Bilingual
En/Fa
Geographical Research is Published in both Persian and English Full-text.
Volume 37, Issue 2 (2022)
GeoRes 2022, 37(2): 231-239 |
Back to browse issues page
Article Type:
Subject:
History
Received: 2021/12/29 | Accepted: 2022/04/11 | Published: 2022/06/11
Received: 2021/12/29 | Accepted: 2022/04/11 | Published: 2022/06/11
How to cite this article
Sherafat M, Yarahmadi D, Fathnia A, Mirhashemi H. Monitoring Snow Cover Changes and the Volume of Snow Water Equivalent Using MODIS and AMSR-2/AMSR-E Sensor Data (Case Study: Karun, Karkheh and Dez Basins). GeoRes 2022; 37 (2) :231-239
URL: http://georesearch.ir/article-1-1280-en.html
URL: http://georesearch.ir/article-1-1280-en.html
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Rights and permissions
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
1- Department of Geography, Faculty of Literature and Humanities, Lorestan University, Khorramabad, Iran
2- Department of Geography, Faculty of Literature and Humanities, Razi University, Kermanshah, Iran
2- Department of Geography, Faculty of Literature and Humanities, Razi University, Kermanshah, Iran
Full-Text (HTML) (151 Views)
Introduction
From the perspective of climatologists and meteorologists who study climatic and atmospheric changes at the global scale, snow monitoring and the availability of accurate information on the spatial distribution of snow-covered areas are essential for weather forecasting and hydrological and climatological modeling [Shan et al., 2005]. In Iran, extensive dam construction projects and large irrigation and drainage networks have been implemented to ensure energy and food security, particularly within the Karun, Karkheh, and Dez basins. This has increased the sensitivity of water resources in these basins to variations in inflow. Moreover, the largest share of the country’s hydropower generation capacity is concentrated in these three basins [Jamali, 2014]. Given that the potential vulnerability of infrastructure can be assessed prior to construction, and considering the direct dependence of production units and hydropower plants on water resources in these basins, it is essential to investigate changes in snow-covered areas as one of the most important sources of water supply for dams and rivers. In addition, as the study area is located within a mountainous basin, the risk of flooding is relatively high; in some cases, the coincidence of precipitation with snowmelt processes leads to floods that cause irreparable damage to agricultural sectors and infrastructure in these regions.
Numerous studies have been conducted on snow cover monitoring and variations in snow water equivalent (SWE, i.e., water derived from snowmelt), some of the most important of which are reviewed here. Langlois et al. [2008] have measured SWE in the Franklin region of Canada from December 7, 2003, to April 30, 2004, using AMSR-E sensor data. In their study, they employed a method based on calculating the brightness temperature of six neighboring pixels of the sensor. The results indicated correlation coefficients of 0.75 and 0.73 between observed and predicted data for smooth ice surfaces in the two pixels. Zhou et al. [2013] have examined changes in snow cover across Central Asia for the period 1980–2008 using AVHRR and MODIS sensor data and a thresholding approach.
Banihabib et al. [2013] used AVHRR imagery from 1986 to 2007 to monitor snow cover in the Shah-Cheraghi Dam basin. Trend analysis of snow cover changes using linear regression and the Mann–Kendall test indicated no significant trend in the snow cover time series. Byun and Choi [2014] have estimated SWE in four regions of South Korea using AMSR-E SWE data and evaluated the results against ground observations for the period December 2002 to February 2011. Their assessment showed that SWE was overestimated in three of the regions.
Bavera et al. [2014] have calculated SWE for Switzerland during 2004–2006 using two statistical models (HS-SWE and SWE-SEM, based on the distribution of snow depth, density, and snow-covered area maps from 32 ground stations) and a physical model (ALPINE3D), and compared the results of these models. The mean difference between the two statistical models is approximately 8%, while the difference between the physical model and the statistical models ranges from about −3% to 10%. Mhawej et al. [2014] have monitored snow cover and SWE in Lebanon using a combination of Terra–Aqua and AMSR-E imagery from 2000 to 2012 and developed an appropriate model for estimating SWE in the study area.
Yang et al. [2015] have estimated SWE in China using passive microwave data from the AMSR-E sensor and have identified topography and vegetation cover as influential factors in the SWE retrieval algorithm. They show that estimation errors are greater in areas with dense vegetation cover compared to sparsely vegetated or non-vegetated areas. Ansari and Marofi [2017] use daily AMSR-E data from the Aqua satellite and the Global Land Data Assimilation System (GLDAS) to estimate daily SWE at snow-gauging stations in northwestern Iran (the Lake Urmia basin and the western Mazandaran watershed) for the period 2006–2010. Their results indicate a statistically significant correlation at the 1% level between estimated and observed SWE. Furthermore, incorporating measured snow density into the AMSR-E data increased the correlation coefficient from 0.27 to 0.55.
