Why has global and regional snow cover declined in recent years? What is the solution? Can humans take action to stop it? What role can international organizations play in this regard?

Snow cover affects the global surface energy balance and, with its high albedo, exerts a cooling effect on the Earth’s climate. Decreases in snow cover alter the flow of solar energy from being reflected away from Earth to being absorbed, increasing the Earth’s surface temperature. To gain a global understanding of snow cover change, in situ measurements are too few and far between, so remotely sensed data are needed. This research used the medium-resolution sensor MODIS on the Terra satellite, which has been observing global snow cover almost daily since the year 2000. Here, the MOD10C2 eight-day maximum value composite time series data from February 2000 to March 2023 were analyzed to detect global and regional trends in snow cover extent for the first 23 years of the 21st century. Trends in snow cover change during different time periods (seasons and snow-year) were examined using the Mann—Kendall test and the univariate differencing analysis. Both methods produced similar results. Globally, snow cover declined two to ten times as much as it increased, depending on the season of analysis, and annually, global snow cover decreased 5.12% (not including Antarctica or Greenland) based on the Mann—Kendall test at the 95th percentile (p < 0.05). Regionally, Asia had the greatest net area decline in snow cover, followed by Europe. Although North America has the second-largest extent of snow cover, it had the least amount of net decreasing snow cover relative to its size. South America had the greatest local decline in snow cover, decreasing 20.60% of its annual (snow-year) snow cover area. The Australia–New Zealand region, with just 0.34% of the global snow cover, was the only region to have a net increase in snow cover, increasing 3.61% of its annual snow cover area.Snow and ice influences the global surface energy balance and hydrologic cycle as well as modifying feedbacks that control these aspects of the world’s climate [1]. When snow cover melts, less solar radiation reflects to space and more of this energy is absorbed by the Earth, and this snow cover feedback is a probable contributor to the polar amplification, which is further warming Arctic regions [2–5]. Various studies have explored the relationship between snow and ice cover, Earth’s albedo, and surface energy [6,7], showing that when snow cover declines, albedo decreases and the Earth’s surface warms [8–10]. Snow also provides a critical short-term water storage mechanism for many people around the world [11]. Snow cover extent (SCE) has been declining in many parts of the world [8,12,13]. These observed changes in SCE are a direct response to climate change, mainly from warming temperatures and changes in precipitation [14–17]. Climate projections indicate that in the future, SCE will continue shrinking [14], enhancing the snow cover feedback and leading to warmer temperatures [12,15]. Even though SCE globally has decreased in the last four decades, there has been considerable inter-annual variability [18]. Anomalously cold periods and large snowfalls in recent winters have been experienced in North America, Asia, and Europe [19], leading to increasing SCE for some areas [20]. Indications are that the quick warming of the Arctic is associated with changes in atmospheric circulation [21,22] and may be responsible for these anomalous events and areas of increasing SCE.Monitoring SCE change in many parts of the world is difficult due to the lack of local observing networks, snow cover’s spatial variability due to local conditions and other physiographic characteristics, frequent cloud cover, and confusion between lake ice and snow cover during the melt season [12]. In situ data provides detailed local data but remains extremely sparse across the globe. With satellite imaging, snow cover recognition is becoming more precise. There are now multiple satellite-based sensors, from thermal and microwave sensors to visible light sensors, that can capture snow cover. This research uses data from the Moderate Resolution Imaging Spectroradiometer (MODIS), which observes snow surface properties using solar illumination (visible and infrared wavelengths) in cloud-free periods and has been shown to be excellent for mapping SCE and duration [23,24]. The MODIS snow products have been validated by many studies [25–29], such as Hall and Riggs (2007), who found an overall absolute accuracy of the base 500 m resolution data to be about 93% [24]. This study uses MODIS data because it now has a greater than 23-year time span of consistent global-scale snow data, which have been used in numerous snow cover studies [8,10,16,20,25,30]. The MOD10C2 data set was used because it is a consistent data set, which minimizes cloud cover contamination, has over 23 years of reliable data, has been used in many other SCE studies [8,31–33] and has been validated, such as by Lei et al. (2011) who evaluated the snow identification accuracies of the MOD10C2 data to station data in northeast China and found the accuracy to be greater than 88%, with cloud cover being the main problem [34]. Different remote sensing studies have shown that SCE is broadly declining in different regions across the globe. Hammond et al. (2018) mapped global snow zones with MODIS