This study evaluated satellite-based precipitation product (SPP), i.e. the Global Precipitation Climatology Project (GPCP), on a daily scale and a monthly scale across the Sunkoshi River Basin. The performance- and magnitude-based indices were used to assess the SPP. Furthermore, the bias correction was carried out based on the quantile mapping technique. The ground data-driven hydrological model was developed with the help of HEC–HMS following the calibration and validation. Then the SPP, both before and after bias correction, was fed to the already calibrated-validated simulation model. The indices, such as Nash–Sutcliffe efficiency, R2, percent bias, Kling–Gupta efficiency, and RSR, were evaluated and compared for simulated data obtained from the observed dataset, before bias correction and after bias correction of SPP. The results showed that after bias correction, GPCP performed comparatively better in the study area across the Sunkoshi River Basin. A mixed result was obtained in the case of energy generation from the ground gauge data and SPP. The energy generation from the ground gauge data was found sensitive to the SPP for the selected hydropower projects in the Sunkoshi River Basin.

  • Quantile mapping is used for the bias correction in satellite-based precipitation product.

  • At the daily and monthly scales, after the bias correction, Global Precipitation Climatology Project (GPCP) estimates show results with the ground gauge observed datasets.

  • GPCP 1DD can replicate the hydrological patterns over the Sunkoshi River Basin in the Hydrologic Engineering Center - Hydrologic Modeling System (HEC-HMS) model.

  • Underestimation and overestimation of precipitation events are shown by GPCP 1DD across the Sunkoshi River Basin.

A hydrological cycle, which can also be termed the water cycle, is referred to as the exchange and movement of water between the underground storage, the atmosphere, and the surface of the Earth. Precipitation, infiltration, runoff, transpiration, evaporation, and condensation are the major processes involved in this cycle. Among these, precipitation is one of the major processes involved. It is necessary to take into account the regular and consistent observation of precipitation patterns for effective flood prediction, forecasting weather trends, climate change studies, disaster risk assessments, and determining extreme events. Also, during the hydrological modelling, the consistent sets of precipitation data play a very crucial role. A river's flow can be impacted by a number of different water balance factors and components of the water balance influence the flow of the river and the basin's entire hydrological cycle (Bajracharya et al. 2018).

Nepal, also known as the Himalayan state or mountainous country, possesses significant climate and topographic variation within a short territory. Numerous major and minor rivers across the country show great potential for energy generation through hydropower projects (HPPs). It has been an undeniable fact that run-of-river (ROR) schemes have dominated the hydropower sector but it is also the current scenario that various other schemes like storage or peak run of river are also receiving more attention. Climate change may result in higher flow in the Kaligandaki Basin, a major hydropower centre, which will be advantageous to ROR hydropower plants like Kaligandaki ‘A’, particularly during the dry season (Bajracharya et al. 2018). Basically, because of the vast geographical variances, economic limitations, and technological limitations, developing nations like Nepal have very few precipitation measuring stations (Talchabhadel et al. 2017). Thus, high-resolution satellite-based precipitation products (SPP) can be an alternate reliable way instead of ground-based gauge data where the terrain is too difficult and remote for the establishment of meteorological stations. When comparing satellite-based rainfall projections to ground-based meteorological data, the former is more economical and offers continuous, reliable data (Han et al. 2010). However, there have been very limited studies regarding SPPs and their applicability in the context of Nepal. However, some SPPs can give good results in certain regions and territories, whereas some may not show good performance. The SPPs may be applied in a range of rainfall–runoff simulations, according to the basin's size (Shrestha et al. 2011).

Although various SPP data are available, which show promising high-resolution precipitation data, it is very important to understand that SPP data contain errors and random biases that can affect the reliability of the output. The hydrological performances of these SPPs need to be verified before using these datasets for further implications. However, the processes of calibrating and validating the hydrological simulation must rely on river discharge observations (Hossain & Katiyar 2006). Numerous experiments have been carried out recently by experts to improve the estimation of SPPs by bias correction. Minimisation of these errors should be done prior to their further use (Belayneh et al. 2020). Quantile Mapping has high correction potential for extreme precipitation and has the potential for new extremes, particularly the extreme events beyond the range observed in the past (Gobiet & Tani 2021). This method effectively corrects biases, such as quantiles, standard deviation, mean, frequency of wet days, and others (Andari et al. 2024). The systematic errors of SPPs were effectively lowered through bias correction, which also increased the SPPs' suitability for use in hydrological modelling (Zhang et al. 2022).

Nepal is located in South Asia, spanning 147,516 km2 between latitudes 26°22′ N to 30°27′ N and longitudes 80°4′ E to 88°12′ E, whereas the average length is 885 km from east to west, with a breadth ranging from 145 to 248 km north to south. In South Asia, from June to September, there is the southwest (summer) monsoon, and from October to December, the northeast (winter) monsoon prevails (Rajeevan et al. 2012). Global Precipitation Climatology Project (GPCP), being one of the satellite precipitation products, is considered more precise than the unadjusted products because they have been coherently adjusted by global ground gauge networks (Le et al. 2020). GPCP 1DD and other SPPs studied across Malaysia showed better results in the northeast monsoon than in the southwest monsoon (Tan et al. 2015). The GPCP exhibits its capacity to predict convective precipitation in the spring and summer and in the winter and autumn seasons, and it is less capable of capturing the precipitation, indicating that GPCP is unable to identify extreme events; however, the result, shown by GPCP 1DD over the United States, is more satisfactory and the results of the analysis of the continuous statistics demonstrate that, for all three zones and seasons, GPCP 1DD can match Land Data Assimilation System estimates with small biases (McPhee & Margulis 2005). The GPCP 1DD is the satellite precipitation data with a daily time scale basis and a spatial resolution of 1°. The GPCP 1DD's ability differs by position, emphasising the requirement for validation studies over bigger regions rather than interpreting outcomes based on a limited sample of grids or a particular area (Gebremichael et al. 2005). Curtis et al. (2007) investigated the El Niño–Southern Oscillation trends in connection with the global monthly and daily precipitation extremes using GPCP and Tropical Rainfall Measuring Mission (TRMM). TRMM is one of the research satellites built to study the distribution and variation of precipitation by covering tropical and sub-tropical regions. In terms of accuracy, the data estimates of GPCP 1DD and from other SPPs are comparable (Skomorowski et al. 2001). The GPCP estimates effectively represent daily precipitation occurrences, and their study concluded that the GPCP daily and monthly datasets are valuable for hydro-meteorological research and work, although there is still potential for development in the recovery of data from satellites and analysis approaches in high-latitude areas (Bolvin et al. 2009). For huge-scale investigations of climate variations, GPCP needs to be chosen over TRMM as it is a worldwide Climate Data Record (Huffman et al. 2023). In order to increase accuracy, particularly across complicated regions, the GPCP has adopted a new climatology-anomaly-based gauge assessment approach (Rudolf & Schneider 2005). When comparing various SPPs in the Himalayan region of Nepal, based on many criteria, the Climate Prediction Centre Morphing Technique (CMORPH), Integrated Multi-satellite Retrievals for Global Precipitation Measurement (IMERG) Final and Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks-Cloud Classification System showed the best overall performance among the SPPs; however, each station's performance varied, and the results showed that CMORPH was the best SPP for both hydrological simulation and precipitation representation, with IMERG Final coming in second. However, CMORPH, IMERG Final, and CHIRPS performed well on a monthly scale, while SPPs often failed to reliably mimic streamflow daily (Kumar et al. 2022).

The GPCP monthly rainfall evaluation, which combines gauge measurements with low-orbit-satellite microwave data and geosynchronous-orbit-satellite infrared information, is one of the most widely used SPPs for research purposes (Adler et al. 2003). Microwaves are electromagnetic waves with wavelengths that vary from roughly 1 m–1 mm, and hence they are also employed in transmission technologies like satellite broadcasts and radar because they can transmit data across large distances with little disruption. Similarly, Infrared is not visible, but it may be perceived as heat. It is electromagnetic radiation having wavelengths greater than visible light but smaller than microwaves. GPCP 1DD is better at identifying rare events than Tropical Multi-satellite Precipitation Analysis (TMPA), and in the study, it is found that the representation of spatial rainfall distribution over India by GPCP 1DD and TMPA is more satisfactory and promising (Joshi et al. 2012). The South Asian monsoon, which is affected by the Bay of Bengal and the Himalayan Mountains, is closely linked to the Sunkoshi River Basin's (Nepal) rainfall trend, whereas about eighty percent of the yearly rainfall occurs in June through September and the spatial and temporal difficulties associated with rainfall are greater over a short distance due to the significant topographic variance (Thapa 2018). The Indian monsoon system typically affects the climate and the summer monsoon governs the weather from May to September, whereas the winter monsoon, which occurs from November to March, is characterised by westerly circulation while the summer monsoon season in Nepal begins in the east, and lasts from mid-June to September (Hossain 2014). The GPCP 1DD accurately reflects the geographical and temporal variability of rainfall, accounting for more than 80% of the variance by effectively recognising rainy days over a broad variety of precipitation criteria (Gebremichael et al. 2005). The timescale available for GPCP is daily, monthly, or pentad (five-yearly). Also, the spatial resolution available for daily and monthly is 1°, whereas the spatial resolution available for pentad (five-yearly) is 2.5°. 1° resolution provides a more accurate and high-resolution representation of hydrological patterns.

