Anticipating Future Hydrological Changes in the Northern River Basins of Pakistan: Insights from the Snowmelt Runoff Model and an Improved Snow Cover Data

Abstract

The water regime in Pakistan’s northern region has experienced significant changes regarding hydrological extremes like floods because of climate change. Coupling hydrological models with remote sensing data can be valuable for flow simulation in data-scarce regions. This study focused on simulating the snow- and glacier-melt runoff using the snowmelt runoff model (SRM) in the Gilgit and Kachura River Basins of the upper Indus basin (UIB). The SRM was applied by coupling it with in situ and improved cloud-free MODIS snow and glacier composite satellite data (MOYDGL06) to simulate the flow under current and future climate scenarios. The SRM showed significant results: the Nash–Sutcliffe coefficient (NSE) for the calibration and validation period was between 0.93 and 0.97, and the difference in volume (between the simulated and observed flow) was in the range of −1.5 to 2.8% for both catchments. The flow tends to increase by 0.3–10.8% for both regions (with a higher increase in Gilgit) under mid- and late-21st-century climate scenarios. The Gilgit Basin’s higher hydrological sensitivity to climate change, compared to the Kachura Basin, stems from its lower mean elevation, seasonal snow dominance, and greater temperature-induced melt exposure. This study concludes that the simple temperature-based models, such as the SRM, coupled with improved satellite snow cover data, are reliable in simulating the current and future flows from the data-scarce mountainous catchments of Pakistan. The outcomes are valuable and can be used to anticipate and lessen any threat of flooding to the local community and the environment under the changing climate. This study may support flood assessment and mapping models in future flood risk reduction plans.

1. Introduction

Earth’s climate system, a dynamic interplay of its atmospheric, cryosphere, terrestrial, and oceanic components, governs global weather patterns. Substantial shifts in this climate system pose a serious environmental challenge to each component. IPCC projections indicate a continuous global temperature rise during this century [1], which is anticipated to intensify the hydrological cycle and elevate flood vulnerability, especially for snow-fed river runoff. Over the past three decades, the Hindukush–Karakoram–Himalaya (HKH) has experienced a rise of 1.5 °C temperature, which is a substantial warming rate that could melt significant glacial and snow cover in the region [2,3,4,5,6,7]. The Indus River, originating from the Himalayas, is the main source of water for the majority of Pakistan’s downstream regions.

The Upper Indus Basin (UIB), located within the HKH region, collects meltwater from snow and glaciers and serves as the drainage basin for the Indus River above the Tarbela Dam (Figure 1). Due to its significant reliance on cryospheric inputs, the UIB’s water resources are especially vulnerable to the impacts of climate change [8]. According to Salerno et al. [9], a 1 °C rise in temperature will result in a 16–17% increase in seasonal runoff in the UIB. The scope, severity, and nature of the hydrological effects of climate change may vary dramatically from one location to another depending on the physical characteristics and geography of a region [10]. To assess how global warming and resulting climate change may affect water supplies in the snow- and glacier-fed watersheds, various methods could be employed. The use of hydraulic and hydrological models is an important task for the investigation of the hydrology of mountainous or data-scarce regions. Snow and glacier hydrological models are reported to assess the climatic impacts on the hydrology of different mountainous watersheds [11].

Hydrological models (HMs) are generally categorized into rainfall-runoff- or snowmelt-runoff-based applications depending on the hydrological attributes of the catchment, i.e., either it is rain fed or snow fed. However, in the case of high-elevation catchments, most rainfall-runoff models are inefficient for daily streamflow prediction due to high snowmelt contribution to waterbodies. Bookhagen and Burbank [12] reported that models relying solely on rainfall (e.g., TRMM 2B31) underpredict discharge by ~40%, especially during snowmelt seasons. Snowmelt represents a critical and seasonally dominant input in high-mountain environments and must be explicitly included for realistic discharge modeling. Snowmelt models can typically be divided into four types, depending on the various ablation algorithms used, such as statistical, conceptual, physical, and data-driven. Statistical snowmelt models anticipate runoff by establishing an association between snow hydrological property parameters, e.g., snow area, using statistical techniques. The empirical connection between snowmelt and temperature is demonstrated by the conceptual snowmelt model [13]. Physical snowmelt models, such as SNOWPACK [14] and SNTHERM [15], use the energy balance of the snow cover to determine snow melting. In regions where ground-based data is limited, coupling hydrological models with remote sensing data offers a practical solution for forecasting, managing water resources, and supporting informed decision-making, particularly for disaster mitigation [16]. Remote sensing technologies (satellites, sensors, etc.) help obtain variable data for hydrological models, such as precipitation, snow cover, soil moisture, vegetation indices, and surface temperature [17].

