1. Introduction
Climate change has altered hydrological processes, adversely influencing global water resources (Abbas et al. 2022; Muir et al. 2018; Field et al. 2012). The impacts of climate change are pronounced in the high-altitude regions, with snow and glacier cover threatening the livelihoods of billions of people (Zierl and Bugmann 2005; Beniston 2003; Becker and Bugmann 2001). High Mountain Asia (HMA) is one such region where the impact of climate change on different hydrological fluxes and its effects on water availability are predominant. HMA is a region of highly complex terrain extending from the Hindu Kush and Tien Shan in the northwest to the eastern Himalayas in the southeast (Bolch et al. 2012) and consists of the largest reserves of glaciers and snow cover outside the polar region, making it vulnerable to climate change (Immerzeel et al. 2020). Increasing glacier retreat and snow-cover changes due to a rise in temperature and shifting spatial and temporal patterns of precipitation will affect the future water availability in the region (Lutz et al. 2014). Hence, it is crucial to advance our understanding of how the changes in climate will impact the hydrological processes in glacierized catchments.
The availability of in situ observations of hydroclimatic variables for facilitating hydrologic research is minimal in HMA due to the complexity of the terrain (Thornton et al. 2022; Mishra et al. 2021; Ward et al. 2011). With the advancements in remote sensing and climate modeling, the scarcity of in situ data in such regions can be resolved to a great extent (Huang et al. 2022, 2020; Avtar et al. 2020; Maswood and Hossain 2016; Yang et al. 2013). The satellite and atmospheric reanalysis-based products (SRP) have been used extensively for research studies over mountainous areas due to their high spatial and temporal coverage (Ougahi and Mahmood 2022; Hafizi and Sorman 2021; Derin et al. 2019; Mei et al. 2016; Derin et al. 2016; Hoyos and Webster 2007). Findings from these past studies highlight that the performance of SRPs compared to ground observations depends on many factors such as the spatial resolution of the product, local climate, and topography of the region. For instance, Nadeem et al. (2022) compared the performance of nine commonly used SRPs in Pakistan and found that the different SRPs perform differently over the same region, influenced by the climatic and topographic conditions. Additionally, SRP is associated with significant uncertainties introduced due to errors in the retrieval algorithm, sampling errors (Nijssen and Lettenmaier 2004; Bell and Kundu 2000), and inconsistencies in convection parameterization (Hersbach et al. 2020; Derin et al. 2019). When used as input to a hydrological model, uncertainties in these products propagate into uncertainties in simulated hydrological variables (Bhuiyan et al. 2019a; Falck et al. 2015; Serpetzoglou et al. 2010; Nikolopoulos et al. 2010). It is worth noting that the pattern of propagation of uncertainties from precipitation to streamflow is also driven by the spatial and temporal scales considered (Nanding et al. 2021; Wu et al. 2017).
For hydrological analysis over complex mountain terrains like HMA, it is crucial to incorporate modeling of snow and ice melting. Two different approaches are typically used for melt modeling: energy balance–based and temperature index–based approaches. Energy balance or physical models are process-based models which introduce complex energy transmission fluxes to the model (Ismail et al. 2023; Meng et al. 2015). They provide a better understanding of the melt rates by integrating the effects of longwave and shortwave radiation, albedo changes, heat transport, etc. While these models provide an accurate estimation of physical processes in the region, they are data intensive and computationally complex. In regions like HMA, where the data availability is deficient, it is common to use temperature-based models for estimating melt rates due to their computational simplicity, fewer data requirements, and reasonable performance (Lutz et al. 2016; Luo et al. 2013; Immerzeel et al. 2009; Zhang et al. 2006; Kayastha et al. 2005; Hock 2003). These models use degree-day factors to establish a relation between air temperature and melting rates for snow and ice. Even though all the physical processes responsible for melting are not incorporated in temperature-based models, their performance is often comparable to energy-based models in a catchment-scale analysis (Hock 2003; WMO 1986). However, temperature-based models are found to be highly sensitive to temperature (Sicart et al. 2008) and have limited spatial and temporal transferability (Wheler et al. 2014; MacDougall and Flowers 2011).
Several past studies have identified that the performance of the hydrological models will be significantly affected by the uncertainties in input data. For instance, Bárdossy et al. (2020) compared the input variables and model parameter uncertainties in a data-sparse region in southwest Germany. The authors found that the input uncertainty is more significant than the model parameter uncertainty and should be taken into consideration while calibrating the models (Bárdossy et al. 2020). An extensive evaluation of eight satellite-based precipitation estimates over the monsoon-prone region in China also confirmed the dominance of input precipitation uncertainty over hydrological model uncertainties (Chen et al. 2020). Pokorny et al. (2021) evaluated multiple sources of uncertainties by considering three models and five satellite-based precipitation products. It was observed that the cumulative effect of multiple uncertainties resulted in larger bounds of simulated uncertainty (Pokorny et al. 2021). A multimodel multiforcing study over the Iberian Peninsula showed that both the input precipitation uncertainties and model structural uncertainties significantly affect the simulation of hydrological fluxes, and the degree of significance varied for different fluxes (Bhuiyan et al. 2019a). Elsner et al. (2014) evaluated the impact of four different meteorological datasets on streamflow simulations using a hydrological model over the western United States and revealed that the choice of forcing dataset influences the model calibration and subsequent streamflow simulations. Assessing and quantifying different sources of uncertainties is essential before using SRP for analysis over regions of complex terrains. Past studies indicate that the uncertainties introduced by SRP are much more prominent in higher elevations than in other areas (e.g., Schreiner-McGraw and Ajami 2022, 2020; Henn et al. 2018; Lundquist et al. 2015, among others). Mei et al. (2016) showed that the systematic error of precipitation products and the model simulations increased with mean basin elevation over the eastern Italian Alps. While several studies evaluated the performance of SRP over glacierized catchments of HMA (He et al. 2021; Dahri et al. 2021; Kanda et al. 2020; Hamm et al. 2020; Li et al. 2018; Hussain et al. 2017; Ceglar et al. 2017; Palazzi et al. 2013; Andermann et al. 2011), investigations on the propagation of precipitation uncertainty on the hydrological response of glacierized catchment are limited (Mimeau et al. 2019).
The studies mentioned, but not limited to the above, have revealed the significant regional dependency on the performance of SRPs, which highlights the continuous need for regional and local scale investigations for assessing the applicability of SRPs for hydrologic analyses. This is further emphasized if one considers that the efficacy of SRP-based hydrologic simulations depends, to a certain degree, on the type of dominant hydrological processes (e.g., rainfall-dominated vs snowmelt-dominated streamflow). The uncertainties in SRPs coupled with model parameter uncertainties pose significant challenges in streamflow estimations in glacierized catchments and require further investigation. Although there are similar studies established in the past, we are presenting this study as an attempt to understand the applicability of satellite and reanalysis products in a data-scarce region like HMA using parsimonious models that are in practical use by local scientists. To the best of our knowledge, assessment of uncertainty propagation from SRP to hydrologic components (including rainfall-runoff, snowmelt, and ice melt) over small glacierized catchments in HMA is limited. Owing to this need, in this paper, we focus on assessing uncertainties introduced by precipitation inputs and model parameters to streamflow simulations over two glacierized catchments in Nepal. To achieve this goal, we analyzed the uncertainty in streamflow simulations based on three widely used SRPs [Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (GPM) (IMERG), Climate Hazards Group Infrared Precipitation with Station (CHIRPS), and the fifth major global atmospheric reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) (ERA5-Land); see details in section 2b] and two glacio-hydrological models. In doing so, the study answers the following specific research questions: 1) How will the uncertainties in SRP propagate to streamflow simulations in glacierized catchments in Nepal? 2) What is the dependence on model structure? 3) How will the uncertainty affect the different streamflow components? Additionally, we investigated the sensitivity of streamflow simulations to temperature to highlight its relative importance for hydrologic simulations in the region.
2. Study area, data, and models
a. Study area
The study focuses on understanding the hydrological characteristics of two glacierized catchments in Nepal, which are a part of the central Himalayas in HMA (Fig. 1a). The propagation of precipitation uncertainty on streamflow is evaluated for two subbasins—the Marsyangdi River basin (MRB) and the Budhigandaki River basin (BRB). MRB and BRB are among the five glaciated subbasins of the Gandaki River basin in Nepal (Bhattarai and Conway 2020). MRB has an area of about 4059 km2 and an elevation ranging from 355 to 7819 m above mean sea level (MSL). Approximately 10.4% of the area of MRB is covered with glaciers [calculated using glacier shapefiles from version 6.0 of Randolph Glacier Inventory (RGI6.0) (RGI Consortium 2017)]. According to the glacier outlines available from RGI6.0, there are 348 glaciers in MRB and 330 glaciers in BRB. BRB extends to around 3881.2 km2, of which 8.1% are glaciers. The elevation ranges from 477 to 7983 m MSL. The permanent snow line in the Nepalese Himalayas is reckoned above 5000 m MSL, where precipitation is in the form of rain and snow (Pangali Sharma et al. 2020). The hypsometric distribution of the two basins revealed that the majority of the basin area is concentrated over the elevation of 4000–6000 m MSL (see Figs. S1a and S1b in the online supplemental material). MRB has an average slope of 27.8°, while BRB has an average slope of 30.6°. The basins are characterized by a temperate climate with dry winters and warm summers in the lower part (Karki et al. 2016). They are dominated by the Indian summer monsoon, bringing precipitation from June to September (Kayastha and Kayastha 2020). Figure 1b shows the subbasin river boundaries of Nepal obtained from the International Centre for Integrated Mountain Development (ICIMOD 2021), with the location of MRB and BRB highlighted. The two catchments are selected based on the availability of ground observations and due to their significance in water resources and energy production. MRB has great hydropower potential with two already existing hydropower plants and three others under planning (Mudbhari et al. 2022; Khadka and Pathak 2016). The feasibility study for the Budhigandaki Hydroelectric Project (BGHEP), located on BRB, has been completed which has a capacity of 1200 MW (Devkota et al. 2017; BGHEP 2015).
