Volume 48, Issue 15 e2021GL092942
Research Letter
Open Access

Rainfall on the Greenland Ice Sheet: Present-Day Climatology From a High-Resolution Non-Hydrostatic Polar Regional Climate Model

M. Niwano

Corresponding Author

M. Niwano

Meteorological Research Institute, Japan Meteorological Agency, Tsukuba, Japan

National Institute of Polar Research, Tachikawa, Japan

Correspondence to:

M. Niwano,

[email protected]

Search for more papers by this author
J. E. Box

J. E. Box

Geological Survey of Denmark and Greenland, Copenhagen, Denmark

Search for more papers by this author
A. Wehrlé

A. Wehrlé

Geological Survey of Denmark and Greenland, Copenhagen, Denmark

Search for more papers by this author
B. Vandecrux

B. Vandecrux

Geological Survey of Denmark and Greenland, Copenhagen, Denmark

Search for more papers by this author
W. T. Colgan

W. T. Colgan

Geological Survey of Denmark and Greenland, Copenhagen, Denmark

Search for more papers by this author
J. Cappelen

J. Cappelen

Danish Meteorological Institute, Copenhagen, Denmark

Search for more papers by this author
First published: 03 July 2021
Citations: 18

Abstract

Greenland ice sheet rainfall is expected to increase under a warming climate. Yet, there have been no active long-term in-situ rainfall records on the ice sheet due to observational difficulties. Here, we utilize the state-of-the-art 5 km polar non-hydrostatic regional climate model NHM-SMAP to evaluate the ice sheet’s rainfall over 40 years (1980–2019). The largest trends include a fourfold increase in annual rainfall for the northwestern ice sheet; 3.1 Gt year−1 or 12 mm m−2 year−1. September ice-sheet-wide rainfall amount and intensity increase by 7.5 Gt month−1 and 20.8 mm h−1 year−1. In the last two decades, the increasing September maximum hourly rainfall rate exceeded 50 mm h−1 six times. The increased surface water delivery has numerous implications, including for snow metamorphism and ice flow dynamics.

Key Points

  • A 5 km horizontal resolution non-hydrostatic regional climate model is used to evaluate Greenland ice sheet rainfall from 1980 to 2019

  • The northwestern sector stands out as a hotspot of increasing rainfall

  • Both the amount and intensity of September ice-sheet-wide rainfall have significantly increased in the 40 years from 1980 to 2019

Plain Language Summary

We find that rainfall has increased over the Greenland ice sheet from 1980 to 2019. Among the eight major ice-sheet drainage areas, the northwestern sector stands out as an increasing rainfall hotspot where the annual rainfall has increased fourfold. For the entire ice sheet, September's total rainfall and its intensity increased by 224% and 54% above the 1981–2010 baseline over the last 40 years. An increase in late melt season rainfall is expected to contribute to an increase in late summer snow and ice melt. Our results provide the first detailed quantification of the state of ice sheet rainfall climatology.

Video Abstract

Rainfall on the Greenland Ice Sheet: Present-Day Climatology From a High-Resolution Non-Hydrostatic Polar Regional Climate Model

by Niwano et al.

1 Introduction

The current increasing trend of Arctic surface air temperatures is more than two times that of the Northern Hemisphere (Cohen et al., 2014; Overland et al., 2017; Serreze et al., 2009) and is driven by several physical processes comprising “Arctic Amplification” of climate change (Cohen et al., 2014; Manabe & Stouffer, 1980; Pithan & Mauritsen, 2014; Serreze et al., 2009). The Intergovernmental Panel on Climate Change (IPCC) Special Report on the Ocean and Cryosphere in a Changing Climate (SROCC; Meredith et al., 2019) assesses high confidence in continued Arctic warming into the foreseeable future. Warming is expected to cause more precipitation in the Arctic to fall as rain (Bintanja & Andry, 2017; Fettweis et al., 2013). In the High Arctic, rain-on-snow events during the wintertime have been observed more frequently in the recent decades, and are expected to become more frequent in the warmer future (e.g., Hansen et al., 2014; Sobota et al., 2020; Wickström et al., 2020), as across the midlatitudes (e.g., Marks et al., 1998; Musselman et al., 2018; Würzer et al., 2016).

