.779*-0.2780.43*-0.484-0.0810.411*0.224*-0.108*-0.046-0.368-0.345*0.960*0.733*-0.3970.957*0.394*-0.1580.177*0.138*0.200*0.097*-0.356-0.488-0.0840.659*0.684*-0.160.682*0.612*0.359*-0.383-0.1250.282*0.109*-0.181*0.144*-0.245-0.3650.698*0.749*-0.4780.705*0.740*0.583*Spearman’s rank correlation coefficients (ρ). * P < 0.05, two-tailedAR allergic rhinitis, AD atopic dermatitis, PM₂.₅ particulate matter ≤ 2.5 μm, PM₁₀ particulate matter ≤ 10 μm, NO₂ nitrogen dioxide, SO₂ sulfur dioxide, CO carbon monoxide, O₃ ozone Spearman correlation coefficients between environmental variables and daily hospital visits for childhood allergic rhinitis, asthma, and atopic dermatitis Spearman’s rank correlation coefficients (ρ). * P < 0.05, two-tailed AR allergic rhinitis, AD atopic dermatitis, PM₂.₅ particulate matter ≤ 2.5 μm, PM₁₀ particulate matter ≤ 10 μm, NO₂ nitrogen dioxide, SO₂ sulfur dioxide, CO carbon monoxide, O₃ ozone Non-linear exposure-response relationships reveal disease-specific environmental susceptibilities Exposure-response analyses revealed disease-specific and exposure-specific curve patterns (Fig. 2). Temperature demonstrated J-shaped associations for AR with minimal risk at cool temperatures (< 15 °C) and sharply increasing risk above 20 °C, U-shaped associations for asthma with lowest risk at moderate temperatures (15–20 °C) and elevated risks at both cold and warm extremes, and inverse associations for AD with progressively decreasing risk as temperature increased. NO₂ showed non-monotonic inverse associations for AR (likely reflecting seasonal confounding), weak positive associations for asthma, and monotonic dose-response relationships for AD. Exposure-response relationships between environmental factors and childhood allergic diseases from distributed lag non-linear models. Cumulative exposure-response curves over 21-day lag period. Panels A-C: temperature effects on AR, asthma, and AD. Panels D-F: NO₂ effects on AR, asthma, and AD. Solid lines: cumulative relative risk (RR); shaded areas: 95% confidence intervals (Monte Carlo simulation, 10,000 iterations). Reference values: temperature 15 °C, NO₂ 40 µg/m³ (RR = 1.0). Models adjusted for humidity, wind speed, O₃, day-of-week, holidays, and long-term trends (8 df/year). AR: allergic rhinitis; AD: atopic dermatitis; NO₂: nitrogen dioxide; RR: relative risk; CI: confidence interval; df: degrees of freedom The exposure-response curve reveals that temperature influences allergy symptoms in a J-shaped manner. This means that at colder temperatures (< 15 °C), the risk of developing allergy symptoms due to environmental exposure is close to baseline and does not increase significantly until temperatures exceed 20 °C at which point the risk is greatly elevated. The range of temperature associated with higher risks of developing allergy symptoms occurs within moderate temperature ranges (20 °C) and reaches maximums at warm temperatures (30 °C) when the relative risk reaches 1.25 (95% CI: 1.16–1.35). For asthma, temperature-exposure-response curves exhibit a biphasic pattern of association with asthma, meaning that temperature exposures at both extremes (low and high) pose a risk to individuals with asthma. Relative risks were highest at moderate external temperatures (18–22 °C) and lowest at corresponding low or warm extremes (32 °C: RR = 1.10, 95% CI: 1.05–1.16). The temperature-exposure-response curves for atopic dermatitis show a directly inverse pattern to the opposite of allergic rhinitis indicating that patients with atopic dermatitis experience the inverse risk for developing an allergic rhinitis response at warm temperatures (30 °C: RR = 0.75, 95% CI: 0.68–0.83). The exposure-response relationships for NO₂ indicate that it has both a nonlinear pattern of exposure-response associations with allergic rhinitis and a direct correlation with atopic dermatitis. The non-linear pattern for allergic rhinitis indicates that there is a risk of developing allergic rhinitis at low-moderate concentrations of NO₂ (RR = 0.92, 95% CI: 0.87–0.98) and increasing the relative risk of allergic rhinitis at higher concentrations (> 70 to < 90 µg/m³, RR = 1.08–1.10, 95% CI: 1.01–1.16). For atopic dermatitis, the relationship between NO₂ and the development of atopic dermatitis was observed to follow a monotonic pattern as demonstrated by the increase in NO₂ concentration in µg/m³ and the corresponding increase of RR = 1.20, 95% CI: 1.15–1.25. Asthma demonstrated modest positive associations with NO₂ (RR = 1.06, 95% CI: 1.02–1.10), substantially smaller than AD-NO₂ associations (RR = 1.20, 95% CI: 1.15–1.25). Meteorological factors demonstrate larger standardized effects than air pollutants Table 3 presents Standardized Effect Estimates from the Multivariate Generalized Linear Model (GLM) Analysis, that quantify the respective environmental contributions to the burden of Pediatric Allergic Disease. All independent variables were standardized (St_β = β × IQR) to remove the influence of scale and facilitate direct comparison of the relative magnitudes of the effect between measures taken on different scales. Table 3Standardized regression coefficients (St_β) from distributed lag non-linear models for environmental factors and childhood allergic rhinitis, asthma, and atopic dermatitisARAsthmaADModel 1Model 2Model 3Model 4Model 5Model 