(a) Extent of HMA. (b) Map of Nepal with the location of MRB (green) and BRB (blue). (c) Land-use map of MRB and BRB.
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
b. Datasets
Daily precipitation and temperature values are required as input to the glacio-hydrological models (discussed in section 2c), and observed streamflow at the outlet of MRB and BRB is used for model calibration and validation. Streamflow observations were obtained for MRB [hydrological station at Bimal Nagar (354 m MSL)] and BRB [hydrological station at Arughat (485 m MSL)] from the Department of Hydrology and Meteorology, Government of Nepal. Ground observations of air temperature and precipitation are obtained from the climatological station at Khudi Bazar (elevation 823 m MSL) for MRB and Gorkha (elevation 1097 m MSL) for BRB (see Table S1). In situ observation of streamflow, precipitation, and temperature is available over the period of 2004–10 for MRB and 1980–2015 for BRB. The land-use/cover data from GlobeLand30 is used to classify the two basins into four major classes, namely, grass/agricultural land, shrub land, barren land, and water bodies (Fig. 1c). In addition to these classifications, the glaciers are also represented as debris-covered and clean glaciers based on information from the ICIMOD glacier inventory (ICIMOD 2014). The glacier cover was derived by combining data from ICIMOD and ice thickness data from Farinotti et al. (2019).
Precipitation data used include two high-resolution satellite precipitation products, 1) the final version 06 release of the IMERG Mission (Huffman et al. 2019) and 2) CHIRPS data (Funk et al. 2014), the ground-based Asian Precipitation–Highly Resolved Observational Data Integration Toward Evaluation (APHRODITE) (Yatagai et al. 2012), and one reanalysis product, the ERA5-Land (Muñoz-Sabater et al. 2021). The IMERG version 06 algorithm uses a combination of precipitation estimates from the Tropical Rainfall Measuring Mission (TRMM) satellite (2000–15) and the recent GPM satellite (2014–present) (Huffman et al. 2019). IMERG precipitation data have a spatial resolution of 0.1°. We used the daily precipitation data from final version 06 of IMERG for this study. The CHIRPS precipitation product is a long-term quasi-global dataset available from 1981 to the present with a spatial resolution of 0.05° (Funk et al. 2014). We used the global daily precipitation from CHIRPS version 2.0 for the analysis. APHRODITE is a continental-scale precipitation dataset based on a dense network of rain gauge data, available from 1951 to 2015 for Asia, with a spatial resolution of 0.25° × 0.25° (Yatagai et al. 2012). Daily grid precipitation over Monsoon Asia from APHRODITE-2 version 1801R1 was used for our analysis. ERA5-Land is a long-term reanalysis dataset providing a consistent view of the evolution of land variables with a spatial resolution of 0.1° (Muñoz-Sabater et al. 2021). Both precipitation and 2-m temperature from ERA5-Land are used in the analysis (Muñoz Sabater 2019). Hourly precipitation and temperature from ERA5-Land was downloaded and aggregated to daily time steps for the analysis. Based on the availability of station data and SRP for the two catchments, the time period of study was selected as 2004–10 for MRB and 2001–15 for BRB. Given that some precipitation products were reporting frequent very low (<0.1 mm day−1) values that manifested as high-frequency noise, we applied a 0.1 mm day−1 value as the rain/no rain threshold across all precipitation datasets used.
c. Hydrological models
1) GDM
The Glacio-Hydrological Degree-Day Model (GDM) is a gridded distributed glacio-hydrological model used to simulate the contribution of hydrological components of streamflow, including rainfall-runoff, snowmelt, ice melt, and baseflow (Kayastha et al. 2020; Kayastha and Kayastha 2020). A melt model based on degree-day factor is used for glacier and snowmelt estimations (Kayastha et al. 2005). The model accepts precipitation and temperature data from a station or a pixel and distributes it to the entire basin using precipitation and temperature lapse rates. The whole basin is divided into grids, and the input data from the station are extrapolated to each grid. A grid size of 6 km was used for both basins. The snowmelt, ice melt, baseflow, and rainfall-runoff components are calculated separately at daily time steps. The range of values for the different parameters used in the model, including positive degree-day factor, recession coefficients, and snow and rainfall-runoff coefficients, is obtained from past studies in the Nepalese Himalayas and calibrated manually. GDM has been successfully used for the glacio-hydrological analysis over many subbasins in Nepal, including MRB, the Trishuli River basin (Kayastha and Kayastha 2020), and the Tamor River basin (Kayastha et al. 2020), in Shigar River basin in Pakistan (Hassan et al. 2021), and in Urumqi River basin in China (Yang et al. 2022).
2) HYMOD_DS
Hydrological Model for Distributed Systems (HYMOD_DS; Wi et al. 2015) is an empirical process-based model with separate modules representing soil moisture, evapotranspiration, snow processes, glacier processes, and flow routing. It is an improved version of the original HYMOD (Boyle 2001) with additional modules for flow routing and snow/glacier processes. HYMOD_DS requires gridded daily precipitation and temperature datasets, digital elevation model (DEM) derived from Shuttle Radar Topography Mission (SRTM), land use–land cover data from GlobeLand30, and glacier thickness derived from Randolph Glacier Inventory (Farinotti et al. 2019) as input. The entire catchment is divided into different hydrological response units (HRUs) based on the precipitation grid and elevation of the region. Specifically, the catchment is first divided into polygons corresponding to 100-m elevation bands and the precipitation grid is overlaid to get the union polygons, which form the HRUs for that particular forcing. Thus, the HRUs vary depending on the precipitation product used for the model simulation. The seasonal temperature lapse rates used to derive the temperature for each HRU are obtained from Kattel et al. (2013). The runoff from the glacierized area is calculated separately and added to runoff generated from the soil moisture accounting module coupled with the snow module, assuming no interchange of water between the soil layers and the glacial area. The runoff from each area is weighted by its area fraction within the basin to obtain the total runoff. The model structure and further details of the HYMOD_DS modeling system are elaborated in Wi et al. (2015).
GDM and HYMOD_DS have been used previously in similar watersheds. One of the main differences between the two models used for the analysis is their ability to accept directly distributed meteorological input. While gridded precipitation and temperature data can be used as input to HYMOD_DS, only station-based or point data can be used as input for analysis using GDM. The gridded data, as in SRPs, can be incorporated into GDM by choosing the grid data closer to the latitude and longitude of the in situ measuring station. The point-based information is then distributed across the entire basin using appropriate precipitation and temperature lapse rates. Additionally, the calibration of parameters for HYMOD_DS utilizes the genetic algorithm (Wang 1991) as the optimization method with Nash–Sutcliffe efficiency (NSE) (Nash and Sutcliffe 1970) as the objective function. The current version of GDM did not include an automatic calibration option and thus was calibrated manually using references for the range of parameters from the past literature (Kayastha et al. 2020; Kayastha and Kayastha 2020) and NSE as the objective function. More details on the theory behind both the models are included in section S1 of the supplementary material.
3. Methodology
a. Analysis framework
This work aims to understand and quantify how the uncertainty in precipitation and its interaction with the model uncertainty (pertaining to model parameters and structure) affect the streamflow estimation in glacierized catchments. We focus on understanding the combined effect of uncertainties introduced by hydroclimatic inputs and model structure by using multiple forcing datasets with two glacio-hydrological models that have distinct differences in parameterization scheme and structure. Figure 2 presents the framework adopted in this study. The different precipitation products mentioned in section 2b were first analyzed and compared using the empirical cumulative distribution function (CDF). The complementary cumulative distribution function (CCDF) (plot on a logarithmic scale to emphasize the right tail of the distribution) was also compared to evaluate differences in the extreme precipitation values from different products. The comparison of CDFs and CCDFs was carried out at point-based and catchment average levels. For point-based comparison, the grid data closer to the latitude and longitude of the measuring station (used as reference) were selected and compared with in situ observations to provide, in addition to the uncertainty among products, an estimate of the bias.
The framework of research methodology [S1, S2, S3, and S4 represent the simulated discharge obtained based on calibrated model parameters for IMERG (I), ERA5-Land (E), CHIRPS (C), and APHRODITE (A) precipitation, respectively. T1, T2, T3, and T4 represent the simulations obtained by varying the temperature inputs].
