Road traffic accidents & climatic factors in an urban area in Kerala, India: A time series approach

Materials & Methods

This retrospective study was undertaken by the department of Public Health, Amrita Institute of Medical Sciences, Amrita Vishwa Vidyapeetham, Kochi, Kerala, India, covering the study period from 2018-2023. The study protocol was reviewed and approved by the Institutional Review Board.

Results

A total of 47852 road traffic accident records and climate data of 2149 days were used for this analysis. The average number of accidents per day was 22±11.3. Characteristics of road traffic accidents (RTAs), road conditions and time range of accidents are presented in table I. Accidents were almost equally distributed between the afternoon (12:01 PM to 6 PM) (33.3%) and morning peak hours (6:01AM to 12 PM) (31.3%).

Accident site characteristics indicated that 65.8 per cent of the sites were uncontrolled, 23 per cent were police-controlled, and 66 per cent of sites lacked dividers. The proportion of accidents decreased during the COVID-19 lockdown period of 2020-2021 but increased to pre-COVID levels in 2022 and further in 2023, accounting for 20.6 per cent of all accidents between 2018-2023 (Supplementary Figure). Driving characteristics and driver demographics indicate that more than half of the drivers involved in accidents were aged 25-49, with the majority being male (n=44721, 93.9%), and over 36 per cent (n=17489) of these drivers were reported for overspeeding. Alcohol use was reported in 2.6 per cent (n=1237) of cases. Safety measures revealed that only 64 per cent (n=6699) of car drivers used seatbelts, and 65.6 per cent (n=18058) of two-wheeler riders wore helmets. The majority (n=33799, 71%) of the injured sustained grievous injuries, while 5.6 per cent (n=2676) of cases were fatal.

The association of climatic factors with RTAs is depicted in table II. The cut-off values for daily maximum temperature, minimum temperature, rainfall and road traffic accidents were determined using median values calculated from the entire dataset. More accidents were recorded, when the temperature fell below 21.71°C (P<0.001) and when the rainfall was less than 2.15 mm (P<0.001). The annual mean (SD) minimum temperature ranged from 21.2°C (0.94) to 22.9°C (1.54) (Figure; A) and maximum temperature varied from 30.1°C (1.6) to 31.8°C (2.2). During this period, the average rainfall ranged from 7.8 mm (13.4) to 10.9 mm (18.2) per yr (Figure; B).

Close

Figure.

Change in Trend of RTAs with Climatic factors. (A) and (B) represent the trend of RTAs alongside minimum, maximum temperature and rainfall from 2018 to 2023, respectively. RTAs declined in 2020 due to COVID-19 restrictions but rebounded thereafter, peaking in 2023. The minimum temperature rose until 2020, stabilising thereafter. Maximum temperature increased steadily until 2020, experienced a slight dip in 2021, then levelled off around 30-31°C. Rainfall exhibited significant variability, with the highest in 2021 and the lowest in 2023, indicating fluctuating precipitation patterns.

Export to PPT

Autocorrelation analysis (Supplementary Table I) identified temporal dependencies in RTA occurrences. In 2020-21, significant positive autocorrelations were observed from lags 1 to 13 (except lag 6), indicating dependency on previous days, whereas 2018-19 and 2022-23 showed weaker autocorrelations, suggesting more random patterns. Seasonal analysis for summer 2020-21 showed strong autocorrelations up to lag 13 (excluding lag 8). The partial autocorrelation function (PACF) highlighted significant lags at 1 and 2, with summer PACF showing additional influence at lag 3. Time range analysis yielded low autocorrelation values, indicating no temporal patterns or persistence in RTAs across different time zones. Cross-correlation analysis (Supplementary Table II) revealed a significant negative correlation between minimum temperature and RTAs across multiple lags, with lower temperatures linked to higher accidents. Maximum temperature showed negative correlations at lags 11 and 12 in 2020-21, while rainfall had no significant association. Seasonal cross-correlation (Supplementary Table III) showed a positive correlation between maximum temperature and RTAs in winter 2018-19 (lags 2 and 3). In 2022-23, minimum temperature had significant negative correlations across multiple lags. During the 2020-21 monsoon, lower temperatures correlated with higher RTAs, while higher rainfall in the 2018-19 monsoon was associated with fewer accidents at specific lags.

