Uniqueness of India's Northeast with respect to climate change impact: an assessment of streamflow variation in the Gomati River basin

Climate change impacts are being realized at different places throughout the globe and on different hydroclimatic variables, such as temperature, precipitation, evapotranspiration, and streamflow. According to the Intergovernmental Panel on Climate Change (IPCC)’s fourth assessment report, ‘climate change refers to a change in the state of the climate that can be identified (e.g., using statistical tests) by changes in the mean and/or the variability of its properties, and that persists for an extended period, typically decades or longer. It refers to any change in climate over time, whether due to natural variability or as a result of human activity’ (IPCC 2013). As per the latest assessments of the IPCC, worldwide climate change is a systematic fact and one of the best challenges faced in recent times (Burgess et al. 2020). One of the important manifestations of climate change is the change in global mean surface temperature. Globally, the increases in surface temperature have caused changes in intensity and temporal trends of precipitation, affecting many regional and local hydrologic systems (Suman & Maity 2019; Dutta & Maity 2020; Sarkar & Maity 2020; Quansah et al. 2021). By the end of the twenty-first century, India's mean surface air temperature is predicted to increase by 1.39–2.70 °C compared to 1976–2005. Due to the gradual increase in the surface temperature, several areas of India, including the Northeast, have become more and more desertified.

India's Northeast is unique in many senses, ranging from the highest monsoon-receiving location and unique topography, including the largest delta in the south to the highest mountains in the north. Though apparently, the high rate of desertification may be expected to suffer by the western and northeast parts of India. According to a recent survey, six states in India's Northeast were among the top 10 regions in the country with the greatest rates of desertification between 2003 and 2018. These six states are Mizoram, Assam, Tripura, Nagaland, Arunachal Pradesh, and Meghalaya (https://www.downtoearth.org.in/news/climate-change/india-s-northeastern-states-desertifying-most-rapidly-78695, accessed in August 2021). In the last few years, the rain patterns over India's Northeast have changed significantly over the previous century, making it drier generally (Das 2016). During dry periods, any increase in temperature will result in a faster rate of soil moisture loss, leading to droughts or even flash droughts. During the monsoon months (June–September), which normally see heavy rainfall in Northeast India, the weather has changed for the worse (Roy et al. 2021). Up to 2021, many states of Northeast India suffer from a serious deficit of rainfall. The highest deficit occurred in Manipur, with 58% less rain than normal. In Mizoram, the deficit was 28%; Nagaland received 23% less rainfall than normal, and Arunachal Pradesh received 21% less rainfall than normal (https://www.downtoearth.org.in/news/climate-change/climate-crisis-in-north-east-india-monsoon-variations-should-ring-alarm-bells-now 78707, accessed in August 2021). Some of the wettest places on the earth are located in Northeast India (e.g., Cherrapunji and Mawsynram). However, during the period 2000–2014, the risk of drought occurring in the region was 54% (Parida & Oinam 2015). Being a hilly state in Northeast India, several deep valleys and rivers are flowing through Tripura, making it a landlocked state. There are mostly low hills covered with dense forests in the state. Despite this, all districts now receive an overall lower rainfall during the monsoon season, which constitutes 60% of annual rainfall. Despite being one of the wettest districts with a mean annual rainfall of 1,549 mm, south Tripura has seen a substantial drop in wet days and heavy rainfall events. There is also a significant rising trend in temperature recorded in Tripura (Tomar et al. 2017).

It is a known fact that changes in precipitation and temperature have a direct impact on runoff (Kim et al. 2021; Quansah et al. 2021). Moreover, human activities, as well as natural forces, have an impact on streamflow generation (Zhang et al. 2012). It is well recognized that changes in climate and human activities are the primary causes of streamflow change over time (Wagener et al. 2010; Ahn & Merwade 2014; Dutta & Maity 2020).

