Projected change in the rainfall behaviour over Odisha

The recent fossil-fuel-driven development has increased greenhouse gas (GHG) emissions and led to a global temperature rise of 1.1 °C in the last 120 years. The rising temperature has global, regional, and local impacts and it is expected there will be a 4 °C rise in global temperature by the end of the century if the present emission trend continues (Legg 2021). The growing frequency of extreme events like floods, droughts, heavy rainfall events, storm surges, heatwaves, cold waves, and wildfires under a changing climate impact the ecology, food security, and socioeconomic conditions of a large global population (Sharma et al. 2016; Prabhakar et al. 2019; Majedul Islam 2022; Shahi et al. 2021). Thus, mitigation-based measures to curb GHG emissions to meet the Paris climate targets of limiting global warming to 1.5 and 2 °C are crucial (Legg 2021). The recent accelerated anthropogenic global warming has influenced the rainfall characteristics across the globe and also has regional implications (Maharana et al. 2018; Swain et al. 2018; Maharana et al. 2021; Anripa et al. 2023; Maharana & Dimri 2024). Indian Summer Monsoon (ISM), which contributes around 80% of the total annual rainfall over the Indian region, has a prominent impact on most of the agrarian economy of South Asia (Kumar et al. 2013; Maharana et al. 2020). Many studies show that the ISM has been impacted in recent times and is projected to change its behaviour under rising temperatures till the end of the century (Jayasankar et al. 2015; Sharmila et al. 2015; Maharana et al. 2020, 2021; Shahi et al. 2021). The impact of climate change on the ISM is reflected in the form of changes in rainfall characteristics such as its mean, extremes, and their spatial distribution across India, which are the ultimate result of the changing land-sea temperature contrast leading to change in the south-westerly bringing moisture to the Indian landmass (Roxy et al. 2015; Maharana et al. 2021; Shahi et al. 2021). The rising temperature has particularly influenced regional energy balance and water balance, thereby altering the atmospheric moisture content, and ultimately leading to the increase in the regional climate extremes in the form of extreme rainfall events, floods, droughts, heatwaves, coldwaves, coastal disasters, etc. (Dore 2005; Trenberth 2011; Mohapatra et al. 2021; Shahi et al. 2021; Bakshi & Kumar Panigrahi 2022; Maharana et al. 2023; Swain et al. 2023; Rani et al. 2024). The changing monsoon behaviour also impacted the rainfall characteristics in a disaster-prone state like Odisha having limited adaptive capabilities.

Odisha, a state on the eastern coast of India, is mostly an agrarian economy with 65% of the total population depending on it (Fig. S1). The agricultural practice is mostly traditional and contributes around 26% of the gross domestic productivity (https://www.odihort.nic.in/agriculturepolicy). Therefore, the timely rainfall and duration become very important for the socioeconomic condition of the people of Odisha (Patra et al. 2012; Mohapatra et al. 2021). Odisha receives a majority of its rainfall during the monsoon period starting from June to September (JJAS, hereafter) and it contributes around 80% of the total annual rainfall similar to the national scenario (Parthasarathy et al. 1993; Kumar et al. 2013; Maharana & Dimri 2014; Odisha Climate Change Action Plan (OCCAP) 2018; Prabhakar et al. 2019). Southwestern and western Odisha are experiencing higher amounts of monsoon rainfall compared to other regions of the state (Swain et al. 2018; Mohapatra et al. 2021). The rainfall during monsoon over Odisha is mainly associated with the north-northwest ward propagating low-pressure system originating over the Bay of Bengal (BoB, hereafter) (Krishnamurthy & Ajayamohan 2010; Rajeevan et al. 2010; Maharana & Dimri 2016). In addition, the interaction of large-scale monsoon flow with the Eastern Ghats also plays a major role in the regional distribution of rainfall (Mohapatra et al. 2003; Swain et al. 2019). The magnitude of the rainfall received during monsoon is around 117 cm (Mohapatra & Mohanty 2004).

