A Canadian River Ice Database from National Hydrometric Program 1 Archives

A Canadian River Ice Database from National Hydrometric Program 1 Archives 2 3 Laurent de Rham, Yonas Dibike, Spyros Beltaos, Daniel Peters, Barrie Bonsal, Terry Prowse 4 5 Environment and Climate Change Canada, Watershed Hydrology and Ecology Research Division, 3800 Finnerty Rd., 6 Victoria, BC V8P 5C2, Canada 7 Environment and Climate Change Canada, Watershed Hydrology and Ecology Research Division, 867 Lakeshore Rd., 8 Burlington, ON, L7S 1A1, Canada 9 Environment and Climate Change Canada, Watershed Hydrology and Ecology Research Division, 11 Innovation Blvd., 10 Saskatoon, SK, S7N 3H5, Canada 11 Correspondence to: Laurent de Rham (laurent.derham@canada.ca) 12 13 14 Abstract 15 16 River ice is a common occurrence in cold climate hydrological systems. The annual cycle of river ice formation, growth, decay 17 and clearance can include low flows and ice jams, as well as mid-winter and spring break-up events. Reports and associated 18 data on river ice occurrence are often limited to site and season-specific studies. Within Canada, the National Hydrometric 19 Program (NHP) operates a network of gauging stations with water level as the primary measured variable to derive discharge. 20 In the late 1990s, the Water Science and Technology Directorate of Environment and Climate Change Canada initiated a long21 term effort to compile, archive and extract river ice related information from NHP hydrometric records. This data article 22 describes the original research data set produced by this near 20-year effort: the Canadian River Ice Database (CRID). The 23 CRID holds almost 73,000 variables from a network of 196 NHP stations throughout Canada that were in operation within the 24 period 1894 to 2015. Over 100,000 paper and digital files were reviewed representing 10,378 station-years of active operation. 25 The task of compiling this database involved manual extraction and input of more than 460,000 data entries on water level, 26 discharge, date, time and data quality rating. Guidelines on the data extraction, rating procedure and challenges are provided. 27 At each location, a time series of up to 15 variables specific to the occurrence of freeze-up and winter-low events, mid-winter 28 break-up, ice thickness, spring break-up and maximum open-water level were compiled. This database follows up on several 29 earlier efforts to compile information on river ice, which are summarized herein, and expands the scope and detail for use in 30 Canadian river ice research and applications. Following the Government of Canada Open Data initiative, this original river 31 ice data set is available at: https://doi.org/10.18164/c21e1852-ba8e-44af-bc13-48eeedfcf2f4 (de Rham et al., 2020) 32 33

requirements if others elect to undertake similar effort and highlight potential uses for this river ice database. The paper begins 114 by describing the Study Area and Hydrometric Monitoring Sites followed by the Methodology covering details of the data 115 extraction procedure. The Discussion section summarizes the data and highlights database utility and future research needs. 116 The paper ends with sections on Data Availability, Data Disclaimer and Conclusion. 117 The locations and characteristics of the near 8,400 active and discontinued NHP stations, including their operation and 121 regulation history, are available (in downloadable .csv format) at: 122 https://wateroffice.ec.gc.ca/station_metadata/reference_index_e.html. The CRID includes data on river ice affected water 123 level, associated channel flows and timing at a subset of 196 gauging stations across Canada (Fig. 1). These select monitoring 124 sites are located within 11 of the 13 provinces and territories, and extend over 10 of the 11 Canadian climate regions (Gullet 125 et al., 1992). In the beginning, the database focused on 143 stations with a minimum 20-year record, drainage area greater than 126 10,000 km 2 , and located north of the mean annual 0 o C isotherm (Prowse and Lacroix, 2001). Thereafter, an examination of 127 spring break-up at 136 northern gauging sites was reported (von de Wall, 2011). For the current study, the geographic criterion 128 was expanded south into a "temperate zone" (Newton et al., 2017) and the drainage area threshold was removed. A review of 129 literature and correspondence with WSC staff and provincial flood authorities identified an additional 60 southern sites prone 130 to mid-winter break-up events. Inclusion of these sites resulted in a network of 196 sites with drainage areas ranging from 131 20.4 km 2 to 1.68 x 10 6 km 2 , including both natural and regulated flow conditions, with the latter distributed throughout this 132 range. 133

134