The snowmelt runoff model (SRM) is one of the most frequently used temperature-index models. The Ganges River Basin in the Himalayas is the biggest basin where the SRM has been applied, covering approximately 917,444 km² [18]. This model is developed to simulate and forecast the daily stream flow in the high-elevation basins, where the chief runoff factor is snowmelt [19]. This model has also been applied to evaluate the effect of changed climate on the seasonal snow cover and runoff in cryosphere catchments [20,21]. By analyzing the effects of climatic variation on snowmelt flow in the UIB, Tahir et al. [22] showed that the SRM can efficiently simulate snowmelt runoff in high-mountain locations. The World Meteorological Organization (WMO) also successfully evaluated the SRM for flow simulations in the basins of various sizes and altitudes [23,24].

Snowmelt runoff modeling plays a vital role in understanding and managing water resources in the data-scarce and climatically sensitive HKH region, where snowmelt and glacier melt contribute significantly to river flows. Numerous studies have demonstrated the effectiveness of the snowmelt runoff model (SRM) in simulating discharge in high-altitude basins by integrating remotely sensed snow cover data, particularly MODIS 8-daily composites, due to their wide coverage and temporal consistency [22,25,26,27,28]. In regions like the UIB, where ground-based precipitation measurements are sparse or limited to lower elevations, satellite-derived snow cover and interpolated precipitation datasets such as APHRODITE and TRMM have been instrumental in improving model accuracy [12,29]. Remotely sensed products offer critical insights into the spatial and temporal dynamics of snow accumulation and snowmelt, thus addressing key data limitations. The novelty of this study lies in the application of an enhanced cloud-free MODIS snow cover product (MOYDGL06), which provides improved snow detection under persistent cloud cover [30], a common challenge in the HKH, thereby increasing the reliability of snowmelt runoff simulations in complex mountainous terrain.

The field estimation of snow- and glacier-melt runoff and prediction of resultant high river discharges, especially in the summer, is a difficult task due to poor gauging in the glacier- and snow-dominated areas of the UIB. This study focuses on estimating the snowmelt runoff under historical and future climates using the SRM in snow- and a glacier-dominated area of the UIB using an improved snow product. The main objectives of the study were (1) to simulate snowmelt runoff from a snow- and glacier-dominated river catchment within the Upper Indus Basin (UIB) using the SRM and a novel improved snow product and (2) to predict future river flows from the study area under mid- and late-21st century RCP climate change scenarios. Two sub-catchments of the UIB were selected as the studied region, the Gilgit (situated in Hindukush–Karakoram) and the Kachura (situated in the Karakoram–western Himalaya) River Basins. These two basins were selected because of their significant contribution towards the mean annual UIB flow [31] but limited climatological data availability. Data scarcity was overcome by using a simpler temperature indexed model, the SRM, and integrating it with remotely sensed MODIS snow cover data, which is the dominant factor for flow from these basins.

2. Material and Methods

2.1. Study Areas (Gilgit and Kachura River Basins)

The Gilgit River and Kachura River Basins are the two significant snow- and glacier-dominated sub-basins of the UIB in the HKH region (Figure 1). A location map of the study areas within the UIB along with other sub-catchments of the UIB is shown in Figure 1. Key characteristics of both the river basins are presented in

Table 1. The key features and characteristics of the study areas.

Characteristics		Catchments
		Kachura	Gilgit
Flow gauge	Latitude (dd)	35.5 N	35.9 N
	Longitude (dd)	75.4 E	74.3 E
Mean elevation (m asl)		~4960	~4200
Total area (km2)		~150,360	~12,671
Glacier cover (km2)		12,630 (8.4%)	1178 (9.3%)
Mean annual snow cover (%)		41	62

. The catchment area and the mean elevation were estimated from the SRTM-DEM, glacier cover from the Randolph Glacier Inventory (RGI 6.0), and the snow cover from the improved MODIS snow product. Details of these datasets are presented in the next section.

