Glacier Recession and Climate Change in Chitral, Eastern Hindu Kush Mountains of Pakistan, Between 1992 and 2022

Abstract

Mountain regions are particularly sensitive and vulnerable to the impacts of climate change. Over the past three decades, mountain temperatures have risen significantly faster than those in lowland areas. The Hindu Kush–Karakoram–Himalaya region, often referred to as the “water tower of Asia”, is the largest freshwater source outside the polar regions. However, it is currently undergoing cryospheric degradation as a result of climatic change. In this study, the Normalized Difference Glacier Index (NDGI) was calculated using Landsat and Sentinel satellite images. The results revealed that glaciers in Chitral, located in the Eastern Hindu Kush Mountains of Pakistan, lost 816 km² (31%) of their total area between 1992 and 2022. On average, 27 km² of glacier area was lost annually, with recession accelerating between 1997 and 2002 and again after 2007. Satellite analyses also indicated a significant increase in both maximum (+7.3 °C) and minimum (+3.6 °C) land surface temperatures between 1992 and 2022. Climate data analyses using the Mann–Kendall test, Theil–Sen Slope method, and the Autoregressive Integrated Moving Average (ARIMA) model showed a clear increase in air temperatures from 1967 to 2022, particularly during the summer months (June, July, and August). This warming trend is expected to continue until at least 2042. Over the same period, annual precipitation decreased, primarily due to reduced snowfall in winter. However, rainfall may have slightly increased during the summer months, further accelerating glacial melting. Additionally, the snowmelt season began consistently earlier. While initial glacier melting may temporarily boost water resources, it also poses risks to communities and economies, particularly through more frequent and larger floods. Over time, the remaining smaller glaciers will contribute only a fraction of the former runoff, leading to potential water stress. As such, monitoring glaciers, climate change, and runoff patterns is critical for sustainable water management and strengthening resilience in the region.

1. Introduction

Mountain glaciers provide essential freshwater for drinking, agriculture, and hydropower generation [1,2,3,4]. They also play a crucial role in regulating regional temperatures and climate while supporting diverse ecosystems vital to environmental health [5,6]. However, mountain regions, particularly their cryosphere, are highly sensitive and vulnerable to the impacts of climate [7,8]. Over the past three decades, mountain temperatures have increased significantly compared to those in lowland areas [9], a phenomenon known as “elevation-dependent warming” (EDW) [10,11]. Globally, the frequency of cold and warm days and nights has shifted in recent decades. Cold days have decreased by one day, and cold nights by 0.5 nights per decade, while warm days have increased by 1.2 days, and warm nights by 1.7 nights per decade [8]. As a result, glaciers are retreating at an accelerated rate in many mountainous regions [12], serving as a clear indicator of climate change [13]. The ablation and recession of mountain glaciers are largely driven by changes in temperature and precipitation [13], where even minor fluctuations can lead to unpredictable consequences. The observed rapid recession of glaciers in many regions has already caused significant impacts, including increased livelihood insecurity, hydrological crises, and more frequent and intense natural disasters such as glacial lake outburst floods (GLOFs), flash floods, soil erosion, and landslides [7].

The Hindu Kush–Karakoram–Himalaya (HKH) region is the largest freshwater resource outside the polar regions, often referred to as the “water tower” of Asia [14]. Most of South Asia’s rivers originate in this region, supporting the livelihoods of over two billion people [15]. However, the HKH is warming at twice the global average, posing a significant threat to the livelihoods of approximately 240 million people within the mountains and 1.65 billion people in the adjacent lowlands [16]. Over the last three decades, the HKH has experienced a sharp temperature increase of 1.5 °C, double the increase observed in other parts of Pakistan [17]. Projections indicate that the trend of elevation-dependent warming (EDW) will continue and intensify, with the region expected to exceed the global average temperature rise by at least 0.3 °C [18]. In the 1980s, the glaciers of the HKH began to show a marked increase in melting and recession, a trend that continues today [3,18,19], although the response of glaciers varies depending on local terrain and climate. Within the HKH, many of the Karakoram glaciers are actually advancing, a phenomenon known as the “Karakoram anomaly” [20], which has been supported by numerous subsequent studies [21,22,23,24,25,26].

