Regional differences in global glacier retreat from 1980 to 2015

LI Yo-Jun, DING Yong-Jin,*, SHANGGUAN Dong-Hui, WANG Rong-Jun

a State Key Laboratory of Cryospheric Science, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou, 730000, China

b Key Laboratory of Ecohydrology of Inland River Basin,Northwest Institute of Eco-Environment and Resources,Chinese Academy of Sciences,Lanzhou,730000,China

c University of Chinese Academy of Sciences, Beijing, 100049, China

Abstract Global warming triggers shrinking and thinning of glaciers worldwide, with potentially severe implications for human society. However,regional differences in glacier retreat and its relationship with climatic characteristics have not been conclusively demonstrated. In this study,regional changes in global glaciers based on two primary features,area change and mass balance,were investigated on the basis of data collected from published research on glacier changes.Results show that during the period 1980-2015,the rate of global glacier area shrinkage was 0.18%per year and that of global glacier mass loss was 0.25 m w.e.per year.Retreat of glaciers located at low and middle latitudes was characterized by severe area shrinkage and mass loss. Correspondingly, in the Arctic, deglaciation was characterized by ice thinning due to a low area reduction but relatively high mass loss rate.However,glaciers in high southern latitudes were in a relatively stable status.High Mountain Asia exhibited the lowest rate of area shrinkage and mass loss among glaciers located at low and middle latitudes, and a slower rate of mass loss compared with the global average.Glaciers in the Tropical Andes exhibited the fastest rate of glacier area shrinkage(-1.6%per year),whereas Antarctic and Subantarctic glaciers showed the lowest rate(-0.11%per year).For mass balance,the most negative occurred at Southern Andes(-0.81 m w.e.per year),followed by Alaska(-0.74 m w.e.per year).Only the Antarctic and Subantarctic experienced small mass gain(0.04 m w.e. per year). High levels of correlation are found between the rates of glacier retreat and annual average temperature and annual total precipitation instead of their trends. The variability of the surface climate conditions in the glacier environment plays a key role in driving these regional differences in global glacier retreat.

Keywords: Glacier retreat; Regional differences; Climate change

1. Introduction

Glaciers are key indicators of climate change because they exhibit high sensitivity to climate fluctuations (Paul et al.,2015). According to the latest Randolph Glacier Inventory(RGI 6.0) (RGI Consortium, 2017), glaciers, including mountain glaciers and ice caps (excluding the Antarctic and Greenland ice sheets), cover an area over 700,000 km2globally, or approximately 0.5% of the Earth's land surface. As a critical component of the Earth system,glaciers are among the most dynamic elements of the solid Earth(Kargel et al.,2014).With regard to global warming, academic attention has focused on how glaciers change and what kind of impacts those changes have. Previous studies revealed that global warming results in shrinking and thinning of glaciers worldwide (Qin and Ding, 2010; IPCC, 2019). Accelerated glacier melting leads to serious implications for downstream systems(Milner et al., 2017; Wang et al., 2019), contributes to sealevel rise (Gardner et al., 2013), and regulates global and regional water resources (Huss and Hock, 2018).

Area changes and mass balance of glaciers reflect prevailing climatic conditions surrounding the glacier (Marzeion et al., 2018; Paul et al., 2015; Qin and Ding, 2010; Roe et al., 2016). Glacier area shrinkage and mass loss have been documented for individual glaciers as well as for global scales using in-situ observations, geodetic methods, satellite derived data, and model simulations (e.g., Marzeion et al.,2012; Shangguan et al., 2015; IPCC, 2019; Zemp et al.,2015). In fact, glacier retreat is determined by the energy balance at the glacier-atmosphere interface,controlled by the surrounding climate conditions and characteristics of the glacier itself(Hock,2005).However,significant heterogeneity exists in global surface climate dynamics (Ying et al., 2015).Bahr et al. (2009) demonstrated that the retreating profile of glaciers is a direct consequence of climate, which can vary regionally and locally as a result of orographic and other meteorological factors.Differences in glacier orientation,size,and topography also lead to variability in glacier sensitivity and response to climate change (Bach et al., 2018; Qin et al.,2018). Systematic regional differences in glacier condition further complicate the understanding of glacier responses to local and regional climate. For example, K¨a¨ab et al. (2012)studied remote-sensing data to reveal contrasting patterns of mass balance in Himalayan glaciers. A large-scale study by Yao et al.(2012)indicated that the evolution of glaciers in the Tibetan Plateau and surrounding area differed systematically across regions. To date, however, studies on regional characteristics of global glacier retreat have focused mainly on mass balance (e.g., Gardner et al., 2013; Kaser et al., 2006;Marzeion et al., 2015), with few studies examining regional differences in glacier area shrinkage (Mu et al., 2018).Furthermore, synthesized research on global glacier retreat,which combines area shrinkage and mass loss,is lacking.The Special Report on the Ocean and Cryosphere in a Changing Climate(SROCC)by the Intergovernmental Panel on Climate Change (IPCC) reviewed available mass-balance estimates of mountain glaciers but did not provide assessments of glacial area change worldwide (IPCC, 2019; Kang et al., 2020).Studies using a synthesis approach,which combines estimates of area change and mass balance to investigate global glacier retreat, are urgently needed to enhance our understanding of the regional differences in glacier response to climate change.

