TermPicks: A century of Greenland glacier terminus data for use in machine learning applications

Marine-terminating outlet glacier terminus traces, mapped from satellite and aerial imagery, have been used extensively in understanding how outlet glaciers adjust to climate change variability over a range of time scales. Numerous studies have digitized termini manually, but this process is labor intensive, and no consistent approach exists. A lack of coordination leads to duplication of efforts, particularly for Greenland, which is a major scientific research focus. At the same time, machine learning techniques are rapidly making progress in their ability to automate accurate extraction of glacier termini, with 5 promising developments across a number of optical and SAR satellite sensors. These techniques rely on high quality, manually digitized terminus traces to be used as training data for robust automatic traces. Here we present a database of manually digitized terminus traces for machine learning and scientific applications. These data have been collected, cleaned, assigned with appropriate metadata including image scenes, and compiled so they can be easily accessed by scientists. The TermPicks data set includes 39,060 individual terminus traces for 278 glaciers with a mean and median number of traces per glacier of 10 136±190 and 93, respectively. Across all glaciers, 32,567 dates have been picked, of which 4,467 have traces from more than 1 https://doi.org/10.5194/tc-2021-311 Preprint. Discussion started: 14 October 2021 c © Author(s) 2021. CC BY 4.0 License.

loss may have been triggered by a warming climate (atmosphere and ocean) that induces sudden rapid retreat of outlet glacier termini (Wood et al., 2021;King et al., 2020). Observations of glacier retreat, however, show a high degree of heterogeneity in the magnitude, timing, and temporal patterns of this retreat across the GrIS across the ice sheet (Moon and Joughin, 2008;Catania et al., 2018;Murray et al., 2015a;Carr et al., 2017;Fahrner et al., 2021), which complicates our understanding of future mass change from outlet glaciers. This suggest that knowledge of past, and the potential for future, terminus change is 25 critical for accurate forecasting of the GrIS contribution to sea level rise (e.g. Felikson et al., 2017;Aschwanden et al., 2019;Slater et al., 2019).
Glacier termini have long been an indicator of climate change and terminus change data have been used to understand a range of processes over multiple time scales (e.g. Warren and Glasser, 1992;Warren, 1991;McNabb and Hock, 2014;Moon et al., 2015;Cook et al., 2005;Howat et al., 2008;Howat and Eddy, 2011). On the long-term (>annual), terminus records are 30 used to inform the timing, regional pattern, and climate controls on marine-terminating glacier retreat (Murray et al., 2015b;Catania et al., 2018;Hill et al., 2018;Bunce et al., 2018;Howat and Eddy, 2011;Wood et al., 2021;King et al., 2020;Fahrner et al., 2021). Outlet glaciers can also change at sub-annual timescales and examination of terminus change on shorter time scales (∼seasonal) aids interpretation of the specific environmental and glaciological processes that influence glaciers Moon et al., 2015;Schild and Hamilton, 2013;Cassotto et al., 2015;Ritchie et al., 2008;Howat et al., 2010;Carr 35 et al., 2014;Moon et al., 2014Moon et al., , 2015Brough et al., 2019;Kehrl et al., 2017;Bevan et al., 2019). Such studies are valuable because glacier termini respond to a diverse set of mechanisms related to the geometry of the glacier-fjord system, inland ice dynamics, and the strength of climate forcing (Moon and Joughin, 2008;Carr et al., 2017;Catania et al., 2018;Bunce et al., 2018;Porter et al., 2018). However, determining the variables controlling seasonal variations can be difficult because changes in the climate system occur simultaneously (e.g. Cowton et al., 2018;Fahrner et al., 2021;Wood et al., 2021). Recent work 40 suggests that the shape of the terminus trace and how it evolves over time may provide additional information about the nature of processes dominating any given glacier Chauché et al., 2014). Such studies demonstrate the need for detailed terminus tracing (map-view, full terminus width) at as high a temporal resolution as possible.
Numerous studies have digitized termini manually (Table 1) for use in interpreting glacier dynamics in response to climate variability; however, the lack of coordination across these studies has resulted in duplicated data and heterogeneity in terms of 45 format, quality, method, location, temporal coverage, and availability. Such factors limit the utility of terminus data to future researchers. In addition, manually picking glacier termini is a laborious process. For example, the data set from Catania et al. (2018) used the entire Landsat record to pick 15 glaciers in central West Greenland and the authors estimate that it took 3 undergraduate researchers nearly 2 summers working 15 hours a week each to download imagery and pick the full width of the terminus, or approximately 48 hours per glacier. Rapidly replacing manual-picking are machine learning techniques, 50 which have recently been developed for automated extraction of glacier termini, across a number of satellite sensors (e.g. Mohajerani et al., 2019;Baumhoer et al., 2019;Cheng et al., 2020;Zhang et al., 2021). Manually-picked data are still needed for validation of machine learning methods and as training data. For example, methods using over 1500 training data inputs result in classification in ∼94% of detectable images, under ideal conditions (Cheng et al., 2020). Further, machine learning methods fail in images where ice conditions do not permit easy delineation of the terminus (e.g. mélange-choked fjords, shadowed termini, 55 etc.) and therefore manually-picked termini will still be needed until machine learning algorithims improve. Importantly, future satellite missions imaging the polar regions are expected to continue for the foreseeable future, suggesting an ongoing need to coordinate terminus data in addition to other important glaciological observations that are highly coordinated (e.g. velocity and elevation). Here we present the most complete set of manually picked terminus data for Greenland's outlet glaciers, reprocessed for use in machine learning methods and scientific analysis. Data have been cleaned, associated with appropriate 60 metadata where possible, and the metadata normalized so they can be easily accessed by scientists.