Coll and Li [2018] assess the accuracy of MODIS snow cover products and data gap-filling methods using 12 years (2001–2012) of observations from 800 snow measurement stations in the western United States. They have concluded that land cover type had the greatest influence on snow cover monitoring accuracy. Soleimani et al. [2018] have investigated changes in snow cover in Kurdistan Province using MODIS satellite imagery and the NDSI method from 2005 to 2017. Their results show that the minimum extent of snow-covered areas occurred in 2007 and 2008, while the maximum extent was observed in 2009, 2012, and 2016. Mohammadi and Khoorani [2019] have analyzed changes in snow cover over the Zagros Mountains using daily MODIS imagery from 2001 to 2016 and found a decreasing trend in snow cover for all months except November from 2009 onward. Guyennon et al. [2019] measure SWE, snow density, and snow depth in the Italian Alps from 2009 to 2018 and used meteorological data to improve their SWE estimation model.
Although recent studies have examined variations in snow cover and SWE, the relationship between these parameters and basin discharge has not been sufficiently investigated. Therefore, the present study aims to analyze changes in snow cover and SWE volume in the Karun, Karkheh, and Dez basins in relation to basin discharge, using MODIS and AMSR-2/AMSR-E sensor imagery in a raster-based framework. The objective is to determine the sensitivity of basin runoff to variations in water derived from snowmelt over snow-covered areas.
Methodology
Based on its nature and objectives, this research is an applied study, and its methodological approach is descriptive–correlational. To monitor snow cover and examine the relationship between SWE and discharge in the Karun, Karkheh, and Dez basins, daily and 8-day MODIS imagery (400 images with 500-m spatial resolution from 2000 to 2020), available 5-day SWE data from AMSR-E and AMSR-2 (400 images with 25-km spatial resolution from 2003 to 2020), and monthly discharge data from the Pay-e-Pol, Gatvand, and Ahvaz hydrometric stations for the period 2000–2018 were used. The primary reason for employing MODIS imagery to extract snow cover was its high temporal and spatial resolution, given the large extent of the study areas, as well as its suitable spectral characteristics for distinguishing snow from clouds. This advantage is particularly evident in 8-day MODIS composites (500-m spatial resolution), which effectively mitigate the transient effects of cloud cover. It should be noted that careful attention was paid to data simultaneity and correspondence between station observations and satellite imagery.
For this study, higher-level (Level 3) data products, already corrected and processed, were primarily used. However, for raw imagery (Levels 0 and 1), geometric and radiometric corrections were applied where necessary. Snow cover extraction from MODIS imagery involved performing geometric and radiometric corrections and mosaicking the images using ENVI 4.7 software (as the study area spanned two scenes). The Normalized Difference Snow Index (NDSI), defined as (Band 4−Band 6) / (Band 4−Band 6), was calculated for all images. A pixel was classified as snow if it satisfied the following conditions [Klein et al., 1998]:
SWE = 4,8 (T18H - T37H)
where SWE denotes snow water equivalent. This algorithm uses the brightness temperature gradient at 18 GHz (T₁₈H) and 37 GHz (T₃₇H) to estimate SWE, which is retrieved through regression between brightness temperature gradients and observed SWE data [Kelly et al., 2004]. After SWE retrieval, the data were stored in digital format with a 25-km grid cell size in HDF-EOS format; pixel values ranging from 0 to 240 represent SWE in millimeters [Ansari & Marofi, 2017]. According to the AMSR-E product documentation, actual SWE values are divided by a factor of 2 for storage [Kelly et al., 2004]; therefore, the values must be multiplied by 2 prior to use [Gao et al., 2010].
Following the extraction of snow cover and SWE and the acquisition of monthly discharge data from the three selected stations, dry months or months with negligible snow were first removed from the dataset. After data normalization, Spearman correlation analysis and the raster-based Mann–Kendall trend test were applied to examine trends in snow cover area and SWE and to assess their correlations with basin discharge. The Mann–Kendall test was originally proposed by Mann [1945] and later extended by Kendall [1970]. The null hypothesis of the Mann–Kendall test assumes randomness and the absence of a trend in the data series, whereas acceptance of the alternative hypothesis (rejection of the null) indicates the presence of a trend in the time series.
Findings
Based on the trend analysis using the Mann–Kendall method, although snow cover in the Karkheh Basin has decreased in recent years, these changes were not statistically significant due to the regional climate and geographic setting (p>0.05). The greatest decreasing trend in snow-covered area in the Karkheh Basin occurred in February, with a value of −1.479. In this basin, considering the discharge recorded at the Pay-e-Pol station, the mean monthly discharge during snow-covered seasons exhibited a decreasing trend; however, according to the Mann–Kendall analysis, these changes were not statistically significant in 50% of the months examined (p>0.05). The largest decrease in discharge in the Karkheh Basin occurred in January, with a value of −3.35, which was statistically significant at the 0.05 level (p<0.05).
In the Dez Basin, snow-covered area showed a noticeable decrease in all months; however, this declining trend was statistically significant only in June, with a value of −2.204. Owing to the mountainous nature of this region, significant changes in snow-covered area were observed in certain months. Discharge at the Gatvand station also indicated an overall decreasing trend in monthly discharge during the study period. The largest decrease in discharge in this basin occurred in May, with a value of −3.26, representing the greatest reduction among all months examined. Overall, due to its relatively limited area and high mountain elevations, the Dez Basin exhibited more pronounced changes compared to the other two basins.
In the Karun Basin, snow-covered area decreased in all months; however, these decreases were statistically significant only in February and June, with values of −2.083 and −2.144, respectively. Discharge at the Ahvaz station also showed a decreasing trend, which was statistically significant only in February and March, with values of −2.35 and −2.12, respectively (p<0.05).