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
First, a calibration/validation exercise was carried out where two glacio-hydrological models were calibrated individually using each precipitation dataset. This resulted in different model parameter/precipitation input pairs for each model per basin. The simulated streamflow obtained from each model setup was compared to the observed streamflow for the basins. Second, each model parameterization obtained from the calibration step was forced with the different precipitation inputs. For instance, model parameters obtained by calibrating the models using IMERG precipitation were used for simulating streamflow with ERA5-Land, CHIRPS, and APHRODITE precipitation inputs. This was repeated using all possible combinations of precipitation products and model parameters. Comparison of the multiple parameterization/forcing output was used to evaluate the combined effect of precipitation and model uncertainties on streamflow estimations. The streamflow simulations thus obtained were compared to a reference simulated streamflow that was chosen based on the performance of precipitation products in the calibration/validation exercise. Using simulated streamflow as a reference (instead of the observed) allows to isolate and quantify the relative contribution of each source of uncertainty across precipitation inputs and model parameters. To maintain consistency between the models, we selected APHRODITE-based simulated streamflow as the reference for both models.
Temperature is also a crucial parameter in determining the hydrological behavior of watersheds, especially in regions such as HMA. The water availability in the form of ice melt and snowmelt is highly influenced by variations in temperature, which implies that the propagation of precipitation uncertainty in streamflow depends, to a certain extent, on uncertainty in temperature. We conducted a sensitivity analysis to investigate the significance of changes in temperature on model-simulated discharge and its dependence on precipitation uncertainty. For the temperature ensitivety exercise, we simply varied temperature inputs to the models by varying the original temperature by ±10% and 20%, respectively. For each precipitation input, the simulated discharge from the sensitivity analysis was compared with the simulated discharge using the reference temperature.
Through this analysis, we try to understand the combined impact of uncertainties introduced by SRPs (as model inputs) and model parameters on streamflow simulations for glacierized catchments. By using two parsimonious models (GDM and HYMOD_DS) for the analysis, we focus on comparing the similarities and differences in model simulations obtained using same precipitation inputs for the same region.
b. Performance evaluation
The performance of the two glacio-hydrological models for the uncertainty propagation analysis was evaluated based on different metrics such as standard deviation (SD), correlation coefficient (CC), and centered root-mean-square error (CRMSE). These metrics were calculated for different model simulations and were plotted in a normalized Taylor diagram for comparison. Taylor diagrams are used as an effective tool to compare complex models in terms of their correlation, root-mean-square differences, and the amplitude of their variances (Taylor 2001). The normalization of SD and CRMSE obtained by dividing them with the corresponding standard deviation value for observation/reference facilitates comparing simulated fluxes of different magnitudes using the same plot (Taylor 2001).
4. Results
This section includes the results obtained for 1) comparison of precipitation products, 2) calibration and validation of models using different precipitation products, 3) precipitation and model parameter uncertainty analysis, and 4) temperature sensitivity analysis.
a. Comparison of precipitation products
Figure 3 illustrates the comparison of CCDF plots for the different precipitation products. The comparison is made based on station pixel and catchment average. The daily precipitation data over the entire study period (2004–10 for Marsyangdi River basin and 2001–15 for Budhigandaki River basin) for both the basins were used for the comparison. ERA5-Land overestimated both the lower and extreme precipitation values. This is found to be true for both station-based and catchment-average precipitation. For MRB, a relative difference of +23.9% was observed between mean precipitation from ERA5-Land and mean station precipitation, while an overestimation of 41.6% was observed in the case of BRB for the respective study periods. On the other hand, IMERG precipitation for MRB was found to be less than the in situ observations for most parts of the distribution spectrum, except for the extremes. An underestimation of 48.1% was found for the mean precipitation from IMERG compared to the mean station precipitation. It was observed that the APHRODITE precipitation data were relatively closer to station data. For BRB, average precipitation from APHRODITE showed a relative difference as low as −1.2%, whereas an underestimation by 18% was observed for MRB. The APHRODITE data are expected to match the station better since it is based on a dense network of rain gauge data (Panthi et al. 2015; Yatagai et al. 2012), and gauges from the Department of Hydrology and Meteorology are part of the sources used in this product. CHIRPS has the highest spatial resolution among other products used in the study and compared well with the station for higher daily (>20 mm day−1) precipitation values although a relative difference of 26% was observed for average precipitation.
CCDF of daily (a) precipitation at station pixel for MRB, (b) catchment average precipitation for MRB, (c) precipitation at station pixel for BRB, and (d) catchment average precipitation for BRB.
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
Regarding average catchment precipitation (Fig. 3b), IMERG, ERA5-Land, and CHIRPS showed a similar distribution, especially for BRB, whereas APHRODITE estimates of extreme precipitation were consistently lower than the other products. This discrepancy in APHRODITE may be accounted for by its coarser spatial resolution and the lack of stations at higher elevations (Andermann et al. 2011). However, we noted that APHRODITE precipitation is comparable with other products over the lower spectrum, matching IMERG data. These findings are consistent with the earlier research by Sunilkumar et al. (2019), in which IMERG data and APHRODITE were compared for three different climate zones, including Nepal. The authors found that the IMERG and APHRODITE showed similar distribution at low precipitation values over Nepal (Sunilkumar et al. 2019). A comparison of the catchment-average precipitation from different products relative to that from APHRODITE revealed a consistent overestimation. Of all products, ERA5-Land precipitation showed significant overestimation for both basins, with a relative difference of 72.0% for MRB and 73.2% for BRB. For MRB, the catchment average IMERG precipitation was found to be comparable to that of APHRODITE with a relative difference of +7.5%, whereas for BRB, an overestimation of 23% was observed. The catchment-average precipitation from CHIRPS was found to be more than APHRODITE by 22% for MRB, while it was relatively close to APHRODITE in the case of BRB (relative difference of +10%).
b. Calibration and validation of the models
The values/ranges of model parameters used for calibrating GDM and HYMOD_DS using different precipitation products are provided in Tables 1 and 2, respectively. All the parameters listed in the tables are not calibrated for each forcing dataset. For instance, temperature lapse rate used for GDM, as mentioned in Table 1 [0.6°C (100 m)−1], was selected based on the past literature (Kayastha et al. 2020; Khadka et al. 2020) and was not calibrated for each forcing. Similarly, the threshold temperature for differentiating precipitation as rain or snow was also kept constant and was not calibrated. The past literature has emphasized the sensitivity and complexity of rain–snow threshold temperature for hydrological model simulations in mountainous regions (Jennings and Molotch 2019; Jennings et al. 2018; Rajagopal and Harpold 2016). However, in this study, we avoided calibrating this parameter and relied on the threshold temperature value used in previous studies where the same models were used in similar catchments (Kayastha et al. 2020; Wi et al. 2015). The list of optimized parameters obtained after calibrating both models using different forcing datasets are provided in Tables S2–S5. Figure 4 shows the time series of 30-day moving average of simulated streamflow obtained for MRB after calibrating the models using different precipitation products. The calibration of the models was done using the daily precipitation and temperature data for the years 2004–07 for MRB and 2001–07 for BRB. The simulated streamflow time series varied considerably in the case of GDM compared to HYMOD_DS. For instance, time series obtained from GDM calibrated using ERA5-Land precipitation resulted in higher early flows from January to March than those obtained from other products. The simulated discharge, NSE values, and the corresponding volume difference with respect to the observed flow for MRB and BRB are tabulated in Tables 3 and 4 and Tables S6 and S7, respectively. We found that for MRB, the best NSE value of 0.8 was obtained for calibration using APHRODITE precipitation in GDM and 0.9 for HYMOD_DS using ERA5-Land precipitation. The corresponding volume differences were 9.9% and 3.4%, respectively. Similar results were obtained for BRB (Fig. S2 and Table S6), except for HYMOD_DS, where the best NSE value was obtained for calibration using IMERG precipitation. The comparison of CCDF of model-simulated streamflow using different precipitation inputs is shown in Fig. 5. The broad spectrum of variation in precipitation products translated to a narrow range in simulated discharge, especially for HYMOD_DS results, demonstrating the effect of model calibration per precipitation input individually. The right tail (higher extreme values) of simulated discharge from both models showed discrepancies compared to observed streamflow, with HYMOD_DS resulting in consistent underestimation.
Model parameters used in the calibration/validation of GDM. Only those parameters listed in bold are calibrated.
Model parameters used in the calibration/validation of HYMOD_DS. Only those parameters listed in bold are calibrated.
Comparison of 30-day moving average of simulated daily discharges for MRB obtained from (a) GDM and (b) HYMOD_DS after calibrating them using different precipitation inputs. The black dashed line separates the time series into calibration and validation periods.
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
Calibration/validation results for MRB using GDM. The average observed discharge for MRB over the years 2004–10 is 213.6 m3 s−1. Negative volume difference indicates an underestimation of model-simulated streamflow with respect to observed streamflow.
Calibration/validation results for MRB using HYMOD_DS.
CCDF of simulated daily discharge: (a) GDM-based results for MRB, (b) HYMOD_DS-based results for MRB, (c) GDM-based results for BRB, and (d) HYMOD_DS-based results for BRB.
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
c. Uncertainty analysis
Having calibrated the models and obtained the best possible set of parameters for each precipitation product, the next step was investigating the combined effect of precipitation and model uncertainties. The simulated discharge obtained using APHRODITE precipitation was chosen as the reference for comparing all other simulations. APHRODITE-based simulations using GDM matched well with the in situ observations. For consistency between the models, the reference simulation for HYMOD_DS was also set as the same, even though ERA5-Land showed better results in the calibration/validation exercise. All possible combinations of precipitation inputs and calibrated model parameters were used to generate a multiforcing/multiparameter simulated streamflow ensemble. The model outputs were compared against the reference simulated output (APHRODITE-based input and calibrated parameter set) using the error metrics mentioned in section 3b and were collectively analyzed using Taylor diagrams.