According to OCCAP (2018), the magnitude of the climatological rainfall is reported to decline after 1950, with 120 days of monsoon being squeezed to 60–70 days with an increase in the intense rainfall events. Panda & Sahu (2019) reported a significant increase in monsoon rainfall over Kalahandi, Bolangir, and Koraput districts of Odisha. This high rainfall variability makes Odisha one of the most disaster-prone states for floods and droughts (Nageswararao et al. 2019). The variability of the rainfall across Odisha with reference to the low-pressure system originating over the BoB and upper air parameters has been examined (Mohapatra et al. 2003; Mohapatra & Mohanty 2006a, 2006b). The trend of rainfall and the frequency of cyclonic disturbances along with their interannual variability in Orissa are reported to have increased (Mohapatra & Mohanty 2007; Swain et al. 2018).

Many studies tried to examine the rainfall characteristics of Odisha in the form of frequency and distribution patterns of extreme rainfall events (Mohapatra & Mohanty 2007; Swain et al. 2018; Mohapatra et al. 2021). Swain et al. (2018) reported that the frequency of very light, light, and moderate rainy days has remained relatively constant; while the trends of heavy, very heavy, and extreme rainy days show an increasing trend. Their study reflected almost 90% of the increase in rainfall extremes associated with monsoon depressions and cyclonic storms which may increase the potential of flood disasters (Goswami et al. 2006). Heavy and extreme rainfall events along with dry days are rising in Odisha in the last 100 years. Meanwhile there is a decline in the number of light-to-moderate rainfall days and wet days are decreasing (Swain et al. 2018; Prabhakar et al. 2019; Swain et al. 2020). The results also corroborate with the findings of Mohapatra et al. (2021), who reported an increase of both dry days and extreme in the form of 1-day maximum precipitation over Odisha, with a maximum rise in mean (extreme) rainfall over western and southwestern regions (western and coastal region). Swain et al. (2019) reported higher Indian Ocean Sea surface temperature and high moisture availability along the Odisha coast during extreme rainfall days and the opposite during dry periods. It is estimated that the major changes in the rainfall variability have taken a major shift around 1945 (Mohapatra & Mohanty 2007; Prabhakar et al. 2019). A few modelling studies have been dedicated to examining the localised prediction of rainfall over Odisha and extreme events such as cyclones and depression as case studies (Mahala et al. 2015, 2021; Maharana et al. 2022; Sisodiya et al. 2022). A few studies tried to explore the change in rainfall characteristics under a climate change scenario over Odisha (Ghosh & Mujumdar 2006; Parihar et al. 2022). Parihar et al. (2022) reported that the mean rainfall has increased from the early to later part of the century by 20–40% and the malaria transmission (mosquito density) will decline (increase) in the future.

Climate change is anticipated to influence every facet of life. There is an urgent need to study and understand the changing rainfall patterns for efficient planning and management of sectors like agriculture, water resources, hydropower generation, and reservoir operation at a regional scale (Patra et al., 2012; Sharma et al. 2016; Prabhakar et al. 2019). At present, studies related to the projected change in the mean and characteristics of rainfall across Odisha are limited and further, we have not come across any study which examines the climate change impact at different warming targets of the Paris Climate Agreement, which can help governments and planners to formulate short-term and long-term strategies to counter the negative impact of climate change. Therefore, the present study focuses on examining the projected rainfall (spatial and temporal) pattern including the extremes over Odisha and its change with respect to the historical period at different specific warming targets (SWTs) of the Paris Climate Agreement for two different scenarios using a robust method. Section 2 provides the details of the dataset used, CORDEX-SA model experiments, and the methodology adopted in the present analysis. Section 3 represents the results, Section 4 deals with the discussion of the study while the summary and conclusion are provided in Section 5.