The rapid depletion of glaciers across the HKH is a growing concern [27,28,29,30], threatening downstream communities. The consequences are already evident, with devastating natural disasters such as avalanches, landslides, and glacial lake outburst floods (GLOFs) becoming more frequent. GLOFs in particular are one of the most dangerous hazards in the HKH, with a high death toll [31]. Additionally, higher temperatures, accelerated glacier melting, shifting seasons, and erratic rainfall are altering the region’s hydrology, further impacting agriculture, food security, and livelihoods [32]. Given these challenges, understanding natural changes and their impacts in mountain environments is critical for managing and mitigating natural hazards and risks. However, there is a dearth of scientific studies on the impacts of climate change on glacier recession in the Eastern Hindu Kush. This study aims to examine climate change and glacier response in Chitral, located in the Eastern Hindu Kush of Pakistan, over the past thirty years.

2. Study Area

Chitral (35–37° N, 71–74° E) is the northernmost valley of the Khyber Pakhtunkhwa province in the northwestern part of Pakistan (Figure 1). It is referred to as “the heart of the Hindu Kush” and “the land of fairies and flowers” [33]. The Hindu Kush Range has several peaks over 7000 m, with Tirich Mir (7708 m asl.) being the highest. It is also the tallest peak in the world outside the Himalayas and Karakoram. Chitral has over 500 glaciers, of which Chiantar (36 km), Tirich Mir (32 km), Booni, and Raman are the most famous ones. The Raman Glacier is situated in the Laspur Valley at an elevation of 5500 m asl. Chitral is administratively divided into Lower Chitral and Upper Chitral districts, covering an area of 14,850 km². In 2023, the valley was home to almost 516,000 people; its population density is thirty-five persons per square kilometer [34].

The climate of Chitral is classified as subtropical continental highland, characterized by long, cold winters and short, warm summers. The semi-arid mountain climate is marked by significant seasonal extremes in both temperature and precipitation [35]. The region has a mean annual temperature of 16 °C and an average annual precipitation of 460 mm, which is unevenly distributed throughout the year [27]. The highest precipitation occurs in the form of snow, driven by western disturbances during the winter months from December to April, accounting for 75% of the total annual precipitation. May sees the highest relative humidity, at 60%, while July experiences the lowest, at 42%. March typically has the highest number of rainy days (12.7), whereas July receives the least rainfall, with only 9.5 mm [36]. Chitral enjoys 3,281 h of sunshine annually, with an average monthly sunshine duration of 273 h [36]. June has the longest daily sunshine, averaging 11.6 h, for a total of 360 h of sunlight in the month. In contrast, January has the shortest daily sunshine, with an average of 7.2 h, totaling 217 h of sunlight for the month. As a result, Chitral’s glaciers are exposed to the maximum solar radiation during the summer months, particularly in June and July.

3. Methods

3.1. Glacier Mapping

To analyze glacial area changes in Chitral, we used satellite imagery and DEM data from 1992 to 2022 (

Table 1. Satellite data utilized in this study.

Acquisition Date	Dataset	Sensor	ID		Spatial Resolution	Cloud Cover
			Row	Path
31 August 1992	LT05	TM	150, 151	034, 035, 036	30 m	<10%
16 July 1997	LT05	TM	150, 151	034, 035, 036	30 m	<10%
30 August 2002	LT05	TM	150, 151	034, 035, 036	30 m	<10%
17 July 2007	LE05	TM	150, 151	034, 035, 036	30 m	<10%
8 July 2012	LE05	TM	150, 151	034, 035, 036	30 m	<10%
15 August 2017	LC08	OLI	150, 151	034, 035, 036	30 m	<10%
23 September 2022	LC08	OLI	150, 151	034, 035, 036	30 m	<10%
8 July 2022	Sentinel-2A	MSI	T42SXE	A035981	10 m	<10%

). The satellite images were downloaded from the U.S. Geological Survey (https://earthexplorer.usgs.gov, accessed on 10 April 2024). Only images with less than 10% cloud cover and minimal snow cover from August and September were selected to optimize glacier delineation. Additionally, glacier polygons from the Global Land Ice Measurements from Space (GLIMS) database were incorporated. The Shuttle Radar Topography Mission (SRTM) DEM of 30 m resolution was used to derive watershed boundaries and extract glacier attribute data. For a more detailed study of glacier dynamics and retreat, we focused on Raman Glacier as a case example. Observations on snow and ice harvesting were made during field surveys from 2020 to 2023.