The flow regime at the 150 natural sites has not been affected by any significant upstream waterworks. At the remaining 46 135 regulated gauging stations, predominantly in southern Canada (Fig. 1), flows were affected by instream waterworks, such as 136 weirs, dams and water diversion/abstraction. The majority of natural sites (120) were in operation up to the end of the study 137 period of Dec 31, 2015, while most of the discontinued (30) stations ceased operating in the mid 1990s (Fig. 2). This late 20 th 138 century reduction in the monitoring network has also been reported by others (Lenormand et al., 2002;Lacroix et al., 2005). 139 The regulated sites include 29 homogeneous (entire period of operation regulated) and 17 heterogeneous (natural then regulated 140 flow during period of operation) hydraulic conditions (Fig. 2). The Peace River system, an example of a heterogeneous 141 hydrometric archive, is affected by both climate and regulation and a system of hydro-ecological focus (e.g. Hall et al., 2018;142 Timoney et al., 2018;Beltaos, 2019). A large number of the older stations have periods of inactive operation during 1920 to 143 1960. A few inactive stations resumed operation since shutdown in the mid-1990s (Fig. 2). After removing the 1,012 years of 144 inactive status, the 196 NHP sites considered represent 10,378 station-years of data prior to 2016. Appendix A1 provides a list 145 of all the stations selected for the CRID, including start and end dates and type. Specific CRID locations within this paper are 146 referenced by gauging site name followed by the NHP alpha-numeric identifier in brackets. The various documents and digital hydrometric archives compiled and reviewed for this study include: (1) continuous water-165 level pen recorder charts (before year ca. 2000) during the freeze-up, mid-winter break-up (if applicable) and spring break-up 166 periods; (2) digital files (after year ca. 2000 onwards) with water level data at discrete 5-to 15-to 60-minute interval, some 167 including minimum and maximum instantaneous water level for entire annual period; (3) station descriptions; (4) site visit 168 survey notes, including ice thickness summary files; (5) gauge and benchmark history; (6) stage-discharge (S-Q) relationship 169 tables; (7) annual station analyses; (8) annual water level tables; (9) discharge measurement summaries; and (10) yearly 170 station summary files (year ca. 2003-2009). Archives since 2009 are primarily in digital format extracted from the Aquarius 171 water data management platform, which simplified the data extraction, as compared to reading hand-written notes and pen 172 charts for prior years. The last year of the CRID is 2015 as finalized NHP archival data can be delayed by up to two years 173 while data control protocol is followed. The NHP works with provincial governments and partner organizations at some 174 network stations; therefore archives also include those provided by the governments of Alberta, Saskatchewan, as well as the 175 Centre d'Expertise Hydrique du Quebec (CEHQ). An earlier report (Groudin, 2001) included baseline break-up and open-176 water river information for 16 Quebec sites. Supplementary digital daily water level data for Quebec stations (Table A1;  177 stations with "RIVIERE" in name) prior to ~ 1997 were limited to first water level recording of the day and, thereafter, 178 summaries of 15 minute and daily average water level were provided. Information on discharge and river ice data qualifiers 179 (such as the B dates, discussed below) were gleaned from the following WSC and CEHQ internet sites: 180 https://wateroffice.ec.gc.ca/index_e.html and http://www.cehq.gouv.qc.ca/hydrometrie/index-en.htm. A final note: the vast 181 majority of historical annual water levels (item (8)) are reported by NHP as preliminary since these values were never 182 published. Similarly, some recent digital water level files (item (2)) were also preliminary since NHP had not yet screened 183 these data. 184

185
The evolution of the CRID was comprised of six data collection campaigns since 2000 (Table 1). Major data archival efforts 186 in the years 2000-2001 and 2010-2011 required a team of two to three people visiting up to 8 WSC regional offices, with each 187 visit lasting up to 2 weeks to photocopy and/or scan hydrometric archives. Following that, all paper based information, except 188 for Quebec stations, was digitally scanned and filed to a central electronic repository. This 0.5 Terabyte digital data consists 189 of over 30,000 folders and 100,000 files that is currently stored on a secure Environment and Climate Change Canada server. 190 The CRID digital archive is available on request.  