Input Data
Terminus traces were collected through email requests to authors who had published papers that made use of such data, or taken from publicly available online databases (Table 1). Since there was no open call for data submission, there may be other 65 sources of terminus trace data that are available and/or unpublished. Authors used a range of image sources (Table 2), but the bulk (∼70%) of terminus traces originate from Landsat images. Collectively, we refer to these collected data as input data to differentiate these data from the output (cleaned, reformatted) training data generated.
All data were provided in ESRI shapefile format (Figure 1) with the bulk of data provided as polylines and a smaller volume of data provided as polygons or polygon-boxes. In these latter cases, the polygons were cropped at the terminus and converted 70 into polylines. All glacier terminus traces were exported into a single ESRI line shapefile format consistent with file formats typically used in machine learning techniques. All shapefiles were re-projected into NSIDC Sea Ice Polar Stereographic North (EPSG:3413).
Glacier termini were commonly traced by importing geographically-rectified images into GIS software (e.g. ArcGIS, ENVI, and QGIS) and manually-digitizing the ice-ocean boundary (terminus). Authors used a range of methods for tracing termini 75 including picking the full width or variations on the Box methods. Box methods consist of using a fix-width rectilinear or curvilinear box along the length of a fjord tracing the terminus within those bounds (for a description of these methods see Lea et al., 2014). For consistency in data format, we exclude termini that were identified with only a center point (e.g. King et al., 2020) because these data do not cover the entire width of termini. Individual terminus trace files are largely indistinguishable between authors, with the exception of those who used the box method for picking the terminus, since this method often 80 produces terminus traces that are truncated before they reach the fjord wall. Across all authors, terminus traces have an average of 23 vertices per kilometer with a median of 10 vertices per kilometer.

Glacier Identification
As the GrIS has several hundred marine-terminating glaciers, proper identification of glaciers is important for data management. Several prior authors have produced identification files (ID files) for GrIS glaciers including Moon and Joughin (2008) 85 (Moon IDs) who created a glacier ID file by identifying all non-stagnant glaciers that terminate in the ocean with terminus widths of roughly 1.5 km or greater. The Moon IDs identify 239 glaciers that are assigned a numerical ID, including 6 ice cap glaciers that are marine-terminating. We received terminus traces for 278 glaciers but subsequently identified 282 glaciers by including all glaciers with a Moon and Joughin (2008) ID and additional glaciers with the following criteria; 1) surface speeds >50 m/yr, 2) grounding lines below sea-level as determined from the BedMachineV3 bed topographic product (Morlighem 90 et al., 2017), and 3) termini greater than or equal to 1 km in width. We excluded glaciers that were not picked by at least two separate authors and those that were land-terminating . Using this new ID file here termed TermPicks ID, we assigned glacier IDs to each glacier in our data base ( Figure 1).
Our TermPicks ID file maintains consistency with the Moon IDs by including the corresponding Moon ID with the TermPicks ID within the metadata. We also include other information in the TermPicks ID file that is relevant for wide community use, 95 including outlet glacier flux gates identified by Mankoff et al. (2019) and glacier naming schemes catalogued by Bjørk et al. (2015) in an ESRI multipoint shapefile so the data can be easily referenced with other data sets.