Gilgit Basin covers an area of ~12,671 km² with a mean elevation of ~4200 m (Figure 2) and is situated between latitudes 35°50′ and 36°50′ N and longitudes 72°30′ and 74°10′ E by considering Gilgit river flow gauge as the outlet point. The catchment area over 5000 m altitude is mostly sheltered in everlasting snow and accommodates most of the glaciers. The glaciers and seasonal snowmelt are key factors in the basin’s river flows. On the other hand, UIB at the Kachura flow gauge (Kachura River Basin, afterward) is found to be around ~150,360 km² with a mean elevation of ~4960 m (Figure 2). The Kachura catchment lies between 31°10′ and 36°00′ N latitude and 75°26′ and 82°00′ E longitude, having diverse topography with glaciers, snow-capped summits, alpine meadows, and steep slopes, which influence the basin’s climate significantly. The climate of Kachura experiences a seasonal fluctuation; summers are comparatively short and pleasant, whereas winters are long and extremely cold. The region experiences the greatest amount of precipitation in the winter and spring seasons, mostly as snow.

The glacier cover percentage is not justified, which is less in Kachura as compared to Gilgit. One reason for this is that the Kachura Basin includes almost 30–40% of the area, which is sometimes called a closed basin [32], within the Shyok catchment. Therefore, practically, if we exclude this area, then the glacier cover percentage would be higher in the Kachura Basin. Also, most parts of the Kachura Basin fall into the cold desert and do not include glaciers. In terms of glacier surface area, Kachura has a higher area ~12,630 km² than the Gilgit Basin ~1178 km², in fact, ten times higher. The percentage of the mean annual snow cover is ~62% and ~41% in Gilgit and Kachura, respectively.

2.2. Dataset Sources and Treatment

In situ daily hydro-meteorological data and the satellite-based remote sensing data of the Digital Elevation Model (DEM) and daily snow cover were to be fed into the snowmelt runoff model (SRM). Daily climate data, which includes temperatures and precipitation from Gilgit and Skardu stations, was collected for the years 2003–2010 (with missing data for 2005) from the Pakistan Meteorological Department (PMD). The daily river flow data for the Gilgit and Kachura gauges were obtained by the Surface Water Hydrology Project (SWHP) of the Water and Power Development Authority (WAPDA), Pakistan for the same data period (2003–2010). This study relied on remotely sensed, satellite-based data for acquiring snow cover area and physical features of the study areas, to overcome the challenges in attaining the field data. For this purpose, the main data sources, including the Moderate Resolution Imaging Spectroradiometer (MODIS) for snow data [33] and the Digital Elevation Model (DEM) from the Shuttle Radar Topography Mission (SRTM) for topography [34,35] were used. These two remote sensing datasets were treated using ArcGIS 10 software developed by Environmental Systems Research Institute (ESRI), California, USA.

The accessed DEM data is SRTM 1 Arc-Second Global [34], which provides global coverage with void-filled data at a resolution of 1 arc-second (~30 m) and is openly accessible as a high-resolution global resource. The United States Geological Survey’s (USGS), data portal was accessed through http://earthexplorer.usgs.gov/ (accessed on 21 January 2023), Earth Explore website, to acquire SRTM-DEM data tiles. The topographical features of both study areas were analyzed using this DEM. The mean elevation of the Gilgit and Kachura river catchments was calculated using the hypsometry (area-elevation curve/hypsometric curve) calculated from this DEM, which was used as an input to the snowmelt model. Area elevation curves of both the Gilgit River Basin and the Kachura Basin are shown in Figure 2a,b, respectively.

Snow cover data was collected from improved cloud-free MODIS (MOYDGL06_2002_2018) product, which prevents uncertainty in the results because cloud distortion can occasionally prevent the sensor from recording an observation in simple MODIS data. Muhammad and Thapa [30] developed cloud-free MODIS data by using the MOD10A2.006 (Terra product), MYD10A2.006 (Aqua product), and MODIS 8-day Composite Collection 6 (C6) as input data. This data is mostly for high-mountain Asia (HMA) and includes the combined MODIS 8-day Terra and Aqua snow cover combination product coupled with Randolph Glacier Inventory (RGI 6.0) version 6 [36]. This data was downloaded from the PANGAEA (data Publisher for Earth & Environmental Science, Germany) web portal https://doi.pangaea.de/10.1594/PANGAEA.918198 (accessed on 12 November 2022) for the period 2003–2010. The spatial and temporal resolution of this data is 500 m and 8 days or 1 day, respectively. Although the spatial resolution is coarser for complex terrain, however, many researchers have validated MODIS data with higher-resolution sensors in mountainous regions [37,38,39,40] and found it a reliable choice to be used for hydrological simulations.