The summertime North Atlantic Oscillation (NAO) index indicates episodes of blocking high pressure over Greenland (e.g., Ahlstrøm et al., 2017; Hanna et al., 20152018). The snow and ice surface energy budget evolves in response to these atmospheric changes in ways that enhance surface melting (e.g., Fausto et al., 2016; Kuipers Munneke et al., 2018; Niwano et al., 2015). Box et al. (2012) analyze the impacts of persistent high pressure over the ice sheet during summer, and highlight how the snow and ice albedo feedback and snow accumulation reduction can amplify surface melt via: (1) increased northward advection of warm and moist air, resulting in increased surface sensible heating and near-surface snow grain growth and further decreased surface albedo; (2) increased surface downward shortwave radiant flux, resulting in more surface heating and further albedo reduction; and (3) reduced snowfall rates to sustain low surface albedo, maximizing near-surface solar heating, progressively lowering albedo over multiple years. As a result, the Greenland ice sheet has undergone amplified changes in the physical structure of its near-surface snow and ice, especially since around 2000. The physical changes of the snow and ice are increasingly decreasing the surface mass balance of the ice sheet, enhancing the mass budget deficit (e.g., The IMBIE team, 2020).

In a warming climate, the precipitation regime on the ice sheet, especially during the lengthening melt season, shifts toward a higher fraction of rain in the total precipitation (Bintanja & Selten, 2014; Box et al., 20122019; Doyle et al., 2015; Lenaerts et al., 2020; Screen & Simmonds, 2012). In the Arctic, it remains difficult to obtain accurate precipitation measurements, primarily due to gauge under-catch of solid precipitation and limited power supply in remote locations (e.g., Serreze & Barry, 2014). Studies that have reported rain observations from ice sheet locations are few (Box et al., 2012; Niwano et al., 2015; Ohmura & Boettcher, 2018). Rainfall can drastically increase the density and temperature of near-surface snow and ice via positive temperatures (Reijmer et al., 2012; Rennermalm et al., 2013) and reduce subsurface cold content (Vandecrux et al., 2020). Changes in rainfall may be important for the future ice sheet mass balance, not only for potential impacts on surface mass balance processes but for ice dynamics. Doyle et al. (2015) found for the southwestern ice sheet that extreme surface runoff from melt and rainfall was linked to short-term accelerations in ice flow speed. Fettweis et al. (2013) highlight that rainfall on the Greenland ice sheet is poised to increase under climate warming, with the ice-sheet rainfall anomaly set to exceed 150 Gt year−1 relative to 1980–1999 by 2100 under a high carbon emissions scenario.

Here, we examine regional and seasonal rainfall patterns and rainfall intensity over the Greenland ice sheet using the polar non-hydrostatic regional climate model NHM-SMAP (Niwano et al., 2018) over a four-decade period (1980–2019). NHM-SMAP is currently one of the few polar regional climate models that utilizes a non-hydrostatic formulation of atmospheric flow. We evaluate the model using in-situ observations, conduct a trend analysis and examine changes in extreme precipitation at regional and ice-sheet scales.