60.405(0.319,0.490)——0.108(0.050,0.166)——-0.286(-0.360,-0.213)——0.018(-0.040,0.075)-0.022(-0.080,0.036)0.032(-0.007,0.071)0.021(-0.018,0.060)-0.180(-0.230,-0.131)-0.148(-0.197,-0.098)-0.068(-0.110,-0.025)-0.072(-0.116,-0.028)0.002(-0.027,0.031)0.000(-0.029,0.030)0.053(0.016,0.089)0.063(0.026,0.101)——-0.158(-0.217,-0.099)——-0.044(-0.084,-0.004)——0.167(0.117,0.217)-0.219(-0.265,-0.173)-0.216(-0.263,-0.169)-0.021(-0.052,0.011)-0.020(-0.051,0.012)0.253(0.214,0.293)0.255(0.215,0.295)0.123(0.050,0.196)0.214(0.144,0.285)-0.003(-0.053,0.046)0.021(-0.027,0.068)-0.119(-0.182,-0.057)-0.168(-0.228,-0.108)Standardized coefficients calculated as β × IQR. Models 1, 3, 5 included temperature; Models 2, 4, 6 included atmospheric pressure (due to collinearity). Values represent cumulative effects over 21-day lagAR allergic rhinitis, AD atopic dermatitis, St_β standardized coefficient, IQR interquartile range, CI confidence interval, df degrees of freedom, NO₂ nitrogen dioxide, O₃ ozone Standardized regression coefficients (St_β) from distributed lag non-linear models for environmental factors and childhood allergic rhinitis, asthma, and atopic dermatitis Standardized coefficients calculated as β × IQR. Models 1, 3, 5 included temperature; Models 2, 4, 6 included atmospheric pressure (due to collinearity). Values represent cumulative effects over 21-day lag AR allergic rhinitis, AD atopic dermatitis, St_β standardized coefficient, IQR interquartile range, CI confidence interval, df degrees of freedom, NO₂ nitrogen dioxide, O₃ ozone Allergic rhinitis The variable ‘Temperature’ was identified as the most dominant predictor among the environmental variables (St_β = 0.405; 95% Confidence Interval 0.319–0.490; p < 0.001), and it represented an approximate 1.9% greater magnitude of effect than the next strongest predictor, ‘Ozone’ (St_β = 0.214 using Model 2). It demonstrated an approximate 1.6% greater magnitude of effect than the strongest air pollutant’s effect across all three allergic disease (NO₂-Allergic Dermatitis, or ‘AD’) (St_β = 0.253). ‘Ozone’ was positively associated with allergic rhinitis (St_β = 0.123; 95% Confidence Interval 0.049–0.197; p = 0.001 using Model 1; St_β = 0.214; 95% Confidence Interval 0.143–0.285; p < 0.001 using Model 2). The pressure-based model resulted in an increase in the magnitude of effect. Nitrogen Dioxide (NO₂) exhibited an inverse association with Allergic Rhinitis (AR) (St_β=-0.219, 95% CI: -0.265 to -0.173). This inverse association should not be interpreted as evidence of protective effects. Multiple lines of evidence indicate it reflects seasonal confounding by unmeasured warm-season aeroallergens: (1) NO₂ concentrations are lowest (summer: 32.4 µg/m³) when AR burden peaks, creating spurious inverse temporal correlation; (2) seasonal stratification reveals strongest inverse associations during spring (St_β=-0.400, 95% CI: -0.497 to -0.302) and summer (St_β=-0.333, 95% CI: -0.421 to -0.245) when aeroallergen exposure is highest, weakening in fall (St_β=-0.066, non-significant) and winter (St_β=-0.186); (3) NO₂ demonstrates positive associations with AD (St_β = 0.253, 95% CI: 0.214–0.293) across most seasons (strongest in winter: St_β = 0.354), consistent with oxidative stress mechanisms; (4) no plausible biological mechanism supports protective effects given NO₂’s established pro-inflammatory properties. Asthma There were generally weak relationships between the environmental variables and asthma, with no one single variable being found to be most influential. ‘Temperature’ exhibited a weak positive association with asthma (St_β = 0.108; 95% Confidence Interval 0.050–0.166; p < 0.001) that was substantially less than its effect on AR (St_β = 0.405). Atopic dermatitis ‘Temperature’ was found to have the greatest negative/magnitude of effect associated with atopic dermatitis (St_β=-0.286; 95% Confidence Interval − 0.360-(-0.213); p < 0.001). It represented a direct inverse correlation with its positive association seen with ‘AR’. Conversely, ‘Nitrogen Dioxide’ (NO₂) demonstrated the largest positive/magnitude of association with atopic dermatitis (St_β = 0.253; 95% Confidence Interval 0.214–0.293; p < 0.001 using Model 5; St_β = 0.255; 95% Confidence Interval 0.215–0.295; p < 0.001 Model 6). The positive association seen with NO₂ is opposite to that found with the association with AR. Cross-disease synthesis Across all three conditions, meteorological variables consistently exhibited larger absolute standardized coefficients compared to air pollutants. Temperature demonstrated the strongest associations for both AR (positive, St_β = 0.405) and AD (inverse, St_β = -0.286), with magnitudes substantially exceeding individual air pollutant effects. Two-pollutant model sensitivity Two-pollutant models assessed co-pollutant confounding. For AR, the inverse NO₂ association persisted after adjusting for PM₂.₅ (St_β = -0.203, 95% CI: -0.251 to -0.155) or CO (St_β = -0.225, 95% CI: -0.273 to -0.177), representing minimal attenuation (2.7–7.3%). For AD, positive NO₂ associations remained robust after adjusting for PM₂.