Data Cleaning
The number of terminus traces included in an input shapefile varied across the input data. Some authors represented multiple dates per glacier within each shapefile while others included single dates per glacier for each shapefile. Our output data merged 100 all terminus traces for all dates together into one shapefile and so input data were re-processed to fit into this format. Some authors included multiple glaciers per date for a shapefile, particularly when glaciers were adjacent to one another. Where possible, these shapefiles were manually split into traces representing separate glaciers, consistent with our output data format ( Figure 2c). This was accomplished using a Greenland MODIS base image for them to be properly sorted into new glaciers along fjord wall boundaries or ice stream where appropriate (Howat, 2017). Traces were also clipped using the GIMP ice mask 105 in order to remove fjord wall traces . The mask was extended where it did not intersect earlier traces. Traces that were picked using the box methods were not interpolated to the fjord wall. In many cases, the box spans nearly the entire width of the fjord, but several datasets use boxes that are much smaller than the width of the fjord (Figure 2a). The lack of data at the edges of glacier termini may lead to differences in total retreat using these data compared to other data (Lea et al., 2014). Thus, terminus traces picked using the box method are flagged in the metadata (Table 3).

Data Formatting
Satellite image scene identifiers (scene IDs) are useful to find the original image from which a glacier terminus was digitized, which is a requirement for these data to be useful for machine learning. These were provided in very few of the input data sets. Where no scene ID was available, Landsat scene identification is assigned to terminus traces that were originally picked using Landsat data. Scenes were assigned by geolocating a Path/Row from the Worldwide Reference Systems (WRS 1 for 115 Landsat 1-3; WRS-2 for Landsat 4 onward) that is closest to the terminus trace, then searching by date using Google Cloud Services. As Landsat scenes are freely available for Level-1 data on Google Cloud Services and most (∼ 70%) of the data are derived from Landsat images, only terminus traces that were known to be picked with Landsat data are assigned IDs ( Figure   1). Some glaciers share multiple overlapping Landsat Path/Row combinations resulting in some terminus traces having two scenes assigned. In these cases, both scene IDs are appended to the metadata. Glaciers with automatically assigned scene IDs 120 have the quality flag of 005 (Table 3). Further, some terminus traces did not have dates that corresponded to a scene ID from Google Cloud Services.

Metadata Creation
Consistent and uniform metadata are critical to the use of training data in machine learning and scientific studies. Feature extraction using image segmentation techniques rely on accurate attribution of training data to the correct time, location and 125 satellite image used for terminus tracing. Input data used for TermPicks suffered from a lack of consistency in the metadata, such as date format, author and satellite identification, scene ID, and digitization techniques. Here we describe the metadata format for the output TermPicks data set ( Figure 1). The TermPicks metadata format was chosen to be consistent with the largest archive of machine-picked terminus traces from Cheng et al. (2020), known as CALFIN. For example, CALFIN includes the date, quality flags, satellite sensor and scene ID, all of which are important for machine learning.

130
Date: The date represents the acquisition time for the image used to pick the terminus for that trace and is formatted into separate columns for year, month, day and decimal date. Date column is a string and the format is "YYYY-MM-DD". Year, month, and day are integers. If a trace included only year information, the date column format is "YYYY-00-00".

135
Satellite: Satellite refers to the original sensor or satellite that produce an image used to pick the terminus. This information was taken from existing attribute tables or file names from the input data and was used to determine the image scene ID where possible.
Author: All people contributing traces have been listed as authors in this paper. Included in the metadata is the Author iden-140 tifier connected to a specific citation using the data provided. We also provide a code block in the code repository to produce citations for the authors of terminus traces that are used in data downloads. This allows for proper attribution to the correct author depending on the location and time span of data downloaded.
Scene ID: Scene ID refers to the image scene identifiers for the original image used to pick the individual glacier trace. This 145 corresponds directly to the sensor. For example, a Landsat Product ID is an example of a scene ID. Certain images (e.g. some aerial images) were used to pick multiple traces. It includes information on the date and location for the original image. If an author included a scene ID, the text was kept the same in case users need to contact the original author for image access.
Glacier IDs: The Glacier ID refers to the TermPicks glacier ID scheme that was created for this project (described in section 150

2.2).
Center X and Y: A centroid point was created for each trace in WGS 84 (EPSG:4326) so that the TermPicks data can be easily referenced with other data sets.