Fig. 3. The CRID 202 includes up to 15 variables that cover the water year (Table 2). These variables are categorized as occurring during one of four 203 seasons: freeze-up, ice cover, break-up, or open-water. For the variables shaded in grey, the objective was to record data on 204 instantaneous water level, associated date and time. These instantaneous values correspond with the water level at the initiation 205 and maximum flood level for ice specific and open water conditions during each calendar year. The procedure for extracting 206 river ice data follows the guidelines of Beltaos (1990), and primarily involves visual examination of water level records. Hence, 207 identification and extraction of river ice data is a subjective process and the resolution to which water level, discharge and 208 event timings were registered is included in Table 2. Depending on the possibility of extracting instantaneous (Table 2, grey  209 shading), mean daily water level or mean daily discharge (HLQ1, HLQ2) based variable, a data quality rating scheme with values 210 of 0, 1 and 2 was used to quantify the continuum of higher to lower data resolution (Table 3). Under some circumstances, 211 judgement was applied to rate data quality higher or lower depending on various circumstances, such as termination of a 212 continuous water level record during the spring break-up season where ice movement, synonymous with variable spring break-213 up initiation (Sect. 3.4.6) damaged the recording instrument. Such data would rate as 0 even though data from the fragmented 214 record rates as 1 on Table 3.  Table 2 for additional information). The variables shaded in grey show the instantaneous water 239 level and associated time when the event occurred Compression of x-axis and vertical exaggeration of y-axis accentuates the 240 water level changes observed during ice conditions. The relative magnitudes of variables and water level pathology should 241 not be considered as typical. 242

Ice Affected Stage-Discharge Relationship and B Dates 261 262
This section highlights challenges related to data collection during the ice season through excerpts from hydrometric program 263 operational manuals, other publications and experience in developing this database. This background information is considered 264 of high value to users when interpreting spatial and temporal characteristics of river ice data. 265 266 A fundamental concept in hydrometry is the stage -discharge (S-Q) relationship. At each NHP monitoring location, a reach-267 specific relationship is established via field surveys. Each year, hydrometric staff complete multiple site visits to measure in 268 situ stream velocity and flow area to calculate discharge for a given water level. This work is ongoing with occasional 269 refinement and adjustment of the S-Q relationship to account for changes in channel morphology and bed roughness -in some 270 cases requiring relocations of station due to loss of stable control section in response to natural and/or anthropogenic impacts. 271 Besides, the open water S-Q relationship is not valid during river ice conditions due to well-known hydraulic effects of ice on 272 flow conveyance. In Canada, ice-influenced flows are identified with a "B" flag to inform the user that the water level is 273 affected by 'Backwater' conditions leading to a higher water level associated with a given discharge on the S-Q curve. The 274 specific river ice condition can take different forms, such as frazil and slush ice, anchor ice, partial ice cover, complete ice 275 cover, ice jams, flowing ice chunks or a mix of these (Poyser et al., 1999). The data user, therefore, has to be aware of these 276 possibilities when using 'B' dates as metric for river ice conditions. In reference to S-Q relationships under ice, Environment showing the ice affected condition is provided in Fig. 4, where the latest time when ice-covered flow can be estimated with a 297 fair degree of confidence is at point A. Under conditions of a stable ice cover, hydrometric staff can apply site-specific 298 methods to estimate the applicable discharge, based in part on sporadic flow measurements during the winter period. Point B 299 in Fig. 4 denotes the last day of backwater, so that after that time discharge can be estimated with very good confidence 300 using the gauge-specific S-Q relationship that applies to open-water conditions. Point C in Fig. 4 approximately delineates 301 the periods of pre-breakup (sheet ice cover, possibly subjected to hinge and transverse cracking) and actual breakup when 302 various events such as ice jams and ice runs generate repeated increases and decreases in the water level that are too sharp to 303 be runoff-generated. For the breakup period, hydrometric staff estimate daily flows by taking into account the general trend 304 of the water level hydrograph, prevailing weather conditions, flows at upstream gauges and tributaries, as well as any in-situ 305 visual observations that may be available. Once the ice cover is fractured, mobilized, and broken up, flow measurement is 306 inhibited by problematic access and safety considerations. Consequently, it is not possible to assign reliable flow estimates 307 during the break-up period, leading to the aforementioned "poor" characterization since there is no way at this time to 308 quantify the reliability of these data. More discussion on this issues are needed to inform the water community of the challenges related to cold-regions 322 hydrometric data collection (Hamilton, 2003) and caution when interpreting study results. The first ever published analysis 323 observations on station locations and the dynamic ice conditions "that the data on river ice should only be considered valid 325 at the gauging station site and may not be transferable to the entire watershed" are applicable to the CRID product. 326 327 328