The improved cloud-free MODIS data developed by Muhammad and Thapa [30] used in this study was developed for the high-mountain Asia (HMA) region. This new data addressed biases in daily MODIS Terra and Aqua snow cover data by developing an improved, combined daily snow product. They corrected under- and overestimations in the daily data through cloud removal, gap filling, and snow extent adjustment. The corrected Terra and Aqua datasets were then merged and integrated with the Randolph Glacier Inventory (RGI 6.0) to produce a more accurate and consistent 500 m resolution daily snow product enhancing snow and glacier monitoring in the region. Accuracy of RGI is reported by Pfeffer et al. [41]. The user may estimate only snow cover, glacier cover, or combination of both components from these images. Therefore, keeping in view the reliability of this improved MODIS data, it was used as an input to SRM in this study.

2.3. Snowmelt Runoff Model (SRM)

The Windows-based snowmelt runoff model i.e., WinSRM version 1.12, was utilized for summer flow (April–October) simulations in the Gilgit and Kachura River Basins to simulate the river discharge. To run the SRM, essential basin characteristics were required, including the basin name, area, mean elevation, measurement units, reference elevation of the flow gauging station, and the geographic coordinates (latitude and longitude) of the associated meteorological station [42]. The basic variables and parameters used to run the SRM are presented in

Table 2. Parameters and variables used in the snowmelt runoff model (SRM).

Parameters	Variables
Snowmelt-Runoff coefficient (Cs)	Temperature (°C)
Rainfall-Runoff coefficient (Cr)	Precipitation (cm)
Degree-day factor (an) (cm °C−1d−1)	Snow cover (%)
Temperature lapse rate (ΔT) (°C/100 m)	Runoff/Flow
Critical temperature (T)
Total rainfall contribution area (A)
Recession coefficient (k)
Lag time (hour)

.

Three years (2003–2006, with missing data for 2005) were selected for model calibration, and the rest of the years (2007–2010) were chosen for validation of the model. The model was calibrated and validated by adjusting the parametric values every fifteen days or a month depending on the season.

In the model, snowmelt and precipitation have been added to the anticipated recession flow in the calculation, where Equation (1) is used to compute the daily discharge.

$${\mathrm{Q}}_{\mathrm{n}+1}=\left[{\mathrm{c}}_{\mathrm{S}\mathrm{n}}.{\mathrm{a}}_{\mathrm{n}}\left({\mathrm{T}}_{\mathrm{n}}+{\mathsf{\Delta }\mathrm{T}}_{\mathrm{n}}\right){\mathrm{S}}_{\mathrm{n}}+{\mathrm{c}}_{\mathrm{R}\mathrm{n}}{\mathrm{P}}_{\mathrm{n}}\right]{\displaystyle \frac{\mathrm{A}·\mathrm{10,000}}{\mathrm{86,400}}}\left(1-{\mathrm{k}}_{\mathrm{n}+1}\right)+{\mathrm{Q}}_{\mathrm{n}}{\mathrm{k}}_{\mathrm{n}+1}$$

(1)

where Q (m³s⁻¹) is the average daily runoff; c is the runoff coefficient, which represents the losses as a ratio; c_s refers to snowmelt and c_R to rain; a (cm °C⁻¹ d⁻¹) is the degree-day factor indicating the snowmelt depth resulting from 1 degree-day; T (°C d) is the number of degree-days; ∆T (°C d) is the adjustment by temperature lapse rate when interpolating the temperature from the station to the average hypsometric elevation of the basin or zone; S is the ratio of the snow-covered area to the total area; P (cm) is precipitation contributing to runoff; A (km²) is the area of the basin or zone; k is the recession coefficient indicating the decline of runoff in a period without snowmelt or precipitation; and n is the sequence of days during the runoff computation period [42].