155
Quality Flag: Quality flagging is used to identify and classify traces that may have issues leading to sources of error. This quality flagging schemed was created in conjunction with Cheng et al. (2020) to enable data synthesis between our data and machine-generated terminus traces. We assign a prefix 'X' for all data defining if the trace was created automatically or manually. All data in the TermPicks dataset has X = 0 while data in the CALFIN data set has X = 1. In addition, shapefiles can have multiple quality flags. We follow the quality flag scheme in Table 3. In this scheme, flags are assigned if there are no 160 issues with the terminus trace (X0), if there is uncertainty in the trace due to environmental or image issues, for example clouds partially obscuring the terminus (X1), if the trace was supplemented (two images were used to digitize the terminus) (X2), if the trace was picked with the Landsat 7 sensor when the Scan Line Corrector was off (X3), if the trace was picked using the box method and is thus incomplete (X4), if the scene ID was automatically assigned because of information provided in the input metadata (X5). The X1 and X2 flags are only used if the trace author indicated this information, and so many traces will 165 not include these flags.

Calculation of Terminus Change
In addition to providing manually-picked terminus traces for glaciers in Greenland, we also compute terminus change for these terminus traces per glacier. First, we calculate terminus change using a method developed in Catania et al. (2018) where equally-spaced points along each terminus trace are projected to the nearest location along the glacier centerline, here named the Interpolation method. The average position of all projected points on the centerline thus becomes the average position of the glacier terminus for that date of the terminus trace. Then, the distance between each of these terminus positions for all dates is used to calculate the distance the glacier terminus moved up or down-fjord over time. This method is most accurate when the glacier traces are all approximately the same length. In addition to the Interpolation Method, we also calculate the fluctuation 175 in terminus position simply by taking the point where the terminus intersects the centerline of each glacier (King et al., 2020), here named the Centerline method.

Results
The TermPicks data set includes 39,060 individual terminus traces for 278 glaciers with a mean and median number of traces per glacier of 136±190 and 93, respectively. However, trace count varies depending on author interest in a specific glacier or

Spatial and Temporal Bias
Heatmaps of the output data demonstrate the temporal coverage and frequency in the training data. We present heatmaps for both regional groups of glaciers ( Figure 4) and individually for each glacier ( Figures A9, A10, A11). These figures demonstrate that terminus data availability is intimately tied to Landsat image acquisition. A combination of U.S.-centric acquisition strategies, ground station coverage, and limitations on data transmission and duty cycles meant that much of the world did 190 not have regular repeat Landsat coverage until 2013 with the launch of Landsat 8, which follows a continental acquisition strategy (Wulder et al., 2016). Further, the failure of Landsat 6 upon launch in October of 1993 meant that imagery was only obtained in a limited capacity (via extension of the Landsat 5 satellite) until the successful launch of Landsat 7 in 1999, when we observe an increase in terminus trace data ( Figure 4). We further compute the percentage of terminus traces for a given glacier compared to all available Landsat images that cover any particular glacier (see Figure 5 for four examples) in order to 195 examine the completeness of the terminus data for all glaciers. All glaciers have an individual coverage figure that is contained in our Google Earth file (Supplementary Information). From this analysis we find that Sermeq Silarleq (ID 288) has traces from 33.1% of all available Landsat images (including cloudy images), the most of any glacier in our data set. The main reason for not picking a terminus for an image is usually dense cloud coverage. However, on average only 5.8% of available Landsat images have been manually traced per glacier.