2 Data and Methods

2.1 Polar Regional Climate Model NHM-SMAP

We apply the polar regional climate model NHM-SMAP (Non-Hydrostatic atmospheric Model with the Snow Metamorphism and Albedo Process model) to the Greenland ice sheet at high spatial (5 km) and hourly output frequency as described in Niwano et al. (2018) and Fettweis et al. (2020). The atmospheric part of NHM-SMAP, JMA-NHM (Japan Meteorological Agency Non-Hydrostatic atmospheric Model, Saito et al., 2006), is forced at its lateral and upper boundaries by the JRA-55 data set (Kobayashi et al., 2015). Additionally, the modeled atmospheric vertical profile is initialized by referring to JRA-55 every day to prevent large deviations between JRA-55 and NHM-SMAP atmospheric fields. A 30-h daily simulation from 18:00 UTC the previous day is composed of a 6-h spin-up with model outputs from the last 24 h. Numerous studies conducted in Japan (e.g., Kato & Aranami, 2005; Tsuguti & Kato, 2014) have confirmed that NHM can be used for heavy rainfall analyses. Surface meteorological properties drive the physical snowpack model SMAP over snow and ice, where no regular initialization is performed unlike the atmospheric part (Niwano et al., 2018). Sea surface albedo and temperature are prescribed from the parent JRA-55 reanalysis data. In this study, we focus on the main ice sheet for regions defined by Zwally et al. (2012, Figure S1) and analyze modeled hourly, monthly, and yearly rainfall data from 1980 to 2019. In this study, the “trend” metric refers to the linear regression temporal slope multiplied by the timespan in years. A measurement for trend confidence is 1−p from a two-tailed Student’s t-test.

NHM-SMAP simulations were evaluated as part of the Greenland ice sheet surface mass balance (SMB) model intercomparison project (Fettweis et al., 2020). Among the intercomparison ensemble, NHM-SMAP is the only member that employs a non-hydrostatic atmospheric model together with a double-moment bulk cloud microphysics scheme.

2.2 In-Situ Rainfall Measurements for Model Evaluation

Although there have been no long-term rainfall records from locations on the ice sheet, precipitation gauge measurements at Greenland coastal sites are available from the Danish Meteorological Institute (DMI) (Cappelen, 2019). Here, we use 5 years of DMI rainfall observations (2014–2018) from V98 type automatic weather stations (Cappelen, 2019) and from manually operated precipitation stations. In Greenland, among eight V98-type weather stations, five measure precipitation: Aasiaat; Nuuk; Qaqortoq; Tasiilaq; and Ikerasassuaq. The data quality at Qaqortoq, Tasiilaq, and Ikerasassuaq was insufficient during the evaluation period. In addition to the V98 sites, manually operated precipitation observations used here are from Sisimiut; Narsarsuaq; Danmarkshavn; and Ittoqqortoormiit.

It is widely recognized that the undercatch of precipitation confounds the accuracy of in-situ measurements (e.g., Nitu et al., 2018). For Greenland studies, undercatch corrections have been applied by, for example, Yang et al. (1999), Johansson et al. (2015), Mernild et al. (2015), and Koyama and Stroeve (2019). The DMI V98-type weather stations employ the Pluvio2 rain gauge (OTT Hydromet), which has similar catch efficiencies as the GEONOR rain gauge (Geonor Inc.) (Kochendorfer et al., 2017). Here, for the Pluvio2 rain gauge installed in the V98-type stations, we use the undercatch corrections developed by Førland et al. (1996) that are used to correct precipitation measured with GEONOR (e.g., Johansson et al., 2015; Morin et al., 2012). To correct the precipitation undercatch of the Danish Hellman rain gauges at manually operated precipitation stations (Cappelen, 2019), we use the method by Yang et al. (1999), also used by Mernild et al. (2015) and Koyama and Stroeve (2019). In these correction methods, given that the catch ratios are defined separately for snowfall and rainfall, it is necessary to discriminate snowfall and rainfall from measured total precipitation. For this purpose, we use the method proposed by Yamazaki (2001), which defines the fraction of rainfall in measured precipitation as a function of surface wet-bulb temperature. For annual rainfall values at Aasiaat, Nuuk, Sisimiut, Narsarsuaq, Danmarkshavn, and Ittoqqortoormiit, the applied corrections increase the raw rainfall values on average by 1.17, 1.20, 1.20, 1.18, 1.23, and 1.20, respectively. These correction factors are lower than those reported by Mernild et al. (2015) for measured total precipitation from 1890 to 2012, because the correction factors are lower for rainfall than for snowfall/total precipitation (Førland et al., 1996; Nitu et al., 2018) which our study exclusively examines.