₅ (St_β = 0.238, 95% CI: 0.199–0.277) or CO (St_β = 0.251, 95% CI: 0.212–0.290), with < 6% attenuation. The persistence indicates co-pollutant confounding does not explain NO₂ effects, though unmeasured seasonal aeroallergens remain potential confounders. Temperature demonstrates the largest attributable disease burden with opposing directional effects Table 4 show that environmental disease burden, as estimated using attributable case estimates based on interquartile range (IQR) exposure variation, was highly heterogeneous for various pediatric allergic conditions. Unlike standardized coefficients that indicate the intensity of effects for each unit of elevation of the IQR, attributable burden accounts for both the magnitude of effect and the actual distribution of the population exposed to the contaminants, and thus represents the total public health impact. Table 4Attributable disease burden from interquartile range (IQR) changes in environmental factors for childhood allergic rhinitis, asthma, and atopic dermatitisVariableARAsthmaADModel 1Model 2Model 3Model 4Model 5Model 6Mean temperature84,640(73145,96548)——58,058(42224,74425)——-58,270(-62045,-54352)——Relative humidity452(-6089,7171)-4425(-11113,2453)6475(-4634,17875)4790(-6440,16319)-20,326(-24477,-16050)-18,406(-22943,-13721)Wind speed-2997(-7487,1578)-5525(-10151,-808)3353(-4210,11051)2013(-5652,9816)5349(1986,8782)9212(5478,13030)Mean air pressure——-31,282(-37724,-24641)——-19,732(-30919,-8230)——51,041(44250,58056)NO2-10,176(-16250,-3941)-17,056(-23037,-10915)6400(-3842,16890)1522(-8536,11822)25,277(20630,30042)29,091(24020,34300)O3-10,164(-18581,-1436)15,350(6261,24765)-19,584(-33795,-4862)3323(-10789,17912)-3089(-9775,3892)-23,611(-29274,-17706)Attributable cases calculated from exposure-response curves (Fig. 2). Positive values indicate excess cases; negative values indicate prevented cases. Percentages show attributable fractions. 95% CI calculated via Monte Carlo simulation (10,000 iterations) AR allergic rhinitis, AD atopic dermatitis, IQR interquartile range, CI confidence interval, AF attributable fraction, NO₂ nitrogen dioxide, O₃ ozone Attributable disease burden from interquartile range (IQR) changes in environmental factors for childhood allergic rhinitis, asthma, and atopic dermatitis Attributable cases calculated from exposure-response curves (Fig. 2). Positive values indicate excess cases; negative values indicate prevented cases. Percentages show attributable fractions. 95% CI calculated via Monte Carlo simulation (10,000 iterations) AR allergic rhinitis, AD atopic dermatitis, IQR interquartile range, CI confidence interval, AF attributable fraction, NO₂ nitrogen dioxide, O₃ ozone For asthma The predominant contributor to asthma cases is temperature, whereby a 13.0 °C elevation in temperature IQR generates 58,058 attributable asthma cases (95% CI: 42,224 − 74,425, P < 0.001) or 7.5% of total asthma cases. Nitrogen dioxide (NO₂) generated 6,400 cases (95% CI: -3,842 to 16,890, P = 0.22) per IQR increment for Model 3, but due to the wide confidence interval the results were statistically nonsignificant. For AR The highest contribution of temperature due to IQR increases, as a result of climate change, for the four allergic conditions studied was AR. In Model 1, the number of AR cases attributable to temperature were estimated to be 84,640 (95% CI: 73,145 − 96,548, P < 0.001), with an increase in temperature of 13.0 °C IQR (or 18.5% of total AR), translating to an estimated annual average of 9,404 cases of AR. Hence, temperature is the most significant modifiable factor in the environment that contributes to the increased prevalence of AR in children living in subtropical climates. Ozone was found to contribute to an additional 15,350 cases of AR (95% CI: 6,261 − 24,765, P < 0.001) in Model 2, representing 3.4% of the estimated total AR burden attributable to temperature IQR increases. NO₂ was inversely related to AR burden. Across models, increases in NO₂ were associated with a decrease of 10,176 − 17,056 attributable cases of AR per IQR increment. For atopic dermatitis AD demonstrated distinctive attribution patterns opposite to respiratory allergic diseases. Temperature was associated with a reduction of 58,270 AD cases (95% CI: -62,045 to -54,352, P < 0.001) per IQR increase in Model 5, representing 20.4% of preventable AD cases. This substantial protective effect directly opposes temperature’s positive associations with AR (84,640 additional cases) and asthma (58,058 additional cases), revealing fundamentally distinct environmental susceptibility profiles. NO₂ contributed 25,277 cases (95% CI: 20,630 − 30,042, P < 0.001) in Model 5, representing 8.9% of total burden. This represents the largest individual air pollutant contribution observed across all disease-pollutant combinations. Cross-disease synthesis and policy implications Temperature demonstrated the largest absolute attributable burden across all conditions, with directional effects varying by disease type: positive associations with respiratory allergic diseases (AR: 84,640 cases, 18.5%; asthma: 58,058 cases, 7.5%) versus inverse association with AD (-58,270 cases, -20.4% preventable). This disease-specific directional divergence has profound implications for climate change health impacts and environmental health policy development. Seasonal attribution analyses revealed summer concentration for AR burden (temperature: 18,896 cases, 22.3% of annual burden) and winter concentration for asthma NO₂ effects (21,558 cases, 11.1% of winter burden). This temporal concentration enables season-specific interventions that may be more cost-effective than year-round programs. Seasonal heterogeneity in NO₂-disease associations Seasonal stratification revealed marked temporal variability, particularly