The Sunkoshi River Basin, a tributary of the Koshi River Basin, has several HPPs, but the basin does not have sufficient ground-based rain gauges in the mountainous region. Also, this study focuses on the reliability and possibility of SPP data for hydropower generation. The evaluation of SPPs on a daily and monthly timescale has not been performed in the Sunkoshi River Basin. Furthermore, information regarding research studies using GPCP 1DD in Nepal is very rare and difficult to find. To address this scenario, this study aims to address the research by evaluating the performance of SPP along with their hydrological simulations and applications in the Sunkoshi River Basin. The objectives of this study can be summarised as follows: (i) evaluating SPP in the Sunkoshi River Basin at daily and monthly timescales (ii) assessing the effectiveness of SPP for further hydrological implications (iii) evaluating the energy production to SPP in selected HPPs (iv) analysing the sensitivity of energy generation of SPP with respect to that of observed data.

This framework evaluates the performance of the SPP before bias correction (raw), SPP after bias correction (bias-corrected), and ground-obtained data, based on their performances in hydrological models prepared in HEC–HMS and various categorical and continuous statistical indices which can be further divided into sequential steps: evaluation of SPP, preparation of hydrological simulation with observed data along with calibration and validation and feeding SPP to the calibrated hydrological model, energy generation analysis and comparison. The diagrammatic representation of the methodology adopted for the research is shown in Figure 1.
Figure 1

Diagrammatic representation of methodology.

Figure 1

Diagrammatic representation of methodology.

Close modal

Study area

Sunkoshi River, which originates from the Himalayas range, is one of the major tributaries of the Saptakoshi River. The Mahabharata range in the south and the Himalayan ridge in the north enclose the majority of the Sunkoshi River Basin, and the ridge that divides the Arun River surrounds the east portion, while the mountain that divides the Bagmati River surrounds the west portion (Thapa 2018). Out of 330 km of the total length of the Sunkoshi River, the length of the Sunkoshi River that flows within the premises of Nepal is 280 km (JICA 1985). In Tibet, the upper Sunkoshi River section is known as Pique (Bhattarai et al. 2024). The Sunkoshi River has five major tributaries: Indrawati, Bhotekoshi, Tamakoshi, Likhu, and Dudhkoshi. Among these major tributaries, the Dudhkoshi River is the largest tributary which meets the Sunkoshi River near Ghurmi. Similarly, the Indrawati, originating from the eastern watershed of Gosaikunda and being the only tributary that resides on the right of the Sunkoshi River, meets the river at Dolalghat. Bhotekoshi originates from the Himalayan region in Tibet and meets the Sunkoshi River at Bahrabise, Sindhupalchowk, whereas Tamakoshi joins the Sunkoshi River at Khurkot, Sindhuli. The total catchment area of the Sunkoshi River Basin at the confluence with the Koshi River is 18,142.11 km2. Figure 2 shows the location of the study area. The river networks within the Sunkoshi River Basin are shown in Supplementary Figure S1.
Figure 2

Location of the Sunkoshi River Basin.

Figure 2

Location of the Sunkoshi River Basin.

Close modal

Data collections

Precipitation and discharge data

The precipitation data were collected from the Department of Hydrology and Meteorology (DHM) from 1997 to 2015. The list of meteorological stations is shown in Supplementary Table S1. The meteorological stations used for this study are also shown in Supplementary Figure S2. The hydrological stations within the Sunkoshi River Basin were identified. The list of hydrological stations is shown in Supplementary Table S2. The hydrological stations are also shown in the Supplementary Figure S1.

Digital elevation model

The ASTER Global digital elevation model (DEM) Version 3 with a resolution of 30 m created by the Sensor Information Laboratory Corporation, Tokyo, was used for this study. The projected coordinate system was taken as WGS 1984 UTZ Zone 45 N.

Satellite precipitation product

The GPCP 1DD available on a daily time scale and GPCP V2.3 available on a monthly time scale, both with 1° spatial resolutions, were collected from the website of the National Centers for Environmental Information from 1997 to 2015. The features of SPP are shown in Supplementary Table S3.

Land cover

Land cover of the Hindu Kush Himalayan region was collected from the regional database system of the International Centre for Integrated Mountain Development. It is a spatial grid data type (ICIMOD 2010).

Hydropower projects

There are many HPPs, from small HPPs to large HPPs, based on power and energy generation capacity located within the Sunkoshi River Basin. For our study, we have selected the HPPs present in the Sunkoshi River Basin on the basis that they are of ROR type, with capacity greater than 20 MW, operational, and under-construction phase projects. The HPPs, which are also shown in Supplementary Figure S1 selected for this study, are as follows: Khimti I HPP (60 MW), Upper Bhotekoshi HPP (45 MW), Khimti II HPP (48.8 MW), Likhu IV HPP (52.4 MW), Madhya Bhotekoshi HPP (102 MW), Solu Khola HPP (86 MW), Upper Lapche Khola HPP (52 MW), Upper Balephi ‘A’ HPP (36 MW), Singati Khola HPP (25 MW) and Khani Khola I HPP (40 MW).

Hydrological modelling in HEC–HMS

HEC–HMS, developed by the US Army Corps of Engineers, simulates the precipitation–runoff process. The design of the model is done in such a way that it can solve a wide range of problems which include huge basin water systems and small basin runoff (Kolekar et al. 2017). HEC–HMS has become very popular and has been adopted in many hydrological studies because of its ability to simulate runoff both in short and long events, its simplicity of operation, and its use of common methods and is adaptable to tropical conditions as the model is based on the hydrological characteristics, topography, soil type and land use of the study area (Tassew et al. 2019).

The ArcGIS tool was used for creating the sub-basins across the Sunkoshi River Basin through the DEM available. Also works like plotting of ground gauges and creation of Thiessen polygons were done in this interface for the study area. The land cover map for the Sunkoshi River Basin was processed in the ArcGIS tool with features available like the spatial analyst tool. Supplementary Figure S3 and Supplementary Table S4 show the land use and land cover of the Sunkoshi River Basin. It is found that forest covers the highest, 41.68% of the total area of the basin. The basin model was then exported using the Create HEC–HMS project tool so that it could be further processed and used in the HEC–HMS interface. The model was divided into 14 sub-basins, which are shown in Supplementary Figure S4.

For the calibration process, the first control specification for the simulation run was assigned from 1st January 1997 to 31st December 2002. Similarly, for the validation process, the control specification was assigned from 1st January 2003 to 31st December 2006. The Thiessen polygon method was used to calculate the time-series precipitation datasets for each sub-basin. The weightage of each meteorological station to the sub-basins was calculated from this approach. Similarly, the streamflow (discharge data) was obtained from DHM and then given as input to the hydrological model prepared.

Similarly, the streamflow (discharge data) was obtained from DHM and then given as input to the hydrological model prepared in HEC–HMS. The precipitation datasets for each sub-basin were assigned to the specified hyetograph as the corresponding gauge to the corresponding sub-basins. These datasets were taken from the time-series datasets. Simple canopy, simple surface, deficit, and constant loss methods, Clark's unit hydrograph, base flow, and Muskingum routing were the parameters used in the HEC–HMS model. The in-built optimisation process was carried out in HEC–HMS for calibration of the model. The deterministic optimisation approach was used for the model optimisation. For the optimisation of the selected parameters, the time interval was chosen as one day, and the method chosen was the Nelder and Mead simplex search algorithm. The Nelder and Mead approach employs the simplex method to analyse every parameter concurrently and determine which parameters need to be tuned up, which is the motive behind this (Gebre 2015). The iterations were done 5,000 times. Peak-weighted root mean square error (RMSE) was selected as the objective function. Although various parameters are available for the simulation only those parameters which are sensitive to this study are chosen. Those selected parameters are described as follows:

  • (a) Simple canopy: Three parameters are to be given in the simple canopy:

    • (i) Initial storage (%): It indicates the percentage of maximum canopy storage filled, and the initial storage value was assigned as 10% at the start of the simulation process. This value was assigned to all the sub-basins initially.

    • (ii) Max storage (mm): The initial value was determined from the land cover of the basin. The advised max canopy storage value for the forest is 2.540, the agriculture area is 1.270 and the grassland is 2.032 (Ahbari et al. 2018). Hence, for convenience the initial value was assigned as 2 and the value was later optimised for better performance of the model.

    • (iii) Crop coefficient: Initially, for all the sub-basins the input value was given as 1. This value was adopted on the basis of various other studies. Later, for better output, optimisation was carried out, and the value was changed accordingly.