Gomati is the most important river basin in Tripura, being the major source of water resources for the state. The only hydropower plant in Tripura, the Gomati Hydropower Project, is located on the west Kalajhari Hill and stores water in the Dumboor Lake, which is close to Tirthamukh. As a result, this river basin serves as one of the major lifelines of the state. However, due to the impact of climate change, the availability of water resources in this catchment has changed drastically in recent years, affecting the lives of its inhabitants directly or indirectly (Das et al. 2022). Changes in water resource availability have also been observed to reduce the output of the Dumboor hydel-power facility. As a result, an understanding of the influence of climate change on the availability of water resources in the Gomati River catchment is essential. Being located in a geographically and hydro-meteorologically unique region, an assessment of climate change impact on the hydrologic systems may reveal new insight. To the best of our knowledge, very few researchers have examined how climate change could affect Gomati's hydrological responses. For example, Das et al. 2022 used the group method of data handling model to simulate the future rainfall–runoff relationship in the Gomati catchment using CMIP5 data, considering the Land Use/Land Cover (LULC) as the same as the historical period. Pal et al. (2019) also used a Hydrologic Engineering Centre-Hydrologic Modeling System (HEC-HMS) model to estimate the water quantity for a small part of the Gomati River catchment (Dumboor Reservoir) for various seasons only for the historical period 2004–2013.

Thus, the objective of the study is to determine the expected hydrologic responses and variations in the average annual and seasonal streamflow variations due to future climate change in the Gomati River basin in the past and future. The future period is divided into three epochs (Epoch 1 or E1: 2015–2039; Epoch 2 or E2: 2040–2069; Epoch 3 or E3: 2070–2099) for an Epoch wise assessment with respect to the reference period (1990–2014). Two conceptual models are used, namely, the HEC-HMS model and the Hydroclimatic Conceptual Streamflow (HCCS) model for streamflow modelling using the climatic inputs. The performance of both models is compared during the reference period for their suitability in simulating streamflow in a changing climate. The scientific information gathered in this research may be used to minimize the adverse effects of climate change on Gomati in several interconnected sectors.

The climate is predominantly warm, with a humid summer and a dry, moderate winter, with plenty of rain from July to October. Most of the annual rainfall occurs in the pre-monsoon (March to May) and monsoon (June to September) seasons.

Daily rainfall, maximum and minimum temperature, streamflow, and solar declination are required for the HCCS model. On the other hand, the daily rainfall, curve number (CN), lag duration, percentage of imperviousness, and initial abstraction value are required for the HEC-HMS model. The historical records of observed daily rainfall, maximum, minimum, and average temperature data are procured from the India Meteorological Department (IMD) in Agartala. Daily records of streamflow data are procured from the Drinking Water and Sanitation (DWS) office at Agartala.

The solar declination for the catchment is estimated for running the HCCS model as outlined in Bhagwat & Maity (2014). For the HEC-HMS model, the curve number, lag duration, and initial abstraction value are calculated by the empirical formula. The percentage of imperviousness is calculated using the Arc-GIS software. Based on the overlapping period of data availability for all the variables, the calibration period is fixed from 1 January 1999, through 31 December 2004, and the validation period is fixed from 1 January 2005, through 31 December 2009. Though the length of available records is a little short, daily values for 6 years are sufficient to calibrate the models. This also helps to validate the models for a longer period of time, i.e., 5 years. If the models perform satisfactorily for a longer period of time, their applicability for future assessment becomes more reliable. Thus, the lengths of the calibration period and the validation period are 6 and 5 years, respectively.

The simulation outputs from two General Circulation Models (GCMs) participating in the Coupled Model Intercomparison Project (CMIP6), namely (i) ACCESS-CM2 (Australian Community Climate and Earth-System Simulator Coupled Model) with a spatial resolution of 1.25° × 1.875° and (ii) IITM-ESM (Indian Institute of Tropical Meteorology Earth system model) with a spatial resolution of 1.9048° × 1.875°, are used to assess climate change effects in the study basin (source: https://esgf-node.llnl.gov/search/cmip6/, accessed in December 2022). Compared with its previous generations, i.e., CMIP5, the CMIP6 simulations feature finer spatial resolutions, improved parameterizations for cloud microphysics, and additional earth system processes and components such as biogeochemical cycles and ice sheets (Kamruzzaman et al. 2021). With respect to the future projections, there is a fundamental difference between CMIP5 and CMIP6. The CMIP5 projections are available based on the radiative forcing values in the year 2100, following four Green House Gas (GHG) concentration pathways. As an alternative to CMIP5, CMIP6 projections are based on the socioeconomic pathways (shared socioeconomic pathways, SSPs). The SSPs are considered more realistic future scenarios. In recent studies, the CMIP6 outputs are established to exhibit the enhanced capacity to simulate both primary and secondary hydrometeorological variables and higher reliability for the future time period (Dutta & Maity 2022).