The two crucial efficiency measures in hydrological modeling are the Nash–Sutcliffe coefficient (NSE) and the difference of volume (Dv, %), which have been used to assess the SRM’s effectiveness, as given in Equations (2) and (3), respectively.

$$\mathrm{N}\mathrm{S}\mathrm{E}=1-{\displaystyle \frac{{\sum }_{i=1}^n{(Qi-Q{i}^{\prime })}^2}{{\sum }_{i=1}^n{\left(Q{i}^{-}\overline{Q}{i}^{\prime }\right)}^2}}$$

(2)

where

Qi—is the measured daily discharge;

Qi′—is the computed daily discharge;

$\overline{Q}{i}^{\prime }$—is the average measured discharge from the past years for each day of the period;

n—is the number of days.

$$\mathrm{D}{\mathrm{v}}_{[\%]}={\displaystyle \frac{V_R-{V^{\prime }}_R}{V_R}}\times 100$$

(3)

V_R denotes the measured seasonal or annual runoff volume, and V′_R is the simulated annual or seasonal runoff volume.

To investigate how climate change may affect snowmelt runoff, scientists and researchers simulate climate change scenarios by estimating future greenhouse gas (GHG) concentrations and aerosol emissions to depict the most likely future weather conditions and their effects. The total amount of carbon dioxide (a greenhouse gas) released and its level of concentration in the atmosphere for the middle- and late-21st century is shown in Representative Concentration Pathways (RCPs) [43]. SRM can be used in combination with climate forecasts produced by regional and global climate models to investigate how climate change may affect snowmelt runoff. The effect of climate change on the mean summer flow (April–October) of the Gilgit and Kachura Rivers was evaluated by applying RCP scenarios to previously computed runoff in a calibrated SRM. For this investigation, the downscaled RCP scenarios for the mid- and late-21st century for the Indus River Basin developed by Su et al. [44] were employed. In their study, the simulation results of CMIP5 (Coupled Model Inter-comparison Project phase 5) multi-model ensemble in the Indus River Basin (IRB) were compared with the CRU (Climatic Research Unit) and APHRODITE (Asian Precipitation-Highly-Resolved Observational Data Integration Towards Evaluation) datasets. The systematic bias between simulations and observations was corrected by applying the equidistant Cumulative Distribution Functions matching method (EDCDFm), and high-resolution simulations were statistically downscaled. It effectively adapted global climate projections to the complex topography and seasonal variability of the Indus Basin, supporting region-specific climate impact assessments. Then, precipitation and temperature were projected for the IRB for the mid-21st century (2046–2065) and late-21st century (2081–2100). Under the RCP scenarios, a shift in average annual precipitation and temperature is displayed in

Table 3. RCP scenarios under mid- and late-21st century used for this study [44].

Time Period	RCP Scenarios	Mean Annual Temperature (°C)	Mean Annual Precipitation (%)
MID-21st century (2046–2065)	2.6	+1.21	+3.2
	4.5	+1.93	+0.1
	8.5	+2.71	+6.2
LATE-21st century (2081–2100)	2.6	+1.10	+3.2
	4.5	+2.49	+0.1
	8.5	+5.19	+6.2

.

3. Results and Discussion

3.1. Snow Cover Area Estimation Using Improved Cloud-Free MODIS Data

The daily snow-covered area of both Gilgit and Kachura catchments for the study period (2003–2010) was estimated from the improved cloud-free MODIS data. The scatter plots shown in Figure 3a,b represent the snow cover area for each 8-day for Gilgit and Kachura Basins.

Data points from eight individual years are illustrated using distinct colored dots, while the thick red line indicates the mean monthly SCA across all these years, presenting the general trend despite yearly fluctuations. Figure 3a,b show a clear seasonal trend by indicating the typically high SCA during the winter and spring months, peaking between February and April. A steady decline begins in late spring, with the lowest snow cover observed in August–September. From September onward, snow cover gradually increases again with the onset of winter. The maximum SCA was recorded in the month of March with ~90% in the Gilgit River Basin and ~65% in the Kachura Basin. The mean SCA of the Gilgit Basin was ~62%, which was far greater than the Kachura Basin, i.e., ~41%, because the Gilgit Basin is snow-dominated and experiences high snow cover annually. The seasonal pattern of SCA observed in the study area highlights the combined effects of temperature and precipitation on snow accumulation and melt processes, with the spread of data points across years (Figure 3) indicating notable inter-annual variability.

3.2. River Flow Simulation by SRM in Gilgit and Kachura River Basin over Historical Data

The basin-wide SRM approach was carried out using the values of a few parameters from different studies [22,45] conducted for mountainous basins and adjustment of some other parameters during the calibration stage. The best values of parameters found for the simulation of river discharge from Gilgit and Kachura River Basins are presented in

Table 4. Parametric values for the basin-wide simulation of summer river discharge (April–October) from the Gilgit and Kachura River Basins using the SRM for the 2003–2010 study period.