The Normalized Difference Glacier Index (NDGI) was used for the spatiotemporal analysis of clean-ice glacier areas [37]. NDGI is a spectral index specifically designed to identify and delineate glacier areas based on their unique spectral properties. It is calculated as: NDGI = (Near Infrared − Green)/(Near Infrared + Green). The green band highlights glacier ice due to its high reflectance, while the shortwave infrared band helps distinguish ice from the surrounding terrain. Pixels with an NDGI value greater than the threshold of 0.4 were classified as glacier ice. Since only clean ice was mapped, the results present the minimum total glacier area. However, debris-covered glaciers are relatively rare in Chitral.

Accuracy assessments for the NDGI-based image classification of clean glacier and non-glacier surface areas were carried out using Google Earth. A total of 300 reference points distributed across all the classified images were verified using Google Earth. User and producer accuracy, commission and omission errors, and the Kappa coefficient were calculated. The key accuracy metrics showed values greater than 75% for all images (

Table 2. Accuracy assessment of the NDGI analysis results.

Year	Overall Accuracy (%)	User’s Accuracy (%)	Producer’s Accuracy (%)	Commission Error (%)	Omission Error (%)	Kappa Coefficient (%)
1992	86.7	83.9	79.4	22.4	12.5	0.84
1997	89.5	86.1	82.1	16.7	10.4	0.87
2002	91.7	89.8	80.6	18.5	13.7	0.89
2007	93.5	91.4	86.2	24.6	18.2	0.91
2012	92.8	80.4	86.7	14.9	10.6	0.88
2017	87.5	91.1	89.1	20.3	15.7	0.92
2022	94.7	89.7	87.7	17.2	9.8	0.95

). The results were also validated using NDSI classification and cross-checked with GLIMS data (https://www.glims.org/maps/glims, accessed on 10 April 2024). However, after inspection, several mapping errors were found in the GLIMS data.

3.2. Climate Data Analysis

Time-series data on temperature and precipitation from 1967 to 2022, collected at meteorological stations in Chitral and Drosh, were analyzed to assess the impact of climatic change on glaciers. The data are archived at the Regional Meteorological Department Peshawar, Pakistan.

3.2.1. Long-Term Temperature and Precipitation Trends

The non-parametric Mann–Kendall test (MKT) was used to assess long-term trends in temperature and precipitation [38,39,40]. The MKT is particularly suitable for climate-related studies because it does not require the data to follow a specific distribution. The test evaluates the presence of a trend in time series data by analyzing the rank correlation between data points. Its robustness and versatility make it a widely used method for detecting monotonic trends in environmental data such as temperature, precipitation, and streamflow. For an observed time series x = x₁, x₂, …, xₙ, the test statistic S is expressed as (Equation (1)) [41]:

$$S=\sum _{i=1}^{n-1}\sum _{j=i+1}^{n-1}Sign(x_j-x_i)$$

(1)

where the value of Sign ($x_j-x_i$) can be computed as (Equation (2)) [41,42]:

$$Sign(x_j-x_i)=\left\{\begin{array}{c}\begin{array}{c} 1\\ 0\\ -1\end{array}\begin{array}{c}{ x}_j-x_i>0\\ { x}_j-x_i=0\\ { x}_j-x_i<0\end{array}\\ \end{array}\right\}$$

(2)

where i and j are the ranks of observation of the x_i, x_j of time series. The variance of S can be expressed as (Equation (3)) [41]:

$$Var \left(S\right)={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)-\sum _{i=1}^n{t}_i\left(t_i-1\right)\left(2t_i+5\right)}{18}}$$

(3)

3.2.2. Magnitude and Direction of Temperature and Precipitation Trends

The Theil–Sen Slope (TSS) method, which is part of the Mann–Kendall test, quantifies the rate of change in temperature and precipitation trends [43,44,45]. Known for its robustness, the TSS calculates the median slope between all pairs of data points, providing a reliable estimate of the trend’s magnitude and direction. Unlike linear regression, TSS is resistant to outliers and effectively handles non-normally distributed data, making it highly suitable for climate dataset analysis. By calculating the slope of the time series, the method reveals how rapidly temperatures or precipitation are increasing or decreasing. The median slope is particularly useful for understanding long-term trends in complex environmental data. The formula for the slope is (Equation (4)) [46]:

F(t) = Qt + B

(4)

where Q represents the slope’s magnitude and B denotes a constant. However, the magnitude of the slope is calculated to assess the slope for the temporal temperature and precipitation data as (Equation (5)) [46]:

$$Q_i={\displaystyle \frac{X_{i }-X_j}{j-k}}$$

(5)

where X_i and X_j represent pairs of temperature data with i = 1, 2, 3, 4, 5, 6, 7, …, N, and among time, denoted by j and k (j > k). The median of the N values of Q are calculated as (Equation (6)) [46]:

$$Q_{med}=\left\{\begin{array}{c}{Q }_{{\displaystyle \frac{[N+1]}{2}}}, \left(If N is odd\right)\\ {\displaystyle \frac{ Q\left[\frac{N}{2}\right]+Q[N+\frac{2}{2}]}{2}}, (If N is even)\end{array}\right.$$

(6)

where Q_med represents the median of the calculated slopes (Q_i). It assists in computing the central tendency for a set of slopes derived from the pairwise differences in the temperature or precipitation data over time. Q_med is then used to characterize the overall trend or direction of change in the temperature or precipitation data: a positive Q_i indicates an increase in temperature or precipitation, whereas a negative Q_i indicates a decrease.

3.2.3. Monthly Temperature and Precipitation Trends

The Autoregressive Integrated Moving Average (ARIMA) model [47] is widely used for the analysis and forecasting of time series climate data, including monthly temperature and precipitation [48,49,50,51,52]. The magnitude of temporal correlation in the time series determines the “autoregressive” (AR, p) and “moving average” (MA, q) terms, while the “integrated” (I, d) term helps transform a non-stationary time series into a stationary one, as shown in (Equation (7)) [47]:

$${ý}_t=c+{\varphi }_1{ý}_{t-1}+\dots {\varphi }_p{ý}_{t-p}+{\theta }_1 {\epsilon }_{t-1}+\dots {\theta }_q{\epsilon }_{t-q}+{\epsilon }_t$$

(7)

where $c$ denotes a constant term in a time series, also known as the drift term, when (d = 1), ${\varphi }_1{ý}_{t-1}+\dots {\varphi }_p{ý}_{t-p}$ represents the AR term with ${\varphi }_1$ to ${\varphi }_{p }$ as a coefficient of p order; ${\theta }_1 {\epsilon }_{t-1}+\dots {\theta }_q{\epsilon }_{t-q}$ represents the MA term with ${\theta }_1$ to ${\theta }_q$ as a coefficient of q order; ${\epsilon }_t$ is an error term for random background noise at time t; and ${ý}_t$ is the differencing series. A first-order differencing is calculated as (Equation (8)) [47]:

$${ý}_t={ý}_t-{ý}_{t-1}$$

(8)

where ${ý}_t,$ is the observation taken at time t. After determining the orders and estimating the coefficients for the temperature and precipitation time series data by fitting the model, point forecasts and interval forecasts were generated using Equation (7).

3.3. Land Surface Temperature

Land Surface Temperature (LST) data were extracted from Landsat-5 TM imagery from 1992 and Landsat 8 OLI imagery from 2022.

4. Results

4.1. Changes in Glacier Area

In 2022, glaciers covered 12.4% of Chitral’s area (Figure 2). Over the last three decades, the total clean-ice glacier area in Chitral decreased by 816 km² (30.8%), from 2650 km² in 1992 to 1834 km² in 2022, at an average rate of 27.2 km² annually (

Table 3. Change in total clean-ice glacier area in Chitral between 1992 and 2022.