for NO₂-AR associations. The inverse NO₂-AR association exhibited strong seasonal heterogeneity: strongest in spring (St_β = -0.400, 95% CI: -0.512 to -0.288) and summer (St_β = -0.333, 95% CI: -0.445 to -0.221), weakening in fall (St_β = -0.156, 95% CI: -0.268 to -0.044), and becoming non-significant in winter (St_β = -0.089, 95% CI: -0.201 to 0.023, P = 0.117). This pattern parallels the inverse temperature-NO₂ correlation where NO₂ is lowest during warm seasons (summer mean: 32.4 µg/m³) when AR burden peaks, and highest during cold seasons (winter mean: 48.7 µg/m³) when AR burden is minimal. Conversely, NO₂-AD associations remained consistently positive across all seasons (spring: St_β = 0.245; summer: St_β = 0.228; fall: St_β = 0.267; winter: St_β = 0.354), demonstrating disease-specific patterns independent of seasonal confounding. Sensitivity analyses confirm robustness Sensitivity analyses examining different degrees of freedom (df = 4, 6, 8, 10, 12) for time trend control confirmed remarkable stability of key environmental associations (Supplementary Table S3). Temperature effects on AR demonstrated exceptional robustness, with standardized coefficients ranging from 0.471 to 0.512 across all df specifications (< 10% variation). Similarly, NO₂ associations maintained directional consistency: inverse for AR (St_β = -0.058 to -0.073) and positive for AD (St_β = 0.178 to 0.194) across all df values. Attributable burden estimates also showed stability, with AR temperature burden ranging from 84,640 to 91,915 cases across specifications (< 10% variation). The consistency of both standardized coefficients and attributable burdens across alternative model specifications indicates genuine environmental-disease relationships rather than statistical artifacts. Our selection of df = 8 (approximately 7 df/year) balanced adequate control for seasonal and long-term trends while avoiding overfitting, as evidenced by optimal AIC/BIC values. Discussion We conducted a large-scale time-series analysis to systematically compare associations between meteorological factors and atmospheric pollutants with childhood allergic diseases in a subtropical urban environment. We found that temperature demonstrated substantially larger associations than individual air pollutants, with disease-specific directional effects: positive associations with respiratory allergies but inverse associations with atopic dermatitis. Our findings reveal complex exposure-response patterns and marked seasonal heterogeneity in disease burdens. Our findings regarding temperature effects on respiratory allergies are consistent with previous studies in subtropical and tropical regions [23, 44]. Temperature affects aeroallergen production and seasonality, with warm conditions extending pollen seasons and enhancing fungal spore proliferation [23]. The positive temperature-AR association aligns with documented increases in subtropical aeroallergen exposure during warm seasons [23]. Similarly, the modest positive temperature-asthma associations reflect established relationships between warm conditions and respiratory allergen burdens [16], though effect magnitudes were smaller than for AR, possibly reflecting asthma’s multifactorial etiology including viral triggers predominating in cooler seasons [45]. In contrast to respiratory allergies, atopic dermatitis demonstrated inverse temperature associations, aligning with skin barrier physiology literature [34, 41]. Warm temperatures enhance skin barrier function through optimized lipid synthesis and reduced transepidermal water loss, while subtropical humidity provides additional barrier protection [24]. These findings are consistent with the well-documented winter predominance of AD exacerbations in temperate climates [46] and the reduced AD burden observed in warm humid environments [34]. Hierarchical mechanistic framework The biological mechanisms underlying these disease-specific environmental susceptibility patterns operate through a hierarchical framework spanning organ systems, cellular processes, and molecular pathways, providing mechanistic coherence to our empirical findings. At the organ system level, respiratory allergies (AR and asthma) are fundamentally driven by aeroallergen-mediated type I hypersensitivity responses [20]. In subtropical monsoon climates, warm temperatures (> 20 °C) substantially increase disease burden through multiple coordinated mechanisms. First, elevated temperatures directly enhance pollen production from subtropical plant species, with peak flowering occurring during summer months when temperatures exceed 25 °C. Grass pollens and tree pollens demonstrate temperature-dependent production patterns, with pollen counts increasing exponentially above 20 °C [23]. Second, warm humid conditions promote fungal spore proliferation, particularly Alternaria and Cladosporium species, extending allergen seasons substantially longer (8–10 months) than temperate regions (4–6 months) [23]. Third, these environmental changes enhance allergen protein expression and increase allergen potency through post-translational modifications [27]. Fourth, warm temperatures may promote Th2-skewed immune responses through heat shock protein (HSP70) upregulation [39], amplifying IgE-mediated allergic inflammation. Finally, elevated ozone formation during warm sunny conditions (through photochemical reactions of NOx and