    • (iv) Uptake method: A simple method was assigned to this parameter.

  • (b) Simple surface: A simple surface contains two more parameters:

    • (i) Initial storage (%): As an initial value, we propose to start the simulation after a no-rainfall period, which makes all the water stored in the basin depression either evaporate or infiltrate, and in this condition, a value of 0% is appropriate (Ahbari et al. 2018). So, the initial value was given as 0%, and for the better performance of the HEC–HMS model, the values were optimised.

    • (ii) Max storage (mm): This parameter shows the maximum depth of water that can be intercepted by the surface. The initial values were taken from the value range between 0.01 and 1,500 mm, and later the optimisation process was carried out.

  • (c) Deficit and constant loss method: This method is used to calculate the amount of moisture or water lost from the basins using a single soil layer. It is a continuous, lumped, empirical, and fitted parameter model category that can be used for longer simulations. There are four parameters within.

  • (d) Clark's unit hydrograph: It falls under the event, lumped, empirical, and fitted parameter model category. It has two parameters:

    • (i) Time of concentration: Time of concentration (Tc, Equation (1)) can be defined as the time taken by the water from the farthest point of the basin to reach the outlet point of the same basin.
      (1)
      where L is the longest flow path (miles); Lc is the centroidal flow path (miles); S is the 10–85% average stream slope (feet/miles)
    • (ii) Storage coefficient: The basin storage coefficient (R, Equation (2)) is an index of the temporary storage of precipitation excess in the watershed as it drains to the outlet point (Dangol et al. 2022).
      (2)

  • (e) Base flow: Constant monthly base flow was assigned as the parameter option, which shows the monthly base flow when there is no runoff from the storm. This base flow model falls under the event, lumped, empirical, and fitted parameter model categories.

  • (f) Muskingum routing: This approach considers the conservation of mass technique and falls under the category of event, lumped, empirical, and fitted parameter models. Considering this approach, it routes the flow from the reach given, there is an inflow hydrograph. The calibration parameter constraints for K are a minimum of 0.1 h to a maximum of 150 h, for X is 0–0.5, and for a number of steps is 1–100. The values of K were calculated using Equation (3), and later the values were optimised. A total of 0.25 was taken as the initial value of X for all reaches and optimised later for better performance.
    (3)
    where K is the Muskingum time travel; L is the length of reach (miles); S is the slope of reach (feet/miles).

Evaluation of SPP

There are four widely used categorical statistical indices (also performance-based indices) critical success index (CSI, Equation (4)), false alarm ratio (FAR, Equation (5)), frequency bias index (FBI, Equation (6)), and probability of detection (POD, Equation (7)).
(4)
(5)
(6)
(7)
where YES is when daily rainfall is indicated by both SPP and observed ground gauge data, NO is when SPP indicates rainfall but the ground gauge does not show any rainfall (generally on non-rainy days) and MISSED is when SPP does not indicate any rainfall but ground gauge detects rainfall (generally on rainy days). Also, when daily rainfall <1 mm for both SPP and observed ground gauge data, then it is considered ‘Ambiguous’.
Similarly, two adopted continuous statistical indices are RMSE and Percent Bias (PBIAS). RMSE shows the deviation of SPPs from the observed ground gauge data by comparing the difference between the two datasets. The larger value of RMSE (Equation (8)) indicates that there is more deviation between SPPs and observed ground gauge data, i.e. more difference in the value between the two datasets. Percent Bias (PBIAS, Equation (9)) is the statistical measure of the mean difference between simulated data and observed data in any hydrological models or any other models. When the value of PBIAS is 0, it shows the model simulation is accurate, whereas a positive value indicates overestimated results and a negative value represents underestimated results.
(8)
(9)
where m is the total number of events and SPPi is the satellite precipitation product.

Evaluation of SPPs’ compatibility for runoff modelling

The model was first prepared with the observed ground datasets. The model was then calibrated and validated. In the same model prepared with HEC–HMS, the GPCP 1DD was fed before bias correction and also after the correction for biases was performed. The simulated results for both the raw GPCP 1DD and biased GPCP 1DD were evaluated with the in-built indices of HEC–HMS: determination of coefficient (R2, Equation (10)), Nash–Sutcliffe efficiency (NSE, Equation (11)) and per cent of volume bias (PBIAS, Equation (12)). The value of R2 ranges from 0 to 1, where 0 denotes without correlation and a value of 1 means ideal correlation. The value of NSE ranges from −∞ to 1.

The NSE index is the most often used performance metric in hydrological activities (Todini & Biondi 2017). The value of NSE less than 0 denotes that the simulated flow is not comparable, whereas the value near to 1 represents the better result, i.e. simulated flow estimates are comparable to the mean of observed flow estimates (Dangol et al. 2022). NSE and Kling–Gupta efficiency (KGE, Equation (16)) are the most common and extensively used performance metrics for the assessment of hydrological studies (Lamontagne et al. 2020). Also, KGE is used for further evaluations. Most commonly employed in water science, the KGE takes into account many forms of model flaws, including mean error, variability and dynamics (Pool et al. 2018). For KGE, correlation, bias and time variation of simulated and observed data are distributed according to Euclidean distance, but because KGE does not make any assumptions about the residuals’ statistical distribution, its ambiguity cannot be explicitly described (Vrugt & Oliveira 2022). KGE value extends from to 1, where the value closer to 1 indicates better performance of the model (Gupta et al. 2009). KGE value >0.9 represents the excellent performance of the hydrological model, KGE value from 0.7 to ≤0.9 shows good performance, KGE range from 0.5 to ≤0.7 indicates satisfactory performance and KGE values ≤0.5 indicate poor performance by the model. KGE variability ratio (, (Equation (13)) is the ratio of the standard deviation of simulated data to the standard deviation of observed data. nearer to the value of indicates a similar variability in observed and simulated data. Also, the ratio of the mean of the simulated data to the mean of observed data is referred to as the bias ratio (, Equation (14)). When the value of the bias ratio is greater than 1, it shows the overestimation of observed data compared to simulated data and less than 1 represents the underestimation of observed data. Pearson correlation coefficient (r, Equation (15)) measures the correlation relationship between the simulated and observed data, with a better indication when the value is nearer to 1. The ratio of RMSE to the standard deviation of the observed data (RSR, Equation (17)) is one of the performance metrics that has been used for the evaluation of the model. RSR can be applied to a variety of different components and provide extra insights (Silva et al. 2015). The performance of the model is good when the values of RSR are lower (Yilmaz & Onoz 2020).

Furthermore, daily discharge hydrograph, scatter plotters, and flow duration curves are generated for observed, simulated, raw GPCP 1DD and bias-corrected GPCP 1DD to assess the SPP compatibility for runoff simulation by visual assessment.
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
where Qi is the observed data, Q is an average observed data, Ri is the simulated data, R is an average simulated data, is the variability ratio, is the bias ratio, r is the Pearson correlation coefficient, is the standard deviation of observed data, is the standard deviation of simulated data, is the mean of simulated data, and is the mean of observed data.

According to Moriasi et al. (2007), the efficiency evaluation of several indicators employed in the model may be summarised, as shown in Table 1.

Table 1

Performance criteria

Performance ratingR2NSEPBIASRSR
Very good 0.65–1 0.65–1 −15 to +15 0.00–0.50 
Good 0.55–0.65 0.55–0.65 ±15 to ±20 0.50–0.60 
Satisfactory 0.40–0.55 0.40–0.55 ±20 to ±30 0.60–0.70 
Unsatisfactory Less than 0.4 Less than 0.4 Greater than ±30 Greater than 0.70 
Performance ratingR2NSEPBIASRSR
Very good 0.65–1 0.65–1 −15 to +15 0.00–0.50 
Good 0.55–0.65 0.55–0.65 ±15 to ±20 0.50–0.60 
Satisfactory 0.40–0.55 0.40–0.55 ±20 to ±30 0.60–0.70 
Unsatisfactory Less than 0.4 Less than 0.4 Greater than ±30 Greater than 0.70 

Catchment area ratio method for discharge calculations

The catchment area ratio (CAR, Equation (18)) method's main assumption is that it is possible to determine the discharge for the required place by multiplying the discharge of the close rivers by the ratio of the required place catchment area to that of a close by discharge station's catchment area. The CAR method gives better results when the flow characteristics of the rivers are comparable (Emerson et al. 2005). Basnet et al. (2018) found that among the four approaches: the rational method, NRCS-CN method, MIP method, and CAR Method, the CAR method gave the most accurate design discharge for Padhu Khola, and also Madi Khola was taken as a reference whose discharge and catchment area were already known.
(18)
where Q2 is the discharge at the required location; Q1 is the discharge at the known location; Area2 is the catchment area of the required location; Area1 is the catchment area of the known location.

In this study, we applied the CAR method to calculate the discharges at the intake locations of HPPs from the obtained simulated discharges generated by the calibrated and validated hydrological model at different locations, particularly the junctions in the HEC–HMS model.