The present study employed first-generation datasets from the Coordinated Regional Downscaling Experiments over the South Asia domain (CORDEX-SA) (Giorgi et al. 2009). These are improved high-resolution datasets prepared from the downscaling of the Coupled Model Intercomparison Project – Phase 5 (CMIP5) Global Climate Models (GCMs) using regional climate models (RCMs) used for impact studies on a regional scale. The program was coordinated by the World Climate Research Program (WCRP). A total of 22 model experiments (GCM–RCM combinations) are considered for the present study (Table S1). The output of these model experiments is available at a horizontal resolution of 0.44° (∼50 km) and downloaded from https://www.esgf-data.dkrz.de/projects/esgf-dkrz and the Center for Climate Change Research of Indian Institute of Tropical Meteorology (CCCR-IITM), Pune, India. The rainfall data at a temporal frequency of 1 month are considered for the analysis. The period 1986 to 2005 (20 years) is considered as the reference period, to compute the changes, which are derived from the simulations based on historical emissions (Taylor et al. 2012). For future projections (from 2006 onwards), the GCM driven by forcing obtained from Representative Concentration Pathway 4.5 and 8.5 (RCP45 and RCP8.5) are considered (Moss et al. 2010; Riahi et al. 2011; Thomson et al. 2011). The RCP8.5 represents the most pessimistic ‘business as usual scenario’, where the GHG concentration constantly increases at the present rate till the end of the century, hence representing the worst climate change scenario. The study also considers the RCP4.5, which represents an intermediate climate change scenario where the emission stabilises in the later part of the century. The simulated rainfall from these model experiments is validated against the observed India Meteorological Department (IMD, hereafter) gridded dataset at a horizontal resolution of 50 km (Rajeevan & Bhate 2009).

The study adopted a noble methodology for the selection of the best models among the 22 model experiments. Earlier, Jayasankar et al. (2015) considered the skills of the model experiments such as the presentation of mean, bias, interannual variability, and seasonal variability (mean annual cycle) to identify the reliability of the rainfall projection. Meanwhile Sharmila et al. (2015) prepared a matrix table based on the performance of individual model experiments on presenting the mean, MAC, variability, and seasonal migration for the selection of best model experiments to examine the future. The present study adapted to the methodology used by Sharmila et al. (2015) by considering a few more relevant skill criteria of the model experiments. For that, five different selection criteria have been considered and a rank of ‘–’ (not good) or ‘ + ’ (good) is assigned based on the performance of the model experiment for each criterion in the historical period. These are the ability to represent

• (i) Spatial distribution of rainfall (spatial distribution rank: based on the spatial distribution of rainfall and area-averaged bias within ± 0.7 mm/day across the study domain is considered as good).

• (ii) Mean annual cycle (MAC) or temporal evolution of the rainfall (MAC rank: based on the time of rise and fall of rainfall over the study domain).

• (iii) Climatological value of rainfall (climatological value rank: based on area-averaged rainfall over the domain and magnitude between the mean ±1 standard deviation of the observed value are considered as good).

• (iv) Variability of rainfall (variability rank: based on the standard deviation values of simulated rainfall, the values between 0.7 and 1.7 are considered as good).

• (v) Overall performance (performance rank: based on the overall performance of the model experiment, where performance index > 1.9 is considered as good).

A matrix table is prepared where the model experiments with rank ‘ + ’ (good) for all criteria are considered as a better model experiment. The criteria employed for the model selection are based on its practical implications for the society of the disaster-prone and vulnerable state of Odisha, which is under the constant threat of climate change with limited adaptive capabilities. Therefore, equal weight is assigned to all criteria. The major reason for this is the time (addressed by MAC rank), spatial distribution (addressed by spatial distribution rank and performance rank), amount (addressed by climatological value rank), variability (addressed by Variability rank and performance rank), and interannual variability (addressed by performance rank) of the rainfall are important in the context of rainfall over Odisha particularly in the field of agriculture planning, water resource management, and disaster management.