3 Results

Evaluation of NHM-SMAP rainfall with observations (Section 2.2) for the time of year with most simulated rainfall (June, July, August, and September) and a comparison of the following statistics with the results by Koyama and Stroeve (2019) reveal acceptable NHM-SMAP rainfall simulations (Figure 1). RMSD is 20 mm month−1 (half of the average observed value) and the coefficient of determination (R2) is 0.78. Bias is insignificant (2 mm month−1). NHM-SMAP disagreement with the observations is due to some combination of imperfect timing and location of the precipitation system that are in part due to the coarse terrain resolution.

Details are in the caption following the image

Evaluation of model-simulated rainfall amount in Greenland during 2014–2018. Comparison between observed and modeled monthly (June–September) values at the six sites (see Figure 2).

The average annual total precipitation together with interannual variability over the ice sheet (excluding peripheral ice caps; Figure S1) is 850.0 ± 86.0 Gt year−1 (510 ± 52 mm m−2 year−1) during the 1981–2010 period, of which 3.0%, 25.4 ± 8.9 Gt year−1 (15 ± 5 mm m−2 year−1) falls as rain. Rainfall is naturally higher at lower elevations and latitudes (Figures 2a and S2a). The average annual rainfall increased by 6.3 Gt year−1 (4 mm m−2 year−1; +25%) in the last decade 2010–2019 compared to the 1981–2010 period. This increase is concentrated at elevations below 2,000 m a.s.l. in the southeast, south, southwest, west, and northwest ice sheet sectors (Figures 2b and S2b). Above 2,000 m a.s.l., the absolute change in the 2010–2019 average annual rainfall with respect to the 1981–2010 climatology (Figure 2a) is generally less than 4 mm, except for the south and southwest sectors (Figure 2c).

Details are in the caption following the image

Climatology and changes of annual rainfall around Greenland simulated by NHM-SMAP. (a) climatological annual rainfall (mm year−1) map (1981–2010), (b) trends of annual rainfall (mm) from 1980 to 2019, and anomalies of (c) 2010–2019 and (d) 2017 average annual rainfall (mm year−1) with respect to the climatological annual value. Trend metrics are the linear regression slope multiplied by the number of years (40). The WMO (World Meteorological Organization) weather station identifiers indicate Aasiaat (04220), Sisimiut (04230), Nuuk (04250), Narsarsuaq (34270), Danmarkshavn (34320), and Ittoqqortoormiit (34339). Solid lines in both figures indicate ice-sheet drainage sectors. Broken lines indicate surface elevations with a contour interval of 1,000 m.

Forty years of hourly 5 km NHM-SMAP simulations (Figure 3a) reveal that annual rainfall amount (hereafter RA) over the entire ice sheet has an increasing trend of 14 Gt year−1, 8 mm m−2 year−1 over the past 40 years with high confidence (1 – p = 0.986). Before 2000, the average annual RA was 21.7 ± 6.1 Gt year−1 or 13.0 ± 3.7 mm m−2 year−1 and then increased by 42% during the most recent two decades (2001–2019) to reach 30.8 ± 12.9 Gt year−1 or 18.5 ± 7.7 mm m−2 year−1. The simulated rain fraction of total precipitation (hereafter fra) has a minor (1.2% in the 40 years) but high confidence (1 – p = 0.948) increasing trend (Figure 3b).