Energy generation

The power (P, Equation (19)) generation capacity is calculated for the HPPs across the Sunkoshi River Basin.
(19)
where P is the power in kilowatt (KW); is an efficiency; is the unit weight of water; Q is the design discharge (m3/s); H is the net head (m).

So, for the calculation of the power generation capacity of HPPs, we follow the Department of Electricity Development (DoED 2018) guidelines, where the design discharge is to be taken as Q40 and the efficiency of the HPPs is taken as 87.39%. Thus, the discharge obtained after the simulation run of the hydrological model from the gauge observed data and discharge obtained by forcing raw and bias-corrected SPP is used after an exceedance flow analysis of 40%. The head was extracted from the salient features of the respective HPPs. The value of the unit weight of water is taken as 9.81 kN/m3.

The power generation capacity of the HPPs is calculated at different dependable flows from 20 to 60% by the simulated discharge from the gauge data that were given as input in the hydrological model and also from the discharge that was generated from the SPP raw and bias-corrected data which were given as input to the calibrated and validated hydrological model and the simulation run. These works were done in order to carry out the sensitivity analysis of energy generation to SPP at dependable flows from 20 to 60%.

Nepal Electricity Authority has categorised the months into two seasons. According to NEA (2017), June to November is grouped as the wet season, whereas December to May is taken as the dry season. The energy generated from the simulated discharge, raw SPP, and bias-corrected SPP is compared.

Performance evaluation of raw SPP

The performance evaluation of SPP before the bias correction was done using categorical statistical indices and continuous statistical indices on a daily and monthly time scale using a point-to-point comparison basis. The boxplots were plotted for the evaluation procedure of SPP.

Categorical statistical indices on a daily time-scale basis

The boxplot of the categorical statistical indices is shown in Figure 3 from 1997 to 2015.
Figure 3

Categorical statistical indices on a daily timescale basis of ground-based stations and a raw SPP in the Sunkoshi River Basin. (a) CSI, (b) POD, (c) FAR and (d) FBI.

Figure 3

Categorical statistical indices on a daily timescale basis of ground-based stations and a raw SPP in the Sunkoshi River Basin. (a) CSI, (b) POD, (c) FAR and (d) FBI.

Close modal

The CSI values of raw SPP (i.e. GPCP 1DD in this case) at the daily scale for all individual ground stations within the Sunkoshi River Basin are between the range of 0 and 1. The highest CSI value is 0.440607 for station number: 1202 (Chaurikhark), whereas the lowest value of CSI is 0.247118 for station number: 1115 (Nepalthok). The rainfall estimating capability of raw GPCP 1DD for all individual ground stations seems satisfactory. Similarly, the highest POD value is 0.64367 for station number: 1322 (Machuwaghat), and the lowest POD value is 0.488881 for station number: 1028 (Pachuwaghat). All the values for all the stations within the basin are within 0–1. Thus, the SPP (GPCP 1DD) for all stations within the basin exhibits the ability to identify the real-time occurrences of rainfall compared to ground stations. The largest value for FAR is 0.620557 for station number: 1115 (Nepalthok), whereas the lowest value of FAR is 0.185439 for station number: 1058 (Tarke Ghyang). All the stations of the Sunkoshi River Basin have a FAR value within the range, but no stations have a FAR value equal to 0. The FBI values for 18 stations are greater than 1, indicating overestimation in precipitation events, whereas the remaining stations within the Sunkoshi River Basin have FBI values of less than 1, representing the underestimation of precipitation event detection. The station with FBI values greater than 1 and with the highest score of FBI 2.071872 is station number: 1115 (Nepalthok), and the lowest value of FBI 1.006466 is station number: 1024 (Dhulikhel). Similarly, the stations with FBI values less than 1 and with the highest FBI value of 0.971702 is station number: 1008 (Nawalpur), and with the lowest value of FBI 0.679174 is station number: 1058 (Tarke Ghyang). Thus, the SPP for individual ground rainfall stations at the daily scale within the Sunkoshi Basin shows the mixed result of overestimation and underestimation of rainfall event detection.

Continuous statistical indices on a daily timescale basis

The RMSE value for station number 1,006 (Gunthang) is highest with a score of 17.0913 mm/day, and the lowest value of RMSE is 6.554052 mm/day for station number 1104 (Melung). Twenty-eight stations present in the Sunkoshi Basin have an RMSE value within the range of 10–14.5, as shown in Supplementary Table S5 and Supplementary Figure S5. So, all the forty stations present in the basin have either positive daily biases or negative daily biases. The perfect score for RMSE is 0, which means the SPP and observed events estimates are fully consistent. Thus, SPP (GPCP 1DD) with the RMSE value from 6.554052 mm/day to 17.0913 mm/day shows promising outcomes. The SPP for ground-based rainfall stations showed negative values of PBIAS, indicating the underestimation of the detection of rainfall events by SPP. The boxplot of continuous statistical indices of daily precipitation events of the Sunkoshi Basin is shown in Figure 4.
Figure 4

Continuous statistical indices on a daily timescale basis of ground-based stations and a raw SPP in the Sunkoshi River Basin. (a) RMSE and (b) PBIAS.

Figure 4

Continuous statistical indices on a daily timescale basis of ground-based stations and a raw SPP in the Sunkoshi River Basin. (a) RMSE and (b) PBIAS.

Close modal

Continuous statistical indices on a monthly time scale basis

The RMSE values for SPP (GPCP 1DD) are from 61.94172 mm/month to 404.9085 mm/month. The PBIAS values are in the negative range from −383.33 to 29.7632. The lower values of RMSE and PBIAS values close to the value zero are preferred. The box plot of the continuous statistical indices of monthly precipitation events in the Sunkoshi basin is displayed in Supplementary Figure S6.

Performance evaluation of bias-corrected SPP

The evaluation of bias-corrected SPP (GPCP) was done on the basis of categorical statistical indices (also known as performance-based indices) and continuous statistical indices (also known as magnitude-based indices).

Categorical statistical indices on a daily time scale basis

The boxplot in Supplementary Figure S7 shows the categorical statistical indices of daily precipitation events of bias-corrected SPP and ground-based stations from 1997 to 2015. For each ground station in the Sunkoshi River Basin, the CSI values of bias-corrected SPP (GPCP 1DD in this example) at the daily scale fall between 0 and 1. Station number 1202 (Chaurikhark) has the greatest CSI value of 0.479403, while station number 1115 (Nepalthok) has the lowest CSI value of 0.224924. For each ground station, the bias-corrected GPCP 1DD's capacity to estimate rainfall appears to be adequate. Similar to this, station number 1017 (Dubachaur) has the greatest POD value of 0.607961, while station number 1115 (Nepalthok) has the lowest POD value of 0.369692. For every station in the basin, the values range from 0 to 1. As a result, in contrast to ground stations, the bias-corrected SPP (bias-corrected GPCP 1DD) for every station in the basin can detect real-time rainfall events. The station number 1115 (Nepalthok) has the highest FAR value of 0.635168, while station number 1028 (Pachuwa Ghat) has the lowest FAR value of 0.190352. The FAR values of all the stations in the Sunkoshi River Basin are within the range; however, none of the stations have a FAR value of zero. Only nine stations in the Sunkoshi River Basin have FBI values larger than 1, suggesting an overestimation of precipitation events, while the remaining stations have FBI values of less than 1, indicating an underestimation of precipitation event detection. The stations with FBI values of greater than 1 include station number 1104 (Melung) which has the highest FBI value of 1.030633, while station number 1207 (Mane Bhanjyang) has the lowest FBI value of 1.002433. These are the stations with FBI scores larger than 1. In a similar vein, the stations with FBI values less than 1 include station number 1023 (Dolalghat), which has the highest FBI value of 0.993222, and station number 1028 (Pachuwa Ghat) has the lowest FBI value of 0.600313. As a result, the SPP for specific ground rainfall stations within the Sunkoshi Basin daily displays a mixed picture of overestimation and underestimation of rainfall event detection.

Continuous statistical indices on a daily time scale basis

The boxplot of continuous statistical indices of daily precipitation events of the Sunkoshi Basin is shown in Figure 5.
Figure 5

Continuous statistical indices on a daily timescale basis of ground-based stations and bias-corrected SPP in the Sunkoshi River Basin. (a) RMSE and (b) PBIAS.

Figure 5

Continuous statistical indices on a daily timescale basis of ground-based stations and bias-corrected SPP in the Sunkoshi River Basin. (a) RMSE and (b) PBIAS.

Close modal

With a value of 18.55811 mm/day, station number 1107 (Sindhuligadhi) has the highest RMSE value, while station number 1104 (Melung) has the lowest RMSE value of 5.301249 mm/day. The RMSE values of the 27 stations in the Sunkoshi Basin fall between 10 and 15 mm/day, as shown in Supplementary Table S6. Thus, the forty stations that are located inside the basin exhibit either positive or negative daily biases. As a result, bias-corrected SPP (bias-corrected GPCP 1DD) exhibits encouraging results, with an RMSE value range from 5.301249 mm/day to 18.55811 mm/day. Compared to raw SPP, bias-corrected SPP has shown improved results in PBIAS indices.