Parameters	Gilgit	Kachura
Lapse Rate (°C/100 m)	0.650	0.650
Tcrit (°C)	0	0
AN or an [degree-day factor (cm/°C/d)]	0.15–0.35	0.10–0.35
Lag Time (hr)	16	16
Cs	0.01–0.40	0.005–0.29
Cr	0.01–0.40	0.005–0.29
RCA	1	1
Xc	1.06	1.06
Yc	0.02	0.02

.

A global temperature lapse rate value of 0.65 °C/100 m was used for this basin-wide approach of flow simulation from the study area using the SRM. The use of multiple stations’ data is preferred to estimate the lapse rate value during the zone-wise application of the SRM. One of the most important parameters in the SRM is the degree-day factor (DDF) for snow. Initially, the value of 0.30 cm/°C/d for the DDF was incorporated and then adjusted during the calibration phase of the model.

Table 5. The efficiency of the basin-wide SRM application to simulate the daily summer flow of the Gilgit River and Kachura River over the period 2003–2010.

		Model Efficiency
Basin	Mean Summer Flow (Apr–Oct)	Volume Difference, Dv (%)	Nash–Sutcliffe Coefficient (NSE)	Pearson Correlation Coefficient
Gilgit	Calibration
	2003	−2.01	0.97	0.98
	2004	−0.5	0.98	0.98
	2006	−4.9	0.97	0.98
	Validation
	2007	−1.0	0.94	0.97
	2008	−1.9	0.96	0.98
	2009	−2.9	0.97	0.97
	2010	−2.2	0.94	0.97
Kachura	Calibration
	2003	1.2	0.96	0.98
	2004	−4.7	0.90	0.95
	2006	−5.0	0.94	0.97
	Validation
	2007	−2.1	0.90	0.95
	2008	−1.9	0.96	0.98
	2009	−9.8	0.93	0.98
	2010	−1.3	0.96	0.98

and Figure 4 show the efficiency of the basin-wide SRM application to simulate the daily summer (April–October) flow of the Gilgit River and Kachura River over the study period during calibration and validation steps. In the case of the Gilgit River Basin, it is found that using the parametric values (Table 4) during the calibration step, the SRM showed the best simulation for the year 2004 with NS coefficient 0.98 and −0.5% difference in volume (Dv). The average NS coefficient values for calibration and validation periods were 0.97 and 0.95, with Dv (%) −2.47 and −1.5, respectively. In addition, the correlation between observed and simulated discharge was also presented by Pearson correlation, which was found to be 0.98 for the calibration period and 0.97 for the validation period.

Similarly, the SRM performed well for the Kachura Basin, with an average NS value of 0.93 for both the calibration and validation period (Table 5 and Figure 4). In addition, the model showed the best simulation for the year 2003 with an NS coefficient of 0.96 and a −1.2% volume difference. This demonstrates the model’s robustness to reproduce the observed discharge patterns. Likewise, the Pearson correlation coefficient, i.e., ≥0.95 across all periods (Table 5), signifies a strong linear relationship between simulated and observed discharges.

Moreover, for both the basins, some minor underestimation in the volume difference (Dv), particularly within ±5% for most years, is observed, indicating acceptable water volume representation. Overall, the consistently high values for NS, Dv, and Pearson coefficient affirm the model’s reliability in discharge simulation across two important basins of the UIB over the specific study period.

Figure 5 presents a comparative analysis of simulated versus measured daily discharge values for the Gilgit and Kachura Rivers during both the calibration and validation periods. The strength of the relationship is indicated by the regression lines and corresponding R² values, which reflect how closely the simulated values match the observed streamflow.

During the calibration period for the Gilgit River (Figure 5a), the model demonstrates good performance, as indicated by an R² value of 0.97. The model has simulated low and high flows with good efficiency. This suggests that the model is well-tuned to the basin’s hydrological characteristics. In the validation period (Figure 5b), the model maintains high accuracy, with R² = 0.959. Similar results were achieved for the Kachura River during calibration (Figure 5c) and validation (Figure 5d), where R² of 0.95 was achieved. Overall, the high R² values, ranging from 0.95 to 0.97, for both river basins and time periods indicate that the SRM is capable of being a useful model for snowmelt runoff simulation in the poorly gauged Hindukush–Himalaya region [45]. The slight deviations observed during validation are expected in hydrological modeling and highlight areas for further improvements, such as the improved spatial resolution of input data [22,42].