In addition, changes in snow water equivalent (SWE) derived from available AMSR-E and AMSR-2 data (2003–2021) were analyzed. The results indicated a decreasing trend in SWE for all months, although these decreases were not statistically significant in some months (p > 0.05). The largest reductions in SWE occurred in January in the Karkheh Basin (−3.18), March in the Dez Basin (−3.33), and February in the Karun Basin (−3.86). Overall, the greatest decreases in SWE across most months were observed in the Karun Basin.
Because most hydrological and climatological data do not follow a normal distribution, the nonparametric Spearman correlation coefficient was used to examine the relationships among snow-covered area, discharge, and SWE. The highest correlation between snow-covered area and discharge across the three basins was observed in the Dez Basin in June, with a coefficient of 0.775 (p<0.05), while the lowest correlation was found in the Karkheh Basin in February, with a value of 0.183. In general, for most months, the correlation between snow-covered area and discharge in the Karun Basin was stronger than in the other two basins. Analysis of the relationship between discharge and SWE derived from AMSR-2/AMSR-E imagery also showed that the highest correlation occurred in the Karun Basin in January, with a coefficient of 0.721 (p<0.05).
Analysis of temporal variations in snow-covered area and discharge over the period 2001–2021 indicated that, in all three basins, the maximum snow-covered area occurred in January 2007.
Based on annual snow cover variations, the Karkheh Basin exhibited a decreasing trend in snow-covered area over the past decade. During the period 2005–2009, extensive areas of the basin were covered by snow each year due to heavy snowfall in mid-winter to late winter; however, after 2010, only one major snowfall event was recorded. Although statistical analysis revealed a significant overall decrease in snow cover in the Karkheh watershed, no significant decreasing trend was observed in the southern part of the basin downstream of the Cham Zab hydrological station. Overall, approximately 34.5% of the southern area did not exhibit a significant decreasing trend, whereas about 38% of the northern and higher-elevation areas showed a decreasing trend over time. These results indicate that areas near Firouzabad and Nourabad cities, located at elevations between 3000 and 3500 m, experienced the greatest reduction in snow cover, with decreases ranging from 26% to 54%, depending on geographic location. In general, the intensity of snow cover reduction in the Karkheh Basin was greater than in the other basins, given its historical snow conditions.
Further analysis showed that the Dez Basin experienced a decreasing trend in snow-covered area over the past 13 years. Between 2002 and 2008, large portions of the basin were snow-covered each year during the winter months; however, after 2007, a declining trend in snowfall was observed. Monthly trend analysis also indicated a reduction in snow cover in this basin. A significant decrease in snow cover was detected in the Dez watershed; however, in the southern part of the basin downstream of the Gatvand hydrological station, no significant decreasing trend was observed. Overall, approximately 47.3% of the southern area did not show a significant decline, whereas about 47% of the northern and higher-elevation areas exhibited a decreasing trend over time. Trend analysis further indicated that at elevations between 2000 and 2500 m, the rate of decrease ranged from 40% to 47% per month, which is higher than in other parts of the basin, although less severe than in the Karkheh Basin.
Finally, analysis of the spatiotemporal patterns of snow cover changes in the Karun Basin, the most important basin in this study, revealed a decreasing trend in snow-covered area over the past 10 years. Between 2000 and 2010, the basin was snow-covered during many months of the year; however, from 2011 onward, a declining trend in snow cover was observed. Detailed trend analysis indicated that no significant decrease in snow-covered areas occurred in the southern part of the basin downstream of the Gatvand hydrological station. Overall, approximately 63.7% of the southern area did not exhibit a significant decreasing trend, whereas much of the basin experienced substantial changes in snow-covered area. About 27% of the northern and higher-elevation regions showed a decreasing trend over time. The most pronounced changes occurred at elevations between 3500 and 4000 m, which account for approximately 2% of the basin area. Similar to the other two basins, no notable changes were observed at lower elevations due to the absence of persistent snow cover.
Discussion
In recent years, the frequency of extreme climatic events such as floods and droughts has increased, highlighting the critical importance of water resources management. Accordingly, the main objective of this study was to investigate changes in snow cover, as one of the most important sources of freshwater, and its relationship with basin discharge using satellite imagery. Various methods are employed to extract snow pixels from satellite images; however, one of the most effective approaches is the Normalized Difference Snow Index (NDSI). This index was first introduced by Dietz in 1984 [Dietz, 2013] and is based on the principle that snow exhibits high reflectance in the visible spectrum and low reflectance in the mid-infrared region [Hüsler et al., 2011].
An important consideration in snow monitoring studies is that snow pixels extracted from satellite imagery primarily represent the spatial extent of snow-covered areas and do not provide sufficient information about the amount of water stored within these areas. In this regard, one of the most reliable methods for estimating snow water equivalent (SWE) is the use of AMSR-E/AMSR-2 sensor data. This sensor is a 12-channel, six-frequency, conically scanning passive microwave radiometer mounted on the Aqua satellite [Lobl et al., 2003; Kawanishi et al., 2003].