Continuous statistical indices on a monthly timescale basis

The box plot of the continuous statistical indices of monthly precipitation events after bias correction in the Sunkoshi Basin is displayed in Supplementary Figure S8. The RMSE values for bias-corrected SPP (bias-corrected GPCP 1DD) are from 56.78993 mm/month to 198.3302 mm/month. The PBIAS values are in the range from −42.5112 to 0.434876. There is a considerable amount of decrease in the values of RMSE for bias-corrected GPCP 1DD compared to raw GPCP 1DD on a monthly timescale basis, indicating that monthly biases have reduced in bias-corrected SPP. Similarly, the PBIAS negative and positive values have shifted towards zero in terms of magnitude for bias-corrected SPP compared to raw SPP on a monthly scale.

Hydrological applications (HEC–HMS-simulated model)

The HEC–HMS model was prepared for the Sunkoshi River Basin with the help of ground gauge datasets. The same model, after calibration and validation, was used to assess the performance of the GPCP 1DD. For the calibration and validation of the simulated model, the ground-based observed discharge data were used. In the calibrated and validated model, SPP data were fed into the model as input and as the outcomes of the model, the performance indices were evaluated to assess the applicability of the SPP. The procedures were carried out for the SPP before the bias correction and also after the bias correction of the SPP (GPCP 1DD). The discharges at the junctions within the basin were also obtained from the simulated model.

Performance of the model with an observed discharge flow

The HEC–HMS hydrological model was calibrated and validated in two hydrological stations: Hampachuwar (Station number: 681) and Pachaur Ghat (Station number: 630).

Hampachuwar (Station number: 681)
Calibration was done from 1st January 1997 to 31st December 2002, and validation was done from 1st January 2003 to 31st December 2006. The calibration was carried out on the basis of the available discharge data of the Hampachuwar hydrological station (Station number: 681). The daily hydrograph of the observed versus simulated discharge flow for the calibration and validation period at Hampachuwar is shown in Figures 6 and 7, respectively. Based on the statistical parameters shown in Table 2, it can be concluded that the performance of the model for the calibration, validation, and the entire period was good. The various parameters during calibration were optimised and updated. These optimised parameters are presented in the Supplementary Tables S7–S12.
Table 2

Performance indices of the model at Hampachuwar (Station number: 681)

PeriodNSE (%)R2PBIAS
Calibration 78.1 0.82 24 
Validation 77 0.8 14.9 
PeriodNSE (%)R2PBIAS
Calibration 78.1 0.82 24 
Validation 77 0.8 14.9 
Figure 6

Daily discharge hydrograph for the model calibration period (1997–2002) at Hamphachuwar (Station number: 681).

Figure 6

Daily discharge hydrograph for the model calibration period (1997–2002) at Hamphachuwar (Station number: 681).

Close modal
Figure 7

Daily discharge hydrograph for the model validation period (2003–2006) at Hamphachuwar (Station number: 681).

Figure 7

Daily discharge hydrograph for the model validation period (2003–2006) at Hamphachuwar (Station number: 681).

Close modal

It can be seen that low flows between observed and simulated discharge demonstrate a good relationship, but the high flows have not been good. One reason may be that NSE values have not attained 100% upper range or R2 values have not attained the perfect score. The other reason may be that during the calibration, the conditions and parameters used in HEC–HMS might have been more biased in certain ways, which may have led to such results. The other cause may be that the HEC–HMS model depends upon the initial assumptions, and the parameter values may not be exactly as they should have been; also, the high flows are associated with complex processes.

The KGE, for the observed data and simulated data at Hampachuwar during the calibration period, is 0.74, and for the validation period it is 0.82, as shown in Table 3. This indicates that the performance of the model is good. The RSR value for the observed and simulated discharge data at the Hampachuwar for the calibration period is 0.47, and for the validation period it is 0.47. This suggests that the model performance is very good, as shown in Table 1.

Table 3

KGE and RSR of the model at Hampachuwar (Station number: 681)

PeriodrKGERSR
Calibration 0.95 1.24 0.9 0.74 0.47 
Validation 0.99 1.15 0.9 0.82 0.47 
PeriodrKGERSR
Calibration 0.95 1.24 0.9 0.74 0.47 
Validation 0.99 1.15 0.9 0.82 0.47 

The scatter plotter shows a positive correlation between observed and simulated data since the slope line is increasing at Hampachuwar (Station number: 681), as shown in Supplementary Figure S9.

Pachaur Ghat (Station number: 630)

The daily hydrograph of observed versus simulated discharge flow for the calibration and validation period at Pachaur Ghat is shown in Supplementary Figures S10 and S11, respectively. Some of the statistical parameters are shown in Supplementary Table S13. Similarly, the performance evaluation through KGE and RSR for the Pachaur Ghat is shown in Supplementary Table S14. The RSR value for the observed and simulated discharge values at the Pachaur Ghat for the calibration period is 0.48, which shows very good performance, and for the validation period it is 0.65, which indicates satisfactory performance, as shown in Table 1.

The scatter plotter shows a positive correlation between observed and simulated data since the slope line is increasing at Pachaur Ghat (Station number: 630), as shown in Supplementary Figure S12.

Performance of SPP in the calibrated-validated hydrological model (HEC–HMS)

Raw SPP and bias-corrected SPP were fed into the hydrological model that was calibrated and validated with the observed discharge datasets. The effectiveness of the raw GPCP 1DD and bias-corrected GPCP 1DD was evaluated by feeding them into the simulated model, respectively, for their hydrological applicability.

GPCP 1DD (Raw)

The performance of the extracted raw GPCP 1DD was obtained by forcing the datasets from 1997 to 2006 into the already calibrated and validated hydrological model in HEC–HMS. The process was simultaneously carried out at two hydrological stations: Hampachuwar (Station number: 681) and Pachaur Ghat (Station number: 630). The daily discharge hydrograph for Hampachuwar and Pachaur Ghat between the observed and simulated raw GPCP 1DD discharge is presented in Supplementary Figures S13 and S14, respectively. From the daily discharge hydrograph of both the stations available, it can be seen that extreme events have not been recorded so well by the simulated raw GPCP 1DD discharge flow since it was unable to generate the high flows similar to that of flows generated by observed from 1997 to 2006, whereas the low discharge flows generated are comparable.

Considering the statistical results of the simulated raw GPCP 1DD model at both the stations, i.e. Hampachuwar and Pachaur Ghat in the Sunkoshi River Basin, the overall performance of the model is good. The statistical indices at Hampachuwar (Station number: 681) are shown in Supplementary Table S15, and the KGE and RSR for Hampachuwar (Station number: 681) are shown in tabulated form in Supplementary Table S16. The RSR value of 0.57 for raw GPCP 1DD shows the good performance of the model at Hampachuwar. Similarly, the indices at Pachaur Ghat (Station number: 630) are shown in Supplementary Table S17 and the KGE and RSR for Pachaur Ghat (Station number: 630) are shown in Supplementary Table S18.

Bias-corrected GPCP 1DD

Bias-corrected GPCP 1DD performed better in identifying the extreme events during high discharge flow compared to the raw GPCP 1DD. Low flows are well captured by bias-corrected data and are comparable to those of the observed discharge flow. However, even after bias correction, the extreme events were not captured compared to the observed flow. The daily discharge hydrograph of bias-corrected GPCP 1DD versus observed flow at Hampachuwar (Station number: 681) and at Pachaur Ghat (Station number: 630) is shown in Supplementary Figure S15 and S16, respectively.

The statistical indices at Hampachuwar (Station number: 681) are shown in Supplementary Table S19 and the KGE and RSR at Hampachuwar (Station number: 681) are shown in Supplementary Table S20. Similarly, the indices at Pachaur Ghat (Station number: 630) are shown in Supplementary Table S21, and the KGE and RSR at Pachaur Ghat (Station number: 630) are shown in Supplementary Table S22. The RSR value for the observed discharge and the bias-corrected GPCP at Hampachuwar (Station number: 681) is 0.58, which shows the good performance of the hydrological model. When comparing the values of various performance indices with the values indicated in Table 1, the overall performance of the model is good for the bias-corrected GPCP 1DD.

Energy calculated and sensitivity to SPP in selected HPPs in the Sunkoshi River Basin

The energy calculation was done from the discharges that were obtained from the ground gauge datasets, the SPP datasets before bias correction, and the SPP datasets after the bias correction. The sensitivity analysis was performed on the basis of different dependable flows.