Figure 6 shows the normalized Taylor diagram for simulated discharge from the different model setups. The shapes (circle—IMERG, square—ERA5-Land, diamond—CHIRPS, and triangle—APHRODITE) represent the model parameters used for the simulation, e.g., circle corresponds to the model parameters obtained when calibration was based on IMERG as precipitation input.
Normalized Taylor diagram for simulated daily discharge: (a) GDM-based results for MRB, (b) HYMOD_DS-based results for MRB, (c) GDM-based results for BRB, and (d) HYMOD_DS-based results for BRB. Note that the colors represent precipitation products and shapes represent corresponding calibrated parameter sets. For example, yellow triangle corresponds to simulations based on parameters calibrated with APHRODITE (triangle) and precipitation input from ERA5-Land (yellow).
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
The colors (red—IMERG, yellow—ERA5-Land, purple—CHIRPS, and green—APHRODITE) represent the different precipitation products. The radial axis (black contours) represents the normalized SD, the angular axis (purple lines) represents the CC, and the red contours represent the normalized CRMSE. The reference-simulated streamflow (APHRODITE-based results) is represented by the data point at the intersection of the CRMSE axis and the curve for SD = 1. These results show an apparent clustering of colors, especially for GDM-based results for BRB, which indicates that even though the parameters were changed, no significant variation was observed for a particular precipitation input. This reveals the strong dominance of precipitation uncertainty over model parameter uncertainty when we use GDM for streamflow simulations in BRB. On the other hand, the different data points are spread over a broader range of CRMSE in the case of GDM-based results for MRB, indicating the significance of both precipitation and model parameter uncertainty. Similarly, the importance of both uncertainties was evident from the HYMOD_DS-based results for both basins, which collectively indicates that the relative importance of precipitation and model parameter uncertainty depends on the model structure. The correlation between the data points of simulated discharge remained consistently high (>0.8), while a significant difference in variability (SD ranging from 0.5 to 2) existed between them with respect to reference streamflow.
Similar to the simulated discharge, Taylor diagrams were used to compare the different hydrological components, including snowmelt and ice melt. The rainfall-runoff and baseflow (hereafter named “RR&Bflow”) simulated using the models are considered as a single component for the analysis. The normalized Taylor diagrams for snowmelt, ice melt, and RR&Bflow from GDM and HYMOD_DS for MRB and BRB are given in Figs. 7 and 8, respectively. The reference simulated components based on APHRODITE are represented as green triangles located at the intersection of the CRMSE axis and the radius SD = 1. The ice melt components for all the model simulations were obtained to be similar to the reference ice melt. This is expected since ice melt is mainly associated with changes in temperature rather than precipitation. The similarity in ice melt contribution was observed for all simulations except for the one obtained from GDM using the ERA5-Land parameter and APHRODITE precipitation for MRB. The scattering of data points, similar to those obtained for total streamflow, was observed for snowmelt and RR&Bflow from GDM-based simulations for MRB, revealing that precipitation and model parameter uncertainties are prominent. The snowmelt obtained using ERA5-Land precipitation and IMERG model parameters was found to have the lowest correlation (0.7 for MRB) and highest CRMSE compared to reference snowmelt.
Normalized Taylor diagram for daily hydrological components of snowmelt, ice melt, and RR&Bflow: (a)–(c) GDM-based results and (d)–(f) HYMOD_DS-based results, respectively, for MRB.
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
As in Fig. 7, but for BRB.
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
For HYMOD_DS, the effect of precipitation uncertainty was found to be more pronounced than parameter uncertainty for the individual streamflow components (manifesting as clustering of colors in the graph) than for the total streamflow. That is, for a given color cluster, even though the parameters used for the simulations were different (represented by different shapes and calibrated based on different precipitation data), the streamflow output is mainly controlled by the precipitation uncertainty rather than model parameter uncertainty. This prominence of precipitation uncertainty is more evident in the case of RR&Bflow (Figs. 7f and 8f), as expected since it is dominated by rainfall-runoff. The different snowmelt components obtained using ERA5-Land precipitation as input to HYMOD_DS showed very high variation and CRMSE despite the high correlation with reference.
d. Temperature sensitivity analysis
Figure 9 presents a comparison of the average monthly contribution of snowmelt, ice melt, and RR&Bflow to the total streamflow obtained using GDM and HYMOD_DS for MRB. GDM showed an increased contribution of ice melt for a 20% increase in temperature compared to HYMOD_DS. The increased ice melt was found to be distributed from March to September, which corresponds to the ice melting season in Nepal. The sensitivity results were consistent across the different precipitation products used. For HYMOD_DS, the relative increase in ice melt is more predominant from June to September and at a lower rate.
Comparison of the contribution of different hydrological components from (a) GDM and (b) HYMOD_DS for MRB. The colored boxes within each figure represent the change in total flow (percentage change with respect to simulated components using original temperature). The blue color box represents a decrease in total flow, with darker colors representing a higher change. The red color box indicates an increase in total flow.
Citation: Journal of Hydrometeorology 25, 6; 10.1175/JHM-D-23-0048.1
The colored boxes in Fig. 9 show the corresponding change in total simulated discharge obtained from the two models for different temperature and precipitation inputs. For GDM, a 20% increase in temperature resulted in a more than 50% increase in simulated discharge, especially for MRB. The high sensitivity of GDM to temperature changes was previously studied (Hassan et al. 2021) and is in line with the results obtained here. For GDM, the temperature from a single station or point is distributed to the entire basin area using a temperature lapse rate of 0.6°C (100 m)−1. When we change the temperature at the station, to say +20%, the intervention of temperature lapse rate along with the increase in temperature may result in a higher relative change over certain parts of the basin. Hence, the resulting change in simulated flow will be more than +20% as obtained. This argument is valid for the decrease in temperature as well. Therefore, it is essential to say that the model structure plays a vital role in the significance of temperature sensitivity, and models, such as GDM, that rely on single station information and the use of lapse rates are more prominent to bias (Khadka et al. 2020) in simulating streamflow due to amplification of bias in temperature induced by the model structure. HYMOD_DS, on the other hand, was observed to be less sensitive to temperature changes, with less than an 8% increase in simulated discharge for a 20% increase in temperature. The change in total flow was found to be directly proportional to the variation in temperature, but the rate of change was lower than that observed for GDM. Note that in contrast to GDM, HYMOD_DS temperature input was changed consistently across the entire basin since the −20% to 20% adjustment was applied across the entire temperature field.
HYMOD_DS simulations using ERA5-Land precipitation showed contrasting results compared to others. As the temperature decreased, the total simulated discharge was found to increase. A possible reason for these contradicting results is the difference in the spatial distribution of different precipitation products across the basin. The average ERA5-Land precipitation was found to be distributed more toward the lower elevations (around 850–3500 m MSL) (see Fig. S3a) where there is no/limited glacier and snow cover. As temperature decreases, evapotranspiration also decreases, which results in higher flow in lower elevations. The decrease in evapotranspiration is also present in higher elevations; still, the total flow is not impacted much because of the simultaneous decline in snow and ice melt due to reduced temperature. The precipitation distribution with elevation from all other products exhibits much less variability compared to ERA5-Land.
The comparison of the average monthly contribution of hydrological components for BRB (Fig. S4) revealed similar characteristics as MRB. However, the ice melt contribution was found to be higher than that for MRB, especially in the case of GDM. For GDM-based simulations, the performance of all precipitation products was consistent across different temperature inputs. For HYMOD_DS, ERA5-Land-based simulations showed a higher contribution of ice melt than others, with an increased contribution from April to June. Also, the effect of evapotranspiration, as seen in the case of MRB, was not observed for BRB.
5. Discussion
Utilization of SRPs for hydrological studies in data-sparse regions like HMA is vital and necessitates the investigation of uncertainties associated with them. The analysis of the different precipitation products considered in the study revealed the range of uncertainty of SRPs in capturing the magnitude and variability of precipitation at both point and catchment scales. ERA5-Land precipitation was found to be overestimating over both catchments considered, which is in line with a previous study on the performance of reanalysis products over the central Himalayas (Chen et al. 2021). An apparent underestimation of IMERG precipitation compared to in situ observations was observed for MRB. The consistent underestimation of IMERG precipitation, especially over the Nepalese Himalayas, has been observed previously (e.g., Nepal et al. 2021; Sharma et al. 2020; Sunilkumar et al. 2019, among others). CHIRPS precipitation showed a significant underestimation compared to all other products for precipitation rates less than 10 mm day−1, which is consistent with other recent studies (Upadhyay et al. 2022). Considerable uncertainty in representing the right tail of precipitation distribution was also apparent for all products considered. Such discrepancies in different aspects of precipitation distribution (low but frequent rates vs extreme values) manifest differently in the hydrologic response of catchments. Lower rainfall rates can play a more critical role in antecedent wetness conditions and snow storage, while higher extreme values are tied more closely to characteristics of flood events. Furthermore, the model parameterization and structure for representing the different hydrologic processes have considerable control over the manifestation of precipitation uncertainty in hydrologic response.