The findings of this study indicated that snow-covered area has exhibited a decreasing trend in all months; however, in some months these changes were not statistically significant. Similar results were reported by Banihabib et al. [2013] for the Shah-Cheraghi Dam basin. In terms of temporal variation, snow-covered area has declined markedly in all three basins since 2009, which is consistent with the findings of Mohammadi and Khoorani [2019]. Overall, the magnitude of snow cover reduction in the Karun and Dez basins was greater than in the Karkheh Basin. From a seasonal perspective, June experienced the most pronounced decrease in snow-covered area.
In contrast, analysis of changes in SWE revealed patterns that differed from those of snow-covered area. The greatest reduction in SWE in all three basins occurred in February, whereas the analysis of snow-covered area did not reflect a comparable level of change during this month. This indicates that in recent years, reductions in snow depth and the volume of water derived from snowmelt have been substantially greater than reductions in the areal extent of snow cover. Correlation analysis between snow-covered area, SWE, and basin discharge showed that the relationship between snow-covered area and discharge was stronger during warmer seasons. For example, in the Dez Basin, the correlation coefficient between snow-covered area and discharge was 0.265 in January but increased to 0.775 in June. This finding suggests that, particularly in the Dez Basin, the sensitivity of basin runoff to snow storage increases with rising temperatures.
Conversely, during colder seasons, this dependency weakens due to reduced snowmelt rates. The strongest correlation between snow-covered area and discharge was observed in the Karun Basin, which can be attributed to the presence of extensive high-elevation and snow-prone mountainous regions in this basin. To further enhance the robustness of the analysis, correlations between SWE derived from AMSR-E/AMSR-2 data and basin discharge were also examined. However, due to the relatively coarse spatial resolution of these datasets, SWE values extracted for June were close to zero in all three basins. A similar discrepancy between estimated and actual SWE values has also been reported by Soleimani et al. [2018]. Therefore, it is recommended that higher spatial resolution datasets be used for SWE analysis during warm seasons, when snow-covered area is limited.
Conclusion
Overall, during the study period, both snow-covered area and snow water equivalent exhibited decreasing trends in all months, with more pronounced reductions observed in the Karun and Dez basins. In addition, for most months, no significant relationship was found between snow-covered area and basin discharge.
Acknowledgements: -
Ethical Permission: -
Conflict of Interest: -
Author Contributions: Sherafat M (First Author), Main Researcher (50%); Yarahmadi D (Second Author), Introduction Writer (30%); Fathnia A (Third Author), Methodologist (10%); Mirhashemi H (Fourth Author), Statistical Analyst (10%).
Funding: -
From the perspective of climatologists and meteorologists who study climatic and atmospheric changes at the global scale, snow monitoring and the availability of accurate information on the spatial distribution of snow-covered areas are essential for weather forecasting and hydrological and climatological modeling [Shan et al., 2005]. In Iran, extensive dam construction projects and large irrigation and drainage networks have been implemented to ensure energy and food security, particularly within the Karun, Karkheh, and Dez basins. This has increased the sensitivity of water resources in these basins to variations in inflow. Moreover, the largest share of the country’s hydropower generation capacity is concentrated in these three basins [Jamali, 2014]. Given that the potential vulnerability of infrastructure can be assessed prior to construction, and considering the direct dependence of production units and hydropower plants on water resources in these basins, it is essential to investigate changes in snow-covered areas as one of the most important sources of water supply for dams and rivers. In addition, as the study area is located within a mountainous basin, the risk of flooding is relatively high; in some cases, the coincidence of precipitation with snowmelt processes leads to floods that cause irreparable damage to agricultural sectors and infrastructure in these regions.
Numerous studies have been conducted on snow cover monitoring and variations in snow water equivalent (SWE, i.e., water derived from snowmelt), some of the most important of which are reviewed here. Langlois et al. [2008] have measured SWE in the Franklin region of Canada from December 7, 2003, to April 30, 2004, using AMSR-E sensor data. In their study, they employed a method based on calculating the brightness temperature of six neighboring pixels of the sensor. The results indicated correlation coefficients of 0.75 and 0.73 between observed and predicted data for smooth ice surfaces in the two pixels. Zhou et al. [2013] have examined changes in snow cover across Central Asia for the period 1980–2008 using AVHRR and MODIS sensor data and a thresholding approach.
Banihabib et al. [2013] used AVHRR imagery from 1986 to 2007 to monitor snow cover in the Shah-Cheraghi Dam basin. Trend analysis of snow cover changes using linear regression and the Mann–Kendall test indicated no significant trend in the snow cover time series. Byun and Choi [2014] have estimated SWE in four regions of South Korea using AMSR-E SWE data and evaluated the results against ground observations for the period December 2002 to February 2011. Their assessment showed that SWE was overestimated in three of the regions.
Bavera et al. [2014] have calculated SWE for Switzerland during 2004–2006 using two statistical models (HS-SWE and SWE-SEM, based on the distribution of snow depth, density, and snow-covered area maps from 32 ground stations) and a physical model (ALPINE3D), and compared the results of these models. The mean difference between the two statistical models is approximately 8%, while the difference between the physical model and the statistical models ranges from about −3% to 10%. Mhawej et al. [2014] have monitored snow cover and SWE in Lebanon using a combination of Terra–Aqua and AMSR-E imagery from 2000 to 2012 and developed an appropriate model for estimating SWE in the study area.