A multi-model ensemble of these selected model experiments is prepared and used for further analysis of the projected change in the rainfall characteristics across Odisha. The climate projections for rainfall are examined at three different SWTs such as 1.5, 2, and 3 °C global temperature rise with respect to the preindustrial period following the Paris Climate Agreement. The present study utilises the time sampling method for the study of different SWTs (James et al. 2017), and the details of the computation of SWTs are deliberated by Maharana et al. (2020) and Maharana & Dimri (2024). This method assumes that the model biases are time-invariant and the future projections are robust to these biases in the model experiments (Maharana et al. 2020; Ballav et al. 2021). The adopted method is computationally inexpensive, allows comparison between different SWTs, and provides climate change information with the time frame needed for the policymakers to formulate long-term and short-term plans on a regional scale (James et al. 2017; Teichmann et al. 2018). The impact of climate change on rainfall characteristics at different SWTs is computed in reference to the historical period.

This section evaluates the performance of the model experiments based on different criteria mentioned in the methodology section. Prior to that, the JJAS rainfall of the model experiments is validated against corresponding IMD observation using the Taylor diagram, which is a widely used matrix for model validation (Maharana & Dimri 2016; Sharma et al. 2023). It simultaneously compared the root mean square error (RMSE), standard deviation, and correlation coefficient (CC). The RMSE of the model experiments lies between 1.5 and 2.5 mm/day while the standard deviation varies between 0.7 and 2.2 mm/day (Fig. S2). Most of the model experiments have a standard deviation value close to the observed one (0.7–1.7 mm/day) except Can_RCA4, GFDL_RegCM4, MIROC_RCA4, MPI_CCAM, and MPI_RegCM4. Interestingly, the magnitude of the CC lies between 0.35 and −0.35 for all the model experiments reflecting that the interannual variability is not well represented; however, this signals that the model experiments are robust in creating their internal environment and rainfall pattern through their dynamics and physical parameterizations schemes.

The spatial variability of rainfall represented by the standard deviation is examined for the historical period and three different SWTs under RCP4.5 and RCP8.5 (Fig. S4). The rainfall variability ranges between 1 and 2 mm/day during the historical period and is higher along the coast as compared to the western region (Fig. S4a). From 1.5 to 3 SWL, the spatial spread of higher value of variability reaches the northern region (Fig. S4b–g). The variability increases to 2–2.5 mm/day at 3 SWT throughout Odisha under both RCPs. This signals that the variability will be higher in the later part of the century and hence pose more threats to the community because of the higher fluctuations in the magnitude of the rainfall received. The higher variability will increase the probability of disasters like floods and drought and also impact the agriculture and water resource management sector and ultimately the local population. The percentage change in the variability provides an idea about the direction of the change. Under RCP4.5, the percentage change is within 5% for all SWTs (Fig. S4h–j). At 1.5 SWT under RCP8.5, the northern region reflects a decline of variability by 10–20%, while the eastern and southern regions show an increase of 5–15% (Fig. S4k). The variability is projected to increase by 1–10% and 5–20% at 2 SWT and 3 SWT, respectively, with a higher spread across Odisha (Fig. S4l–m).

The VWD is around 15–21% of the total wet days during the historical period (Fig. S15a), which intensifies with the rise of global temperature from 1.5 to 3 SWTs to 33–45% under RCP4.5 (Fig. S8b–d), while the rise is between 27 and 36% under RCP8.5 (Fig. S15e–g). Under RCP4.5, VWDs intensify in the range of 6–12%, 14–22% and 18–26% at 1.5, 2 and 3 SWTs, respectively (Fig. S15h–j). The rise is shallow under RCP8.5, ranging between 6 and 12% at 1.5 and 2 SWT, which intensifies to 10–16% at 3 SWTs across Odisha (Fig. S15k–m). With the rise in VWDs, the PRVWD also increases from the historical period (18–25%) to 3 SWT (35–50%) under RCP4.5 (Fig. S16a–d). The projected rise in PRVWD ranges between 10 and 20% at 1.5 SWT and 15–25% at 2 and 3 SWTs, with a maximum increase across Odisha at higher SWT (Fig. S16h–j). While the same under RCP8.5 is 25–35% (30–40%) at 1.5 and 2 SWT (3 SWT) (Fig. S16e–g). The increase is between 8 and 10% (10–15%) at 1.5 and 2 SWT (3 SWT) (Fig. S16k–m). The intermodel variability of VWD and PRVWD among selected model experiments is also illustrated (Fig. S17 and S18).