volatile organic compounds) [16] enhances airway hyperresponsiveness and allergen penetration through epithelial barrier disruption. In contrast, atopic dermatitis operates through fundamentally distinct skin barrier dysfunction mechanisms. Warm temperatures protect against AD through multiple barrier-enhancing pathways. First, optimal temperatures (25–30 °C) promote ceramide and filaggrin synthesis in the stratum corneum [41], strengthening barrier integrity. Second, warm conditions reduce transepidermal water loss through improved lipid lamellae organization [24]. Third, subtropical humidity during warm seasons provides additional barrier protection by maintaining stratum corneum hydration at optimal levels (> 60%) [34]. Fourth, reduced indoor heating exposure during warm seasons minimizes xerosis (dry skin), a major AD trigger in temperate climates. Finally, decreased heavy clothing and reduced friction minimize mechanical barrier disruption. At the molecular level, NO₂ exposure exerts disease-specific effects through oxidative stress cascades. NO₂ generates reactive oxygen species (ROS) while depleting antioxidant defenses such as glutathione [47], triggering pro-inflammatory pathways through NF-κB activation and inducing cytokine release (IL-1β, IL-6, IL-8), with these inflammatory responses exacerbated under extreme temperature conditions [48]. For atopic dermatitis specifically, this oxidative stress directly damages skin barrier proteins including filaggrin and tight junction components (claudins, occludins) [24, 49]. Similar TRPV1 pathway activation, which amplifies neurogenic inflammation, has been demonstrated in allergic asthma under high humidity and NO₂ co-exposure [50], suggesting common molecular pathways across allergic diseases. The progressive dose-response relationship we observed (RR = 1.20 at 80 µg/m³ NO₂) aligns with this cumulative oxidative damage mechanism, providing biological plausibility for the associations we documented. The disease-specific directional inconsistency between NO₂-AR (inverse) and NO₂-AD (positive) associations, despite identical analytical methods, strongly suggests the AR pattern represents unmeasured seasonal aeroallergen confounding rather than genuine biological effects. This hierarchical framework not only explains our empirical findings but also provides a unified conceptual model for understanding how global environmental changes differentially impact organ systems through distinct pathways. Our findings regarding early-life environmental exposures align with accumulating evidence from retrospective cohort studies. Recent research has demonstrated that intrauterine and early postnatal exposure to traffic-related air pollutants, particularly during critical windows such as the first and second trimesters, significantly influence the development of childhood allergic rhinitis [37], asthma [35], and eczema [10]. These studies identified cumulative exposure effects and emphasized the importance of periconceptional and prenatal periods in shaping pediatric allergic disease trajectories. Additionally, meteorological parameters during pregnancy, particularly temperature and wind speed in early gestation, have been shown to play crucial roles in childhood asthma development [36], supporting our observation of temperature’s dominant influence on allergic disease burden. The inverse NO₂-AR association warrants careful interpretation and should not be interpreted as evidence of protective effects. The observed inverse association contrasts with positive associations reported in some temperate cities [51] and likely reflects residual seasonal confounding rather than genuine protective effects. Multiple lines of evidence support this interpretation. First, seasonal pattern analysis reveals that the inverse NO₂-AR association exhibits strong seasonal heterogeneity, with the strongest inverse associations during spring and summer when NO₂ concentrations are lowest and AR burden peaks, suggesting NO₂ serves as a proxy for seasonal aeroallergen patterns. Second, two-pollutant model sensitivity analyses demonstrate that the inverse association persisted after adjusting for PM₂.₅ and CO with minimal attenuation, indicating co-pollutant confounding does not explain the pattern, though unmeasured seasonal aeroallergens remain potential confounders. Third, disease-specific contrast provides critical evidence: for AD, NO₂ demonstrated consistent positive associations across all seasons, aligning with toxicological expectations of NO₂’s oxidative stress effects on skin barrier function [26]. This directional inconsistency between NO₂-AR and NO₂-AD associations, despite identical study design and analytical methods, strongly suggests the AR pattern reflects confounding by unmeasured warm-season aeroallergens rather than genuine biological effects. Finally, mechanistic implausibility argues against protective effects, as NO₂’s pro-inflammatory properties and epithelial barrier disruption [52] provide no plausible protective mechanism. Throughout this manuscript, we interpret the inverse NO₂-AR association as likely reflecting seasonal confounding rather than causal effects, which does not diminish our core findings regarding temperature effects and disease-specific environmental susceptibilities. Season-stratified analyses revealed striking temporal heterogeneity enabling targeted prevention. Summer emerged