Yang et al. [2015] have estimated SWE in China using passive microwave data from the AMSR-E sensor and have identified topography and vegetation cover as influential factors in the SWE retrieval algorithm. They show that estimation errors are greater in areas with dense vegetation cover compared to sparsely vegetated or non-vegetated areas. Ansari and Marofi [2017] use daily AMSR-E data from the Aqua satellite and the Global Land Data Assimilation System (GLDAS) to estimate daily SWE at snow-gauging stations in northwestern Iran (the Lake Urmia basin and the western Mazandaran watershed) for the period 2006–2010. Their results indicate a statistically significant correlation at the 1% level between estimated and observed SWE. Furthermore, incorporating measured snow density into the AMSR-E data increased the correlation coefficient from 0.27 to 0.55.
Coll and Li [2018] assess the accuracy of MODIS snow cover products and data gap-filling methods using 12 years (2001–2012) of observations from 800 snow measurement stations in the western United States. They have concluded that land cover type had the greatest influence on snow cover monitoring accuracy. Soleimani et al. [2018] have investigated changes in snow cover in Kurdistan Province using MODIS satellite imagery and the NDSI method from 2005 to 2017. Their results show that the minimum extent of snow-covered areas occurred in 2007 and 2008, while the maximum extent was observed in 2009, 2012, and 2016. Mohammadi and Khoorani [2019] have analyzed changes in snow cover over the Zagros Mountains using daily MODIS imagery from 2001 to 2016 and found a decreasing trend in snow cover for all months except November from 2009 onward. Guyennon et al. [2019] measure SWE, snow density, and snow depth in the Italian Alps from 2009 to 2018 and used meteorological data to improve their SWE estimation model.
Although recent studies have examined variations in snow cover and SWE, the relationship between these parameters and basin discharge has not been sufficiently investigated. Therefore, the present study aims to analyze changes in snow cover and SWE volume in the Karun, Karkheh, and Dez basins in relation to basin discharge, using MODIS and AMSR-2/AMSR-E sensor imagery in a raster-based framework. The objective is to determine the sensitivity of basin runoff to variations in water derived from snowmelt over snow-covered areas.
Methodology
Based on its nature and objectives, this research is an applied study, and its methodological approach is descriptive–correlational. To monitor snow cover and examine the relationship between SWE and discharge in the Karun, Karkheh, and Dez basins, daily and 8-day MODIS imagery (400 images with 500-m spatial resolution from 2000 to 2020), available 5-day SWE data from AMSR-E and AMSR-2 (400 images with 25-km spatial resolution from 2003 to 2020), and monthly discharge data from the Pay-e-Pol, Gatvand, and Ahvaz hydrometric stations for the period 2000–2018 were used. The primary reason for employing MODIS imagery to extract snow cover was its high temporal and spatial resolution, given the large extent of the study areas, as well as its suitable spectral characteristics for distinguishing snow from clouds. This advantage is particularly evident in 8-day MODIS composites (500-m spatial resolution), which effectively mitigate the transient effects of cloud cover. It should be noted that careful attention was paid to data simultaneity and correspondence between station observations and satellite imagery.
For this study, higher-level (Level 3) data products, already corrected and processed, were primarily used. However, for raw imagery (Levels 0 and 1), geometric and radiometric corrections were applied where necessary. Snow cover extraction from MODIS imagery involved performing geometric and radiometric corrections and mosaicking the images using ENVI 4.7 software (as the study area spanned two scenes). The Normalized Difference Snow Index (NDSI), defined as (Band 4−Band 6) / (Band 4−Band 6), was calculated for all images. A pixel was classified as snow if it satisfied the following conditions [Klein et al., 1998]:
- NDSI > 0.4;
- Band 2 reflectance > 11%;
- Band 4 reflectance < 10%.
SWE = 4,8 (T18H - T37H)
where SWE denotes snow water equivalent. This algorithm uses the brightness temperature gradient at 18 GHz (T₁₈H) and 37 GHz (T₃₇H) to estimate SWE, which is retrieved through regression between brightness temperature gradients and observed SWE data [Kelly et al., 2004]. After SWE retrieval, the data were stored in digital format with a 25-km grid cell size in HDF-EOS format; pixel values ranging from 0 to 240 represent SWE in millimeters [Ansari & Marofi, 2017]. According to the AMSR-E product documentation, actual SWE values are divided by a factor of 2 for storage [Kelly et al., 2004]; therefore, the values must be multiplied by 2 prior to use [Gao et al., 2010].
Following the extraction of snow cover and SWE and the acquisition of monthly discharge data from the three selected stations, dry months or months with negligible snow were first removed from the dataset. After data normalization, Spearman correlation analysis and the raster-based Mann–Kendall trend test were applied to examine trends in snow cover area and SWE and to assess their correlations with basin discharge. The Mann–Kendall test was originally proposed by Mann [1945] and later extended by Kendall [1970]. The null hypothesis of the Mann–Kendall test assumes randomness and the absence of a trend in the data series, whereas acceptance of the alternative hypothesis (rejection of the null) indicates the presence of a trend in the time series.