CRID Variables 329 330
The following sub sections, corresponding to the four seasons of occurrence (Table 2)  (HF) occurred November 9 and these images were obscured by clouds. River channel open water is green and ice cover is 367 white on these true colour images. 368 369 Formation of a channel-wide ice cover is the culmination of various processes that include frazil ice growth, ice pan 370 development, juxtaposition and upstream progression taking place. When the ice cover 'bridges' or is present 'bank to bank' 371 across the river channel the increasing frictional resistance causes a rise in the water level. This initial ice cover progression 372 upstream past the gauge will cause a gradual increase to a maximum in the water level chart and is depicted as HF (freeze-373 over water level) in Fig. 3. The CRID includes transcription of the NHP recorded instantaneous water level, up to the minute 374 timing, date and associated daily discharge, as available are manually extracted and given a '0' rating. Instantaneous discharge 375 during ice conditions is not a NHP data product since the open water S-Q relationship is invalid. If no instantaneous record 376 was available, the lower-resolution daily water levels are used to identify the maximum water level occurring after the First B 377 meteorological site with: "Rainfall or temperature records used for estimating the missing periods or the ice affected periods". 384 It was generally observed, though not recorded, that maximum freeze-over water level tend to occur when temperatures 385 dropped to -10 °C. While ice jamming at freeze-up is a known occurrence (e.g. Jasek, 1999), there was no attempt to 386 distinguish these events in the current exercise due to the complex hydrological and hydraulic conditions affecting these 387 processes. Beltaos (1990) discussed the unlikelihood that a complete ice cover forms at the instant of HF. A later 388 recommendation was to define the freeze-up water level as the average water level for one week after formation of a complete 389 ice cover (Beltaos, 1997). Following this methodology, the CRID includes all available daily water level at HF and the 390 following 29 days to: (1) allow for calculation of a 7-day average to parameterize a water level threshold of exceedance for 391 the ice to detach from channel banks at break-up (Beltaos, 1997) and (2)  level records to determine if they are results of ice cover break-up is a challenge (Beltaos, 1990), especially in the absence of 426 other supporting evidence (e.g. site observations, new reports, flood summaries). Similar to freeze-over interpretation (Sect. 427 3.4.1), the review of daily climate data from nearby stations informs if temperatures exceed 0 o C and associated rainfall 428 occurred. During data extraction it was often observed that mid-winter break-up occurrence corresponded with 10's of cm 429 reductions in daily snow on ground for day(s) prior to the event. A review of the discharge measurement summary (item 9, 430 Sect. 3.1) also increased interpretation confidence towards when station visit remarks were available days before or after the 431 "winter peak" alluding to channel ice condition or if discharge measurements were collected from the ice cover or wading. 432

433
The instantaneous HMWB represents the onset of ice cover movement at a site during the winter season and is identified as a 434 spike on the rising limb of the water level record. The cause of this spike is a rapid decrease in hydraulic resistance as the ice 435 cover breaks and starts moving downstream. This variable cannot be determined from mean daily summaries of water levels. 436 Following the initial break-up event, the water level will typically continue to rise until it reaches a maximum value represented 437 by instantaneous HMWM. For some stations, HMWB and HMWM can occur more than once during a single ice season (e.g. Beltaos, 438 2002). In such cases, only the first HMWB and the highest HMWM are included in the CRID. In some cases, a mid-winter breakup 439 event is followed by a dramatically cold period during which frazil generation is significant. The result may be a very thick The occurrence of ice cover season maximum water levels, not associated with the freeze-up or break-up of the ice cover were 460 identified from the hydrometric archive and input to the CRID. If there was mid-winter break-up event, an attempt was made 461 to extract the first of the 7-day maximum average winter water level (HF2) after the event. As with HF (Sect. 3.4.1), these data since the CRID archive does not have historical pen recorder charts (Sect. 3.2) much beyond the HMWM event. Examination 464 of