Year	Area (km2)	Change (km2)	Change (%)	Average Annual Change (km2)
1992	2650	---	---	---
1997	2600	−50	−1.9	−10.0
2002	2501	−99	−3.8	−19.8
2007	2462	−39	−1.6	−7.8
2012	2236	−226	−9.2	−45.2
2017	2009	−227	−10.2	−45.4
2022	1834	−175	−8.7	−35.0
1992–2022	---	−816	−30.8	−27.2

). Glacial recession accelerated from 1997 to 2002, slowed over the following five years, and then accelerated significantly again after 2007. While the five-year loss in the glacier area remained below 100 km² before 2007, it rose to approximately 200 km² in the periods that followed. As an example, the Raman Glacier shrank by 3.25 km² (33.6%), from 9.66 km² in 1992 to 6.41 km² in 2022 (

Table 4. Change in clean-ice area of the Raman Glacier in Upper Chitral between 1992 and 2022.

Year	Area (km2)	Change (km2)	Change (%)	Average Annual Change (km2)
1992	9.66	---	---	---
1997	9.50	−0.16	−1.7	−0.03
2002	7.52	−1.98	−20.8	−0.40
2007	7.30	−0.22	−2.9	−0.04
2012	7.00	−0.30	−4.1	−0.06
2017	6.73	−0.27	−3.9	−0.05
2022	6.41	−0.32	−4.8	−0.06
1992–2022	---	−3.25	−33.6	−0.67

, Figure 3). Similar to the overall trend in Chitral, substantial glacier recession occurred particularly from 1997 to 2002 and after 2007.

4.2. Changes in Temperature

For both Chitral city and Drosh, MKT and TSS indicate a weak increasing trend in MMT_max between 1967 to 2022, particularly during June, July, August, September, December, and January (

Table 5. Mann–Kendall test results for the monthly mean maximum temperature (MMT_max) in Chitral city and Drosh between 1967 and 2022. (TSS: Theil–Sen Slope, H_o: null hypothesis).

Month	Chitral City						Drosh
	Kendall’s Tau (τ)	S′ (S Prime)	Var (S′)	p	TSS	Ho	Kendall’s Tau (τ)	S′ (S Prime)	Var (S′)	p	TSS	Ho
Jan	0.32	23	149	0.043	0.75	Rejected	0.06	4	104	0.036	0.48	Rejected
Feb	0.21	15	144	0.021	0.64	Rejected	0.06	4	104	0.769	0.29	Accepted
Mar	0.33	24	155	0.541	0.55	Accepted	0.10	7	103	0.554	0.42	Accepted
Apr	0.31	22	146	0.049	0.50	Rejected	0.08	6	102	0.621	0.05	Accepted
May	0.21	15	93	0.119	0.54	Accepted	0.13	9	103	0.043	0.41	Rejected
Jun	0.03	2	106	0.015	0.01	Rejected	0.14	10	102	0.042	0.02	Rejected
Jul	0.14	10	126	0.043	0.46	Rejected	0.04	3	101	0.032	0.30	Rejected
Aug	0.03	2	148	0.052	0.30	Rejected	0.04	3	103	0.034	0.34	Rejected
Sep	0.03	2.	59	0.045	0.10	Rejected	0.23	16	102	0.037	0.20	Rejected
Oct	0.24	17	145	0.048	0.30	Rejected	−0.06	−4	104	0.769	−0.05	Accepted
Nov	0.18	13	113	0.021	0.24	Rejected	−0.06	−4	97	0.761	−0.25	Accepted
Dec	0.20	14	162	0.027	0.45	Rejected	0.07	5	103	0.039	0.17	Rejected

, Figure 4 and Figure 5). Additionally, TSS showed a strong increasing trend in MMT_max for Chitral in August, while a weak decreasing trend was observed for Drosh in September. Both MKT and TSS suggested a strong increase in MMT_min for Chitral city in June and a very strong increase for Drosh (

Table 6. Mann–Kendall test results for the monthly mean minimum temperature (MMT_min) in Chitral city and Drosh between 1967 and 2022. (TSS: Theil–Sen Slope, H_o: null hypothesis).