In 2000, the World Climate Research Program started its core project called “Climate and Cryosphere” with a key goal of increasing quantitative assessments of glacier changes caused by increasing temperatures (Qin et al., 2018). To understand glacier vulnerabilities and the potential impacts of glacier retreat, it's vital to investigate the effects on regional and local scales on account of the significant connection between glacier change and human society (Qin et al., 2018;Xiao et al., 2015). Therefore, the main objective of this study is to provide a comprehensive analysis of worldwide glacier retreat characteristics at regional scales.This objective is achieved by summarizing existing research on the regional magnitudes of glacier area shrinkage and mass balance from 1980 to 2015. The corresponding temporal and spatial differences over the investigated period are determined for each region and worldwide, and the relationships between glacier retreat and climatic factors(temperature and precipitation)are analyzed.This study aims to increase the understanding of the ongoing glacier retreat, and aid efforts at finding reasonable ways to adapt to climate change. Besides, one major issue in early sea-level assessments and hydrological effects is the ignorance of the impact of glacier area changes on regional mass-change estimates (Kaser et al., 2010; Pritchard, 2019;Zemp et al., 2019). The results of area change rates for all 19 RGI glacial regions contribute toward improvement of mass-change assessments of glaciers worldwide.

2. Materials and methods

2.1. Glacier inventory data

RGI is a collection of digital outlines of global glaciers,excluding the Greenland and Antarctic ice sheets (RGI Consortium, 2017), as a supplement to the Global Land Ice Measurements from Space (GLIMS) project (Raup et al.,2007), which is motivated by the IPCC Fifth Assessment Report (IPCC, 2013). In this study, we obtained RGI 6.0 with 0.5°× 0.5°grid data as shown in Fig. 1 from GLIMS (www.glims.org)to exhibit glacier distribution worldwide as well as to calculate selected regional average climatic factors (i.e.,temperature and precipitation). The grid data, which are provided as plain-text DAT files, hold zonal records of blankseparated glacierized areas in km2in each 0.5°× 0.5°cell of a latitude/longitude grid (RGI Consortium, 2017).

2.2. Glacial area change

Glaciers are often inaccessible due to their remote locations and complex distribution. Thus, remote-sensing images are widely used to analyze changes in glacier areas (Paul et al.,2015). We searched the literature to find changes in glacier areas in 19 RGI regions(as shown in Table 1).Following Yao et al. (2012), we used weighted averages to calculate the global and regional average rates of glacier area change during the period 1980-2015, as follows:

where R was the regional average change rate in the glacier area, Piwas annual change rate in glacier area in each study(%),and Aiwas the area of glaciers in each research region in 1980 (km2) (Table 1). All selected studies for analysis should satisfy the criteria that the time series had to include years beyond 2000 to ensure that the result could reflect the latest status of the glacier area change.Given different time spans in selected studies, we first assumed that the glacier area experienced a linear change during the investigated period in every study following Zemp et al. (2013). Then, we obtained the glacier area for 1980 (Table 1) in every selected region through calculation based on the most recent area after 2000 and the changing rate. Finally, we obtained the regional average change rate following Eq. (1). However, evaluating uncertainty in the regional average results was difficult because data were obtained using a wide range of available sensors and archives from ongoing and historical research efforts.