The JJAS climatological averaged rainfall over Odisha is projected to intensify with the rising temperature and is maximum at 3 SWT (4–20%), except for the northern hilly region with a decline of 4%. The intensification of the rainfall over Odisha can be attributed to the increase in land–ocean temperature contrast, which will strengthen the south-westerly and enhance the moisture supply leading to higher rainfall (Roxy 2017; Maharana et al. 2020). The intensification of simulated rainfall mostly will occur during September, while the monthly averaged rainfall from June–August remains similar in magnitude. Interestingly, the intense rainfall continues till October indicating an intense prolonged rainy season over Odisha under a higher global warming scenario. The lengthy rainy season with intense rainfall may bring a lot of disaster in the form of post-monsoon floods and also impact the agriculture sector. Odisha has three major rainfall peaks (north, southeast, and southern extreme) where CWD is much higher. With the increase in the projected monsoon wind and corresponding rainfall, the CWD shows higher variability across Odisha. The higher CWD are projected over southwestern, north, and northeastern at 1.5 SWT which shift to southern and coastal areas at 2 SWT, while it declines across Odisha except the coastal region at 3 SWT. The increase in CWD and their merger possibly leads to a decline in the CWDS by 1 across Odisha at 1.5 SWT. Whereas, the projected increase in the length of the rainy seasons has escalated the number of CWDS at 2 and 3 SWTs with the rise of global temperature. The increase in rainy days and projected elongation of the rainy season over Odisha leads to a decline of the CDD by 2–6 days except for southern Odisha where it increases by the same number of days at all SWTs. The CDDS is projected to increase by 1 at 1.5 and 2 SWTs, particularly over western and central Odisha. As the CWDS and CDDS both are projected to increase, it reflects that the rainfall will mostly be confined to elongated rainy periods and the remaining part of the year will become drier, particularly at 2 SWT. However, CDDS will decline across the whole domain with the increasing strength of monsoon at 3 SWT (Maharana et al. 2020). The MDWs are much higher in western Odisha as compared to other parts, which further intensify across Odisha with rising global temperature. The highest escalation is projected over the western, northeastern, and eastern borders, which is comparable to the findings of Mohapatra et al. (2021). The PRMDWs contribute the major part of the total annual rainfall around 50–60% of the historical period. Due to its intensification, the contribution to the total rainfall increases significantly by 2–10% under different SWTs. The VWDs, which constitute 6–8% of the total wet days also increase by 6–12% along with its associated rainfall by 8–24% across Odisha at different SWTs. While the MWDs are higher in the western part, the R20 is much higher in the coastal region which increases significantly and spreads towards the central Odisha at higher SWTs, indicating the rise in rainfall extremes under global warming scenarios. Similar rising trends of heavy and extreme rainfall events in IMD observed data (1901–2013) have also been reported over Odisha (Swain et al. 2018). Even though the results are statistically significant, the variability of rainfall increases with rising temperature. This indicates higher rainfall variability under higher SWT would pose a threat of floods due to torrential rainfall as well as drought due to prolonged dry spells and make society more vulnerable.