Month	Chitral city						Drosh
	Kendall’s Tau (τ)	S′ (S Prime)	Var (S′)	p	TSS	Ho	Kendall’s Tau (τ)	S′ (S Prime)	Var (S′)	p	TSS	Ho
Jan	0.12	−9	103	0.037	0.14	Rejected	0.06	4	104	0.025	0.01	Rejected
Feb	0.03	2	102	0.025	0.02	Rejected	0.08	2	104	0.413	−0.01	Accepted
Mar	0.20	14	102	0.203	0.10	Accepted	−0.07	−1	103	0.437	−0.01	Accepted
Apr	0.11	8	102	0.492	−0.22	Accepted	0.20	16	104	0.014	0.03	Rejected
May	0.22	16	104	0.014	−0.57	Rejected	0.08	5	103	0.026	0.02	Rejected
Jun	0.54	38	102	0.017	1.07	Rejected	0.29	20	102	0.001	0.06	Rejected
Jul	0.39	28	104	0.012	0.44	Rejected	0.38	26	104	0.0001	0.06	Rejected
Aug	0.35	25	103	0.024	0.16	Rejected	0.40	27	103	0.0001	0.06	Rejected
Sep	0.01	1	103	0.017	0.02	Rejected	0.33	15	103	0.001	0.07	Rejected
Oct	0.13	9	103	0.430	0.04	Accepted	0.22	13	103	0.024	0.04	Rejected
Nov	0.40	28	100	0.012	−0.53	Rejected	0.21	5	103	0.028	0.04	Rejected
Dec	0.11	8	104	0.049	0.10	Rejected	0.12	7	101	0.034	0.02	Rejected

, Figure 6 and Figure 7). For both cities, significant increases in MMT_min were found in July and August. The ARIMA model indicated a warming trend for both Chitral city and Drosh, particularly from May to September, with projections suggested that this trend will continue from 2022 to 2042 (Figure 4, Figure 5, Figure 6 and Figure 7).

In 1992, land surface temperature (LST) ranged from a maximum of 31.0 °C to a minimum of −11.0 °C (Figure 8). By 2022, it had increased to a maximum of 38.3 °C (by 7.3 °C) and a minimum of −7.4 °C (by 3.6 °C). Over the study period, a significant warming trend has been observed in elevations above 3000 m asl. While in 1992 the LST ranged from −6.4 °C to −11.0 °C, it had increased to a minimum temperature of −3.8 °C by 2022.

4.3. Changes in Precipitation

Both Chitral city and Drosh experienced decreases in precipitation between 1967 and 2022. In Chitral city, the most significant reduction was observed in December and January (

Table 7. Mann–Kendall test results for the monthly mean precipitation in Chitral city and Drosh between 1967 and 2022. (TSS: Theil–Sen Slope, H_o: null hypothesis).

Month	Chitral city						Drosh
	Kendall’s tau (τ)	S′ (S Prime)	Var (S′)	p	TSS	Ho	Kendall’s tau (τ)	S′ (S Prime)	Var (S′)	p	TSS	Ho
Jan	−0.57	−4.0	104	0.017	−2.68	Rejected	−0.46	−4	104	0.007	−7.26	Rejected
Feb	−0.17	−12.0	104	0.028	−8.29	Rejected	−0.62	−16	104	0.014	−8.48	Rejected
Mar	−0.03	−2.0	104	0.032	−4.72	Rejected	−0.06	−4	104	0.037	−3.30	Rejected
Apr	−0.06	−4.0	104	0.769	0.81	Accepted	−0.25	−18	104	0.102	−9.92	Accepted
May	0.06	4.0	104	0.769	4.33	Accepted	−0.08	−6	104	0.624	−5.63	Accepted
Jun	0.15	11.0	103	0.324	−0.16	Accepted	0.24	17	103	0.110	4.50	Accepted
Jul	0.22	−16.0	104	0.141	−1.22	Accepted	0.08	−6	104	0.104	−2.90	Accepted
Aug	0.01	1.0	103	0.154	−1.03	Accepted	0.08	6	104	0.620	0.31	Accepted
Sep	0.36	26.0	104	0.014	1.97	Rejected	−0.01	−1	103	1.001	0.32	Accepted
Oct	0.08	6.0	104	0.624	1.66	Accepted	−0.11	−8	104	0.490	−2.27	Accepted
Nov	0.22	16.0	104	0.141	6.29	Accepted	0.19	14	104	0.200	4.82	Accepted
Dec	−0.76	−4.0	104	0.039	−1.28	Rejected	−0.25	−18	104	0.001	−10.58	Rejected

). A more pronounced decrease was observed in Drosh, particularly in December. In contrast, precipitation slightly increased in both cities during the summer months of June to August. However, the p-values indicate that these increases were statistically insignificant, suggesting that precipitation likely remained stable during the summer. Nevertheless, it is important to continue monitoring changes in precipitation amounts.