Supplementary table IV presents key parameters of the SARIMA model. The autoregressive terms (ar.L1, ar.L2, ar.L3) reflect past monthly injury influences, while seasonal terms (ar.S.L12, ar.S.L24, ar.S.L36) capture periodic patterns. All parameters were statistically significant (P< 0.05), indicating that grievous injury patterns follow three-month trends and recurring three-yr cycles, influenced by short-term seasonal, medium-term biennial, and long-term triennial factors. The diagnostic tests confirm the SARIMA model’s adequacy, with no autocorrelation (P=0.33), normal residuals (P=0.92), stable variance (P=0.18), near-symmetrical skewness (-0.08), and normal kurtosis (3.22), ensuring reliability in analysing grievous injury patterns (Supplementary Table V).

Discussion

Cross-correlation analysis revealed that lower minimum temperatures had a significant negative correlation with RTAs during winter (-0.21 to -0.28) in 2022-23, and the monsoon season (-0.20 to -0.26) in 2020-21. The SARIMA model indicated that the grievous injury patterns were predicted by the past three months and also recurred every three yr suggesting that injury patterns are influenced by a complex mix of short-term seasonal factors, medium-term biennial patterns, and longer-term triennial cycles.

The findings of our study indicate a strong positive autocorrelation in daily RTAs over the 2020–21 period, suggesting noticeable patterns in the occurrence of RTAs. During the summer of 2020–21, autocorrelation coefficients were significantly higher (0.21 to 0.41), with strong seasonal patterns observed up to 13-day lags. In contrast, the monsoon season showed moderate autocorrelations (0.20 to 0.21) at specific lags, indicating lower persistence. The observed autocorrelation patterns may be influenced by COVID-19-related changes in traffic flow and mobility, potentially impacting RTA trends. These findings underscore the importance of understanding seasonal influences on RTAs for effective safety planning

This indicates that with fall in minimum temperature there is an increase in number of RTAs. Similarly, higher rainfall was negatively correlated (-0.2 to -0.22) with RTAs during the monsoon in 2018-19. These variations in correlations may be influenced by behavioural factors, traffic, and road conditions throughout different seasons. Studies by Daanen et al¹², Walker et al¹³, and Wyon et al¹⁴ show that drivers performance can be rapidly impacted by extreme temperatures.

Our findings indicate that 71 per cent of road traffic accident victims sustained grievous injuries. Using a SARIMA model, we forecasted trends in grievous injuries, capturing both seasonal and non-seasonal patterns. A similar SARIMA-based approach was used by Deretic et al¹⁵ to analyse traffic accidents in Belgrade, showing the model’s utility for identifying high-risk periods and supporting road safety interventions.

The highest proportion of accidents reported was in 2023, accounting for 20.6 per cent of the total. This finding is in line with the general pattern seen in many Indian cities and urban areas elsewhere, where a rise in the number of vehicles on the road, rapid urbanisation, and population growth have all contributed to an increase in RTAs over time^16,17. Motorcyclists are particularly susceptible to collisions, contributing to more severe injuries due to higher speeds, insufficient protective gear, and increased exposure to risks on the road¹⁸.

Our findings show that one-third of the incidents happened between 12.01 and 6 PM, while a study in Kerala revealed that road traffic accident fatalities occurred more frequently in the evening and night, particularly between 18.00 and 21.00 PM¹⁹. These disparities may be attributed to regional variations in traffic patterns, road conditions, and commuting behaviours.