Khimti I HPP (60 MW)

At 40% of dependable flow, the simulated bias GPCP 1DD generates energy of 140.115 GWh in the dry season, which is more than that generated by the simulated gauge: 133.019 GWh in that particular season. Similarly, during the wet season, the simulated bias GPCP 1DD has generated 285.8 GWh of energy, meanwhile the simulated gauge has an energy production of 286.074 GWh. The energy generated by simulated raw GPCP 1DD has less energy generation during both seasons compared to simulated bias GPCP 1DD and simulated gauge. The energy generated from the simulated gauge and simulated bias GPCP 1DD is the same from June to November, whereas the energy generated from the simulated bias GPCP 1DD is more than that of the simulated gauge from March to May. The monthly energy generated at Khimti I HPP at a 40% dependable flow is displayed in Figure 8.
Figure 8

Monthly energy generated by the Khimti I HPP (60 MW).

Figure 8

Monthly energy generated by the Khimti I HPP (60 MW).

Close modal

Upper Bhotekoshi HPP (45 MW)

The simulated bias GPCP 1DD generates an energy of 92.753 GWh in the dry season, which is more than that generated by the simulated gauge, which is 84.792 GWh in that particular season at 40% of the dependable flow. Similarly, the simulated bias GPCP 1DD has generated 178.053 GWh of energy, meanwhile the simulated gauge has an energy production of 166.527 GWh during the wet season. The energy generated by simulated raw GPCP 1DD has less energy generation during both seasons compared to that of energy generated by simulated bias GPCP 1DD and simulated gauge. The energy generated from the simulated bias GPCP 1DD is more than the energy generated from the simulated gauge. The monthly energy generated at the Upper Bhotekoshi HPP at a 40% dependable flow is displayed in Supplementary Figure S17.

Khimti II HPP (48.8 MW)

At 40% of dependable flow, the simulated bias GPCP 1DD generates energy of 59.834 GWh in the dry season, which is less than that generated by the simulated gauge, which is 61.151 GWh in that particular season. Similarly, during the wet season, the simulated bias GPCP 1DD generated 125.92 GWh of energy, meanwhile, the simulated gauge has an energy production of 135.654 GWh. The energy generated by simulated raw GPCP 1DD has less energy generation during both seasons compared to the simulated bias GPCP 1DD and the simulated gauge. The energy generated from the simulated gauge is more than the energy generated from the simulated bias GPCP 1DD. The monthly energy generated at Khimti II HPP at a 40% dependable flow is displayed in Supplementary Figure S18.

Likhu IV HPP (52.4 MW)

The energy generated after the bias correction of GPCP 1DD, which refers to the simulated raw GPCP 1DD, is 90.607 GWh, and the energy generated by the simulated gauge in the dry season is 82.793 GWh. In the wet season, the simulated bias GPCP 1DD generates 189.399 GWh of energy, meanwhile, the energy generated by the simulated gauge is 188.763 GWh. The GPCP 1DD before the bias correction, i.e. the simulated raw GPCP 1DD, has an energy generation of 74.179 GWh in the dry season and 177.42 GWh in the winter season. The annual energy generated by the simulated gauge, the simulated raw GPCP 1DD, and the simulated bias GPCP 1DD is 271.566, 251.599 , and 280.006 GWh, respectively. The energy generated from the simulated bias GPCP 1DD is more than the energy generated from the simulated gauge. The energy generated from the simulated raw GPCP 1DD is less than that from the simulated bias GPCP 1DD and the simulated gauge. The monthly energy generated at Likhu IV HPP at a 40% dependable flow is displayed in Supplementary Figure S19.

Madhya Bhotekoshi HPP (102 MW)

At 40% of dependable flow, the simulated bias GPCP 1DD generates the energy of 172.118 GWh in the dry season, which is more than that generated by the simulated gauge, which is 157.342 GWh in that particular season. Similarly, during the wet season, the simulated bias GPCP 1DD generated 330.402 GWh of energy, meanwhile the simulated gauge discharge has energy production of 309.015 GWh. The energy generated by simulated raw GPCP 1DD has less energy generation during both seasons compared to the simulated bias GPCP 1DD and the simulated gauge. The monthly energy generated at Madhya Bhotekoshi HPP at a 40% dependable flow is displayed in Supplementary Figure S20.

Solu Khola HPP (86 MW)

The bias-corrected GPCP, i.e. simulated bias GPCP 1DD, generates an energy of 180.129 GWh in the dry season, which is more than that generated by the simulated gauge, which is 148.033 GWh in that particular season at 40% of dependable flow. Similarly, the simulated bias GPCP has generated 397.665 GWh of energy, meanwhile the simulated gauge has energy production of 390.741 GWh during the wet season. The energy generated from the simulated bias GPCP 1DD is more than the energy generated from the simulated gauge. The energy generated by simulated raw GPCP 1DD has less energy generation during both seasons compared to the energy generated by the simulated bias GPCP 1DD and the simulated gauge. The monthly energy generated at the Solu Khola HPP at a 40% dependable flow is displayed in Supplementary Figure S21.

Upper Lapche Khola HPP (52 MW)

The energy generated by the simulated bias GPCP 1DD is 146.442 GWh, and the energy generated by the simulated gauge in the dry season is 146.683 GWh. In the wet season, the simulated bias GPCP 1DD generates 255.159 GWh of energy, meanwhile, the energy generated by the simulated gauge is 292.995 GWh. The simulated raw GPCP 1DD has an energy generation of 139.639 GWh in the dry season and 216.165 GWh in the winter season. The annual energy generated by the simulated gauge simulated raw GPCP 1DD and simulated bias GPCP 1DD is 439.678, 355.804, and 401.601 GWh, respectively. The energy generated from the simulated gauge is more than that of the energy generated from the simulated bias GPCP 1DD from May to December. Similarly, the energy generated from the simulated raw GPCP 1DD is less than the other. The monthly energy generated at the Upper Lapche Khola HPP at a 40% dependable flow is displayed in Supplementary Figure S22.

Upper Balephi ‘A’ HPP (36 MW)

At a 40% dependable flow, the simulated bias GPCP 1DD generates an energy of 28.44 GWh in the dry season, which is more than that generated by the simulated gauge, which is 25.999 GWh in that particular season. Similarly, during the wet season, the simulated bias GPCP 1DD has generated 54.594 GWh of energy, meanwhile the simulated gauge has an energy production of 51.057 GWh. The energy generated from the simulated bias GPCP 1DD is more than the energy generated from the simulated gauge. The energy generated by simulated raw GPCP 1DD has less energy generation during both seasons compared to the energy generated by simulated bias GPCP 1DD and simulated gauge. The monthly energy generated at Upper Balephi ‘A’ HPP at a 40% dependable flow is displayed in Supplementary Figure S23.

Singati Khola HPP (25 MW)

The energy generated after the bias correction of GPCP 1DD, i.e. the simulated bias GPCP 1DD, is 33.406 GWh and the energy generated by a simulated gauge in the dry season is 33.46 GWh. In the wet season, the simulated bias GPCP 1DD generates 58.206 GWh of energy; meanwhile, the energy generated by the simulated gauge is 66.837 GWh. The simulated raw GPCP 1DD has an energy generation of 31.855 GWh in the dry season and 49.311 in the winter season. The annual energy generated by the simulated gauge simulated raw GPCP 1DD and simulated bias GPCP 1DD is 100.297, 81.166, and 91.612 GWh, respectively. The energy generated from the simulated gauge is more than the energy generated from the simulated bias GPCP 1DD. The monthly energy generated at the Khimti II HPP at a 40% dependable flow is displayed in Supplementary Figure S24.

Khani Khola I HPP (40 MW)

The energy generated by the simulated bias GPCP 1DD is 48.877 GWh, and the energy generated by the simulated gauge in the dry season is 48.956 GWh. In the wet season, the simulated bias GPCP 1DD generates 85.161 GWh of energy, meanwhile, the energy generated by the simulated gauge is 97.791 GWh. The simulated raw GPCP 1DD has an energy generation of 46.608 GWh in the dry season and 72.15 GWh in the wet season. The energy generated from the simulated bias GPCP 1DD is less than the energy generated from the simulated gauge. The annual energy generated by the simulated gauge, simulated raw GPCP 1DD, and simulated bias GPCP 1DD is 146.747, 118.758, and 134.038 GWh, respectively. The monthly energy generated at the Khani Khola I HPP at a 40% dependable flow is displayed in Supplementary Figure S25.

Sensitivity

The variation in energy generation before bias correction at Q20 dependable flow is more than other dependable flow percentages. However, after the bias correction, the variation in energy generation reduced significantly at all dependable flows. The energy generation from GPCP is sensitive to various dependable flows, mostly at Q20 and Q60. At Khimti I HPP, the percentage change in annual energy at Q20 is 170.85% and at Q60 it is 31.08%, whereas the percent change dropped to 8.16% at Q20 and to −7.37% at Q60 after bias correction. Similarly, Singati Khola HPP has a percent change in annual energy of 2.31% at Q20 and 3.39% at Q60, but after the bias correction, it is 0.12% at Q20 and 0.34% at Q60. The percentage change in annual energy generation at different dependable flows is shown in Supplementary Table S23.