3.3. Impact of Climate Change Scenarios on the Mean Summer Flows of Gilgit and Kachura

Figure 6 illustrates the mean monthly discharge for the summer season (April–October) from the Gilgit and Kachura Basins, simulated under different Representative Concentration Pathway (RCP) climate scenarios for two future periods: mid-21st century (2046–2065) and late-21st century (2081–2100) by using the year 2009 and 2008 as reference (base) years for the Gilgit and Kachura Basins, respectively.

Generally, both mid-century (Figure 6a) and late-century (Figure 6b) projections for the Gilgit Basin show a steadily increasing pattern, with discharge generally increasing during the entire summer season. This pattern likely reflects the influence of increased snowmelt and monsoon precipitation. For both periods, RCP 4.5 and RCP 8.5 consistently project a greater rise in mean monthly discharge compared to RCP 2.6 and the base-year flow. In the mid-21st century, the increase in flow ranges from +0.30% (RCP 2.6) to +9.4% (RCP 8.5), while by the late-21st century, these increases in flow become more obvious, ranging from +3.4% (RCP 2.6) to +10.8% (RCP 8.5). This indicates a stronger hydrological response to projected increasing temperature due to higher greenhouse gas concentrations [46]. The seasonality of the flow is projected to be altered, such as the flow shift, 15 days earlier in the months of May–September. Interestingly, the general timing of peak flows remains parallel; however, the magnitude of these peaks is projected to increase, especially under RCP 8.5, indicating potentially higher flood risks or altered water availability during the peak discharge season in the Gilgit Basin.

For the Kachura Basin, a seasonal discharge pattern like the Gilgit Basin, with an increased summer flow, is projected (Figure 6c,d). The higher RCP scenarios (4.5 and 8.5) in Kachura also lead to increased mean monthly discharge, but the percentage increases in flow are generally lower than those observed for the Gilgit Basin. It is found that in the mid-21st century, the rise in flow ranges from +0.6% (RCP 2.6) to +2.0% (RCP 8.5), while by the late-21st century, the increases are slightly higher, ranging from +1.2% (RCP 2.6) to +3.3% (RCP 8.5). The peaks in the Kachura River Basin also show a mounting trend under higher RCPs; however, the relative increase is less compared to Gilgit. Changes in river flows (%) for both study areas under mid- and late-21st-century RCPS climate scenarios are also shown in

Table 6. Change in mean summer river flows (%) for both study areas under RCP mid- and late-21st century climate scenarios.

Period	RCP Scenarios	Change in River Flow (%)
		Gilgit River	Kachura River
MID-21st century (2046–2065)	2.6	+0.3	+0.6
	4.5	+3.6	+0.9
	8.5	+9.4	+2.0
LATE-21st century (2081–2100)	2.6	+3.4	+1.2
	4.5	+9.3	+1.9
	8.5	+10.8	+3.5

.

The overall findings suggest a general trend of rising mean monthly discharge in both the Gilgit and Kachura Basins under future climate scenarios due to rising temperatures and precipitation and the resulting excessive melting of snow and glaciers. The Gilgit Basin appears to be more sensitive to these changes, showing larger percentage increases in flow compared to the Kachura Basin. The sharper increase in climate change-induced runoff sensitivity in the Gilgit Basin (+10.8%) compared to the Kachura Basin (+3.3%) under RCP 8.5 (and under all scenarios generally) can be attributed to several key factors. First, Gilgit’s lower mean elevation (~4200 m) places a larger portion of its snowpack near the melting threshold, making it more responsive to rising temperatures, while Kachura’s higher elevation (~4960 m) keeps much of its snow and glacier mass above the 0 °C isotherm, limiting melt under the same warming scenario. Second, Gilgit’s elevation range likely spans more climate-sensitive zones, while Kachura contains a larger proportion of high-altitude accumulation zones (Figure 1 and Figure 2), which remain insulated from rapid changes. Third, Gilgit exhibits a higher mean snow cover area (~62%) and greater seasonal variability (~25–90%), indicating a dominant and dynamic seasonal snowpack that rapidly responds to temperature fluctuations, whereas Kachura’s lower mean (~41%) and more stable SCA (~20–65%) suggests a more persistent snow and glacier cover with slower response (Figure 3). Fourth, Gilgit likely contains more transient seasonal snow, which melts quickly under warming, and less glacier cover, while Kachura’s cryosphere is more glacier dominated and debris covered [47,48,49], reducing immediate melt acceleration. Lastly, warming causes the earlier snowmelt phases in Gilgit, shifting and concentrating runoff to earlier seasons, as is evident from its mean SCA variation in Figure 3, thereby increasing total discharge. Together, these factors explain Gilgit’s significantly higher hydrological sensitivity to climate change.