data and found that between 2001 and 2016 for areas of snow cover, 5.8% declined in snow persistence while 1.0% increased in snow persistence. They also found that declining trends were greatest in the winter months [30]. Notarnicola (2020), using MODIS snow products, studied hot spots of snow cover change in mountain regions across the globe between 2000 and 2018 and found that 78% of observed areas were affected by snow decline and a snow cover area decrease of up to 13%, while above 4000 m only negative changes were observed [35]. In another publication, Notarnicola (2022) used a combination of snow cover data sets including MODIS and found that between 1982 and 2020 over global mountain areas an overall negative trend of −3.6% ± 2.7% for yearly SCE [36]. The season most affected by negative trends was winter. While most mountain ranges had negative trends in SCE, like the Alps, some mountain ranges, such as those in the northern highmountain Asia region (Karakorum, Kunlun Shan, Pamir Mountains) had positive trends. Because the Northern Hemisphere has the highest percent (98%) of the world’s snow cover between the Arctic and Antarctic circles [30], most snow cover studies have focused on the Northern Hemisphere. Kunkel et al. (2016) studied trends in extreme SCE in the Northern Hemisphere based on satellite observations for 1967 to 2015 and found an overall negative trend in SCE [37]. Using a multi-source remote sensing data set, Wang et al. (2018) found that between 2000 and 2015, the maximum, minimum, and annual average SCE in the Northern Hemisphere exhibited a fluctuating downward trend [38]. Using MODIS and AVHRR data, Hori et al. (2017) found an average decrease in SCE of 10 days/decade in the Northern Hemisphere since 1978 [39]. Using MODIS data, Eythorsson et al. (2019) estimated a decrease in Arctic snow cover frequency of 9.1 days/decade since 2001 [40]. Hernández-Henríquez et al. (2015) examined different latitudes and elevations of SCE declines in the Northern Hemisphere between 1971 and 2014 based on the NOAA snow chart climate data record and found the majority of statistically significant negative trends in the mid- to high latitudes [41]. Brown et al. (2021) analyzed snow cover trends for Canada (1955–2017) using the daily snow-depth-observing network of Environment and Climate Change Canada (ECCC) where results are broadly similar to previously published assessments showing long-term decreases in annual snow cover duration and snow depth over most of Canada [42]. Some large regions of snow cover have stayed relatively stable. Wang et al. (2017) used MODIS data and found no widespread decline in snow cover over on the Tibetan Plateau from 2000 to 2015 [43].Many SCE studies have focused on snow onset dates and snow end dates [15] as well as the Northern Hemisphere’s spring season [12,14,18,44] because with a high sun angle, spring snow in northern Canada, Alaska, and Siberia reflects extensive energy back to space that would otherwise potentially be absorbed and heat the planet further [2]. Shi et al. (2013), using the NOAA weekly snow cover maps, found that late spring/early summer SCE significantly decreased over the Arctic between 1972 and 2006 [45]. Derksen and Brown (2012), using the NOAA snow chart climate data records from April to June, found the Eurasia region set successive records for the lowest June SCE every year from 2008 to 2012 while North America set the June record 3 out of the 5 years (2008-2012) [46]. They also found the rate of loss of June snow cover extent between 1979 and 2011 (−17.8% decade−1 ) is greater than the loss of September sea ice extent (−10.6% decade−1 ) over the same period [8]. Brown and Robinson (2011), using the NOAA weekly SCE dataset, found that Northern Hemisphere spring SCE has undergone significant reductions over the past 90 years and that the rate of decrease has accelerated over the past 40 years. They also found that Eurasia had a significant earlier spring snow melt (March) than North America [14]. Musselman et al. (2021), using station data, found that in western North America (30 years+) snowmelt is increasing during the snow accumulation season [11]. This research differs from previous studies in that it analyzes changes in persistent SCE decline and increase globally and regionally. Although snow cover regions are quickly warming, changes in snow cover vary by region [14,18,21,47] and it is important to study the regional variation in a global context. There are few global scale SCE studies which focus on annual and seasonal areas increasing and decreasing the fastest. This study looks at areas of persistent decline and persistent increases in SCE globally and regionally from 2000 through 2022, and 2023 for the winter season. Most SCE change studies have generally focused on the spring season, when higher snow albedo feedbacks occur [14,46,48], but this study analyzes all four seasons and the snow-year (Northern Hemisphere: September to August of the following year, Southern Hemisphere: March to February of the following year). This study uses the Mann—Kendall test to analyze persistent changes in SCE globally and regionally as well as applying the univariate differencing analysis to determine SCE change between the beginning and end of the study period. The Mann—Kendall test is a analytical tool commonly used to analyze changes in snow cover [8,11,35,40,41] while the univariate