The ARIMA results revealed variations in annual precipitation. In Chitral city, the highest annual precipitation of 664 mm occurred in 2022, while the lowest, 73 mm, was recorded in 1997, indicating a relatively large range (Figure 9). The 56-year mean annual precipitation was 258 mm, with a standard deviation of 130 mm. The analysis identified a trend of decreasing annual precipitation from 1967 to 2022. In Drosh, the highest annual precipitation of 952 mm occurred in 1972, while the lowest, 302 mm, was recorded in 1971. The mean annual precipitation was 577 mm, with a standard deviation of 135 mm. Annual precipitation fluctuated significantly and showed a decreasing trend between 1967 and 2022, with projections suggesting that this trend will continue from 2022 to 2042.

5. Illegal Ice and Snow Harvesting

Over the past decades, illegal harvesting and commercial exploitation of snow and glacier ice have become significant economic activities in Chitral (Figure 10). According to interviews and focus group discussions with local ice sellers, forty to fifty vehicles transport harvested snow and ice to Chitral town daily from May to September. This activity is managed by various contractors, who collect the snow and ice from different locations around Lawari Pass. Each vehicle carries between 45 and 55 nuks (approximately 40 kg), resulting in a total daily harvest of between 72 and 110 tons of snow and ice. Over the five-month harvesting season, this amounts to between 10,800 and 16,500 tons, or roughly 12 to 18 million cubic meters. Snow and ice are also harvested from other areas in Upper Chitral, such as Attakh and Tirich. With rising temperatures during the summer, more frequent heatwaves, and electricity load shedding, the demand for snow and ice has increased. Local vendors sell the snow and ice for domestic use at a rate of PKR 100 (~35 US cents) per kg.

6. Discussion

Our glacier mapping based on Sentinel-2 imagery and the Normalized Difference Glacier Index (NDGI) method revealed a total glacier area of 1834 km² in 2022, covering 12.4% of the Chitral Valley’s landscape. This closely aligns with the 1850 km² in the GLIMS database and the 1822 km² reported by the Pakistan New Glacier Inventory for the same year [53]. While both studies utilized Sentinel-2 imagery, the inventory employed a different approach, combining the Normalized Difference Snow Index (NDSI), Normalized Difference Vegetation Index (NDVI), and Band 4 (Red). The NDSI is calculated as NDGI = (Green − SWIR)/(Green + SWIR), whereas the NDVI is determined as NDVI = (Near Infrared − Red)/(Near Infrared + Red). Despite this consistency in total glacier area, discrepancies arise when comparing our results with those of an earlier study that followed a different method that delineates glaciers through variations in grey scale values in panchromatic satellite images [54]. Using Landsat-7 images, that study estimated a glacier area of 1903 km² in 2001, whereas our analysis found a significantly larger area of 2501 km² in 2002. The nearly 600 km² difference likely reflects advancements in glacier mapping techniques. Notably, it remains unclear why the 2014 study employed a methodology dating back to 1977. The glacier recession pattern follows a typical trend, with an increasing number of glaciers resulting from the progressive disintegration of larger glaciers into smaller ones over time. For 2001, different studies identified 542 and 546 glaciers, respectively [54,55]. By 2013, the count had risen to 806 [55], and in 2022, it reached 1872 [53]. However, some caution is necessary interpreting these numbers, as defining the exact boundaries of a glacier and distinguishing individual glaciers within a larger glacierized area remain challenging.

Our study highlights significant glacier recession in the Eastern Hindu Kush of Chitral over the past three decades. Between 1992 and 2022, glaciers lost 816 km² (31%) of their total area, with an average annual loss of 27 km². Accelerated glacier recession occurred between 1997 and 2002, when a severe drought occurred, and again after 2007. A study on six larger glacier networks (>20 km²) revealed a mean glacier terminus retreat of 130 m from 2001 to 2018 [56].