Findings
Based on the trend analysis using the Mann–Kendall method, although snow cover in the Karkheh Basin has decreased in recent years, these changes were not statistically significant due to the regional climate and geographic setting (p>0.05). The greatest decreasing trend in snow-covered area in the Karkheh Basin occurred in February, with a value of −1.479. In this basin, considering the discharge recorded at the Pay-e-Pol station, the mean monthly discharge during snow-covered seasons exhibited a decreasing trend; however, according to the Mann–Kendall analysis, these changes were not statistically significant in 50% of the months examined (p>0.05). The largest decrease in discharge in the Karkheh Basin occurred in January, with a value of −3.35, which was statistically significant at the 0.05 level (p<0.05).
In the Dez Basin, snow-covered area showed a noticeable decrease in all months; however, this declining trend was statistically significant only in June, with a value of −2.204. Owing to the mountainous nature of this region, significant changes in snow-covered area were observed in certain months. Discharge at the Gatvand station also indicated an overall decreasing trend in monthly discharge during the study period. The largest decrease in discharge in this basin occurred in May, with a value of −3.26, representing the greatest reduction among all months examined. Overall, due to its relatively limited area and high mountain elevations, the Dez Basin exhibited more pronounced changes compared to the other two basins.
In the Karun Basin, snow-covered area decreased in all months; however, these decreases were statistically significant only in February and June, with values of −2.083 and −2.144, respectively. Discharge at the Ahvaz station also showed a decreasing trend, which was statistically significant only in February and March, with values of −2.35 and −2.12, respectively (p<0.05).
In addition, changes in snow water equivalent (SWE) derived from available AMSR-E and AMSR-2 data (2003–2021) were analyzed. The results indicated a decreasing trend in SWE for all months, although these decreases were not statistically significant in some months (p > 0.05). The largest reductions in SWE occurred in January in the Karkheh Basin (−3.18), March in the Dez Basin (−3.33), and February in the Karun Basin (−3.86). Overall, the greatest decreases in SWE across most months were observed in the Karun Basin.
Because most hydrological and climatological data do not follow a normal distribution, the nonparametric Spearman correlation coefficient was used to examine the relationships among snow-covered area, discharge, and SWE. The highest correlation between snow-covered area and discharge across the three basins was observed in the Dez Basin in June, with a coefficient of 0.775 (p<0.05), while the lowest correlation was found in the Karkheh Basin in February, with a value of 0.183. In general, for most months, the correlation between snow-covered area and discharge in the Karun Basin was stronger than in the other two basins. Analysis of the relationship between discharge and SWE derived from AMSR-2/AMSR-E imagery also showed that the highest correlation occurred in the Karun Basin in January, with a coefficient of 0.721 (p<0.05).
Analysis of temporal variations in snow-covered area and discharge over the period 2001–2021 indicated that, in all three basins, the maximum snow-covered area occurred in January 2007.
Based on annual snow cover variations, the Karkheh Basin exhibited a decreasing trend in snow-covered area over the past decade. During the period 2005–2009, extensive areas of the basin were covered by snow each year due to heavy snowfall in mid-winter to late winter; however, after 2010, only one major snowfall event was recorded. Although statistical analysis revealed a significant overall decrease in snow cover in the Karkheh watershed, no significant decreasing trend was observed in the southern part of the basin downstream of the Cham Zab hydrological station. Overall, approximately 34.5% of the southern area did not exhibit a significant decreasing trend, whereas about 38% of the northern and higher-elevation areas showed a decreasing trend over time. These results indicate that areas near Firouzabad and Nourabad cities, located at elevations between 3000 and 3500 m, experienced the greatest reduction in snow cover, with decreases ranging from 26% to 54%, depending on geographic location. In general, the intensity of snow cover reduction in the Karkheh Basin was greater than in the other basins, given its historical snow conditions.
Further analysis showed that the Dez Basin experienced a decreasing trend in snow-covered area over the past 13 years. Between 2002 and 2008, large portions of the basin were snow-covered each year during the winter months; however, after 2007, a declining trend in snowfall was observed. Monthly trend analysis also indicated a reduction in snow cover in this basin. A significant decrease in snow cover was detected in the Dez watershed; however, in the southern part of the basin downstream of the Gatvand hydrological station, no significant decreasing trend was observed. Overall, approximately 47.3% of the southern area did not show a significant decline, whereas about 47% of the northern and higher-elevation areas exhibited a decreasing trend over time. Trend analysis further indicated that at elevations between 2000 and 2500 m, the rate of decrease ranged from 40% to 47% per month, which is higher than in other parts of the basin, although less severe than in the Karkheh Basin.
Finally, analysis of the spatiotemporal patterns of snow cover changes in the Karun Basin, the most important basin in this study, revealed a decreasing trend in snow-covered area over the past 10 years. Between 2000 and 2010, the basin was snow-covered during many months of the year; however, from 2011 onward, a declining trend in snow cover was observed. Detailed trend analysis indicated that no significant decrease in snow-covered areas occurred in the southern part of the basin downstream of the Gatvand hydrological station. Overall, approximately 63.7% of the southern area did not exhibit a significant decreasing trend, whereas much of the basin experienced substantial changes in snow-covered area. About 27% of the northern and higher-elevation regions showed a decreasing trend over time. The most pronounced changes occurred at elevations between 3500 and 4000 m, which account for approximately 2% of the basin area. Similar to the other two basins, no notable changes were observed at lower elevations due to the absence of persistent snow cover.