This study evaluated the performance of SPP over the Sunkoshi River Basin. The evaluation of raw SPP was statistically done against the observed data. Also, the bias correction was applied for the raw SPP using the quantile mapping method. After that, the bias-corrected SPP was evaluated statistically against the same observed data over the Sunkoshi River Basin. The conclusions of this study can be summarised as follows: The CSI values showed more improvement after the bias correction of GPCP 1DD than those of the raw one. This shows the improvement of capability after bias correction of GPCP 1DD to correctly estimate the rainfall events of the ground-based gauges. The POD values showed no significant changes in values before and after the bias correction done to GPCP 1DD. The study showed that raw and bias-corrected GPCP 1DD both capture the real-time occurrence of rainfall in a similar fashion to that captured by ground gauges within satisfactory ranges. Also, GPCP 1DD, both raw and bias-corrected, cannot detect the day-to-day rainfall events precisely. The FAR values had no significant improvement even after the bias correction on a daily timescale. The FBI values before bias correction of GPCP 1DD showed mixed results of over- and underestimation of precipitation over the Sunkoshi River Basin. However, after the bias correction, the GPCP 1DD showed an underestimation of precipitation detection for the majority of regions of the Sunkoshi River Basin. The values of RMSE showed improvement after the bias correction was done to GPCP, both on the daily and monthly scales. So, there can be uncertainty issues if the SPP is used without carrying out bias corrections.

The values for NSE, R2 and PBIAS for the observed gauge, raw GPCP 1DD, and bias-corrected GPCP 1DD all showed good results. The analysis of the scatter plotter showed good results. SPP before bias correction performed well. However, after the bias correction of SPP, the overall performance was better with respect to the performance indicators. Thus, this study suggests that the SPP (GPCP) can replicate the hydrological patterns over the Sunkoshi River Basin in the HEC–HMS model, which is calibrated and validated based on ground gauge datasets. Thus, SPP is used further for the evaluation of energy generation in various HPPs within the Sunkoshi River Basin. The energy generated by SPP before bias correction has less energy generation during dry and wet seasons compared to bias-corrected SPP and ground gauge simulated discharge. This study shows that energy generation from the raw SPP was agreeable; however, after the bias correction of the GPCP 1DD, the energy generated by the selected HPPs improved and was much more comparable to the energy generated from the ground gauge datasets. So, with the scientific improvement in GPCP products in the future from the concerned parties, it can be predicted that this SPP product will definitely yield better results. This study suggests that the GPCP 1DD was sensitive to energy generation at different dependable flows when compared with the observed datasets. The different dependable flows taken as the parameter for the sensitivity analysis of the generated energy of selected HPPs in this study showed that GPCP 1DD is not reliable when there are certain changes in the parameters. However, the bias-corrected GPCP showed improvement in the sensitivity analysis with a considerable decrease in variations of energy at different dependable flows compared to that of energy generated from the ground gauge data. Various previous studies have shown that the effectiveness of GPCP products varies with the place, indicating the need for extensive research in smaller places. This study was carried out in a large basin like the Sunkoshi River Basin, which may be the reason for the mixed outcomes.

The usefulness of SPPs in hydrological modelling and energy generation evaluations is clarified by this work, especially in areas with limited data, such as Nepal's Sunkoshi River Basin. The results have important findings for a number of stakeholders, such as disaster management agencies, hydropower creators, and strategists of water resources.

The potential of bias-corrected SPP data in calculating energy generation for HPPs is demonstrated in this study. For hydropower designers and developers in Nepal, where ground-based precipitation data are scarce and dispersed unevenly, this is very helpful. The findings can help improve the planning and functioning of hydropower plants, particularly in isolated and hilly areas where setting up ground gauges is difficult.

After bias correction, SPP can reproduce hydrological patterns in the HEC–HMS model, making it a dependable substitute for water resource management. In the Sunkoshi River Basin and other comparable areas, this can aid in river flow forecasts, agricultural water allocation management, and sustainable water usage.

This work adds to the expanding amount of research on the use of SPPs in hydrological modelling. It offers a methodology for assessing and correcting for bias in SPPs that may be applied to other basins with comparable data constraints.

The results are especially meaningful to underdeveloped nations with comparable topography and climates, where logistical and financial obstacles frequently limit the ability to collect data on the ground.

We express our gratitude to the Department of Civil Engineering, Pulchowk Campus, Lalitpur, Nepal, for their continuous support. We are thankful to the concerned organisations for making the SPP data available freely.

No funding was received.

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

The authors declare there is no conflict.