Through this study, we demonstrated how the uncertainty in precipitation magnitude and model parameters impacts streamflow simulations in small glacierized catchments in Nepal. The calibration of glacio-hydrological models using different precipitation products resulted in different hydrological behaviors for the same catchment. Our investigation of the combined effect of precipitation and model parameter uncertainty on streamflow generation revealed the significant importance of both sources of uncertainties. However, depending on the model and hydrologic component examined, the relative significance of each source of uncertainty varies. The Taylor diagrams (Figs. 6–8) showed an apparent clustering of precipitation forcings, indicating that the simulated streamflow components obtained using the same precipitation and different model parameters do not vary significantly in terms of correlation, SD, and CRMSE. The clustering was found to be more evident for HYMOD_DS than GDM, implying a substantial impact of precipitation uncertainty on HYMOD_DS-based simulations, specifically for RR&Bflow. The automated calibration of parameters for HYMOD_DS, compared to the manual calibration of GDM, might influence the reduced parameter uncertainty for HYMOD_DS. On the other hand, GDM-based results for MRB (Figs. 6a,b) were spread across a broader range of CC, revealing the combined effect of both precipitation and model parameter-induced uncertainties.
Simulation results in certain cases (Figs. 6a,b) also highlighted the fact that models can be well calibrated in a particular precipitation forcing but can exhibit significant reduction in performance for other forcings. This precipitation input dependence of the model performance implies “overfitting” of parameters, and it is a known problem in hydrologic model calibration (Zhang et al. 2016; Skinner et al. 2015). To mitigate this issue, past works have proposed approaches that involve calibration based on ensemble forcing (Skinner et al. 2015) and utilization of additional (to in situ streamflow) sources of hydrologic observations from remote sensing dataset (e.g., soil moisture) to better constrain parameter optimization (Silvestro et al. 2015). Investigating additional calibration approaches is beyond the scope of this work, but using our findings, we wish to emphasize the importance of validating model parameters based on auxiliary hydrologic information and local knowledge particularly in such remote regions like High Mountain Asia.
GDM was found to be more sensitive to temperature variations than HYMOD_DS due to its dependence on temperature lapse rates. The comparison of the average monthly contribution of streamflow components indicated that as temperature increases, GDM simulates more ice melt than HYMOD_DS and is distributed over the entire summer months in Nepal. The temperature sensitivity of GDM is consistent across all the precipitation products and for both basins. In the case of HYMOD_DS, although the rate of the change in ice melt contribution is lower than GDM for both the basins, ERA5-Land-based simulations showed a higher amount of snow and ice melt than other products. The higher amounts of flow contribution using ERA5-Land may be attributed to the overestimation of ERA5-Land precipitation (Sun et al. 2021). The contribution of different streamflow components varied significantly for both models. GDM simulations for MRB with actual temperature resulted in a meager contribution of snow and ice melt, whereas, for HYMOD_DS, snow and ice melt was significant from April to August. The model simulations for BRB indicated an increased amount of snow and ice melt compared to MRB. The explanation for such differences in the contribution of hydrological components lies in the inherent disparities in the model structure (Kayastha et al. 2020). This highlights the need for extra caution when relying on parsimonious empirical models for evaluating changes in hydrological fluxes. Overall, results on the sensitivity of temperature suggest that in addition to precipitation forcing uncertainty, uncertainty in temperature forcing has an important control on streamflow partitioning in glacierized catchments, and its level of importance depends strongly on the model structure. Even though MRB and BRB are two adjacent rainfall-dominated catchments which are a part of central Himalayas, the results obtained from the study revealed significant discrepancies between them. The dominance of precipitation uncertainty over model parameters was found to be more pronounced for GDM-based streamflow simulations for BRB than MRB. This indicates a potential strong dependence of our findings on catchment characteristics, which calls for similar investigations in an extended region of HMA to be able generalize our findings.
6. Conclusions
The study presents an analysis of the hydrological response of meteorological forcing and model parameter/structure uncertainty that serves as a testimony to the complexity and challenges involved in simulating hydrological fluxes in glacierized catchments at data-scarce regions of the world, such as High Mountain Asia.
Circling back to the original questions that this study aims to answer, we provide the following concluding remarks based on our findings:
-
How will the uncertainties in SRP propagate to streamflow simulations in glacierized catchments?
Over the period of study considered for each basin, uncertainty in daily precipitation appears to dampen through the precipitation-to-streamflow transformation (i.e., decreasing range of disagreement of daily streamflow). For a given model, propagation of precipitation uncertainty in streamflow is primarily controlled by the uncertainty in precipitation and less by the model parameters. This, however, varies among basins (e.g., GDM results for MRB vs BRB) which suggests that differences in the topographic and land surface characteristic (even from basins within the same region) exert certain control on the relative importance of the precipitation/parameter uncertainty.
-
What is the dependence on model structure?
The overall finding of reduced streamflow uncertainty relative to precipitation forcing uncertainty is consistent between models. However, the results in terms of the clustering observed on the Taylor diagrams, as well as the relative deviation among the various precipitation/parameter pairs examined, exhibit significant differences between the two models. Therefore, the absolute magnitude of uncertainty in streamflow is highly model dependent, and thus, caution should be taken on the hydrologic evaluations of SRPs, especially when single models are considered. Interestingly, while HYMOD_DS utilizes directly gridded precipitation input, an advancement relative to the lapse rates used in the GDM, and reports higher NSE values during calibration/validation per product; simulations using the APHRODITE-based parameters and the other precipitation products resulted in relatively poor performance (low correlation values). On the other hand, GDM exhibited a more consistent response in terms of streamflow simulation for the different precipitation products. Models that are well calibrated in a particular forcing but are more sensitive to forcing uncertainty imply “overfitting.” This is an important consideration during model selection and model calibration procedures given that all forcing data are associated with some degree of uncertainty and emphasizes the need to consider auxiliary hydrologic information (e.g., soil moisture, snow cover, and land surface temperature) for better constraining model parameters. Sensitivity to temperature depends significantly on the model structure. The lapse rate–based distribution of temperature over the basin (used in GDM) leads to much higher sensitivity of the streamflow simulations than the HYMOD_DS which uses directly gridded temperature information.
-
How will the uncertainty affect the different streamflow components?
Precipitation uncertainty, manifested as scatter of colors on Taylor diagrams, affects predominantly the RR&Bflow component and to a less degree the snowmelt components of streamflow. Runoff contributed from glacier melting is primarily affected by uncertainty in temperature. Our results showed that for models that are based on temperature lapse rates, a 20% increase in temperature input can lead to as high as 50% increase in streamflow due to increased contributions of glacier melting. Uncertainty in temperature and its impact on streamflow simulations has received less attention in previous studies, but for high mountain areas and glacierized catchments, it is an important contributing factor in streamflow uncertainty and has to be accounted for.
We note that the comparison between the two models is driven by our objective to examine model dependence on the uncertainty analysis and not to declare which of the models should be considered as best. Here, we provided a reality check of how similar or different the model simulations can be when using two structurally different models over the same area. Both models have advantages, which can be effectively utilized in a data-sparse region like HMA. However, caution should be taken when attempting to simulate different hydrologic components without prior knowledge of uncertainty sources and their impact.
Results presented for the two catchments in Nepal provide important insight but cannot be generalized for the entire HMA given the strong regional dependence of SRP across complex terrains, as well as the role of dominant hydrological processes in precipitation uncertainty propagation to streamflow. Future work should be carried out across more complex terrain regions to provide a more integrated view on the dependence on hydrologic regimes and basin characteristics. For advancing the performance of glacio-hydrologic model simulations, the use of blended products that utilize the advantages of satellite estimates and reanalysis products (e.g., Maina et al. 2022; Bhuiyan et al. 2019b) may offer a better (reduced uncertainty) option for hydrological applications in HMA.
Reliable precipitation products are key elements for improving our understanding of land surface hydrologic variables and for assisting the decision-making process of relevant stakeholders. Thus, continuous efforts from the relevant agencies and the scientific community are needed for improving precipitation estimation over complex terrain areas and for developing datasets on additional hydrologic variables (such as streamflow and snow/glacier melt) that can help constrain/improve model parameterizations.
Acknowledgments.
This work was supported by the NASA High Mountain Asia Program (Grants 80NSSC20K1300 and 80NSSC20K1299).
Data availability statement.
The data from ERA5-Land, IMERG, CHIRPS, and APHRODITE used in this study are openly available in https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-land?tab=overview, https://gpm.nasa.gov/data/directory, https://data.chc.ucsb.edu/products/CHIRPS-2.0/, and http://aphrodite.st.hirosaki-u.ac.jp/download/, respectively. Due to their proprietary nature, the in situ precipitation and streamflow data used in this study cannot be made openly available. Further information about the data and conditions for access are available from the Department of Hydrology and Meteorology of Nepal. The software for GDM and HYMOD_DS models used are available upon reasonable request from RBK and SW, respectively.
REFERENCES
Abbas, M., L. Zhao, and Y. Wang, 2022: Perspective impact on water environment and hydrological regime owing to climate change: A review. Hydrology, 9, 203, https://doi.org/10.3390/hydrology9110203.
Andermann, C., S. Bonnet, and R. Gloaguen, 2011: Evaluation of precipitation data sets along the Himalayan front. Geochem. Geophys. Geosyst., 12, Q07023, https://doi.org/10.1029/2011GC003513.