Discussion
In recent years, the frequency of extreme climatic events such as floods and droughts has increased, highlighting the critical importance of water resources management. Accordingly, the main objective of this study was to investigate changes in snow cover, as one of the most important sources of freshwater, and its relationship with basin discharge using satellite imagery. Various methods are employed to extract snow pixels from satellite images; however, one of the most effective approaches is the Normalized Difference Snow Index (NDSI). This index was first introduced by Dietz in 1984 [Dietz, 2013] and is based on the principle that snow exhibits high reflectance in the visible spectrum and low reflectance in the mid-infrared region [Hüsler et al., 2011].
An important consideration in snow monitoring studies is that snow pixels extracted from satellite imagery primarily represent the spatial extent of snow-covered areas and do not provide sufficient information about the amount of water stored within these areas. In this regard, one of the most reliable methods for estimating snow water equivalent (SWE) is the use of AMSR-E/AMSR-2 sensor data. This sensor is a 12-channel, six-frequency, conically scanning passive microwave radiometer mounted on the Aqua satellite [Lobl et al., 2003; Kawanishi et al., 2003].
The findings of this study indicated that snow-covered area has exhibited a decreasing trend in all months; however, in some months these changes were not statistically significant. Similar results were reported by Banihabib et al. [2013] for the Shah-Cheraghi Dam basin. In terms of temporal variation, snow-covered area has declined markedly in all three basins since 2009, which is consistent with the findings of Mohammadi and Khoorani [2019]. Overall, the magnitude of snow cover reduction in the Karun and Dez basins was greater than in the Karkheh Basin. From a seasonal perspective, June experienced the most pronounced decrease in snow-covered area.
In contrast, analysis of changes in SWE revealed patterns that differed from those of snow-covered area. The greatest reduction in SWE in all three basins occurred in February, whereas the analysis of snow-covered area did not reflect a comparable level of change during this month. This indicates that in recent years, reductions in snow depth and the volume of water derived from snowmelt have been substantially greater than reductions in the areal extent of snow cover. Correlation analysis between snow-covered area, SWE, and basin discharge showed that the relationship between snow-covered area and discharge was stronger during warmer seasons. For example, in the Dez Basin, the correlation coefficient between snow-covered area and discharge was 0.265 in January but increased to 0.775 in June. This finding suggests that, particularly in the Dez Basin, the sensitivity of basin runoff to snow storage increases with rising temperatures.
Conversely, during colder seasons, this dependency weakens due to reduced snowmelt rates. The strongest correlation between snow-covered area and discharge was observed in the Karun Basin, which can be attributed to the presence of extensive high-elevation and snow-prone mountainous regions in this basin. To further enhance the robustness of the analysis, correlations between SWE derived from AMSR-E/AMSR-2 data and basin discharge were also examined. However, due to the relatively coarse spatial resolution of these datasets, SWE values extracted for June were close to zero in all three basins. A similar discrepancy between estimated and actual SWE values has also been reported by Soleimani et al. [2018]. Therefore, it is recommended that higher spatial resolution datasets be used for SWE analysis during warm seasons, when snow-covered area is limited.
Conclusion
Overall, during the study period, both snow-covered area and snow water equivalent exhibited decreasing trends in all months, with more pronounced reductions observed in the Karun and Dez basins. In addition, for most months, no significant relationship was found between snow-covered area and basin discharge.
Acknowledgements: -
Ethical Permission: -
Conflict of Interest: -
Author Contributions: Sherafat M (First Author), Main Researcher (50%); Yarahmadi D (Second Author), Introduction Writer (30%); Fathnia A (Third Author), Methodologist (10%); Mirhashemi H (Fourth Author), Statistical Analyst (10%).
Funding: -
Keywords:
References
1. Ansari H, S Marofi (2017). Snow water equivalent estimation using AMSR-E and GLDAS model (case study: basins of northwestern Iran). Journal of Water and Soil. 31(5):1497-1510. [Persian] [Link]
2. Banihabib MI, Jamali FS, Saghafian B (2013). Detection of the snow cover area using NOAA-AVHRR in Shahcheraghi Dam basin. Physical Geography Research Quarterly. 45(3):13-29. [Persian] [Link]
3. Bavera D, Bavay M, Jonas T, Lehning M, De Michele C (2014). A comparison between two statistical and a physically-based model in snow water equivalent mapping. Advances in Water Resources. 63:167-178. [Link] [DOI:10.1016/j.advwatres.2013.11.011]