Adler
R. F.
,
Huffman
G. J.
,
Chang
A.
,
Ferraro
R.
,
Xie
P.-P.
,
Janowiak
J.
,
Bruno
R.
,
Schneider
U.
,
Curtis
S.
,
Bolvin
D.
,
Gruber
A.
,
Susskind
J.
,
Arkin
P.
&
Nelkin
E.
(
2003
)
The Version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–Present)
,
Journal of Hydrometeorology
,
4
(
6
),
1147
1167
.
https://doi.org/10.1175/1525-7541(2003)004 < 1147:TVGPCP > 2.0.CO;2
.
Ahbari
A.
,
Laila
S.
,
Agoumi
A.
&
Serhir
N.
(
2018
)
Estimation of initial values of the HMS model parameters: application to the basin of Bin El Ouidane (Azilal, Morocco)
,
Journal of Materials and Environmental Sciences
,
9
(
1
),
305
317
.
https://doi.org/10.26872/jmes.2018.9.1.34
.
Andari
R.
,
Nurhamidah
N.
,
Daoed
D.
&
Marzuki
M.
(
2024
)
Evaluation of bias correction methods for multi-satellite rainfall estimation products
,
IOP Conference Series: Earth and Environmental Science
,
1317
(
1
),
012008
.
https://doi.org/10.1088/1755-1315/1317/1/012008
.
Bajracharya
A. R.
,
Bajracharya
S. R.
,
Shrestha
A. B.
&
Maharjan
S. B.
(
2018
)
Climate change impact assessment on the hydrological regime of the Kaligandaki Basin, Nepal
,
Science of The Total Environment
,
625
,
837
848
.
https://doi.org/10.1016/j.scitotenv.2017.12.332
.
Basnet
K.
,
Baniya
U.
&
Karki
S.
(
2018
)
Comparative study of design discharge calculation approaches: a case study of Padhu khola, Kaski, Nepal
,
OODBODHAN
,
5
(
5
),
41
49
.
Belayneh
A.
,
Sintayehu
G.
,
Berhanu
K. G.
&
Muluken
T.
(
2020
)
Evaluation of satellite precipitation products using HEC–HMS model
,
Modeling Earth Systems and Environment
,
6
(
2
),
1
18
.
https://doi.org/10.1007/s40808-020-00792-z
.
Bhattarai
R.
,
Mishra
B. K.
,
Bhattarai
D.
,
Khatiwada
D.
,
Kumar
P.
&
Meraj
G.
(
2024
)
Assessing hydropower potential in Nepal's Sunkoshi river basin: an integrated GIS and SWAT hydrological modeling approach
,
Scientifica
,
2024
,
1
19
.
https://doi.org/10.1155/2024/1007081
.
Bolvin
D. T.
,
Adler
R. F.
,
Huffman
G. J.
,
Nelkin
E. J.
&
Poutiainen
J. P.
(
2009
)
Comparison of GPCP monthly and daily precipitation estimates with high-latitude gauge observations
,
Journal of Applied Meteorology and Climatology
,
48
(
9
),
1843
1857
.
https://doi.org/https://doi.org/10.1175/2009JAMC2147.1
.
Curtis
S.
,
Salahuddin
A.
,
Adler
R. F.
,
Huffman
G. J.
,
Gu
G.
&
Hong
Y.
(
2007
)
Precipitation extremes estimated by GPCP and TRMM: eNSO relationships
,
Journal of Hydrometeorology
,
8
(
4
),
678
689
.
https://doi.org/10.1175/JHM601.1
.
Dangol
S.
,
Talchabhadel
R.
&
Pandey
V. P.
(
2022
)
Performance evaluation and bias correction of gridded precipitation products over Arun river basin in Nepal for hydrological applications
,
Theoretical and Applied Climatology
,
148
(
3–4
),
1
20
.
https://doi.org/10.1007/s00704-022-04001-y
.
DoED
(
2018
)
Guidelines for Study of Hydropower Projects
.
Kathmandu
:
Department of Electricity Development
. .
Emerson
D. G.
,
Vecchia
A. V.
&
Dahl
A. L.
(
2005
)
Evaluation of drainage-area ratio method used to estimate streamflow for the Red river of the North Basin, North Dakota and Minnesota
.
Scientific Investigations Report
. 2005–5017. https://doi.org/10.3133/sir20055017
.
Gebre
S. L.
(
2015
)
Application of the HEC-HMS model for runoff simulation of Upper Blue Nile River Basin
,
Hydrology: Current Research
,
6
(
2
),
1
.
https://doi.org/10.4172/2157-7587.1000199
.
Gebremichael
M.
,
Krajewski
W. F.
,
Morrissey
M. L.
,
Huffman
G.
&
Adler
R. F.
(
2005
)
A detailed evaluation of GPCP 1° daily rainfall estimates over the Mississippi river basin
,
American Meteorological Society
,
44
(
5
),
665
681
.
https://doi.org/10.1175/JAM2233.1
.
Gobiet
A.
&
Tani
S. N.
(
2021
)
Quantile mapping for improving precipitation extremes from regional climate models
,
Journal of Agrometeorology
,
21
(
4
),
434
444
.
https://doi.org/10.54386/jam.v21i4.278
.
Gupta
H. V.
,
Kling
H.
,
Yilmaz
K. K.
&
Martinez
G. F.
(
2009
)
Decomposition of the mean squared error and NSE performance criteria: implications for improving hydrological modeling
,
Journal of Hydrology
,
377
(
1–2
),
80
91
.
https://doi.org/10.1016/j.jhydrol.2009.08.003
.
Han
W. S.
,
Burian
S. J.
&
Shepherd
M.
(
2010
)
Assessment of satellite-based rainfall estimates in urban areas in different geographic and climatic regions
,
Natural Hazards
,
56
(
3
),
733
747
.
https://doi.org/10.1007/s11069-010-9585-7
.
Hossain
K.
(
2014
)
Satellite Based Flood Forecasting for the Koshi River Basin, Nepal
.
Master of Engineering thesis
.
Asian Institute of Technology
.
Hossain
F.
&
Katiyar
N.
(
2006
)
Improving flood forecasting in international river basins
,
Eos, Transactions American Geophysical Union
,
87
(
5
),
49
54
.
https://doi.org/10.1029/2006EO050001
.
Huffman
G. J.
&
Pendergrass
A.
&
National Center for Atmospheric Research Staff
(
2023
)
The Climate Data Guide: TRMM: Tropical Rainfall Measuring Mission
.
Boulder, CO
:
Climate Data Guide
. .
ICIMOD
(
2010
)
Land Cover of HKH Region 2010
.
Kathmandu
:
ICIMOD
. .
JICA
(
1985
)
Master Plan Study on the Kosi River Water Resources Development
.
Tokyo
:
Japan International Cooperation Agency
.
Joshi
M. K.
,
Rai
A.
&
Pandey
A. C.
(
2012
)
Validation of TMPA and GPCP 1DD against the ground truth rain-guage data for Indian region
,
International Journal of Climatology
,
33
(
12
),
2633
2648
.
https://doi.org/10.1002/joc.3612
.
Kolekar
S. V.
,
Muthappa
K. J.
,
Gowtham Prasad
M. E.
,
Shruthi
H. G.
,
Shivaprasad
H.
&
Ram
N. S.
(
2017
)
Applicability of HEC-HMS tool to Western Ghats-Nethravathi river
,
International Journal Of Advanced Research in Engineering & Management (IJAREM)
,
3
(
4
),
70
79
.
https://doi.org/10.13140/RG.2.2.12323.09763
.
Kumar
S.
,
Amarnath
G.
,
Ghosh
S.
,
Park
E.
,
Baghel
T.
,
Wang
J.
,
Pramanik
M.
&
Belbase
D.
(
2022
)
Assessing the performance of the satellite-based precipitation products (SPP) in the data-sparse Himalayan terrain
,
Remote Sensing
,
14
(
19
),
4810
.
https://doi.org/10.3390/rs14194810
.
Lamontagne
J. R.
,
Barber
C. A.
&
Vogel
R. M.
(
2020
)
Improved estimators of model performance efficiency for skewed hydrologic data
,
Water Resources Research
,
56
(
9
),
1
24
.
https://doi.org/10.1029/2020WR027101
.
Le
M.-H.
,
Lakshmi
V.
,
Bolten
J.
&
Bui
D. D.
(
2020
)
Adequacy of satellite-derived precipitation estimate for hydrological modeling in Vietnam basins
,
Journal of Hydrology
,
586
,
124820
.
https://doi.org/10.1016/j.jhydrol.2020.124820
.
McPhee
J.
&
Margulis
S. A.
(
2005
)
Validation and error characterization of the GPCP-1DD precipitation product over the contiguous United States
,
Journal of Hydrometeorology
,
6
(
4
),
441
459
.
https://doi.org/10.1175/JHM429.1
.
Moriasi
D.
,
Arnold
J.
,
Liew
M. W.
,
Bingner
R.
,
Harmel
R.
&
Veith
T. L.
(
2007
)
Model evaluation guidelines for systematic quantification of accuracy in watershed simulations
,
Transactions of the ASABE
,
50
(
3
),
885
900
.
https://doi.org/10.13031/2013.23153
.
NEA
(
2017
)
NEA Board Decisions on the Power Purchase Rates
.
Kathmandu
:
Nepal Electricity Authority
. .
Pool
S.
,
Vis
M. J.
&
Seibert
J.
(
2018
)
Evaluating model performance: towards a non-parametric variant of the Kling-Gupta efficiency
,
Hydrological Sciences Journal
,
63
(
13
),
1941
1953
.
https://doi.org/10.1080/02626667.2018.1552002
.
Rajeevan
M.
,
Unnikrishnan
C. K.
,
Bhate
J.
,
Kumar
K. N.
&
Sreekala
P. P.
(
2012
)
Northeast monsoon over India: variability and prediction
,
Meteorological Applications
,
19
(
2
),
129
264
.
https://doi.org/10.1002/met.1322
.
Rudolf
B.
&
Schneider
U.
(
2005
). '
Calculation of gridded precipitation data for the global land-surface using in-situ gauge observations
’,
Proceedings of the 2nd Workshop of the International Precipitation Working Group IPWG
.
Shrestha
M.
,
Artan
G.
,
Bajracharya
S. R.
,
Gautam
D. K.
&
Tokar
S. A.
(
2011
)
Bias-adjusted satellite-based rainfall estimates for predicting floods: narayani Basin
,
Journal of Flood Risk Management
,
4
(
4
),
271
373
.
https://doi.org/10.1111/j.1753-318X.2011.01121.x
.
Silva
M. G.
,
Netto
A. O.
,
Neves
R. J.
,
Vasco
A. N.
,
Almeida
C.
&
Faccioli
G. G.
(
2015
)
Sensitivity analysis and calibration of hydrological modeling of the watershed Northeast Brazil
,
Journal of Environmental Protection
,
6
(
8
),
837
850
.
https://doi.org/10.4236/jep.2015.68076
.
Skomorowski
P.
,
Rubel
F.
&
Rudolf
B.
(
2001
)
Verification of GPCP-1DD global satellite precipitation products using MAP surface observations
,
Physics and Chemistry of the Earth
,
26
(
5–6
),
403
409
.
https://doi.org/10.1016/S1464-1909(01)00026-0
.
Talchabhadel
R.
,
Karki
R. C.
&
Parajuli
B.
(
2017
)
Intercomparison of precipitation measured between automatic and manual precipitation gauge in Nepal
,
Measurement
,
106
,
264
273
.
https://doi.org/10.1016/j.measurement.2016.06.047
.
Tan
M. L.
,
Ibrahim
A. L.
,
Duan
Z.
,
Cracknell
A. P.
&
Chaplot
V.
(
2015
)
Evaluation of six high-resolution satellite and ground-based
,
Remote Sensing
,
7
(
2
),
1504
1528
.
https://doi.org/10.3390/rs70201504
.
Tassew
B. G.
,
Belete
M. A.
&
Miegel
K.
(
2019
)
Application of HEC-HMS model for flow simulation in the Lake Tana basin: the case of Gilgel Abay catchment, Upper Blue Nile Basin, Ethiopia
,
Hydrology
,
6
(
1
),
21
.
https://doi.org/10.3390/hydrology6010021
.
Thapa
S.
(
2018
)
Development of Hydropower Plants on Sunkoshi River in Nepal
.
MSc thesis
.
UNESCO-IHE: Institute for Water Education
,
Delft
.
Todini
E.
&
Biondi
D.
, (
2017
)
Calibration, parameters estimation, uncertainty, data assimilation, sensitivity analysis and validation
. In:
Singh
V. P.
(ed.)
Handbook of Applied Hydrology
,
New York
:
McGraw-Hill Education
, pp.
22-1
22-15
.
Vrugt
J. A.
&
Oliveira
D. Y.
(
2022
)
Confidence intervals of the Kling-Gupta efficiency
,
Journal of Hydrology
,
612
(
Part A
),
127968
.
https://doi.org/doi.org/10.1016/j.jhydrol.2022.127968
.
Zhang
L.
,
Xin
Z.
,
Zhang
C.
&
Song
C.
(
2022
)
Exploring the potential of satellite precipitation after bias correction in streamflow simulation in a semi-arid watershed in northeastern China
,
Journal of Hydrology: Regional Studies
,
43
(
2–4
),
265
274
.
https://doi.org/10.1016/j.ejrh.2022.101192
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).

Supplementary data