The present study employed a robust methodology to select the best model experiments among all the available model experiments prior to the analysis of the projected change in the rainfall characteristic across Odisha. The best model experiments are selected based on the rank assigned by the analysis of the spatial distribution of rainfall, the temporal evolution of rainfall, mean climatological value averaged across the domain, variability and overall performance with respect to the IMD observed value during the historical period. The multi-model ensemble is prepared from the selected model experiments for the analysis of the change in the rainfall climatology and various rainfall indices across Odisha at different SWTs. The recent increasing land-sea temperature contrast has intensified the south-westerly which brings in moisture to the Indian landmass, thereby influencing the rainfall characteristics of Odisha. The major conclusions of the study are as follows:

• The climatological rainfall over Odisha is projected to increase from 1.5 to 3 SWT with the highest increase over coastal regions.

• The model ensemble indicates an increase in the length of the rainy period from June to October. The extended rainy season with intense rainfall may lead to significant disasters, such as post-monsoon floods, and also negatively impact the agriculture sector.

• The prolongation of the projected rainfall increases the count of CWD at different SWTs. The CWDS is also projected to increase under higher SWTs, while the possible merger of CWDS reduces its number at 1.5 SWT.

• The CDD is projected to decline with the strengthening of monsoon except over southern Odisha.

• The CDDSs and CWDSs both are increasing over western and central Odisha at 2 SWT, making them more vulnerable to strong seasonality (very wet during monsoon and very dry during the remaining part of the year).

• The major contribution of the rainfall over Odisha comes from MWDs (particularly over western Odisha), which further intensify under the warming climate along with VWDs. R20 is also projected to escalate, particularly over the eastern coast and central Odisha. This indicates that the entire Odisha will be prone to rainfall extremes in the future particularly during monsoon and post-monsoon.

• The rising rainfall variability from 1.5 to 3 SWT would increase the risk of flood (due to extreme rain) and drought (prolonged dry period) and the overall vulnerability of the region to rainfall extremes.

The results of the present study will help the state Government, planners, and disaster management authorities to formulate short-term and long-term mitigation plans and to take appropriate measures for its enforcement to counter the negative impact of climate change such as increasing drought, floods, and impact on agriculture. The results provide us with an idea of the projected climate and time if a certain GHG trajectory is followed. That time period can be considered as the ‘climate change action period’, where the Government, local administration, and planners formulate and implement short-term and long-term adaptation and mitigation strategies to avoid the global warming target of 1.5 and 2 °C. Many of the plans are also discussed in the OCCAP (2018). The following are a few suggestions:

• (i) Conservation of water resources integrating modern and traditional knowledge. Promoting rainwater harvesting and use of sprinklers during irrigation instead of surface or energy-intensive lift irrigation in the drought-prone areas.

• (ii) Disaster risk reduction by promoting early warning systems for floods, droughts, and coastal disasters. Construction of multipurpose flood and cyclone shelters and improvement of drainage system for its effective management of flood water.

• (iii) Incorporate risk-sensitive land use planning in the city's Master Plan like promotion of urban storm water and drainage management for urban flood control.

• (iv) Encourage climate-resilient agriculture practices, such as growing drought-resistant crops and adopting sustainable land use practices.

• (v) Promote the development and implementation of climate-resilient infrastructure to better withstand and adapt to changing climate conditions.

• (vi) Enhance forest cover to support environmental sustainability and climate resilience.

• (vii) Encourage the adoption and use of clean energy sources and energy-efficient devices to reduce environmental impact and enhance sustainability.

• (viii) Facilitate the exchange of knowledge and information, as well as technology transfer, with national and international agencies to enhance collaboration and innovation in addressing global challenges.

• (ix) Integrate robust scientific information to formulate effective climate change policies at the local and regional scale.

The present study employed the first-generation CORDEX RCM experiments over South Asia, which are available at 50 km horizontal resolution. The relatively coarser resolution may be a possible limitation of the present study. The author will compare the results of the present study with the recently available high-resolution datasets from second-generation CORDEX-CORE datasets available at 25 km horizontal resolution as the future assignment.

The author thanks the World Climate Research Program's Working Group on Regional Climate, the Working Group on Coupled Modelling and Center for Climate Change Research (CCCR), Indian Institute of Tropical Meteorology for making the CORDEX South Asia data freely available for the scientific community.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.