4. Byun K, Choi M (2014). Uncertainty of snow water equivalent retrieved from AMSR-E brightness temperature in northeast Asia. Hydrological Processes. 28(7):3173-3184. [Link] [DOI:10.1002/hyp.9846]
5. Coll J, Li X (2018). Comprehensive accuracy assessment of MODIS daily snow cover products and gap filling methods. ISPRS Journal of Photogrammetry and Remote Sensing. 144:435-452. [Link] [DOI:10.1016/j.isprsjprs.2018.08.004]
6. Chang ATC, Foster JL, Hall DK, Goodison BE, Walker AE, Metcalfe JR, et al (1997). Snow parameters derived from microwave measurements during the BOREAS winter field campaign. Journal of Geophysical Research. 102(D24):29663-29671. [Link] [DOI:10.1029/96JD03327]
7. Dietz A (2013). Central Asian snow cover characteristics between 1986 and 2012 derived from time series of medium resolution remote sensing data [dissertation]. Universität Würzburg. [Link]
8. Hall DK, Riggs GA (2016). MODIS/Terra Snow Cover 8-Day L3 Global 500m SIN Grid, Version 6 [Internet]. Boulder: National Snow and Ice Data Center; [Unknown Cited]. Available from: https://nsidc.org/data/MOD10A2/versions/6 [Link]
9. Hüsler F, Fontana F, Neuhaus C, Jan Musial M, Wunderle S (2011). AVHRR archive and processing facility at the university of bern: a comprehensive 1 Km satellite data set for climate change studies. EARSeL eProceedings. 10(2):83-101. [Link]
10. Gao Y, Xie H, Lu N, Yao T, Liang T (2010). Toward advanced daily cloud-free snow cover and snow water equivalent products from Terra-Aqua MODIS and Aqua AMSR-E measurements. Journal of Hydrology. 385(1-4):23-35. [Link] [DOI:10.1016/j.jhydrol.2010.01.022]
11. Guyennon N, Valt M, Salerno F, Bruna A, Romano E (2019). Estimating the snow water equivalent from snow depth measurements in the Italian Alps. Cold Regions Science and Technology. 167:102859. [Link] [DOI:10.1016/j.coldregions.2019.102859]
12. Jamali S (2014). Hydropower vulnerability assessment in the face of climate change impacts case study: Karkheh river basin. Iranian Dam and Hydroelectric Powerplant. 1(2):25-37. [Persian] [Link]
13. Kawanishi T, Sezai T, Ito Y, Imaoka K, Takeshima T, Ishidoet Y, et al (2003). The advanced microwave scanning radiometer for the earth observing system (AMSR-E), NASDA'S contribution to the eos for global energy and water cycle studies. IEEE Transactions on Geoscience and Remote Sensing. 41(2):184-194. [Link] [DOI:10.1109/TGRS.2002.808331]
14. Kendall MG (1970). Rank correlation methods. 2nd Edition. New York: Hafner. [Link]
15. Kelly R, Foster J, Tedesco M (2004). AMSR-E/Aqua Daily L3 Global Snow Water Equivalent EASE-Grids, Version 2 [Internet]. Boulder: NASA National Snow and Ice Data Center; [Unknown Cited]. Available from: https://nsidc.org/data/ae_dysno/versions/2 [Link]
16. Klein AG, Hall DK, Riggs GA (1998). Improving snowcover mapping in forests through the use of a canopy reflectance model. Hydrological Processes. 12(10-11):1723-1744.
https://doi.org/10.1002/(SICI)1099-1085(199808/09)12:10/11<1723::AID-HYP691>3.0.CO;2-2 [Link] [DOI:10.1002/(SICI)1099-1085(199808/09)12:10/113.0.CO;2-2]
17. Langlois A, Scharien R, Geldsetzer T, Iacozza J, Barber DG, Yackel J (2008). Estimation of snow water equivalent over first-year sea ice using AMSR-E and surface observations. Remote Sensing of Environment. 112(9):3656-3667. [Link] [DOI:10.1016/j.rse.2008.05.004]
18. Lobl ES, Spencer RW, Shibat A, Imaoka K, Sasaki M, Kachi M (2003). Global climate monitoring with the Advanced Microwave Scanning Radiometer (AMSR and AMSR-E). Microwave Remote Sensing of the Atmosphere and Environment III. 4894. [Link] [DOI:10.1117/12.466518]
19. Mann HB (1945). Nonparametric tests against trend. Econometrica. 13:245-259. [Link] [DOI:10.2307/1907187]
20. Mhawej M, Faour G, Fayad A, Shaban A (2014). Towards an enhanced method to map snow cover areas and derive snow-water equivalent in Lebanon. Journal of Hydrology. 513:274-282. [Link] [DOI:10.1016/j.jhydrol.2014.03.058]
21. Mohammadi Ahmadmahmoudi P, Khoorani A (2019). Snow cover changes of zagros range in 2001-2016 using daily data of MODIS. Journal of Earth and Space Physics. 45(2):355-371. [Persian] [Link]
22. Shan LU, Kazuo OKI, Kenji OMASA (2005). Mapping snow cover using AVHRR NDVI 10-daycomposite data. Journal of Agricultural Meteorology. 60(6):1215-1218. [Link] [DOI:10.2480/agrmet.1215]
23. Soleimani K, Darvishi S, Shokrian F, Rashidpour M (2018). Spatial-temporal monitoring of snow cover in Kurdistan province using MODIS images. ranian Remote Sensing & GIS Society. 10(3):104-77. [Persian] [Link]
24. Yang J, Jiang L, Ménard CB, Luojus K, Lemmetyinen J, Pulliainen J (2015). Evaluation of snow products over the Tibetan Plateau. Hydrological Processes. 29(15):3247-3260. [Link] [DOI:10.1002/hyp.10427]
25. Zhou H, Aizen E, Aizen V (2013). Deriving long term snow cover extent dataset from AVHRR and MODIS data: Central Asia case study. Remote Sensing of Environment. 136:146-162. [Link] [DOI:10.1016/j.rse.2013.04.015]