2.3. Mass balance data

Glacier mass balance is the key link between glaciers,hydrology, and water resources. Until recently, records of insitu measurements of annual mass balance existed for only≤0.5% of the glaciers worldwide. Our analysis of the global and regional mass balance was based on the latest dataset containing all available direct and geodetic mass balance measurements developed by Zemp et al. (2019). This dataset has been used as a benchmark to evaluate global glacier mass loss in SROCC (IPCC, 2019). Uncertainty estimates for the dataset were obtained from a comprehensive error analysis(Zemp et al., 2019). The density of ice is assumed to be 850 kg m-3.

2.4. Meteorological data

Monthly ERA-Interim(Dee et al.,2011)output was used to investigate the trends in annual mean temperature and total precipitation in 1980-2015.Following Jones et al.(2017),we regridded the ERA-Interim data onto 0.5°× 0.5° resolution grid using the bilinear interpolation method by Climate Data Operator (Schulzweida, 2019). We derived the trends in annual average temperature and annual total precipitation globally based on linear regression.

Previous studies have investigated the response of glaciers to climate regime on 0.5°× 0.5°resolution, in which they assumed that the climatic data is consistent in each grid(Sakai and Fujita,2017;Sakai et al.,2015).Averages of each climatic variable for each glacial region were calculated using weights dependent on the glacierized areas (in km2) in each 0.5°× 0.5°cell (color depth in Fig. 1), as follows:

where ΔW is the regional average of each climatic variable(annual mean temperature, annual total precipitation, linear trend of annual mean temperature and annual precipitation),and aiis the glacierized area in each cell in km2(Fig. 1).

3. Results

3.1. Global patterns of glacier area change on regional scale

Our findings indicate that the overall glacier area in the world has been shrinking at an average rate of 0.18%per year during 1980-2015. The magnitude of the decrease in the glacier area for different regions varied considerably (Fig. 1).Generally,glaciers located at low-latitudes(0°-30°)exhibited a higher rate than those at mid-latitudes (30°-60°), and highlatitude (＞60°) glaciers exhibited the lowest rates in the Northern and Southern Hemispheres.

Fig. 1. Annual rates of glacier area change during 1980-2015.

Table 1 Retreating rate of glacier areas for each RGI region (References are shown in Table S1).

Table 1 (continued)

The Tropical Andes,defined as the Low Latitudes region in RGI,is the location of more than 99%of all glaciers found in tropical areas. In this region, glaciers are critical for local water budgets, agricultural production, hydropower generation, and ecosystem health (Vergara et al., 2007). Andean cities at altitudes above 2500 m are highly dependent on water supply from the glacier to offset water shortage during the dry season(Bradley et al.,2006).In fact,the majority(almost twothirds) of Peru's population living in the relatively dry Pacific side basin rely on the Andean glaciers for most of the water resources (Drenkhan et al., 2015). These glaciers were retreating at the fastest rate (-1.6% per year) in the world.

The middle latitude in the Northern Hemisphere, including Central Europe, Western Canada and U.S., also experienced severe glacier shrinkage. Our findings show a rate of approximately 1.2% per year in glacier area shrinkage, indicating that the Alpine glacier retreat in Europe occurred at one of the highest rates worldwide since 1980. The annual area shrinkage rate in Western Canada and U.S. was 0.53% per year during the period 1980-2015. However, some subregions, such as Central Coast and Wind River Range,reached over 1%per year(Bolch et al.,2010;Thompson et al.,2011).In the Southern Hemisphere,the glacier area located in New Zealand experienced a shrinkage rate of 0.69% per year.Although the glacial area of the Southern Andes decreased at a rate of 0.18% per year,all sub-region glaciers shrank at a rate over 0.5% per year, except those in the Patagonian ice field,which is the third largest ice body in the world (Table 1).

High-latitude glaciers, especially at the pole regions,exhibited the lowest rates of glacier area shrinkage. Thus far,the glacier area in Greenland Periphery,Arctic Canada North,Arctic Canada South, and Antarctic and Subantarctic has decreased by 0.1%per year.In addition,the glacier area in the high latitudes of the Northern Hemisphere, including Alaska,Scandinavia, Iceland, Russian Arctic, and Svalbard, has exhibited a shrinkage rate of no more than 0.3% per year.

