(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Modeling snowmelt-driven streamflow dynamics in a Himalayan Basin under climate warming scenarios [1]

['Romana Jamshed', 'Department Of Environmental Sciences', 'Comsats University Islamabad', 'Cui', 'Abbottabad Campus', 'Abbottabad', 'Esha Naeem', 'Adnan Ahmad Tahir', 'Muhammad Irshad', 'Faizan-Ur-Rehman Qaisar']

Date: 2025-11 

The zone-wise mean monthly temperature variation is presented in Fig 6b . Mean (minimum-maximum) annual temperature for zone A, B, C and D were found to be ~ 17.5 °C (~6.6–27 °C), ~ 12.2 °C (~1.4–22 °C), ~ 5.3 °C (~ - 5.6–14.8 °C), and ~ - 1.96 °C (~ - 12.8–7.5 °C), respectively, over data period 2005–2015. Zone A (lowest altitude) is found to have the highest mean annual temperature while Zone D (highest altitude) has the lowest mean annual temperature. However, the variation in temperature during the year shows maximum monthly temperature in July and minimum monthly temperature in January for all the zones.

Fig 6a clearly depicts the snow accumulation phases, evident from October through December, and peaking in January and February. Conversely, the significant decline from March onwards indicates the melt season, accelerating into late spring and early summer (May-July). High flow values from June-September also indicate an overlapped monsoon season. This signifies the seasonal transition from snowmelt-dominated runoff to rainfall-driven or baseflow contributions [ 62 ]. Such distinct zone-wise stratification indicates the significant influence of elevation or climatic differences influencing snow accumulation and melting within the basin [ 53 ].

A clear seasonal snow cover pattern across all the zones is evident in Fig 6a with distinct variations in magnitude and duration among them. Zone D maintains the highest mean monthly snow cover (~97.5–100%) from November to May. Zone D is the highest elevation zone where the colder temperatures, governed by the environmental lapse rate, are responsible for the precipitation in the form of snow and its prolonged retention over months [ 2 , 61 ]. On the other hand, Zone A demonstrates minimal mean snow cover, with a peak slightly exceeding ~12% in January signifying its warmer, lower-elevation status. Zone B and Zone C demonstrate peak SCA in February, i.e., ∼68% and ∼100%, respectively. Mean (minimum-maximum) annual SCA for zone A, B, C and D were found to be ~ 5% (~1–12%), ~ 31% (~12–67%), ~ 65% (~29–100%), and ~88% (~57–100%), respectively, over data period 2005–2015.

In zone-wise application of SRM, daily zonal SCA and zonal mean temperature are used as basic input. The average of the basin’s daily total precipitation may be used in each zone if the zonal precipitation data is not available knowing that the SRM is less sensitive to precipitation input. Zonal SCA was estimated from improved MODIS data, while the zonal mean temperature at the hypsometric mean elevation of each zone was estimated by using the basin’s temperature lapse rate value for all four elevation zones (A, B, C, and D) of the Swat River Basin over the study period. The radar graph ( Fig 6 ) illustrates the monthly mean SCA and mean temperature variation for all elevation-based zones.

Temporal variations in basin-wide mean monthly SCA of Swat River Basin are shown in Fig 3 (mean monthly of entire data period) and Fig 4 (mean of each data year). Mountainous basins like Swat River Basin are characterized by gradual decline in spatial extent of seasonal snow cover during snowmelt season. Overall, it is found that in the SRB, snow accumulates from October to February, followed by melting from March to September, showing a declining trend. Analysis indicates a mean annual SCA of ~50%, peaking at ~72% in February and reaching a low of ~31% in September. February’s maximum SCA is due to westerly winds causing heavy snowfall ( Fig 3 ) at higher elevations and persistent cold that slows down the melting process. Conversely, summer’s high temperatures drive extensive melt, leading to September’s minimum SCA. Spatially, northern slopes in the basin have a consistent snow cover year-round ( Fig 5 ) due to higher elevation zones in the north as apparent from the DEM ( Fig 1 ) and thus lower mean temperatures, while southern slopes have lower elevation zones experiencing higher mean temperatures causing snow to melt faster hence lower mean annual snow coverage. Maximum mean annual SCA of ~54% was observed during 2005 whereas 2010 was the year with least mean annual SCA of ~48% ( Fig 4 ).

The mean annual discharge over the study period was estimated to be ~ 253 m 3 /s. It is found that the highest discharge occurs in the months from May-August with highest mean flow in the month of June (~474 m 3 /s). The rising temperatures in summers cause rapid melting of snow which ends up magnifying the discharge in the Swat River. Conversely, months with minimum flow, span from November to February with lowest mean discharge in January (~100 m 3 /s). Freezing minimum temperatures, excessive snowfall due to westerlies and reduced snowmelt decelerate the water flow from the mountains, thereby resulting in less discharge during winters. The river flow begins to increase as the temperature starts rising in late winter and peaking in summer (June-July), demonstrating the primary discharge of SRB driven by snowmelt from spring to early summer [ 60 ], and by sustaining the summer discharge due to remaining snow-glacier melt and monsoon precipitation. Such hydro-climatological profile of SRB is aligned with the defining characteristics of mountainous and snow dominated river basins.