Simultaneously, both maximum and minimum monthly mean air temperatures increased between 1992 and 2022, particularly during the summer months of June, July, and August. Our findings are supported by a seasonal trend analysis study indicating a significant increase in temperature [57]. In contrast, an analysis of temperature data from the Chitral weather station for the period 1990–2013 described the mean temperature as stable [56]. Furthermore, our study found that both maximum (+7.3 °C) and minimum (+3.6 °C) land surface temperatures significantly increased from 1992 to 2022 across Chitral, and they increased significantly in elevations above 3000 m asl. This warming trend is projected to continue at least until 2042, which is further supported by RCP4.5 and RCP 8.5 scenarios projecting a temperature increase across Chitral by the end of the century [58,59].

Between 1967 and 2022, a trend of decreasing annual precipitation was observed, primarily due to reduced precipitation during the winter months, although there was a slight (but statistically insignificant) increase in summer precipitation. Similar to the temperature findings, an analysis of precipitation data from the Chitral city weather station from 1990 to 2013 also reported stable annual precipitation [56].

Nevertheless, our results suggest that less snow fed the glacial accumulation zones at higher elevations during winter, while increased rainfall in the summer may have contributed to glacier melt, particularly at lower elevations. Additionally, the earlier onset of the snowmelt season further exacerbated the glacier recession. Our results suggested a continuation of decrease in annual precipitation, which is supported by some other studies [58,59,60]. A decline in precipitation by the end of the century has been predicted in estimates using the RCP4.5 scenario, by 16%, and using the RCP8.5 scenario, by 35% [59]. Furthermore, the RCP8.5 scenario predicted a decline in winter precipitation by almost 10% and an increase in summer precipitation by almost 29%. In contrast, other studies predicted a substantial increase, predominantly during the winter months [61,62].

Our findings on glacier recession align with trends observed regionally [63,64,65,66] and globally [67,68,69], which have been primarily attributed to climate change [70]. Some studies also highlight the deposition of black carbon and dust on glacier surfaces, which lowers albedo and accelerates melting, as an additional factor driving the accelerated glacier retreat [71,72]. While the recent harvesting of snow and ice is not currently a major contributor to the retreat of Chitral’s glaciers, the volumes harvested are significant and may increase. This could further impact smaller glaciers, especially considering the population growth of approximately 200,000 people (>60%) over the last twenty-five years.

The melting of glaciers may initially lead to increased runoff and water availability [73] but also poses a significant risk of flooding, particularly through glacial lake outburst floods (GLOFs) [74,75,76,77,78]. For the Chitral River Basin, estimates based on RCP4.5 scenarios indicate that snowmelt currently contributes 95.1% of the total stream flow, followed by glacier melt at 4.6% and rainfall at 0.2% [58]. By the end of the century, snowmelt contributions are projected to decline to 87.4%, while glacier melt is expected to rise to 12.2%. However, the study does not clarify whether this increase in glacier melt represents an absolute increase or merely a relative shift due to reduced snowfall and declining snowmelt. A separate study estimated that the mean summer discharge of the Chitral River will increase by 14–19% under the three mid-century RCP scenarios and by 13–37% under the three late-century RCP scenarios compared to that in the reference period of 2000–2005 [32]. While these projections suggest a long-term increase in meltwater runoff, other studies indicated that once glaciers retreat to higher elevations—where conditions may still support ice accumulation—water availability for ecosystems and human activities will eventually diminish, which could lead to severe water shortages and potential crises [76,79]. Secondary consequences, such as agricultural decline and food insecurity, may also emerge as a result [77].

7. Conclusions

The increase in land surface and summer air temperatures, coupled with decreasing winter precipitation, has driven significant glacial recession in the Eastern Hindu Kush of Chitral, Pakistan over the past thirty years. Modeling results project that this trend will continue over the next two decades. The expected impacts of glacier recession include the formation of new glacial lakes, increased flooding and sedimentation. Additionally, while an initial phase of increased runoff and heightened flood risk is anticipated, it is likely to be followed by a long-term runoff deficit and worsening water scarcity. Therefore, monitoring Chitral’s snow and ice resources, along with their careful management, is crucial for sustainable development that reduces vulnerabilities and enhances resilience. Future investigations should explore the impact of black carbon from anthropogenic industrial and transportation activities on the accelerated recession of Chitral’s glaciers.