The simulation outputs with the initial condition, r1i1p1f1, are selected for both ESMs to maintain uniformity in initial conditions. We used these simulations without any bias correction as the systematic bias is ignorable. In fact, there are two schools of thoughts. Some studies use a bias correction method to match the observed and simulated values in the historical period and apply it in the future periods/epochs. However, some researchers argue against it (e.g., Knutti et al. 2010; Ehret et al. 2012). The bias correction methods assume that the magnitude of bias in climate model simulations is the same for historical and future periods. As a consequence, the bias correction methods assume that the simulation biases of climate models are constant over time (Maraun 2016). Following this, and noticing an ignorable systematic bias, bias correction is not applied.

Both historical and future simulations with four SSP scenarios, i.e., SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5, are used for the analysis. The future periods were defined as three epochs in this study to assess future hydroclimatic responses: Epoch 1 (2015–2039), Epoch 2 (2040–2069), and Epoch 3 (2070–2099). The SSPs are projections of future worldwide socioeconomic trends through the year 2100. They are applied to create scenarios for greenhouse gas emissions under various climate policies. The SSP1-2.6 scenario, with 2.6 W/m² radiative forcing by 2100, is a reimagining of the optimistic RCP2.6 scenario as in CMIP5, intending to replicate a development that is compatible with the 2 °C targets. The SSP2-4.5, an update of RCP4.5, reflects the medium route of future greenhouse gas emissions, leading to a radiative forcing of 4.5 W/m² by the year 2100. With a radiative forcing of 7 W/m² by the year 2100, SSP3-7.0 is in the upper-middle range of emission scenarios. It is a newly introduced scenario to bridge the gap between RCP6.0 and RCP8.5. The scenario SSP5-8.5 represents the highest limit of the emission scenarios, with an increased 8.5 W/m² radiative forcing by the year 2100. The SSP5-8.5 is an updated version of the RCP8.5 scenario with socioeconomic factors added.

The HEC-HMS is a physically based and conceptual semi-distributed model which is developed by the U.S. Army Corps of Engineers Hydrologic Engineering Center (HEC). With many objectives, such as flood research, streamflow forecasts, water quality, and sediment transport, the HEC-HMS model simulates the rainfall–runoff processes in a single outlet watershed (U.S. Army Corps of Engineers 2008). The HEC-HMS, version 4.2.1, is utilized in this study to compute daily stream flows. To simulate the hydrologic response, data pertaining to watershed characteristics and meteorological influences are used in HEC-HMS. These are catchment area, land use pattern in the catchment areas, daily rainfall data, daily river flow data, percentage of imperviousness, standard lag, time of concentration, initial abstraction value, and curve number are all needed for hydrological modelling.

A soil conservation system (SCS) unit hydrograph was used to compute direct surface runoff hydrographs based on the HEC-HMS model, while SCS losses were used as a method for separating the base flow. The CN value and initial abstraction (I_a) are estimated based on the land use and antecedent moisture content (AMC) under the AMC-II condition (Subramanya 2017). The SCS approach was actually developed for the watershed with 15 km² area. However, it has been modified to be used for bigger watersheds by weighing curve numbers concerning watershed/land cover area (Pancholi et al. 2015).

The HEC-HMS model is performed for the historical period (also referred to as the reference period) using the daily rainfall data and the basic information of the watershed. To highlight the influence of climate change on future streamflow relative to the historical/reference period, the model is re-run for the future period with various alternative assumptions (see Table 1). These assumptions indicate potential changes in the watershed features in the future. It may be noted here that the watershed characteristics may change over time as the land use pattern may be modified in the future. As a consequence, other parameters (e.g., CN value, lag time, initial abstraction) may also be modified in the future as they all directly or indirectly depend on the land use pattern. However, it is extremely difficult to assess the future land use pattern that depends on various developmental factors, including government policies. We adopt different possibilities of changes through different assumptions, as shown in Table 1. In the very first assumption (A1), the same values of the percentage of imperviousness as in the reference period are used for all future periods (Epochs). Next, assumptions 2–4 assume a 10, 20, and 30% increase in imperviousness in the future. However, the increment is assumed to be the same for all future periods. The last assumption (A5) adopts a 10% increase in Epoch 1, a 20% increase in Epoch 2, and a 30% increase in Epoch 3. The models are run based on all these assumptions, and the results are compared with each other and against the reference period.