differencing analysis for change analysis is also used, but not as much as the Many SCE studies have focused on snow onset dates and snow end dates [15] as well as the Northern Hemisphere’s spring season [12,14,18,44] because with a high sun angle, spring snow in northern Canada, Alaska, and Siberia reflects extensive energy back to space that would otherwise potentially be absorbed and heat the planet further [2]. Shi et al. (2013), using the NOAA weekly snow cover maps, found that late spring/early summer SCE significantly decreased over the Arctic between 1972 and 2006 [45]. Derksen and Brown (2012), using the NOAA snow chart climate data records from April to June, found the Eurasia region set successive records for the lowest June SCE every year from 2008 to 2012 while North America set the June record 3 out of the 5 years (2008-2012) [46]. They also found the rate of loss of June snow cover extent between 1979 and 2011 (−17.8% decade−1 ) is greater than the loss of September sea ice extent (−10.6% decade−1 ) over the same period [8]. Brown and Robinson (2011), using the NOAA weekly SCE dataset, found that Northern Hemisphere spring SCE has undergone significant reductions over the past 90 years and that the rate of decrease has accelerated over the past 40 years. They also found that Eurasia had a significant earlier spring snow melt (March) than North America [14]. Musselman et al. (2021), using station data, found that in western North America (30 years+) snowmelt is increasing during the snow accumulation season [11]. This research differs from previous studies in that it analyzes changes in persistent SCE decline and increase globally and regionally. Although snow cover regions are quickly warming, changes in snow cover vary by region [14,18,21,47] and it is important to study the regional variation in a global context. There are few global scale SCE studies which focus on annual and seasonal areas increasing and decreasing the fastest. This study looks at areas of persistent decline and persistent increases in SCE globally and regionally from 2000 through 2022, and 2023 for the winter season. Most SCE change studies have generally focused on the spring season, when higher snow albedo feedbacks occur [14,46,48], but this study analyzes all four seasons and the snow-year (Northern Hemisphere: September to August of the following year, Southern Hemisphere: March to February of the following year). This study uses the Mann—Kendall test to analyze persistent changes in SCE globally and regionally as well as applying the univariate differencing analysis to determine SCE change between the beginning and end of the study period. The Mann—Kendall test is a analytical tool commonly used to analyze changes in snow cover [8,11,35,40,41] while the univariate differencing analysis for change analysis is also used, but not as much as the Mann—Kendall test [8,49,50]. The objectives of the study are: (1) to analyze persistent trends in SCE throughout the four different seasons and annually (snow-year) at both the global and regional scales; and (2) to analyze changes in SCE through univariate differencing between the beginning (average of 2000 to 2004) and end (average of 2018 to 2022) of the period. The novelty of this research, in addition to studying the four seasons and snow-year, is that it uses the Z-values of the Mann—Kendall test to show the intensity of significant changes (p < 0.05, 0.01) as well as using the Mann—Kendall test significant values to filter the results of the univariate differencing analysis. 2. Materials and Methods 2.1. MOD10C2 Data Set The MODIS/Terra Snow Cover 8-Day L3 Global 0.05 Degree Climate Modelling Grid (CMG), Version 61 (MOD10C2) data were used to map snow cover in this study. The MOD10C2 data have a MODIS sinusoidal projection, which is an equal-area projection and is appropriate for the global and regional analysis of SCE. The MOD10C2 data set has a spatial resolution of 0.05◦ and is an aggregation of MOD10A2 products with 500 m spatial resolution, which is an 8-day composite of MOD10A1 daily SCE maps [26]. The data were downloaded [6 January 2023 and 21 April 2023] from NASA’s Earth Data web site (https://search.earthdata.nasa.gov/search). Version 61 data were downloaded and processed for the whole period from 24 February 2000 to 1 March 2023.The 8-day composite is considered useful because persistent cloudiness limits the number of days available for surface observations in many regions, particularly at high latitudes [1]. However, MODIS snow products have some issues concerning cloud cover, the difficulty of detecting snow in forest areas, and topography that may impact the accuracy of the results [25,51,52]. Concerning extreme topographic changes, such as in mountainous terrain, several studies suggest that the use of MODIS snow cover products is particularly suitable to address the challenges in mountain areas [30,43,53]. The MOD10C2 product was chosen because the data set was developed for global studies and has been used in numerous studies [8,31,54–56]. The alternative MODIS global product is MOD10A1, which has a finer spatial resolution (500 m) and temporally resolution (daily), but this data set still has large data gaps and other issues [57] that require further processing. MOD10C2 pixel values are the maximum percentage of snow cover (0% to 100%) for the pixel’s area for eight continuous days [23]. The snow cover