Details are in the caption following the image

Ice-sheet rainfall (1980–2019) and trend metrics from NHM-SMAP output. Trend metrics are the linear regression slope multiplied by the number of years (40). (a), (c), (e), and (g) are Rainfall amount (RA) and (b), (d), (f), and (h) are rainfall fraction of total precipitation (fra). (a) and (b) represent the ice sheet over the whole year, (c) and (d) the northwest sector over the year, (e) and (f) the ice sheet during September, and (g) and (h) the northwest sector for July. Bold values indicate high confidence (1 – p > 0.9) trends shown with broken lines. Monthly RA and fra for all other ice-sheet sectors appear in Figures S3 and S4.

4 Discussion

4.1 Annual Rainfall

NHM-SMAP indicates that the maximum annual RA (64 Gt year−1) and fra (7.0%) occurred in 2012, when the record surface melting occurred (Bonne et al., 2015; Nghiem et al., 2012; Niwano et al., 2015) (Figures 3a and 3b). In July 2012, continuous heavy rainfall was measured at the SIGMA-A site on the northwestern ice sheet with accumulated precipitation between July 10 and 14 was 100 mm (Niwano et al., 2015). Similarly, in summer 2019, exceptional surface melt has been recognized to result in extreme runoff (Sasgen et al., 2020; Tedesco & Fettweis, 2020; van As et al., 2020). However, in 2019, annual RA and fra were not extremely high (Figures 3a and 3b). This is because the atmospheric humidity around Greenland was smaller in 2019, relative to 2012, despite higher July air temperature anomalies over the ice sheet (Tedesco & Fettweis, 2020). This indicates that extreme rainfall years do not always coincide with years of strong melt and runoff. We also compared RA and fra with modeled 2 m air temperature from June to September, and found no obvious correspondence at any sector, suggesting that RA and fra on the ice sheet are not governed by local surface air temperature.

van den Broeke et al. (2016) report that annual average RA and fra over the ice sheet from 1991 to 2005 estimated by the polar regional climate model RACMO2.3 forced by the ERA-Interim reanalysis (Dee et al., 2011) are 28 Gt year−1 and 3.9%. These values agree well with the estimation by NHM-SMAP over the same period (28 Gt year−1 and 3.2%). Fettweis et al. (2013) find that annual rainfall over the ice sheet from 1980 to 1999 simulated by the polar regional climate model MAR, also forced by ERA-Interim is 25 ± 4 Gt year−1. This value is also close with the estimation by NHM-SMAP of 21 Gt year−1 over this period.

4.2 Regional Characteristics

The largest increasing rate of RA that has high statistical confidence (1 – p > 0.9) is the 3.1 Gt year−1 (12 mm m−2 year−1) increase, or 151% above the 1981–2010 baseline for the northwest sector (Figure 3c). RA in this area has increased fourfold over the past four decades. The increasing fra in this sector (2.1% in the 40 years) (Figure 3d) is higher than that for the whole ice sheet value (1.2% in the 40 years) (Figure 3b). Together, these trends in RA and fra highlight the northwestern ice sheet as a regional hotspot of increasing rainfall in addition to where the largest mass imbalance is found (Kjeldsen et al., 2015; Mouginot et al., 2019; Rignot et al., 2008; The IMBIE team, 2020; van den Broeke et al., 2009). This northwestern rainfall increase can be partly attributed to the northward shift of the maximum latitude of the summer upper tropospheric ridge (Tedesco et al., 2016) that accelerates southerly to westerly warm air advection in north Greenland. This regime shift allows relatively warm atmospheric rivers originating from the Atlantic Ocean (Bonne et al., 2015; Mattingly et al., 2020; Neff et al., 2014) to reach further north in Greenland. In addition, increased cloud cover and downward longwave radiation in the northern part of the ice sheet in recent years (Orsi et al., 2017; Noël et al., 2019) could support the rainfall increase in this sector.