The aforementioned model runs using the meteorological data from the reference period for calibration and validation and the same from climate model simulation for future periods, as explained earlier.

Four statistical metrics are utilized to assess the performance of the two hydrological models. These are the coefficient of determination (R²), Nash–Sutcliffe efficiency (NSE), RMSE (root mean square error)–observation standard deviation ratio (RSR), and percent bias (PBIAS). A brief description of each statistical parameter is presented as follows (Moriasi et al. 2007; Maity 2022).

The HEC-HMS and HCCS hydrological models are used for the daily streamflow simulation in the study basin. After the model calibration and validation with historical data (1990–2014), a month-wise assessment of future streamflow is carried out for the future period (2015–2099).

Next, before proceeding to the future assessment, it is worthwhile to mention here that the credibility of future projections depends on the scenario to be actually followed in the future, which is not possible to ascertain beforehand. Thus, results are presented for multiple scenarios. A comparison between the observed data and GCM simulations during the historical period indicates a reasonably reliable association (NSE = 0.90–0.86, R² = 0.87–0.88, RSR = 0.32–0.33, and PBIAS = −3.41–10.16%). This may be considered acceptable performance for the models to be used for future assessment.

The maximum decrease in streamflow is found under scenario SSP5-8.5, which recorded a reduction of 21.12, 23.11, and 21.24% for Assumptions 2–4, respectively. Furthermore, in the case of Assumption 5, where the increase in percentage change of the parameter is considered as 10, 20, and 30% for three Epochs, the highest decrease amount of average annual discharge is 23.11% (SSP5-8.5) for Epoch 2, and the lowest decrease amount is 1.68% (SSP1-2.6).

According to the aforementioned observations, if the three Epochs are compared with each other, the most considerable decrease in streamflow is found for Epoch 2 under the scenario SSP5-8.5, and the least decrease in streamflow is found for Epoch 3 under the scenario SSP1-2.6 for all other assumptions. Also, for the assumption-wise comparison, the maximum decrease in the future streamflow value is found for Assumption 1, where the parameter is taken as the same as the reference period. Furthermore, the assumption-wise comparison emphasizes the impact of climate change; in the SSP5-8.5 scenario, the study area experiences a more significant change in the average annual streamflow value than under the SSP1-2.6 scenario.

Overall, the model performs satisfactorily during the historical period. The model performances are similar to some previous studies, e.g., Bhagwat & Maity (2014), Suman & Maity (2019), and Pal et al. (2019). For instance, Bhagwat & Maity (2014) and Suman & Maity (2019) compared the HCCS and Least squares-Support vector machine for regression (LS-SVR) model's performances in the Narmada and Mahanadi River Basins and concluded that HCCS was the better model. Pal et al. (2019) also reported similar results in case of the Dumboor Reservoir of Gomati River catchment. They reported the mean absolute percentage deviation value of 2.2%, which is found to be satisfactory and R² value of 0.97 for the HEC-HMS model.

The future assessment indicates that the future streamflow magnitude reduces with respect to the reference period (1990–2014). This is true for all the climate change scenarios and worst for SSP5-8.5. Many studies indicated that the runoff patterns are expected to change due to changes in the climate system as a result of the climate change. For developing countries, climate adaptation related to water-related disasters may be more important than other aspects to minimize damage. Thus, the findings of this study are helpful in investigating the potential changes in hydrological response under future climate scenarios. Although hydrological impact assessments under climate change are important, it has received little attention in the northeast parts of India. In fact, hydroclimatic studies in the northeast part of India suffer from the lack of available observed data. There is a lack of reliable data as well as missing data in the case of the Gomati River catchment also. Thus, the limitation lies in the availability of the discharge data. It is available for a relatively short period of time (1999–2009), based on which the models are developed. Longer dataset would have been better for more reliable model development and reduced uncertainty in streamflow simulations.