If these scenarios come true, the glaciers may retreat significantly after generating high flows for a specific period. However, the SRM does not provide quantification of the glacier cover under future climate projections. Some other modeling techniques may be incorporated in future studies to project the future status of glacier cover in the region. These projected changes have significant implications for regional water resource management, potentially affecting agricultural practices, hydropower generation, and the frequency and intensity of flood hazards in the downstream areas. The shift towards higher discharge during the summer season needs adaptive strategies for water infrastructure and disaster preparedness in these regions [22,45].

A limitation of the SRM is the lack of incorporation of glacier-melt runoff. Glacier melt is influenced by multiple physical processes, including ice thickness, surface albedo, and debris/sediment cover. However, the use of a simplified degree-day method with constant DDFs is a well-established and widely applied approach in data-scarce high-mountain regions such as the UIB, especially where physically based energy balance modeling is infeasible due to the lack of detailed input data (e.g., glacier albedo, ice thickness, or debris distribution). In this study, we aimed to assess snowmelt runoff using a simple temperature indexed model suitable for large-scale and remotely located basins. To reduce uncertainty typically associated with DDF-based models, we incorporated cloud-free MODIS MOYDGL06 snow cover data, which improved the accuracy of snow extent and depletion patterns over time. This enhancement helps constrain the timing and magnitude of snowmelt more realistically, thereby mitigating the potential overestimation of runoff under warming scenarios. We acknowledge that a spatially and temporally varying DDF or an energy balance model could provide a more physically detailed representation of glacier-melt processes. However, this was beyond the scope of our current study, which focuses on integrating improved remotely sensed snow cover data into a parsimonious modeling framework suitable for long-term hydrological simulation under climate change.

4. Conclusions

This study utilized the SRM coupled with an improved cloud-free snow and glacier composite product to simulate snowmelt runoff from both a snow-dominated (Gilgit Basin) and a glacier-dominated (Kachura Basin) river catchment within the Upper Indus Basin (UIB). The research aimed to predict future flow increases by evaluating how rising temperatures and precipitation, as projected under 21st-century RCP climate scenarios, influence snow and glacier behavior.

This study concludes that the SRM is one of the most effective and successful models for the simulation of daily discharge from the Gilgit and Kachura River Basins, which are snow- and glacier-fed mountainous basins. Improved remote sensing snow and glacier data products seemed to enhance the SRM’s performance compared to previous studies that used traditional MODIS snow products. The highly sensitive parameters observed in the SRM were degree-day factor (cm/°C/d), rainfall-runoff coefficient (Cr), and snow-runoff coefficient (Cs) based on adjustments completed during the calibration and validation periods. The projected increased snow/ice melt variations will impact downstream hydropower, crop production, and the mountain ecosystem. Both regions anticipate increased mean summer flows under future RCP scenarios, with the Gilgit Basin (snow-dominated) showing a greater increase than the highly glaciated Kachura catchment. The comparatively higher flow from the Gilgit Basin is likely because it is snow-dominated, making its runoff more immediately responsive to temperature increases than the more heavily glaciated Kachura Basin. Gilgit Basin’s lower elevation, more variable and seasonally dominant snow cover, and greater exposure to temperature-induced melt compared to the Kachura Basin, which is colder, higher, and likely glacier-buffered, are the dominant factors for its varying response to climate change. The probability of flooding could grow due to the rate at which snow and ice are melting because of rising temperatures in the region.

Assessing future flood risks can be effectively achieved through the robust combination of snowmelt runoff modeling and climate change projections. However, a significant improvement is required in the uniform integration of remote sensing data with hydrological models, especially for areas where ground-based data is sparse. Given the vulnerability of flood-prone areas in northern Pakistan, there is an urgent need to design a comprehensive flood mitigation strategy by enhancing the flow prediction capabilities and establishing effective management practices. Emphasizing non-structural approaches, such as raising awareness about the floods among the local community and ensuring the disaster preparedness before its occurrence, is also a significant part of flood mitigation.