The average of total annual precipitation was ~ 122 cm (1220 mm) as estimated from the data. The maximum of total precipitation occurs during the February‒April months (~49 cm) followed by June-September (~43 cm). Maximum value of monthly total precipitation of ~43 cm was recorded in July of 2010. This was the exceptional case and caused catastrophic floods in 2010 in the study area. Otherwise, the maximum monthly precipitation values were found in February 2011 (~30 cm) and March 2014 (~29 cm). For the total monthly precipitation ( Fig 3 ), it is observed that there are two precipitation seasons influencing the study area: the westerly weather pattern bringing precipitation from February to April and summer monsoon from June to September [ 3 , 26 ].

The lowest mean monthly temperature and ultimately the lowest discharge is found in January. This is because the winter discharge from November-January is generated through low melting induced by low temperatures with no significant contribution by the rainfall-runoff. On the other hand, the peak monthly mean temperature and discharge was found in June-July. The reason for this maximum discharge is higher temperature causing accelerated melting overlapped by the summer monsoon rainfall-runoff.

The overall basin-wide hydro-meteorological characteristics of SRB over 2005‒2015 are demonstrated in Fig 3 revealing the variations in monthly mean temperature, total precipitation, discharge and snow cover area. The mean annual temperature of the basin was estimated to be ~ 17 °C, with overall the highest mean monthly temperature observed in the month of July (~27 °C) and lowest mean monthly temperature in the month of January (~6.5 °C) over the data period of 2005‒2015 ( Fig 3 ). Over the course of the year, two distinct phases of change are observed in the mean monthly temperature. The first is a sharp rise beginning in February, with temperatures increasing at a rate of ~4.8 °C per month until May. After this, the rate of increase slows, and temperatures peak in July. From August to December, temperatures decline rapidly at a rate of ~5 °C per month, followed by a slower decrease that continues until the lowest temperatures are reached in January. This temperature variation plays a key role in the runoff generation from the snow and glacier melting in the study area. Lowest specific mean monthly temperature value was recorded as ~4.13 °C in January 2012, and the highest temperature as ~28.12 °C in July 2014.

Hydro-climatological data of the Swat River Basin is presented in Figs 3 - 5 , which shows mean monthly discharge, total monthly precipitation, mean monthly temperature and mean monthly snow cover area (SCA) over a period of 2005‒2015.

4.2. Basin-wide and zone-wise simulation of historical daily discharge (2005‒2015)

Daily river flow from Swat River Basin was simulated by applying basin-wide and zone-wise SRM over the data period of 2005‒2015 for the melt season (April to September). Both the basin-wide and zone-wise SRM approaches were carried out using the daily hydro-meteorological variables including the SCA. Several parameter values were adopted from previous studies [2,3,5,63] on mountainous basins of the UIB, while others were adjusted during calibration to reflect seasonal variations in temperature, precipitation, and snow conditions in specific months. These adjustments were constrained within logical ranges consistent with values reported in the literature for the hydrological response of mountain catchments. The best calibrated parameter values used for simulating river discharge from Swat River Basin are presented in Table 7.

A range of DDF (0.3‒0.65 cm.°C−1.d−1) was obtained for SRB due to certain factors like seasonal variation (more ablation during hotter months) and difference in characteristics of snow and ice (snow density, albedo effect, thermal properties). The lag time refers to the time taken between the occurrence of a rainfall or snowmelt event and the resulting runoff to reach the farthest point in the watershed. Therefore, time lag is usually small for a smaller basin and more for larger ones [53]. A range of time lag, (10‒15 hours) was taken for the Swat River Basin to ensure that the model accurately captures various factors influencing water movement in this mountainous basin. The runoff coefficients, C s and C r , were adjusted according to the months with a dominant pattern of contribution from the snowmelt or rainfall. Similarly, RCA (Rainfall Contributing Area) was taken 1 in the months when rainfall events are frequent in the basin, i.e., February-April and July-August while it was taken 0 in the months when snow is dominant. Lastly, X c and Y c recession coefficient values were used to adjust the time and magnitude of peaks for fine tuning of the simulated curve.

Table 8 presents SRM’s efficiency over calibration and validation periods using basin-wide and zone-wise application for the simulation of river discharge. This approach ensured that the SRM parameters efficiently captured inter-annual variability in runoff response. Overall, the SRM showed strong predictive performance, aligning with its recognized effectiveness in simulating the discharge for the snow and glacier dominated mountainous catchments [3,26]. For the basin-wide approach, the highest NSE of 0.95 was achieved particularly in 2005 and 2014 indicating significantly a high alignment between simulated and observed runoff during these years. While the lowest NSE (0.82) for the basin-wide approach occurred in 2006, still expressing a highly acceptable model performance [53]. In case of volume difference, the basin-wide approach showed its best performance in 2012 with D v = 0.03%, however the major deviation as an underestimation was indicated in 2011 with D v = ˗3.64%.