The glacier area changes are greatly governed by the glacier size, which means that small glaciers tend to suffer higher area reduction than large ones (Wang et al., 2009; Ye et al., 2003). These small glaciers are located in Central Europe and Tropical Andes, but the larger glaciers, such as those located in the polar region, tend to exhibit smaller rates of area reduction.

Fig.2.Annual mass balance of global glaciers in 1980-2015 based on Zemp et al. (2019).

3.2.Global patterns of glacier mass balance on regional scale

The compilation of direct and geodetic observations showed negative glacier mass balance around the world except in Antarctic and Subantarctic during 1980-2015 (Figs. 2 and 3). Glacier mass loss reached an average rate of 0.25 m w.e.per year globally.Low mass loss rate in the 1980s(0.07 m w.e.per year) was followed by a continual acceleration to 0.25 m w.e. per year in the 1990s and 0.37 m w.e. per year in the 2000s. The maximum mass loss was 0.52 m w.e. per year in 2015.

Except for Antarctic and Subantarctic, which exhibited indications of stability or small mass gain in glaciers, all regions had a glacier mass deficit (Fig. 3) in 1980-2015, according to Zemp et al. (2019). The change in glacier mass balance was highly variable across the globe. The most negative mass balances occurred in Southern Andes (0.81 m w.e. per year), followed by Alaska (0.74 m w.e. per year) and Tropical Andes (0.67 m w.e. per year), whereas the loss rate was lowest in South Asia West (0.04 m w.e. per year). However, the Antarctic and Subantarctic glaciers, which include most of the global glaciers (Farinotti et al., 2019), showed a slight mass gain at a rate of 0.04 m w.e. per year during the period 1980-2015. The rate of mass loss in High Mountain Asia (HMA, including Central Asia, South Asia West, and South Asia East) glaciers was the lowest at the mid-low latitudes and a much slower rate of glacier thinning compared with the global average.

Fig. 3. Annual mass balance of global glaciers in 1980-2015.

In contrast to the shrinkage in the glacier area, the worldwide glacier mass balance did not exhibit regional characteristics. Mass balance differences among RGI regions were relatively lower than those for the rate of glacier area shrinkage. Alaska, which is located at high latitudes in the Northern Hemisphere, showed the second fastest mass loss globally, a more negative rate compared with the glaciers in Western Canada and U.S.,which are in the middle latitudes of the Northern Hemisphere. In the Southern Hemisphere, the Southern Andes also showed a more negative mass balance compared with the glaciers in Tropical Andes.

3.3. Regional differences in glacier change

Our findings reflect the fact that glaciers have lost a considerable portion of their area and mass on a global scale.At low to middle latitudes, glacier retreat is characterized by severe area shrinkage and mass loss. Glaciers in the Tropical Andes showed the fastest rate of area shrinkage,whereas those in the Southern Andes has the most negative mass balance worldwide at a rate of 0.9 m w.e. per year. High rates of area shrinkage were also observed in the Alps,Western Canada and U.S., where the glacier mass loss rate reached approximately 0.6 m w.e. per year; this rate is relative higher to the global average level. At high latitudes, especially in the pole regions and Alaska, deglaciation was characterized by thinning or a low-level decrease in area but a high loss deficit.The Antarctic and Subantarctic, which account for almost one-fifth of the total glaciers worldwide, exhibited a minimum rate of glacier area shrinkage (-0.11% per year) and a slight mass gain(0.04 m w.e. per year) over the period 1980-2015.

Marked by broad societal concern due to the effect on water resources,which are essential to hundreds of millions of people inhabiting the regions(Chen et al.,2016;Immerzeel et al.,2010,2013; Sorg et al., 2012; Wang et al., 2019), glaciers in HMA featured a moderate reduction degree.However,compared with other regions in the middle to low latitudes,HMA exhibited the lowest rate of glacier area retreat(Fig.1).Moreover,the rate of mass loss of HMA glaciers was also the lowest among glaciers in the mid-low latitudes(Fig.3)and a much slower thinning rate compared with the global mean level.