200
Regional differences in data availability also exist (Figures 3 and 4). Higher latitude glaciers experience more frequent coverage by satellite image sensors than lower latitude glaciers due to increased scene overlap at high latitudes (e.g. Figure 5b after 2013). However, there are fewer traces in Southwest Greenland simply due to the lack of marine-terminating glaciers in this region, which is primarily drained through land-terminating ice. There are also fewer overall traces in North and Northeast Greenland than Central West Greenland, a region with a similar number of glaciers, potentially due to less interest in tracing in 205 North and Northeast Greenland (Figure 3). The densest coverage is in Central West and Northwest Greenland (IDs 279 to 3) where nearly every available image from Landsat and other sensors were traced  to create as complete a record as possible of regional glacier change. Other glaciers of interest include Helheim, Kangerlussuaq, and Sermeq Kujalleq (Jakobshavn; IDs 181, 152, and 278), which also have dense coverage.

210
As a metric of error between data sets, we calculated the Hausdorff distance (commonly used in pattern recognition (Huttenlocher et al., 1993)), the greatest minimum distance between two lines. A larger Hausdorff distance indicates two lines are less similar to each other; however, large Hausdorff distances could also indicate that two otherwise identical lines have different endpoints (different lengths). To avoid this latter issue, we trimmed each terminus trace to a glacier reference box, modified from those used by Moon and Joughin (2008), before computing Hausdorff distances. We also excluded traces that did not 215 span the width of these glacier boxes. Excluding short traces reduced the dataset to 25,355 (65% of the original TermPicks dataset). Then, we calculated the Hausdorff distance between every pair of traces for traces that were picked at the same glacier and on the same date by multiple authors. We identify 2,671 individual traces where multiple authors picked them on the same dates (sometimes more than two authors). This resulted in a total of 5,748 duplicated traces. The overall median error between pairs in this reduced dataset is 107 m, which is comparable to that obtained in most machine learning studies when comparing 220 machine-traced termini to manually-traced termini (Cheng et al., 2020).
The median error between any given pair of authors varies with the greatest median error (7,350 m) between Cheng and Hill, and the least median error (58.6 m) between Fahrner and TermPicks (Figure 7). The magnitude of errors are not necessarily due to inaccurate digitisation by authors, but can be explained by Hill and other authors focusing on northern glaciers (that can be difficult to trace due to terminus crevasses), and Farhner focusing on late summer observations where the glacier margin is 225 often most clear. The mean and median of the median errors for each author are presented in Table 4, and there was no clear distinction in error based on methodology used (box vs. full-width tracing). Glaciers with >500 m error between traces were manually checked for errors. If two traces are on the same date but the trace was not equivalent (e.g. the trace did not appear to be from the same front), then the trace with more complete metadata (e.g. includes the original scene ID) was kept. If a trace has 3 authors and one is not equivalent, it is removed if 2/3 match. Only 0.4% of total traces were removed from the data set 230 through this manual checking. In some cases, there are glaciers that have higher errors than other glaciers (e.g. IDs 39, 73, 86, 99, 100, 101) due to the fact that they appear to have highly fractured ice tongues and they develop long, linear cracks that authors may or may not trace in their entirety.

Greenland-wide Glacier Retreat
Terminus retreat is calculated for each glacier using both the Interpolation and Centerline methods. The retreat time-series using the Interpolation method reveals small errors that are present as anomalous spikes in the retreat record possibly due to poor quality traces. Centerline retreat as an average over each decade of the observational record  where sufficient data permit) show regional patterns of retreat ( Figure 6). Traces with only year information were placed approximately at the middle of the year. We find similar ice-sheet wide retreat patterns as previously published sources. For example, total retreat for 2000-2010 is ∼252 km in 225 glaciers, which is comparable to Murray et al. (2015a) who found ∼267 km in 199 glaciers.