Avtar, R., and Coauthors, 2020: Assessing sustainable development prospects through remote sensing: A review. Remote Sens. Appl., 20, 100402, https://doi.org/10.1016/j.rsase.2020.100402.
Bárdossy, A., F. Anwar, and J. Seidel, 2020: Hydrological modelling in data sparse environment: Inverse modelling of a historical flood event. Water, 12, 3242, https://doi.org/10.3390/w12113242.
Becker, A., and H. Bugmann, 2001: Global change and mountain regions: The Mountain Research Initiative. International Geosphere-Biosphere Programme Rep. 49, 89 pp., http://www.igbp.net/download/18.1b8ae20512db692f2a680006378/1376383119394/report_49-BAHC.pdf.
Bell, T. L., and P. K. Kundu, 2000: Dependence of satellite sampling error on monthly averaged rain rates: Comparison of simple models and recent studies. J. Climate, 13, 449–462, https://doi.org/10.1175/1520-0442(2000)013<0449:DOSSEO>2.0.CO;2.
Beniston, M., 2003: Climatic change in mountain regions: A review of possible impacts. Climatic Change, 59, 5–31, https://doi.org/10.1023/A:1024458411589.
Bhattarai, K., and D. Conway, 2020: Contemporary Environmental Problems in Nepal: Geographic Perspectives. Advances in Asian Human-Environmental Research, Springer, 764 pp.
Bhuiyan, M. A. E., and Coauthors, 2019a: Assessment of precipitation error propagation in multi-model global water resource reanalysis. Hydrol. Earth Syst. Sci., 23, 1973–1994, https://doi.org/10.5194/hess-23-1973-2019.
Bhuiyan, M. A. E., E. I. Nikolopoulos, and E. N. Anagnostou, 2019b: Machine learning–based blending of satellite and reanalysis precipitation datasets: A multiregional tropical complex terrain evaluation. J. Hydrometeor., 20, 2147–2161, https://doi.org/10.1175/JHM-D-19-0073.1.
Bolch, T., and Coauthors, 2012: The state and fate of Himalayan glaciers. Science, 336, 310–314, https://doi.org/10.1126/science.1215828.
Boyle, D. P., 2001: Multicriteria calibration of hydrologic models. Ph.D. dissertation, The University of Arizona, 138 pp.
Ceglar, A., A. Toreti, G. Balsamo, and S. Kobayashi, 2017: Precipitation over monsoon Asia: A comparison of reanalyses and observations. J. Climate, 30, 465–476, https://doi.org/10.1175/JCLI-D-16-0227.1.
Chen, J., Z. Li, L. Li, J. Wang, W. Qi, C.-Y. Xu, and J.-S. Kim, 2020: Evaluation of multi-satellite precipitation datasets and their error propagation in hydrological modeling in a monsoon-prone region. Remote Sens., 12, 3550, https://doi.org/10.3390/rs12213550.
Chen, Y., S. Sharma, X. Zhou, K. Yang, X. Li, X. Niu, X. Hu, and N. Khadka, 2021: Spatial performance of multiple reanalysis precipitation datasets on the southern slope of central Himalaya. Atmos. Res., 250, 105365, https://doi.org/10.1016/j.atmosres.2020.105365.
Dahri, Z. H., and Coauthors, 2021: Spatio-temporal evaluation of gridded precipitation products for the high-altitude Indus basin. Int. J. Climatol., 41, 4283–4306, https://doi.org/10.1002/joc.7073.
Derin, Y., and Coauthors, 2016: Multiregional satellite precipitation products evaluation over complex terrain. J. Hydrometeor., 17, 1817–1836, https://doi.org/10.1175/JHM-D-15-0197.1.
Derin, Y., and Coauthors, 2019: Evaluation of GPM-era global satellite precipitation products over multiple complex terrain regions. Remote Sens., 11, 2936, https://doi.org/10.3390/rs11242936.
Devkota, R. P., V. P. Pandey, U. Bhattarai, H. Shrestha, S. Adhikari, and K. N. Dulal, 2017: Climate change and adaptation strategies in Budhi Gandaki River Basin, Nepal: A perception-based analysis. Climatic Change, 140, 195–208, https://doi.org/10.1007/s10584-016-1836-5.
Elsner, M. M., S. Gangopadhyay, T. Pruitt, L. D. Brekke, N. Mizukami, and M. P. Clark, 2014: How does the choice of distributed meteorological data affect hydrologic model calibration and streamflow simulations? J. Hydrometeor., 15, 1384–1403, https://doi.org/10.1175/JHM-D-13-083.1.
Falck, A. S., V. Maggioni, J. Tomasella, D. A. Vila, and F. L. R. Diniz, 2015: Propagation of satellite precipitation uncertainties through a distributed hydrologic model: A case study in the Tocantins–Araguaia basin in Brazil. J. Hydrol., 527, 943–957, https://doi.org/10.1016/j.jhydrol.2015.05.042.
Farinotti, D., M. Huss, J. J. Fürst, J. Landmann, H. Machguth, F. Maussion, and A. Pandit, 2019: A consensus estimate for the ice thickness distribution of all glaciers on Earth. Nat. Geosci., 12, 168–173, https://doi.org/10.1038/s41561-019-0300-3.
BGHEP, 2015: Feasibility study and detailed design of Budhi Gandaki HPP. Phase 3: Draft detailed design report. Volume 1, Main Rep. BG-DDR-Vol.1-Rev.0, 181 pp., https://www.scribd.com/document/510538135/Budhi-Gandaki-Final-Detailed-Desing-Report.
Field, C. B., V. Barros, T. F. Stocker, and Q. Dahe, Eds., 2012: Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. Cambridge University Press, 582 pp., https://doi.org/10.1017/CBO9781139177245.
Funk, C. C., and Coauthors, 2014: A quasi-global precipitation time series for drought monitoring. USGS Data Series 832, 12 pp., https://doi.org/10.3133/ds832.
Hafizi, H., and A. A. Sorman, 2021: Assessment of satellite and reanalysis precipitation products for rainfall–runoff modelling in a mountainous basin. Environ. Sci. Proc., 8, 25, https://doi.org/10.3390/ecas2021-10345.
Hamm, A., and Coauthors, 2020: Intercomparison of gridded precipitation datasets over a sub-region of the central Himalaya and the southwestern Tibetan Plateau. Water, 12, 3271, https://doi.org/10.3390/w12113271.
Hassan, J., X.-q. Chen, R. B. Kayastha, and Y. Nie, 2021: Multi-model assessment of glacio-hydrological changes in central Karakoram, Pakistan. J. Mt. Sci., 18, 1995–2011, https://doi.org/10.1007/s11629-021-6748-9.
He, Q., J. Yang, H. Chen, J. Liu, Q. Ji, Y. Wang, and F. Tang, 2021: Evaluation of extreme precipitation based on three long-term gridded products over the Qinghai-Tibet Plateau. Remote Sens., 13, 3010, https://doi.org/10.3390/rs13153010.
Henn, B., A. J. Newman, B. Livneh, C. Daly, and J. D. Lundquist, 2018: An assessment of differences in gridded precipitation datasets in complex terrain. J. Hydrol., 556, 1205–1219, https://doi.org/10.1016/j.jhydrol.2017.03.008.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.
Hock, R., 2003: Temperature index melt modelling in mountain areas. J. Hydrol., 282, 104–115, https://doi.org/10.1016/S0022-1694(03)00257-9.
Hoyos, C. D., and P. J. Webster, 2007: The role of intraseasonal variability in the nature of Asian monsoon precipitation. J. Climate, 20, 4402–4424, https://doi.org/10.1175/JCLI4252.1.
Huang, Q., D. Long, M. Du, Z. Han, and P. Han, 2020: Daily continuous river discharge estimation for ungauged basins using a hydrologic model calibrated by satellite altimetry: Implications for the SWOT mission. Water Resour. Res., 56, e2020WR027309, https://doi.org/10.1029/2020WR027309.
Huang, Q., D. Long, Z. Han, and P. Han, 2022: High-resolution satellite images combined with hydrological modeling derive river discharge for headwaters: A step toward discharge estimation in ungauged basins. Remote Sens. Environ., 277, 113030, https://doi.org/10.1016/j.rse.2022.113030.
Huffman, G. J., and Coauthors, 2019: NASA Global Precipitation Measurement (GPM) Integrated Multi-satellitE Retrievals for GPM (IMERG). Algorithm Theoretical Basis Doc., version 06, 39 pp., https://gpm.nasa.gov/sites/default/files/2020-05/IMERG_ATBD_V06.3.pdf.
Hussain, S., X. Song, G. Ren, I. Hussain, D. Han, and M. H. Zaman, 2017: Evaluation of gridded precipitation data in the Hindu Kush–Karakoram–Himalaya mountainous area. Hydrol. Sci. J., 62, 2393–2405, https://doi.org/10.1080/02626667.2017.1384548.
ICIMOD, 2014: Glaciers of Nepal 2010. ICIMOD, accessed 11 June 2021, https://doi.org/10.26066/RDS.9352.
ICIMOD, 2021: Sub-sub-basins of Hindu Kush Himalaya (HKH) Region. ICIMOD, accessed 1 July 2022, https://doi.org/10.26066/RDS.7952.