There are still major challenges associated with hydrological modelling in some areas of Northeast India, including the Gomati River catchment, primarily due to data availability (e.g., a lack of reliable data as well as missing data) and sparsely gauge data across time and space. This could be one bottleneck and limitation that the discharge data are available for a relatively short period of time (1999–2009), based on which the models are developed. Longer dataset would have been better for more reliable model development and reduced uncertainty in streamflow simulations.

It is worthwhile to note that the analysis is focused on the streamflow variation of the Gomati river basin. Initially, we portrayed the uniqueness of the northeast part of India owing to its geographical location from the perspective of climate change. It may be noted that variables like streamflow are sparsely recorded in this region. However, the importance of the region from the hydroclimatic perspective is immense. It may also be noted that geographically, inclusion of Bangladesh might also be necessary for certain studies. Considering all these issues, this study may be considered as an initiation of the new domain in hydroclimatology for this least understood region, and similar studies that may include multiple basins and other hydrological variables may be a potentially important direction of the future scope of this study.

India's Northeast exhibits a unique feature with respect to climate change impacts on hydroclimatic variables owing to its geographical and hydrometeorological characteristics. This study highlights some unique features of climate change impacts on streamflow variation of a river basin located in Tripura, one of the states in India's Northeast. Future changes in the characteristics are assessed based on CMIP6 simulation from two GCMs (ACCESS-CM2 and IITM-ESM) and four SSP scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5). The major findings of the study are as follows:

• (i)

  There is a significant increase in temperature (∼1.4–2.1 °C) along with a large seasonal variation in the annual precipitation. However, the future streamflow magnitude reduces with respect to the reference period (1990–2014). This is true for all the climate change scenarios and worst for SSP5-8.5.

• (ii)

  Validation with the historical data reveals that both HCCS and HEC-HMS hydrological models performed well in capturing the temporal variation in streamflow, with a marginally better performance for the HCCS model. Both the models agree on a common fact that the future change is maximum in the case of SSP5-8.5, which is generally a common finding in many parts of India as well as the world.

• (iii)

  It is noticed that the decrement in the annual mean of streamflow is maximum under the scenario SSP5-8.5 during Epoch 2 compared to Epochs 1 and 3. This is in accordance with maximum temperature and minimum precipitation in Epoch 2 compared to Epochs 1 and 3.

Indicated changes in precipitation and temperature extremes may cause an increased risk to socioeconomic aspects. Furthermore, as per the results of this study, the maximum daily runoff is likely to reduce the water availability in the future. Besides this, the change in the LULC of the study area is also responsible for reducing the streamflow of the Gomati River catchment. In brief, the projected climate change may significantly impact the water resources availability and socioeconomic aspects of the Gomati River basin. It may therefore serve as crucial information for the water resource management authorities and policymakers for the Gomati River catchment with respect to the agricultural practices and hydroelectric power generation, as Tripura's only hydroelectric power station is located at Dumboor on the Gomati River. In addition, it may also be useful for future water resource availability and transboundary cooperation with neighbouring countries.

The soil map of the Gomati River catchment is downloaded from https://data.apps.fao.org/map/catalog/srv/eng/catalog.search#/metadata/cc45a270-88fd-11da-a88f-000d939bc5d8, accessed in October 2022. The SRTM DEM is downloaded from https://earthexplorer.usgs.gov/, accessed in October 2022. Similarly, the LULC map from ESRI Land Cover with 10 m resolution is downloaded from https://www.arcgis.com/apps/instant/media/index.html?appid=fc92d38533d440078f17678ebc20e8e2, accessed in October 2022. The rainfall and temperature data are collected from the India Meteorological Department (IMD) in Agartala. The daily streamflow data are procured from the Drinking Water (DWS) office at Agartala. The simulated daily precipitation and temperature data sets are obtained from two GCMs, which are downloaded from https://esgf-node.llnl.gov/search/cmip6/, accessed in October 2022. All the figures in the manuscript were prepared using either Q-GIS software or Microsoft Excel 2016.

The authors declare there is no conflict.