product is a Normalized Difference Snow Index (NDSI) created from MODIS band 4 (green) (0.545 µm to 0.565 µm) and band 6 (near-infrared) (1.628 µm to 1.652 µm). Cloud-cover contamination is the greatest deficiency of the MOD10C2 data set, while other issues, such as errors of commission, have been found to be very low [24]. Data were processed into seasonal and annual averages for December–January– February (12-01-02), March–April–May (03-04-05), June–July–August (06-07-08), September– October–November (09-10-11), and annually based on the snow-year or hydrologic-year (Northern Hemisphere: September to August of the following year, Southern Hemisphere: March to February of the following year) [20]. Because persistent cloud cover is the major issue with the MOD10C2 data, for this research snow analysis masks were created for each season and annually. An “analysis mask” is a means of identifying areas to be included in analysis. For each season (and annually), the maximum value for all 8-day snow cover files from 2000 to 2023 were calculated. The resulting maximum value snow cover file for each season (and annually) was then turned into a snow analysis mask where values of 100 = 1 and all other values = 0. The resulting snow analysis mask shows where there was at least one day of 100% snow cover in each season, and annually, between 2000 and 2023. These snow analysis masks were created so that for each season, only data within the masked area were analyzed; this way, clouds outside of the potential snow areas would not contaminate the data. Clouds can cover areas of snow cover which can contaminate the data. To reduce cloud contamination, the MODIS cloud cover layer was added to the MODIS snow cover layer and when the 8-day cloud cover pixels plus 8-day snow cover pixels equaled a value of 100% and were completely surrounded by snow cover pixels with a value of 100%, it was assumed that clouds were covering 100% snow cover and for that 8-day period the pixels were classified as 100% snow cover. All other cloud pixels within the seasonal (annual) snow analysis masks were considered cloud contamination. A review of cloud contamination was performed on every 8-day snow–cloud combined file for the first and last five years of the data set and found the average of cloud contamination to be 2.9%. Because both Greenland and Antarctica are both primarily ice covered, these two regions were dropped out of the data set. Also, Greenland and Antarctica are in the high latitudes and thus a low sun angle during much of the year poses problems for global optical satellite data and analysis. 2.2. Mann—Kendall Test This research used the Mann—Kendall test to determine if SCE values were increasing, decreasing, or staying the same for the global area of snow cover. The Mann—Kendall test is used to statistically assess if there is a monotonic upward or downward trend of SCE over time (2000–2023) [58–60]. A monotonic upward (downward) trend means that the variable consistently increases (decreases) through the period of the data set. The Mann—Kendall test analyzes the sign of the difference between later-measured data and earlier-measured data. There is no requirement that the measurements be normally distributed or that the trend, if present, be linear [61].From this research and that of others, it is clear that the world is quickly losing its snow cover. This study used global scale snow cover data (MOD10C2) from February 2000 to March of 2023 to analyze how snow cover has changed over the first 23 years of the 21st century. Two methods were used to analyze the data, univariate differencing and the more commonly used Mann—Kendall test. Novel to this research is the use of the Mann—Kendall Z-value (standard deviations of change from the norm) to determine intensity of snow cover change and the use of the Mann—Kendall p-value (p < 0.05) to filter the results of the univariate differencing.The first 23 years of the 21st century show that changes in snow cover continue along a path of decline, which has been happening for the past 100 years. During the past 23 years, snow cover has been broadly disappearing around the world, especially in Asia, Europe, and South America, along with areas in North America and Africa. Snow cover in Asia is declining faster than the global average, and snow cover loss is extensive throughout much of the region. Europe and South America are also losing snow cover faster than the global average. North America, the world’s second-largest area of snow cover, is experiencing a slower decline with multiple areas of increase. Despite its slower pace, snow cover is also declining in North America and quickly in the New England region. Snow cover in Africa is not as concentrated as in other parts of the world, but the snow cover here is also declining. The only region with more increasing snow cover than decreasing was the Australia–New Zealand region, but the increases were slight and with the regional variation the future direction of change can go in either direction. With global greenhouse gas emissions continuing at a record pace [104], the trend of decreasing SCE will continue. The next step in this research will be to explore the relationship between land surface temperature and changes in SCE, using the MOD11C3 land surface temperature data with the MOD10C2 snow cover data used in this paper, and potentially also exploring the use of the MOD10A1 data.