as the critical period for AR, with temperature contributing 18,896 cases representing 22.3% of the annual burden. Winter demonstrated concentrated asthma burden from NO₂, accounting for 21,558 cases or 11.1% of winter burden. Spring represented a natural protective period for AD, with warm temperatures preventing 9,006 cases. These findings necessitate dynamic, season-specific prevention strategies that concentrate resources during high-burden periods rather than year-round interventions, potentially improving cost-effectiveness of public health programs. Our findings have important implications for environmental health policy in subtropical urban settings. The substantial disease burden attributable to temperature (allergic rhinitis: 84,640 cases [18.5%]; asthma: 58,058 cases [7.5%]) necessitates integrating climate adaptation into pediatric health policies. Health authorities should develop temperature-health alert systems with disease-specific thresholds and coordinate with meteorological services for risk forecasting. The marked seasonal concentration enables targeted interventions. Summer allergic rhinitis prevention (18,896 cases, 22.3% of annual burden) should concentrate resources during May-August, while winter asthma management should prioritize NO₂ reduction during atmospheric inversion periods (21,558 cases, 11.1% of winter burden). The directional divergence in temperature effects (positive for respiratory allergies, inverse for atopic dermatitis) has implications for climate change assessments. Projected warming may simultaneously increase respiratory burden while reducing dermatological burden, requiring adaptation strategies balancing competing outcomes. Finally, temperature’s dominance over air pollutants suggests environmental policies focusing exclusively on air quality may miss the primary driver in subtropical settings. Long-term strategies should integrate climate projections into pediatric allergy forecasting and establish coordination between meteorological, environmental, and health departments. This study has several strengths. First, we provide the first systematic quantitative comparison of meteorological factors versus air pollutants within the same pediatric population, analyzing over 1.5 million hospital visits across three interconnected allergic diseases representing the complete allergic march spectrum. Second, we employed DLNMs capturing complex exposure-response patterns that linear models would miss, with comprehensive sensitivity analyses confirming robustness across alternative specifications. Third, we provide critically needed evidence from subtropical climates representing 40% of the global population, addressing a major geographical gap in environmental health research [53]. Several limitations should be acknowledged. First, our ecological study design relies on ambient monitoring data from fixed stations, which may not fully capture individual-level exposure variations including indoor air quality, personal mobility patterns, and school microenvironments. However, such exposure misclassification would likely be non-differential and thus would attenuate rather than inflate effect estimates, suggesting our findings may underestimate true associations. Second, we lacked direct measurements of aeroallergen concentrations (pollen counts, fungal spore concentrations), which are key mechanistic intermediates linking temperature to respiratory allergies. This represents a critical limitation affecting interpretation of our findings. Temperature may serve as a proxy for unmeasured aeroallergen exposure in subtropical climates, where warm temperatures directly increase pollen production and fungal spore proliferation. Consequently, our finding that temperature accounts for 18.5% of AR burden may partly reflect aeroallergen-mediated pathways rather than direct physiological effects. The inverse NO₂-AR pattern likely reflects residual confounding by seasonal aeroallergen patterns that our spline-based temporal adjustment could not fully capture. Future studies incorporating direct aeroallergen measurements are needed to validate these associations. Third, our findings from Chongqing’s subtropical basin environment may not directly extrapolate to all settings. Our results are most applicable to regions sharing key environmental characteristics including subtropical monsoon climate with extended warm-humid seasons, basin topography restricting air circulation, and high ambient humidity (year-round mean > 75%). Disease-specific temperature patterns (opposing effects for respiratory versus dermatological allergies) may differ in temperate climates with shorter warm seasons altering aeroallergen exposure duration, arid climates where low humidity modifies skin barrier physiology, or coastal regions with different air circulation patterns. While our findings may inform environmental health strategies in similar subtropical basin settings, replication studies in diverse environments are needed to assess generalizability. Despite these limitations, our study provides robust evidence through large sample size (> 1.5 million visits), advanced statistical methods (DLNM), and comprehensive sensitivity analyses demonstrating consistency of findings across alternative specifications. The time-series design offers advantages over cross-sectional approaches by controlling for time-invariant confounders through