In our study, 53 per cent of drivers involved in RTAs were aged between 25–49 yr, and most of them were males (93.9%). Females have consistently been underrepresented in traffic fatalities worldwide. Although their percentage in India is among the world’s lowest²⁰. This might be because women make up around 6.8 per cent of all motor vehicle licence holders, with men making up the remaining license holders²¹. Furthermore, more than 25 per cent of drivers were private employees, with professional drivers coming in second (18.8%). According to an analysis of data from the French SUMER survey (Medical Monitoring of Occupational Danger Exposure), non-professional drivers have more autonomy when operating a motor vehicle and professional drivers often face job demands, such as time schedules and extended work hours, which makes driving less self-regulated²².

Results from other studies conducted globally are in step with our finding that sizable fraction (36%) of drivers involved in collisions are penalised for over speeding^23-25. Our results revealed that alcohol consumption was recorded in 2.6 per cent of road accident instances. Underreporting of alcohol-related road collision fatalities in official crash data is a common occurrence²⁶. Regarding safety measures, our study reports that around 64 per cent of car riders used seatbelts, while 65.6 per cent of two-wheeler riders wore helmets. Following the adoption of strict rules in Kerala, the use of seat belts and helmets has grown significantly. These measures are critical for reducing the severity of injuries and fatalities

Many accidents (88.4%) happened on paved roads, which is consistent with other research findings. Paved roads, while facilitating higher vehicle speeds, may increase accident risks if proper safety precautions are not taken. Our study also found that accidents on national highways accounted for 24.8 per cent, while those on minor roads constituted 59.4 per cent of cases. Minor roads in metropolitan areas often experience higher traffic volumes, more crossings, and greater pedestrian activity, increasing the risk of accidents²⁷. Interestingly, 67 per cent of accidents occurred on straight roads, emphasising that road safety precautions should not be disregarded, even on roads perceived as less dangerous²⁸.

In our study, 52 per cent of accidents occurred in the areas with speed limit 40-60 km/h. In qualitative research from Melbourne, most drivers admitted to exceeding the speed limit at least occasionally by 10 km/h or more²⁹. Promoting strict adherence to current speed limits and improving driver education regarding the dangers of speeding are essential measures in lowering the number of accidents. Unsignalised intersections are another area of concern, with one-fifth of junction collisions occurring at such sites ending in significant injuries. Unsafe spacing conditions are the primary cause of crashes at unsignalised junctions³⁰.

Our study demonstrates that lower temperatures and rainfall are significantly associated with the occurrence of accidents. These findings highlight the influence of mild environmental changes on accident rates. This contrasts with findings by Malyshkina et al³¹, which identified a positive link between road accidents and extreme weather conditions, such as high summer temperatures or very low winter temperatures. The discrepancy may stem from differences in climatic settings, methodologies, or definitions of extreme conditions^32,33. These findings underscore the importance of incorporating weather-based forecasting and seasonal safety measures into road traffic injury prevention strategies in urban settings.

The study period overlaps with external disruptions such as the COVID-19 lockdowns and major flood events in Kerala, which may have influenced traffic volume and altered accident trends, posing a considerable limitation to the study. However, data imputation was used to address missing values and preserve continuity in the time series. Additionally, certain climatic parameters such as humidity and wind speed could not be included due to inconsistent data availability. Despite these constraints, the study utilised a substantial sample size and applied robust time series techniques, enhancing the reliability of the observed trends and associations.

This study highlights that road traffic accidents are influenced by behavioural, environmental, and seasonal factors, with two-wheeler crashes on minor roads during afternoon hours being the most common. Over-speeding and grievous injuries were frequent, while lower temperatures and reduced rainfall were associated with increased accidents. Hospitals and trauma care centres need to anticipate seasonal surges in severe injuries, especially during colder months and specific multi-yr cycles, to enable better resource allocation, preparedness, and targeted preventive strategies in collaboration with urban planners and meteorological services. Targeted interventions such as seasonal speed enforcement, pre-monsoon awareness campaigns, and weather-responsive advisories are warranted. Close collaboration between meteorological services and urban planners is essential for developing adaptive strategies to reduce accidents and their severity.