Discussion
This is the first published study of manually-traced Greenland-specific marine terminating glacier traces with consistent metadata and formatting across multiple data sets from different authors. Glacier terminus traces have been a staple indicator of 245 glacier change for decades (e.g. Weidick, 1958;Higgins, 1990;Warren and Glasser, 1992;Murray et al., 2015a). From this paper alone, 22 sources have digitized and interpreted terminus positions in Greenland, with many more using these data to aid interpretation of GrIS change. However, all of these efforts have happened independently, with duplicate efforts and lack of consistency across data format and accessibility. Figure 8 shows a time-series of Glacier 116 (F. Graae Gletscher) with authors labeled at each trace to demonstrate the utility of combining data sources, which enables a more complete view of the change 250 at this glacier than previous individual studies. Additionally, these data set can be combined with the CALFIN data set from existing machine learning techniques (Cheng et al., 2020) to provide a more complete record of glacier terminus change for Greenland, or alone to aid improved machine-enabled tracing of glacier termini.
While this data set is the most comprehensive to-date, there remain limitations in drawing large scale conclusions on retreat patterns with these data alone. There are many glaciers with more traces than others that may lead to bias in retreat rates and 255 timing compared to glaciers with less complete information (Figure 3). Therefore, we encourage users to incorporate both TermPicks and CALFIN data (Cheng et al., 2020) to aid their analysis. A combined version of TermPicks and CALFIN is available in our data repository.
Termini traced with different methods or widths of the glacier may have some systemic differences in terminus retreat over time (Lea et al., 2014). For example, Figure 9 shows Glacier 152 (Kangerlussuaq Gletsjer) on 8/11/2006. This date was picked 260 by 3 separate authors (Bunce, Cheng, and ESA) at different extents of the glacier front. When the Interpolation method is used, there is a 0.5 km difference in terminus position change because the end points for each trace are different. Bunce and Cheng will show a higher retreat compared to ESA because the Interpolation method accounts for the entire width of the glacier, therefore the other traces' mean positions will be further up-glacier. While there is no systemic difference between retreats calculated from box method vs. full width picks, users of these data should be aware of this potential misfit between picks based on end points. For example, Bunce traces use the box method while Cheng traces uses the full width method; however, they both end before the fjord wall. Glacier 152 has dead ice on its northern margin and, as shown in the image, the scan line errors in the Landsat 7 imagery block some of the ice, so some authors may or may not pick the entire front for numerous reasons.
Although machine-enabled terminus tracing has made great strides in the past few years, there will be a continued need for 270 manually-tracing glacier termini. This is because certain environmental conditions, such as heavy shadows, cloud cover, ice mélange, and low solar illumination, make it difficult for current machine learning algorithms to accurately trace all available images. The data provided here will aid improvements in machine learning that will ultimately reduce the need for future manual-tracing. Ideally, machine and manual-tracing efforts would work in concert, with data gaps or large errors reported by machine learning quickly identifying where need is the greatest for the manual-tracing team. For example, both the data 275 presented here and the data in CALFIN (Cheng et al., 2020) are not extended beyond 2020 and there is no funding in place to provide continued coordinated (between machine-and manual-picked authors) updates to terminus positions in the future.
Coordinated effort between machine-and manual-tracing teams is warranted to ensure regular delivery of future data, given its importance to the wider scientific community.
Until fully-automated, frequently-updated and publicly available terminus traces are available for Greenland and elsewhere, 280 we anticipate that authors will continue to manually-trace in studies that are spatially or temporally limited. Ideally, future efforts would occur in conjunction with this work, producing data with similar format, metadata, and visibility. To that end, we recommend the use of a bespoke version of the Google Earth Engine Digitisation Tool (GEEDiT; Lea (2018)) within Google Earth Engine's (GEE) API (Gorelick et al., 2017). This GEEDiT-TermPicks version builds substantially on the original GEEDiT, with improvements made to both the digitisation interface, metadata options, sensor availability, and image acces-285 sibility. A user guide is provided as an appendix to this paper. A major advantage of GEEDiT-TermPicks over traditional repository download and visualisation approaches is that it accesses the archive of Landsat, Sentinel-1 and Sentinel-2 and ASTER images on the Google Cloud servers within a standard web browser. It therefore allows for much faster access to imagery compared to the alternative of downloading, extracting, processing and each individual image. This is combined with an interface for easy digitisation of margins that now uses GEE's DrawingTools functions to improve both speed and flexibility 290 of digitisation for users.
To ensure that future data generated using this tool will be consistent with our dataset, the GEEDiT-TermPicks interface visualises the TermPicks ID locations, allowing the user to easily identify the glaciers present and access relevant imagery.
Once a glacier is chosen, GEEDiT-TermPicks provides rapid access to all available satellite images of that glacier, which can be pre-filtered by date and satellite. If the image is clear, the termini can be extracted by simply clicking on the screen along the 295 glacier margin. Images with glacier termini that are low in quality can be compared with previous or subsequent images that are nearby in date to help better determine the location of the terminus for a specific date/time. If this is done, it will automatically be flagged in the image metadata, though this (and other) image quality flag options can be manually selected, including options to provide a written note as to why the image is inadequate. Data exported from GEEDiT-TermPicks will therefore include as standard all metadata required for easy inclusion into future TermPicks data releases. Finally, we recommend a minimum of 11 https://doi.org/10.5194/tc-2021-311 Preprint. Discussion started: 14 October 2021 c Author(s) 2021. CC BY 4.0 License. vertices per km of trace for quality that is consistent with this database. We also recommend tracing across the entire width of the glacier terminus as previous studies have shown that information about mass loss processes can be obtained from studying the map-view change in trace morphology at high levels of detail Chauché et al., 2014).