4. Discussion

4.1. Relationship between glacier retreat and climate change

Earlier research(Chen et al.,2016;Qiu,2008)revealed that the climate in HMA has been warming since the mid-1950s,with temperature increasing by up to 0.3°C per decade,which was approximately three times that of the global rate. The relatively moderate glacier reduction, given such a significant temperature increase, can be attributed to low temperature(Fig. 4) caused by unique geographical and topographic conditions of the largest mountain region in the world with an average elevation over 4000 m.The Arctic region has warmed more than the rest of the world in what is known as Arctic amplification (Pithan and Mauritsen, 2014; Screen and Simmonds, 2010). However, Arctic glaciers have exhibited the lowest rate of area shrinkage in the world (Fig. 1). By contrast,the Tropical Andes,which experienced less warming than the Arctic and HMA (Fig. 5b), exhibited a higher rate of glacier shrinkage and glacier mass loss than any other region on Earth. The Tropical Andes has the highest air temperature among all the glaciated regions(Fig.4a)and this characteristic leads in part to the greatest glacier shrinkage in the world. A relatively warm environment can also explain the high rate of loss of the glacier area in the Alps and western North America.

Fig. 4. Surface climate and climate change across RGI regions during 1980-2015: (a) annual mean temperature; (b) temperature trend; (c) annual total precipitation; (d) precipitation trend.

Our findings indicate that the surface climate and climate change are heterogeneous across the RGI regions (Fig. 4).Earlier studies have shown that no simple relationship existed between the rate of glacier retreat and climate change over a long time scale (Leclercq et al., 2011; Marzeion et al., 2014).Furthermore, no significant difference has been observed in the mass loss of glaciers under greatly different warming scenarios (Marzeion et al., 2018). No simple linear relationship is found among any of the parameter comparisons between glacier retreat and climate change (Fig. 5). Therefore,we conclude that in the recent past, glacier retreat (including area shrinkage and mass loss) and climate change rates(temperature and precipitation) were not related on large regional scales(Fig.5b,d,f,and h).Moreover,we found high levels of correlation between the rates of glacier retreat and annual mean temperature and annual total precipitation instead of their trends, especially between the rate of glacier area shrinkage and annual mean temperature(Fig.5a).Notably,the change rate of area and mass balance have a significant negative correlation with annual total precipitation because glaciers in a wetter regime are more sensitive to climate warming (Oerlemans and Fortuin, 1992). Generally, glaciers under humid and warming conditions are characterized with high mass turnover. That's why the mass loss rates are positively correlated with both air temperature and precipitation(Fig. 5e and g). We conclude that glacier fluctuations are primarily driven by their surface climate conditions rather than the changing rate of climatic conditions. On a decadal and centurial time scale, glacier fluctuations are determined by meteorological conditions of the glacial environment(Marzeion et al., 2018; Oerlemans, 2005). Glacier melt is determined by the energy balance, which is controlled by the climatic conditions surrounding them, especially the temperature (Hock, 2003, 2005). Climate warming would affect the glacier fluctuations by changing the meteorological conditions of the glacial environment.

Furthermore,annual mean temperature is highly correlated with the average ice thickness significantly in each RGI region(Fig. 6), although almost all the glaciers are not in balance with the 1980-2015 climate regime. Glaciers in cold regions,such as those in the Arctic and Antarctic, are relatively thick(Farinotti et al., 2019); the retreat is characterized by ice thinning with a lower rate of area shrinkage.However,glaciers in warm regions, such as those in Low Latitudes and central Europe,are much thinner(Farinotti et al.,2017)so that glacier retreat involves mass loss and area shrinkage. In addition, the relationship between ice thickness of glaciers and their surface slope would be useful to understand the response of glaciers to climate change. Unfortunately at this point there are not enough sample data.In the future,employ the determine factor of ice thickness, more than climate background and topographic factor,is the research front.This study focuses on the regional differences in glacier retreat by combining estimates of area change and mass balance and their response to climate change.

Fig.5.Relationship between glacier retreat and climate change in 1980-2015,(a)annual mean temperature vs.area loss rate,(b)temperature trend vs.area loss rate,(c)annual total precipitation vs.area loss rate,(d)precipitation trend vs.area loss rate,(e)annual mean temperature vs.mass loss rate,(f)temperature trend vs.mass loss rate,(g)annual total precipitation vs.mass loss rate,and(h)precipitation trend vs.mass loss rate(For mass loss rate,a positive number indicates that the mass balance of glaciers is negative and vice versa).