Immerzeel, W. W., P. Droogers, S. M. de Jong, and M. F. P. Bierkens, 2009: Large-scale monitoring of snow cover and runoff simulation in Himalayan River basins using remote sensing. Remote Sens. Environ., 113, 40–49, https://doi.org/10.1016/j.rse.2008.08.010.
Immerzeel, W. W., and Coauthors, 2020: Importance and vulnerability of the world’s water towers. Nature, 577, 364–369, https://doi.org/10.1038/s41586-019-1822-y.
Ismail, M. F., W. Bogacki, M. Disse, M. Schäfer, and L. Kirschbauer, 2023: Estimating degree-day factors of snow based on energy flux components. Cryosphere, 17, 211–231, https://doi.org/10.5194/tc-17-211-2023.
Jennings, K. S., and N. P. Molotch, 2019: The sensitivity of modeled snow accumulation and melt to precipitation phase methods across a climatic gradient. Hydrol. Earth Syst. Sci., 23, 3765–3786, https://doi.org/10.5194/hess-23-3765-2019.
Jennings, K. S., T. S. Winchell, B. Livneh, and N. P. Molotch, 2018: Spatial variation of the rain–snow temperature threshold across the Northern Hemisphere. Nat. Commun., 9, 1148, https://doi.org/10.1038/s41467-018-03629-7.
Kanda, N., H. S. Negi, M. S. Rishi, and A. Kumar, 2020: Performance of various gridded temperature and precipitation datasets over Northwest Himalayan Region. Environ. Res. Commun., 2, 085002, https://doi.org/10.1088/2515-7620/ab9991.
Karki, R., R. Talchabhadel, J. Aalto, and S. K. Baidya, 2016: New climatic classification of Nepal. Theor. Appl. Climatol., 125, 799–808, https://doi.org/10.1007/s00704-015-1549-0.
Kattel, D. B., T. Yao, K. Yang, L. Tian, G. Yang, and D. Joswiak, 2013: Temperature lapse rate in complex mountain terrain on the southern slope of the central Himalayas. Theor. Appl. Climatol., 113, 671–682, https://doi.org/10.1007/s00704-012-0816-6.
Kayastha, R. B., and R. Kayastha, 2020: Glacio-Hydrological Degree-Day Model (GDM) useful for the Himalayan River basins. Himalayan Weather and Climate and their Impact on the Environment, A. P. Dimri et al., Eds., Springer, 379–398, https://doi.org/10.1007/978-3-030-29684-1_19.
Kayastha, R. B., Y. Ageta, and K. Fujita, 2005: Use of positive degree-day methods for calculating snow and ice melting and discharge in glacierized basins in the Langtang Valley, Central Nepal. Climate and Hydrology in Mountain Areas, John Wiley & Sons, Ltd, 5–14, https://doi.org/10.1002/0470858249.ch2.
Kayastha, R. B., N. Steiner, R. Kayastha, S. K. Mishra, and K. McDonald, 2020: Comparative study of hydrology and icemelt in three Nepal River basins using the Glacio-Hydrological Degree-Day Model (GDM) and observations from the Advanced Scatterometer (ASCAT). Front. Earth Sci., 7, 354, https://doi.org/10.3389/feart.2019.00354.
Khadka, D., and D. Pathak, 2016: Climate change projection for the Marsyangdi River basin, Nepal using statistical downscaling of GCM and its implications in geodisasters. Geoenviron. Disasters, 3, 15, https://doi.org/10.1186/s40677-016-0050-0.
Khadka, M., R. B. Kayastha, and R. Kayastha, 2020: Future projection of cryospheric and hydrologic regimes in Koshi River basin, Central Himalaya, using coupled glacier dynamics and glacio-hydrological models. J. Glaciol., 66, 831–845, https://doi.org/10.1017/jog.2020.51.
Li, H., J. E. Haugen, and C.-Y. Xu, 2018: Precipitation pattern in the Western Himalayas revealed by four datasets. Hydrol. Earth Syst. Sci., 22, 5097–5110, https://doi.org/10.5194/hess-22-5097-2018.
Lundquist, J. D., M. Hughes, B. Henn, E. D. Gutmann, B. Livneh, J. Dozier, and P. Neiman, 2015: High-elevation precipitation patterns: Using snow measurements to assess daily gridded datasets across the Sierra Nevada, California. J. Hydrometeor., 16, 1773–1792, https://doi.org/10.1175/JHM-D-15-0019.1.
Luo, Y., J. Arnold, S. Liu, X. Wang, and X. Chen, 2013: Inclusion of glacier processes for distributed hydrological modeling at basin scale with application to a watershed in Tianshan Mountains, northwest China. J. Hydrol., 477, 72–85, https://doi.org/10.1016/j.jhydrol.2012.11.005.
Lutz, A. F., W. W. Immerzeel, A. B. Shrestha, and M. F. P. Bierkens, 2014: Consistent increase in high Asia’s runoff due to increasing glacier melt and precipitation. Nat. Climate Change, 4, 587–592, https://doi.org/10.1038/nclimate2237.
Lutz, A. F., W. W. Immerzeel, P. D. A. Kraaijenbrink, A. B. Shrestha, and M. F. P. Bierkens, 2016: Climate change impacts on the upper Indus hydrology: Sources, shifts and extremes. PLOS ONE, 11, e0165630, https://doi.org/10.1371/journal.pone.0165630.
MacDougall, A. H., and G. E. Flowers, 2011: Spatial and temporal transferability of a distributed energy-balance glacier melt model. J. Climate, 24, 1480–1498, https://doi.org/10.1175/2010JCLI3821.1.
Maina, F. Z., S. V. Kumar, I. J. Dollan, and V. Maggioni, 2022: Development and evaluation of ensemble consensus precipitation estimates over high mountain Asia. J. Hydrometeor., 23, 1469–1486, https://doi.org/10.1175/JHM-D-21-0196.1.
Maswood, M., and F. Hossain, 2016: Advancing river modelling in ungauged basins using satellite remote sensing: The case of the Ganges–Brahmaputra–Meghna basin. Int. J. River Basin Manage., 14, 103–117, https://doi.org/10.1080/15715124.2015.1089250.
Mei, Y., E. I. Nikolopoulos, E. N. Anagnostou, and M. Borga, 2016: Evaluating satellite precipitation error propagation in runoff simulations of mountainous basins. J. Hydrometeor., 17, 1407–1423, https://doi.org/10.1175/JHM-D-15-0081.1.
Meng, X.-Y., D.-L. Yu, and Z.-H. Liu, 2015: Energy balance-based SWAT model to simulate the mountain snowmelt and runoff—Taking the application in Juntanghu watershed (China) as an example. J. Mt. Sci., 12, 368–381, https://doi.org/10.1007/s11629-014-3081-6.
Mimeau, L., M. Esteves, H.-W. Jacobi, and I. Zin, 2019: Evaluation of gridded and in situ precipitation datasets on modeled glacio-hydrologic response of a small glacierized Himalayan catchment. J. Hydrometeor., 20, 1103–1121, https://doi.org/10.1175/JHM-D-18-0157.1.
Mishra, S. K., and Coauthors, 2021: Grand challenges of hydrologic modeling for food-energy-water nexus security in high mountain Asia. Front. Water, 3, 728156, https://doi.org/10.3389/frwa.2021.728156.
Mudbhari, D., M. L. Kansal, and P. Kalura, 2022: Impact of climate change on water availability in Marsyangdi River basin, Nepal. Quart. J. Roy. Meteor. Soc., 148, 1407–1423, https://doi.org/10.1002/qj.4267.
Muir, M. J., and Coauthors, 2018: Effects of climate change on hydrology, water resources, and soil. Climate change vulnerability and adaptation in the Intermountain Region: Part 1, J. E. Halofsky, et al., Eds., U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, General Tech. Rep. RMRS-GTR-375, 60–88.
Muñoz Sabater, J., 2019: ERA5-Land hourly data from 2001 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS), accessed 10 May 2021, https://doi.org/10.24381/CDS.E2161BAC.
Muñoz-Sabater, J., and Coauthors, 2021: ERA5-Land: A state-of-the-art global reanalysis dataset for land applications. Earth Syst. Sci. Data, 13, 4349–4383, https://doi.org/10.5194/essd-13-4349-2021.
Nadeem, M. U., A. A. J. Ghanim, M. N. Anjum, D. Shangguan, G. Rasool, M. Irfan, U. M. Niazi, and S. Hassan, 2022: Multiscale ground validation of satellite and reanalysis precipitation products over diverse climatic and topographic conditions. Remote Sens., 14, 4680, https://doi.org/10.3390/rs14184680.
Nanding, N., H. Wu, J. Tao, V. Maggioni, H. E. Beck, N. Zhou, M. Huang, and Z. Huang, 2021: Assessment of precipitation error propagation in discharge simulations over the contiguous United States. J. Hydrometeor., 22, 1987–2008, https://doi.org/10.1175/JHM-D-20-0213.1.
Nash, J. E., and J. V. Sutcliffe, 1970: River flow forecasting through conceptual models part I—A discussion of principles. J. Hydrol., 10, 282–290, https://doi.org/10.1016/0022-1694(70)90255-6.
Nepal, B., D. Shrestha, S. Sharma, M. S. Shrestha, D. Aryal, and N. Shrestha, 2021: Assessment of GPM-Era Satellite Products’ (IMERG and GSMaP) ability to detect precipitation extremes over mountainous country Nepal. Atmosphere, 12, 254, https://doi.org/10.3390/atmos12020254.