Conclusions
We present a new compilation of outlet glacier terminus traces for the GrIS spanning a time period from 1916 to 2020 obtained 305 through manual tracing of the ice-ocean boundary. Data were cleaned, reformatted, assigned to image scene IDs, and quality controlled for use in machine learning algorithms that will enable semi-automated terminus tracing. Termini are provided in the same format and with similar metadata to ongoing machine learning-based terminus tracing. We find errors on the order of ∼100 m, similar to machine-identified termini, but biases in terms of data coverage with well-studied glaciers heavily covered with terminus traces, and other glaciers devoid of consistent terminus trace identification. We provide tools for future tracing 310 efforts and include software to enable the use of these data for the broader scientific community.
Code availability. This work was performed using freely-available software primarily Google Earth, Google Earth Explorer, Python, and QGIS. Code to generate an text file that includes the Digital Object Identifier of citations for users is available on the GitHub site (https://github.com/sgoliber/TermPicks). GEEDiT TermPicks can be accessed through Google Earth Engine Code Editor ( https://code.earthengine.google.com/90fc8d8ec49ddeea5ead6779f120cd2)

315
Data availability. Terminus trace data will be made available at NSIDC, a NASA DACC. Until the data submission is approved, data are currently available on Zenodo (https://doi.org/10.5281/zenodo.5512724). A shapefile of combined CALFIN and TermPicks data is included in this repository. We also provide terminus retreat data, the TermPicks ID shapefile, and kmz file that can be viewed in Google Earth and provides a quick look of temporal coverage (compared to imagery availability) for all glaciers.
Author contributions. SG collected, cleaned, and formatted the data and metadata and performed most of the analyses. TB performed some data analysis and synthesis. SG, TB, and GC wrote the manuscript. JL created GEEDiT TermPicks. DC provided expertise on machine learning. All authors contributed to the editing and refining of the manuscript and data contribution. Clicking on the glacier ID gives an image pop up that is the same format as Figure 5. This can be used to get a overview of what the data coverage is for each glacier; however, it does not include other data sources, such as Sentinel data. This file can be found in the TermPicks GitHub repository.

340
GEEDiT is written within Google Earth Engine's (GEE) API (Gorelick et al., 2017). This bespoke version of GEEDiT (Lea, 2018) provides much the same functionality as the original, though represents a significant re-writing of its structure to allow for several improvements and TermPicks specific requirements, including: 1. Changing the digitisation interface so it operates using the Google Earth Engine DrawingTools functions (Google, 2021, link: https://developers.google.com/earth-engine/tutorials/community/drawing-tools, last accessed 5/21/2021). This al-345 lows even more rapid digitisation due to data being temporarily stored 'client side', rather than in the original tool where vertices were submitted 'server side' for subsequent visualisation (see Google, 2021, link: https://developers.google. com/earth-engine/guides/client_server for more information, last accessed 5/21/2021).
2. User skipping of images by date as well as image number 3. Inclusion of ASTER L1T Radiance image archive (https://lpdaac.usgs.gov/products/ast_l1tv003/, last accessed 5/21/2021) 4. Easy user access to imagery of glaciers that make up the TermPicks database and their respective locations.
5. Automatic appending of glacier, imagery and digitisation metadata, allowing any future versions of TermPicks to be easily and quickly generated.
6. Compulsory fields for user names and email addresses that are appended in metadata to ensure that those who digitise the data are properly acknowledged if and when they are subsequently shared/published, and (where necessary) to enable 355 user inter-comparisons.