The responses of glaciers to climate change are not identical in different regions of the world.Fig.5 provides a statistic that measures the degree to which glacier retreat and climate regime move in relation to each other on a global scale.Further work is needed, including higher-resolution (spatial and temporal) datasets on glacier change, reliable climatic data for glacial regions, and new methods such as model simulation, to determine mechanisms that affect the response of glaciers to climate change.

Fig. 6. Ice thickness (Farinotti et al., 2019) plotted with annual mean temperature in each RGI region.

4.2. Uncertainty analysis

Despite considerable progress in observation methods and spatial coverage,lack of data remains the greatest constraint in our knowledge about glacier retreat and its relationship with climate change (Gruop, 2018; Marzeion et al., 2017). Our efforts to detect regional differences in glacier retreat involve uncertainty due to the following factors:

- Uncertainty in the rate of glacier area shrinkage remains high because data originated from studies using a wide range of available sensors and archives from ongoing and historical research efforts. However, evaluating uncertainty in the regional average results was difficult because data were obtained using a wide range of available sensors and archives from ongoing and historical research efforts.Furthermore, although we used the results of studies with the largest possible research regions so that the results would be more representative and reduce system errors for each RGI region, information on glacier area reduction in some regions, especially in the Russian Arctic and Greenland periphery, are still lacking.

- Glacier mass-change assessments need to be improved across global and regional scales. For example, recent estimates of glacier mass balance are relatively divergent in HMA (Brun et al., 2017; Maurer et al., 2019; Yao et al.,2012). Uncertainty in glacier mass balance estimates are dominated by the limited number of regularly in-situ observed glaciers (450 out of more than 200,000 glaciers have at least one measurement). Details on the uncertainty estimates of the glacier mass balance dataset used in this study can be found in the work of Zemp et al. (2019).

- Very small glaciers(with area ≤1 km2)account for 80%of the total number of glaciers in the world. The 0.5°× 0.5°gridded map of the RGI glacier inventory,which has a low resolution, is inadequate to demonstrate the distribution of glaciers worldwide.Moreover,the resolution of the current global climate dataset is still limited.The low resolution of the ERA-Interim also adds uncertainty in detecting means and trends in annual mean temperature and annual total precipitation for each RGI region.

- Glacier retreat worldwide was due to a combination of various factors such as climate, debris coverage, debris thickness, glacier size, ice thickness, terminus elevation,topography,black carbon,aerosols,and albedo(Barry,2006;Davaze et al.,2018;Qin et al.,2018;Radi´c and Hock,2014).Differences in climate are not the only factor related to the regional differences in worldwide glacier retreat. Considering the lack of relevant data, we failed to detect the relationship between glacier retreat and all the aforementioned related factors except climate. Extensive work is still necessary to associate regional differences in glacier retreat.

5. Conclusion

Using the data from published research on glacier changes,we investigated the regional differences in glacier response to climate change by combining the mass balance and area loss in 19 regions and relating the results to temperature and precipitation observations.We summarize the findings as follows:

1) Global glaciers showed an area shrinkage rate of 0.18%per year and a mass loss rate of 0.25 m w.e. per year in 1980-2015.The highest area shrinkage occurred at Low Latitudes with a rate of 1.6% per year while the most negative mass balances occurred in the Southern Andes(-0.81 m w.e. per year).

2) Retreat of glaciers located at low and middle latitudes was characterized by severe area shrinkage and mass loss. Glaciers in the Arctic were characterized by ice thinning due to a low area reduction but relatively high mass loss rate. On the other hand, glaciers in high southern latitudes were in a relatively stable status.

3) The rate of area change and mass balance had a significant negative correlation with annual mean temperature and annual total precipitation instead of their variable trends. Glacier fluctuations were primarily driven by their surface climate conditions rather than the changing rate of climatic conditions.

However, further work is necessary to produce more reliable estimates of glacier retreat, which would reduce the uncertainty in detecting regional differences and their relationship with climatic characteristics.

Conflict of interest

The authors declare no conflict of interest.

Acknowledgments

This research was supported by the Strategic Priority Research Program of the Chinese Academy of Sciences(XDA19070501), the National Natural Science Foundation of China (41730751 & 41671066), the International Partnership Program of Chinese Academy of Sciences(131C11KYSB20160061, Y560L01001).

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.accre.2020.03.003.