Nijssen, B., and D. P. Lettenmaier, 2004: Effect of precipitation sampling error on simulated hydrological fluxes and states: Anticipating the global precipitation measurement satellites. J. Geophys. Res., 109, D02103, https://doi.org/10.1029/2003JD003497.
Nikolopoulos, E. I., E. N. Anagnostou, F. Hossain, M. Gebremichael, and M. Borga, 2010: Understanding the scale relationships of uncertainty propagation of satellite rainfall through a distributed hydrologic model. J. Hydrometeor., 11, 520–532, https://doi.org/10.1175/2009JHM1169.1.
Ougahi, J. H., and S. A. Mahmood, 2022: Evaluation of satellite-based and reanalysis precipitation datasets by hydrologic simulation in the Chenab River basin. J. Water Climate Change, 13, 1563–1582, https://doi.org/10.2166/wcc.2022.410.
Palazzi, E., J. von Hardenberg, and A. Provenzale, 2013: Precipitation in the Hindu-Kush Karakoram Himalaya: Observations and future scenarios. J. Geophys. Res. Atmos., 118, 85–100, https://doi.org/10.1029/2012JD018697.
Pangali Sharma, T. P., J. Zhang, N. R. Khanal, F. A. Prodhan, B. Paudel, L. Shi, and N. Nepal, 2020: Assimilation of Snowmelt Runoff Model (SRM) using satellite remote sensing data in Budhi Gandaki River basin, Nepal. Remote Sens., 12, 1951, https://doi.org/10.3390/rs12121951.
Panthi, J., and Coauthors, 2015: Spatial and temporal variability of rainfall in the Gandaki River basin of Nepal Himalaya. Climate, 3, 210–226, https://doi.org/10.3390/cli3010210.
Pokorny, S., T. A. Stadnyk, G. Ali, R. Lilhare, S. J. Déry, and K. Koenig, 2021: Cumulative effects of uncertainty on simulated streamflow in a hydrologic modeling environment. Elementa, 9, 431, https://doi.org/10.1525/elementa.431.
Rajagopal, S., and A. A. Harpold, 2016: Testing and improving temperature thresholds for snow and rain prediction in the western United States. J. Amer. Water Resour. Assoc., 52, 1142–1154, https://doi.org/10.1111/1752-1688.12443.
RGI Consortium, 2017: Randolph glacier inventory—A dataset of global glacier outlines, version 6. National Snow and Ice Data Center, accessed 11 June 2021, https://doi.org/10.7265/4M1F-GD79.
Schreiner-McGraw, A. P., and H. Ajami, 2020: Impact of uncertainty in precipitation forcing data sets on the hydrologic budget of an integrated hydrologic model in mountainous terrain. Water Resour. Res., 56, e2020WR027639, https://doi.org/10.1029/2020WR027639.
Schreiner-McGraw, A. P., and H. Ajami, 2022: Combined impacts of uncertainty in precipitation and air temperature on simulated mountain system recharge from an integrated hydrologic model. Hydrol. Earth Syst. Sci., 26, 1145–1164, https://doi.org/10.5194/hess-26-1145-2022.
Serpetzoglou, E., E. N. Anagnostou, A. Papadopoulos, E. I. Nikolopoulos, and V. Maggioni, 2010: Error propagation of remote sensing rainfall estimates in soil moisture prediction from a land surface model. J. Hydrometeor., 11, 705–720, https://doi.org/10.1175/2009JHM1166.1.
Sharma, S., N. Khadka, K. Hamal, D. Shrestha, R. Talchabhadel, and Y. Chen, 2020: How accurately can satellite products (TMPA and IMERG) detect precipitation patterns, extremities, and drought across the Nepalese Himalaya? Earth Space Sci., 7, e2020EA001315, https://doi.org/10.1029/2020EA001315.
Sicart, J. E., R. Hock, and D. Six, 2008: Glacier melt, air temperature, and energy balance in different climates: The Bolivian tropics, the French Alps, and northern Sweden. J. Geophys. Res., 113, D24113, https://doi.org/10.1029/2008JD010406.
Silvestro, F., S. Gabellani, R. Rudari, F. Delogu, P. Laiolo, and G. Boni, 2015: Uncertainty reduction and parameter estimation of a distributed hydrological model with ground and remote-sensing data. Hydrol. Earth Syst. Sci., 19, 1727–1751, https://doi.org/10.5194/hess-19-1727-2015.
Skinner, C. J., T. J. Bellerby, H. Greatrex, and D. I. F. Grimes, 2015: Hydrological modelling using ensemble satellite rainfall estimates in a sparsely gauged river basin: The need for whole-ensemble calibration. J. Hydrol., 522, 110–122, https://doi.org/10.1016/j.jhydrol.2014.12.052.
Sun, H., and Coauthors, 2021: General overestimation of ERA5 precipitation in flow simulations for High Mountain Asia basins. Environ. Res. Commun., 3, 121003, https://doi.org/10.1088/2515-7620/ac40f0.
Sunilkumar, K., A. Yatagai, and M. Masuda, 2019: Preliminary evaluation of GPM-IMERG rainfall estimates over three distinct climate zones with APHRODITE. Earth Space Sci., 6, 1321–1335, https://doi.org/10.1029/2018EA000503.
Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res., 106, 7183–7192, https://doi.org/10.1029/2000JD900719.
Thornton, J. M., N. Pepin, M. Shahgedanova, and C. Adler, 2022: Coverage of in situ climatological observations in the world’s mountains. Front. Climate, 4, 814181, https://doi.org/10.3389/fclim.2022.814181.
Upadhyay, S., P. Silwal, R. Prajapati, R. Talchabhadel, S. Shrestha, S. Duwal, and H. Lakhe, 2022: Evaluating magnitude agreement and occurrence consistency of CHIRPS product with ground-based observations over medium-sized river basins in Nepal. Hydrology, 9, 146, https://doi.org/10.3390/hydrology9080146.
Wang, Q. J., 1991: The genetic algorithm and its application to calibrating conceptual rainfall-runoff models. Water Resour. Res., 27, 2467–2471, https://doi.org/10.1029/91WR01305.
Ward, E., W. Buytaert, L. Peaver, and H. Wheater, 2011: Evaluation of precipitation products over complex mountainous terrain: A water resources perspective. Adv. Water Resour., 34, 1222–1231, https://doi.org/10.1016/j.advwatres.2011.05.007.
Wheler, B. A., A. H. MacDougall, G. E. Flowers, E. I. Petersen, P. H. Whitfield, and K. E. Kohfeld, 2014: Effects of temperature forcing provenance and extrapolation on the performance of an empirical glacier-melt model. Arct. Antarct. Alp. Res., 46, 379–393, https://doi.org/10.1657/1938-4246-46.2.379.
Wi, S., Y. C. E. Yang, S. Steinschneider, A. Khalil, and C. M. Brown, 2015: Calibration approaches for distributed hydrologic models in poorly gaged basins: Implication for streamflow projections under climate change. Hydrol. Earth Syst. Sci., 19, 857–876, https://doi.org/10.5194/hess-19-857-2015.
WMO, Ed., 1986: Intercomparison of Models of Snowmelt Runoff. WMO Series, Vol. 23, Secretariat of the World Meteorological Organization, 440 pp.
Wu, H., R. F. Adler, Y. Tian, G. Gu, and G. J. Huffman, 2017: Evaluation of quantitative precipitation estimations through hydrological modeling in IFloodS River basins. J. Hydrometeor., 18, 529–553, https://doi.org/10.1175/JHM-D-15-0149.1.
Yang, J., and Coauthors, 2013: The role of satellite remote sensing in climate change studies. Nat. Climate Change, 3, 875–883, https://doi.org/10.1038/nclimate1908.
Yang, M., Z. Li, M. N. Anjum, R. Kayastha, R. B. Kayastha, M. Rai, X. Zhang, and C. Xu, 2022: Projection of streamflow changes under CMIP6 scenarios in the Urumqi River head watershed, Tianshan Mountain, China. Front. Earth Sci., 10, 857854, https://doi.org/10.3389/feart.2022.857854.
Yatagai, A., K. Kamiguchi, O. Arakawa, A. Hamada, N. Yasutomi, and A. Kitoh, 2012: APHRODITE: Constructing a long-term daily gridded precipitation dataset for Asia based on a dense network of rain gauges. Bull. Amer. Meteor. Soc., 93, 1401–1415, https://doi.org/10.1175/BAMS-D-11-00122.1.
Zhang, J. L., Y. P. Li, G. H. Huang, C. X. Wang, and G. H. Cheng, 2016: Evaluation of uncertainties in input data and parameters of a hydrological model using a Bayesian framework: A case study of a snowmelt–precipitation-driven watershed. J. Hydrometeor., 17, 2333–2350, https://doi.org/10.1175/JHM-D-15-0236.1.
Zhang, Y., S. Liu, C. Xie, and Y. Ding, 2006: Application of a degree-day model for the determination of contributions to glacier meltwater and runoff near Keqicar Baqi glacier, southwestern Tien Shan. Ann. Glaciol., 43, 280–284, https://doi.org/10.3189/172756406781812320.
Zierl, B., and H. Bugmann, 2005: Global change impacts on hydrological processes in Alpine catchments. Water Resour. Res., 41, W02028, https://doi.org/10.1029/2004WR003447.