C GEEDiT-TermPicks walkthrough
Link to GEEDiT-TermPicks: https://code.earthengine.google.com/90fc8d8ec49ddeea5ead6779f120cd2 Step 1: 1. Define date range, months of interest, maximum image cloud cover limit, and satellites to visualise imagery from. Note 360 that maximum image cloud cover limit uses the metadata values indicating cloud cover across the entire image that are provided with Landsat, Sentinel 2, and ASTER imagery. See Figure A1 for overview of menu screen.
2. Zoom to the glacier of interest and click on its blue dot.
Step 2: 1. The tool will automatically zoom to the selected glacier, and the blue dot will turn red. Imagery in the background is the 365 standard Google Earth base imagery. If you have selected the incorrect glacier, click the 'Go Back' button and this will return you to the previous screen. See Figure A2.
2. Enter your name and email address in the boxes provided, and click the 'Go to images' button to continue. These are compulsory fields to ensure that data can be appropriately acknowledged where they are shared/published.
Step 3: 370 1. Imagery for the selected glacier, satellite and date is displayed. Zoom to the desired level to allow accurate digitisation of the terminus, and click on the screen to start digitising, and double click to end. It is possible for multiple lines to be digitised per image, though if users are seeking consistency with the TermPicks dataset this should be avoided. See 'Remove added images' will remove any images that have been added by the user, leaving only the original satellite image on the screen. The 'Edit' button can be used where a line has been finished, but needs to be subsequently modified or deleted. To do this, click the 'Edit' button and then click on the line that needs to be modified. This will allow its vertices to be moved, while the line can be deleted by pressing the delete or backspace key while the line 380 is selected. To switch back to drawing mode, press the 'Draw new line' button. This will allow a line to be digitised by clicking on the screen as before. See Figure A4. (c) Panel displaying glacier name, TermPicks ID and satellite that collected the displayed image. Text boxes display the date of the displayed image in YYYY-MM-DD format, and the image number of the total available of the glacier. Users can also skip to different images, by date or image number. Where users choose to enter dates, they 390 must be given in YYYY-MM-DD format, and the image shown will be the image that is the closest available in time to the entered date. If a user defined image number falls outwith the range of valid values the map will be cleared and a panel requesting the user to enter a valid number will appear. Once a date/image number has been entered, the user can skip to that image by pressing the enter key. See Figure A6.
(d) Panel that allows the user to skip to the next/previous image number, or export the entire set of digitised margins.

395
By pressing any of these buttons, the user will log the digitised margins for export. Once any of these buttons have been pressed, subsequent modification of the data via the geometry imports bar will result in duplicate margins in the exported dataset. Pressing 'Export' will set up an export task that can be accessed through the 'Task Manager' tab in the top right of the screen (Google, 2021, https://developers.google.com/earth-engine/guides/playground, last accessed: 5/21/2021). To avoid the possibility of data loss through failure of internet connection and/or browser 400 crashes, it is recommended that users regularly export their data. See Figure A7.
(e) By hovering the cursor over the Geometry Imports panel, users can view all previous temini that have been digitised.
The name of each geometry is given in the format t_YYYY_MM_DD_HHmm, where the date and time are derived from the image. Margins from previous images are visualised in blue, and those for the current image in black by default. Note that any modifications to previous margins (i.e. blue lines) will not be logged. See Figure A8.  Common issues addressed in data cleaning and labeling. a) Box method glacier traces are contained within a box that is smaller than the full terminus width at Glacier #224 b) Landsat 7 ETM+ Scan Line Corrector-off image line artifacts at Glacier #291 and c) A single shapefile containing several different glaciers (#27-30) that need to be split manually into separate glaciers to be consistent with the ID scheme. Additionally, all 3 images show varied levels of obstruction of the terminus in the fjord due to ice mélange. Landsat-7 and Landsat-8 images courtesy of the U.S. Geological Survey.       Table 1. Original sources for terminus traces for the TermPicks data set. Spatial coverage describes the number of glaciers and name/region(s) of the traces. Date range are the years covered by the data set. Resolution is the temporal resolution; Annual is approximately one trace per year, sub-annual is more than one trace per year, decadal is approximately one trace every ten years, sub-decadal is more than one trace every 10 years, but not each year. Method is the tracing method used by the author to digitize the terminus. The Author key is the label given to that data set in the TermPicks data set. Automatically assigned scene ID Table 3. Flags assigned to output terminus trace data.

Author
Vertices per km Mean Median Error (m) Median Median Error (m)