Increasing trends for RA and fra are found in all the sectors (Figures S3 and S4). Both average RA and fra are highest in south Greenland with the majority (74%) of the total ice sheet rainfall occurring in the southeast (3.7 Gt year−1), south (6.5 Gt year−1), and southwest (9.0 Gt year−1) sectors, where the impact of topography on precipitation is important (van de Berg et al., 2020). The mean annual fra in the south and southwest sectors (7.7% and 7.2%, respectively) are more than two times the average for the entire ice sheet (3.0%). The two southern ice sheet areas have the largest increase in RA between the 1980–2010 period and the last decade (Figure 2c): the ice-sheet east upstream of Nuuk and the southern ice sheet’s Qagssimiut ice lobe. The latter, located at the southern edge of the ice sheet, has been designated as a mass loss hotspot (Hermann et al., 2018) with rainfall, in addition to strong net downward turbulent heat fluxes, playing a role in extreme 2012 surface melting episodes (Fausto et al., 2016).

According to NHM-SMAP results (Figure 1), the 2017 August rainfall amount measured at Nuuk was exceptionally high during 2014–2018. In 2017, the simulated annual rainfall amount near Nuuk was higher than the 1981–2010 climatology (Figure 2a) by more than 18% or 20 mm (Figure 2d). Both observations and NHM-SMAP indicate that one-third of the monthly rainfall during August 2017 at Nuuk was recorded from the 21st to the 22nd of the month. Synoptic meteorological conditions from JRA-55 on August 22, 2017 (Figure S5) include an atmospheric river, which is defined as the area with the column integrated water vapor transport exceeding 150 kg m−1 s−1(Mattingly et al., 2018), impinged along southwestern Greenland.

4.3 Seasonal Characteristics

The majority (89%) of annual rainfall (23.1 Gt) occurs from June to September, with July and August receiving most rainfall (both 7.9 Gt month−1) although only September has a clear increasing trend. The uncertainty of the monthly rainfall estimated from the bias value (Section 3) using the method presented by Noël et al. (2017) is 3.3 Gt month−1. During these two summer months, the ice-sheet-wide average fra reached 11%. For the ice sheet as a whole, an increasing rate of 7.5 Gt month−1 in the 40 years (224% of the 1981–2010 baseline) is found in September (Figure 3e), when significant increasing trends are present in all sectors except the north and northwest. During September, fra is also simulated to be increasing rapidly, by 3.4% in the 40 years from 1980 to 2019 (Figure 3f). These results corroborate the suggestion of Schuenemann and Cassano (2010), who find that climate change is increasing the magnitude of cyclonic precipitation events and with a shift later in the season, especially in the southwest of the ice sheet.

In contrast to the whole ice sheet pattern, for the north and northwest ice-sheet, the highest magnitude of rainfall trends is in July and August. For the northwest, the increasing trend is higher in July (1.3 Gt month−1 in the 40 years; 177% of the 1981–2010 baseline) than August (1.2 Gt month−1 in the 40 years; 128% of the 1981–2010 baseline). The uncertainty according to the monthly rainfall estimated from the bias value (Section 3) is 0.5 Gt month−1 in the northwest ice sheet. This trend in the northwest ice sheet increased from the year 2000 (Figure 3g) as is the case for fra (Figure 3h), for which a strong increasing trend of +10.7% in the 40 years from 1980 to 2019 is found.