within-city temporal comparisons. In the context of climate change and ongoing air quality challenges, our findings suggest that temperature demonstrated stronger short-term associations with childhood allergic disease burden than individual air pollutants in subtropical urban settings. These findings necessitate integrating climate adaptation measures alongside air quality management in pediatric environmental health policies [21], moving beyond pollution-centric approaches to address dual environmental challenges with region-specific strategies. Future studies should incorporate direct aeroallergen measurements to disentangle temperature’s direct physiological effects from allergen-mediated pathways. Conclusion This study provides the first systematic quantitative comparison of climatic factors versus atmospheric pollutants as determinants of childhood allergic diseases in a subtropical environment. Temperature demonstrated substantially larger short-term associations than individual air pollutants, with disease-specific directional effects: 18.5% of allergic rhinitis cases attributed to temperature, versus 20.4% reduction in atopic dermatitis cases with warmer temperatures. Seasonal analyses revealed marked temporal heterogeneity, with summer concentrating allergic rhinitis burden and winter concentrating asthma NO₂ burden. These findings necessitate integrating climate adaptation measures alongside air quality management in pediatric environmental health policies, with season-specific interventions during high-burden periods. Our results from Chongqing’s subtropical basin setting may inform environmental health strategies in similar subtropical environments, though replication studies are needed to assess generalizability. Future studies should incorporate direct aeroallergen measurements to further disentangle temperature’s mechanistic pathways and inform targeted prevention strategies. Supplementary Information Supplementary Material 1: Supplementary Table S1. Sensitivity of standardized coefficients to degrees of freedom (df) for long-term trend control. Results from sensitivity analyses testing model robustness to alternative specifications for long-term trend control. We systematically varied df from 4 to 12 per year (approximately 3.5–10.5 df/year over the 2,922-day study period). Standardized coefficients (St_β = β × IQR) shown for temperature and NO₂ effects across all three diseases. Bold values indicate statistical significance (P < 0.05). The consistency of estimates across df specifications (variation < 20% for key associations) confirms robustness of primary findings to alternative model specifications. AR: allergic rhinitis; AD: atopic dermatitis; St_β: standardized coefficient; IQR: interquartile range; NO₂: nitrogen dioxide; df: degrees of freedom. Supplementary Material 2: Supplementary Table S2. Season-stratified standardized coefficients for environmental factors and childhood allergic diseases. Standardized coefficients from season-specific models fitted separately for spring (March-May), summer (June-August), fall (September-November), and winter (December-February). Each seasonal model included natural splines for within-season time trends (4 df per season) to control for residual temporal patterns. Values represent cumulative effects over 21-day lag period. Bold values indicate statistical significance (P < 0.05). Notable seasonal heterogeneity observed for temperature-AR association (strongest in summer) and NO₂-AR association (strongest in spring, non-significant in winter), suggesting seasonal modification by unmeasured aeroallergens. AR: allergic rhinitis; AD: atopic dermatitis; St_β: standardized coefficient; NO₂: nitrogen dioxide; O₃: ozone; df: degrees of freedom. Supplementary Material 3: Supplementary Table S3. Sensitivity of attributable burdens to degrees of freedom (df) for long-term trend control. Attributable disease burden estimates across different df specifications (4–12 per year) for long-term trend control. Values represent excess or prevented cases associated with IQR changes in environmental factors. The consistency of burden estimates across df specifications confirms robustness of primary findings. AR: allergic rhinitis; AD: atopic dermatitis; IQR: interquartile range; CI: confidence interval; df: degrees of freedom. Supplementary Material 4: Supplementary Table S4. Season-stratified attributable burdens for environmental factors and childhood allergic diseases. Attributable disease burden estimates from season-specific models. Values represent excess or prevented cases associated with IQR changes in environmental factors within each season. Seasonal heterogeneity patterns align with standardized coefficient findings (Supplementary Table S4). AR: allergic rhinitis; AD: atopic dermatitis; IQR: interquartile range; CI: confidence interval. Supplementary Material 5: Supplementary Table S5. Model selection criteria (quasi-AIC and quasi-BIC) for alternative degrees of freedom specifications for long-term time trend control. Quasi-AIC and quasi-BIC values for distributed lag non-linear models with varying degrees of freedom (df = 4–12 per year) for time trend control. Models used temperature-based specifications with identical exposure-response (4 df), lag-response (3 df, 21-day maximum), and covariate structures. Lower values indicate better fit. Selected specification (df = 8, ✓) balanced fit