The calibration and validation phase demonstrated satisfactory outcomes as can be seen from the basin-wide and zone-wise values of D v and NSE in Table 8. Likewise, for the zone-wise approach, the highest NSE (0.93) was observed in 2005 (during calibration), with the lowest (0.80) was recorded in 2011 and 2013, i.e., during the validation. Volume differences for the zone-wise method were negligeable in 2007 (~0.03%), however the model showed large underestimation of -4.56% in 2013. The model efficiently simulated the year 2007 discharge with basin-wide (zone-wise) approach presenting a D v = 0.61% (0.03%) and NSE = 0.93 (0.91), hence it was chosen as base-year for simulation of future discharge under RCPs. The minor year-to-year fluctuations in model performance likely result from varying quality of input data or the occurrence of extreme weather events and shifts in how snowmelt contributed to runoff each year may not be fully captured by SRM.

Figs 7 and 8 illustrate the SRM’s model performance across the calibration (2005‒2009) and validation period (2011‒2015) by comparing the simulated and observed discharge for the basin-wide and zone-wise application, respectively. The model’s ability to capture the temporal dynamics of discharge is evident from the close agreement between the simulated (red line) and measured (blue line) hydrographs.

For the basin-wide application (Fig 7), the average value of NSE was 0.90 and 0.87 with volume differences (D v ) of 0.51% and ˗1.31% during calibration and validation, respectively. Similarly, for the zone-wise application (Fig 8), the average NSE was 0.91 and 0.84 with D v values of ˗0.83% and ˗2.27% during calibration and validation, respectively. Hence, for the basin-wide (zone-wise) application, the average NSE and D v over the entire data period (2005‒2015) were 0.88(0.87) and ˗0.40% (˗1.5%), respectively. These metrics indicate a reliable fit between observed and simulated discharge during the calibration for both applications and demonstrate that the model’s predictive competence remained substantial, though slightly reduced, when applied to data periods not used during its initial calibration. The D v values suggest a slight overestimation of total water volume during basin-wide calibration and a slight underestimation in both validation periods, with a more noticeable underestimation in the zone-wise validation. Overall, the SRM model demonstrates a convincing ability to simulate discharge across both basin-wide and zone-wise scales.

A correlation (Coefficient of Determination, R2) between the observed and simulated runoff of the Swat River Basin over the entire data period of 2005–2015 (excluding 2010) is shown in Fig 9. This correlation demonstrates the strong ability of the SRM to simulate discharge in the Swat Basin, at the basin-wide and zone-wise scale.

The simulated discharge shows a close alignment between R2 = 0.91 for basin-wide (Fig 9a) and R2 = 0.91 for zone-wise (Fig 9b). However, some scatters, particularly at higher discharge rates, suggest a degree of variability in the model’s accuracy at peak flows. Fig 9c demonstrates the performance of the SRM model by illustrating the relationship between basin-wide and zone-wise simulated Swat River discharges. The red line represents the best-fit linear regression between these two approaches depicting the strong positive correlation (R2 = 0.93). These results highlight the model’s robustness in accurately simulating hydrological responses, upholding its suitability for both long-term discharge analysis and understanding spatial variations in discharge within the high-altitude mountainous basin such as SRB at basin-wide and zone-wise scales. However, lack of gauged weather data in such areas has contributed to the limitations in precisely capturing all discharge fluctuations which eventually affected the runoff simulation results.

Throughout the study, the basin-wide approach consistently delivered slightly better NSE values and showed more stable runoff volume differences compared to the zone-wise method, particularly during the validation period. This indicates that basin-wide SRM approach proves to be remarkably consistent and reliable in offering deeper insights into specific local processes as compared to the approach of dividing the basin into zones [3] for simulating the total discharge of the Swat River Basin. The improved performance of the basin-wide simulation over the zone-wise approach in this study can be attributed to several key factors. Basin-wide modeling offers a more integrated representation of hydrological processes, reducing complexity by requiring fewer parameters (only for one zone) and avoiding inconsistencies that may arise from subdividing the basin. It ensures continuity in snowmelt and runoff processes, minimizes the influence of localized input data anomalies or scarcity, error accumulation, data quality [3,57] and avoids irregular zonal boundaries that can distort flow dynamics. As a result, it delivers more stable and reliable discharge predictions, particularly evident during the validation period, as reflected in the higher NSE values and consistent runoff volumes. Other reasons could be that the hydrological processes can vary significantly with scale, that may cause a model to perform well at basin level but compromise in capturing the distinctions of unique hydrological characteristics of zones (smaller sub basins) [53]. Such as, the factors like precipitation, snowmelt, and land cover can exhibit more variability at the zone level. The model might struggle to capture the complexities and accurately represent this erraticism with the same set of variables used for the entire basin.

[END]
---
[1] Url: https://journals.plos.org/climate/article?id=10.1371/journal.pclm.0000739

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/