4.4 Extreme Rainfall

Extreme rainfall is increasing in many regions of the world as a consequence of warming and the related increasing atmospheric moisture capacity (Fujibe, 2015; Fujibe et al., 2006; IPCC, 2013; O'Gorman & Schneider, 2009; Pfahl et al., 2017). Fujibe (2015) demonstrates that the seasonally maximum hourly rainfall is a reasonable proxy for warming. Here, the hourly meteorological conditions simulated by NHM-SMAP reveal that the maximum hourly rainfall rates (taken from grid point values, hereafter RR1hr_max) in July and August have not significantly increased during the study period (Figure S6), even in the northwest sector. In September, however, RR1hr_max is increasing over the entire ice sheet significantly by 20.8 mm h−1 in the 40 years (54% of the 1981–2010 baseline) (Figure 4). During the first half of the record (1980–1999), the average RR1hr_max was 32 mm h−1, with September RR1h_max never exceeding 50 mm h−1. Over the second half of the record (2000–2019), average summer RR1h_max was 48% higher (47 mm h−1), and exceeded 50 mm h−1 six times, peaking at ∼120 mm h−1 in 2000. The majority (78%) of the September maximum hourly rainfall events were recorded in the southeast and south ice-sheet sectors (38% and 40%, respectively), where September RA are increasing at the higher rates of 1.5 and 1.9 Gt month−1 over the 40 years, respectively. The highest rates of increase in September RR1h_max in the 40 years are found for the southeast (18.6 mm h−1), south (17.6 mm h−1), and southwest (9.3 mm h−1) sectors. The RR1h_max trend is only of high confidence (1 – p > 0.9) for the south sector. The majority of the southeast sector is covered by a thick snow and firn layer (Burgess et al., 2010). Infiltrating water in these areas can be refrozen or temporarily retained in perennial firn aquifers (Forster et al., 2014; Vandecrux et al., 2020). These processes potentially reduce and dampen the delivery of rainwater to the ice sheet base in the accumulation area. As a result, less rainfall-induced glacier speed-up can be expected in the southeast sector. In contrast, the south sector has a more extensive bare ice area, where water transfer to the bed can be more rapid (Hermann et al., 2018), and may be more prone to rainfall-induced ice accelerations in response to more extreme September rainfall in a warming world.

Details are in the caption following the image

Interannual variations of September maximum hourly rainfall rates over the ice sheet and each sector (Figure S1) simulated by NHM-SMAP (solid lines). For the entire ice sheet (GrIS), as well as southeast (SE), south (S), and southwest (SW) sectors, linear trends are indicated together (dashed lines): 20.8 (GrIS), 18.6 (SE), 17.6 (S), and 9.3 mm h−1 in the 40 years (SW), respectively. For the GrIS and S, these trends are significant (1 – p > 0.9).

5 Conclusions

Our analysis reveals that Greenland ice sheet rainfall has increased in the 40 years between 1980 and 2019. While rainfall is greatest on average in warmer south Greenland, it is the lower temperature northwest ice sheet where the most pronounced rainfall increases has occurred. For the northwest, the annual total rainfall has increased fourfold in 40 years (1980–2019), where the increase occurred especially during July and August. Seasonal maximum hourly rainfall and annual total rainfall in September are both increasing over the ice sheet. A pronounced increasing trend in September maximum hourly rainfall is also evident for the southernmost ice sheet sector. Increased efforts to establish long-term on-ice rainfall measurements are warranted given model uncertainty and the apparent increasing rainfall trend.

Acknowledgments

This study was supported in part by (1) the Japan Society for the Promotion of Science through Grants-in-Aid for Scientific Research numbers JP17KK0017, JP17K12817, JP18H05054, JP18H03363, and JP20H04982; (2) the Ministry of the Environment of Japan through the Experimental Research Fund for Global Environment Conservation MLIT1753; and (3) the Arctic Challenge for Sustainability II (ArCS II), Program Grant number JPMXD1420318865. J. E. Box, A. Wehrlé, B. Vandecrux, and W. T. Colgan were supported by the Program for Monitoring of the Greenland Ice Sheet (www.promice.dk) of the Danish Ministry of Climate, Energy and Utilities. The authors thank Atsumu Ohmura for helpful discussions on this study. The authors thank two anonymous referees for providing constructive comments and suggestions, which improved the quality of this manuscript.

    Data Availability Statement

    The NHM-SMAP model simulation data from 1980 to date are available from https://ads.nipr.ac.jp/dataset/A20210517-005. The DMI in-situ AWS data are available from the following link in which English descriptions are provided under the year tabs/breaks, although the main text in the link is written in Danish: https://www.dmi.dk/publikationer/.