with parsimony, demonstrated minimal residual autocorrelation, and showed < 10% variation in effect estimates across df = 6–10 (Supplementary Table S1). AR: allergic rhinitis; AD: atopic dermatitis; df: degrees of freedom; AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion. Supplementary Material 6: Supplementary Figure S1. Sensitivity of exposure-response curves to degrees of freedom for time trend control. Exposure-response curves showing temperature-AR (A), temperature-AD (B), and NO₂-AD (C) associations under different degrees of freedom (df) specifications for long-term trend control. Each panel shows five curves representing df = 4, 6, 8, 10, and 12 per year. Line colors indicate df specifications (df = 4: light blue; df = 6: blue; df = 8: dark blue [primary analysis]; df = 10: orange; df = 12: red). The consistency of curve shapes and risk estimates across df specifications demonstrates robustness of primary findings to alternative model specifications. AR: allergic rhinitis; AD: atopic dermatitis; NO₂: nitrogen dioxide; RR: relative risk; df: degrees of freedom. Supplementary Material 7: Supplementary Figure S2. Season-stratified exposure-response curves for environmental factors and childhood allergic diseases. Exposure-response curves from season-specific distributed lag non-linear models. Panels show temperature effects in spring (A: AR, B: asthma, C: AD), summer (D: AR, E: asthma, F: AD), fall (G: AR, H: asthma, I: AD), and winter (J: AR, K: asthma, L: AD). Solid lines represent relative risk (RR); shaded areas indicate 95% confidence intervals. Reference value: temperature 15°C. The marked seasonal heterogeneity, particularly for AR (strongest effects in summer) and AD (strongest protective effects in spring), indicates seasonal modification of temperature-disease relationships, likely mediated by seasonal variations in aeroallergen exposure and humidity patterns. AR: allergic rhinitis; AD: atopic dermatitis; RR: relative risk; CI: confidence interval. Acknowledgements The authors thank the staff of Intelligent Application of Big Data in Pediatrics Engineering Research Center of Chongqing Education Commission of China, the China Meteorological Administration for meteorological data, and the Chongqing Municipal Ecology and Environment Bureau for air quality data. Declaration of generative AI and AI-assisted technologies in the writing process During the preparation of this work, the authors used ChatGPT to improve readability and language clarity. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article. Abbreviations Atopic Dermatitis Akaike Information Criterion Allergic Rhinitis Bayesian Information Criterion Confidence Interval Carbon Monoxide Coefficient of Variation Distributed Lag Non-linear Model Degrees of Freedom Generalized Linear Model International Classification of Diseases, 10th Revision Interquartile Range Nitrogen Dioxide Ozone Particulate Matter ≤ 2.5 μm Particulate Matter ≤ 10 μm Relative Risk Sulfur Dioxide Variance Inflation Factor World Health Organization Authors’ contributions RS, LN, and DH conceived and designed the study. RS, LN, and BP performed thestatistical analyses. RS, BP, and JL collected and curated the data. TX, RS and LNwrote the original draft. DH, TX and RS supervised the study and acquired funding. All authors contributed to manuscript revision, read, and approved the submitted version. Funding This research was supported by the Natural Science Foundation of Chongqing (Grant number: cstc2021jcyj-msxmX0069 awarded to Rong Sun and cstb2022nscq-msx0090 awarded to Xinye Tang). The funding agency had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data availability The raw clinical data that support the findings of this study are not publicly available due to patient privacy protection and institutional policies of Chongqing Children’s Hospital, but are available from the corresponding author upon reasonable request and with appropriate ethical approval from the hospital’s Ethics Committee. Environmental monitoring data (meteorological variables and air pollutant concentrations) are publicly accessible from: China Meteorological Administration (http://data.cma.cn/); Chongqing Municipal Ecology and Environment Bureau (http://sthjj.cq.gov.cn/). Declarations Ethics approval and consent to participate The studies involving humans were approved by the Ethics Committee of Children’s Hospital of Chongqing Medical University (approval number: 2025 − 530). The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation was not required from the participants or the participants’ legal guardians/next of kin in accordance with the national legislation and institutional requirements, as this was a retrospective study using de-identified aggregated health data. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The raw clinical data that support the findings of this study are not publicly available due to patient privacy protection and institutional policies of Chongqing Children’s Hospital, but are available from the corresponding author upon reasonable request and with appropriate ethical approval from the hospital’s Ethics Committee. Environmental monitoring data (meteorological variables and air pollutant concentrations) are publicly accessible from: China Meteorological Administration (http://data.cma.cn/); Chongqing Municipal Ecology and Environment Bureau (http://sthjj.cq.gov.cn/).