more recent continuous digital water level records reveals that after mid-winter break-up, limited 'stage up', synonymous 465 to HF was usually observed. This may be due to the lack of complete ice flush down the channel after HMWM. Since large, 466 fragmented ice blocks likely remain in the channel, the hydraulic resistance and refreezing of the ice cover is probably a less 467 dynamic event. Daily water level values after mid-winter break-up generally reveal a pattern of steadily declining daily water 468 levels. Notably, this patterns is likely typical on relatively flat river channels, while on steep river sections, progressive frazil 469 accumulation produced in newly open section exposed to cold could increase water levels even during receding flows. If HMWM 470 was followed by days with no 'B' data flag, HF2 was restricted to days when 'B' data flag appear again. As with the first freeze-471 up events, HF2 and the following 29 days of daily water level were recorded. Water levels within the first 7 days after HF2 were 472 also assessed to extract a maximum (HF2 MAX) daily water level exceeding HF2. This variable may more closely match the 473 instantaneous processes resulting in the HF occurrence 474 475 Maximum winter water level was also recorded at select locations with no mid-winter break-up event. In this situation, the 7 476 day average water level beginning at HF2 exceeds that commencing of HF. This may correspond with a secondary stage up 477 during extreme cold events described by (Hamilton, 2003)  Whenever an HF2 variable was identified, the ice cover period was examined for a second winter-low water level (HLW2) and 508 discharge (HLQ2) event. These data were only added to the CRID if HLW1 or HLQ1 were before HF2. At some locations, several 509 months may have lapsed between the first and second occurrences of winter-low events as shown in Fig. 8. The incident of a 510 second winter-low is probably one of the most understudied events in ice covered channels, while it can have all the water 511 quality and navigation related implications as that of the first winter-low events described in Sect. 3.4.2 above. 512 513 3.4.5 Ice Cover: ITHICK 514 515 instrument maintenance approximately six to eight times per year, which include both open-water and ice-covered conditions. 517 During the latter, a measure related to the solid portion of the ice cover thickness is recorded on the site survey note (item 4, 518 Sect. 3.1). End of ice cover season measurements quantify ice thickness prior to the spring break-up and some cases this may 519 represent a pre-melt ice thickness, a relevant factor in break-up initiation and potential severity (Beltaos, 1997). Measurements 520 prior to ~1995 are generally limited to water surface elevation to bottom of ice cover, thus may underestimate the actual 521 thickness of the ice cover since the specific gravity of river ice is commonly taken as 0.92. Nevertheless, these measurements 522 are assumed to represent the actual ice cover thickness. WSC Regional office and provincial partner protocols for collection 523 and summary of this ancillary ice thickness data differ, while some of the more recent digital data archives may have actual 524 ice thickness measurements. Figure 10  include slush. These five measurements are removed when calculating average river ice thickness. 536 safety considerations. As an example, Fig. 11 shows a time series of 47 average ice thickness data points at one CRID location. 539 Over the time series, the measurement dates range over a 10-week (72 day) time window. In addition to data collection timing, 540 incomplete archival and scanning for the database may also be a reason for missing or wide ranges in time series. Thus, any 541 time series analysis of ITHICK needs to account for this year-to-year sample date variability. While an attempt was made to 542 compile the time series of final (season's end) ice thickness measurements, a more detailed climatological analysis will be 543 required to establish if this measurement was collected prior to the ice cover beginning to melt. ice cover. The associated decrease in resistance to flow registers as a spike on the rising limb of the water level hydrograph 556 (see Fig. 3). Beltaos (1990) indicated that identification of break-up initiation can be uncertain and that it is not possible to 557 establish HB from a record of mean daily water level. Therefore, the timing and magnitude of HB may be less accurate than 558 HM, the maximum instantaneous or daily water level established following HB. Data ratings are provided to indicate the 559 resolution of these events. The Last B Date was the final day with a B data flag (R data flag for CEHQ sites). 560 the ice cover to decay. A slow increase in channel flow will prolong the decay period and resulting water levels do not reach 564 magnitudes much beyond those with similar flow indicated by the open water S-Q relationship. Conversely, a mechanical 565 break-up is characterized by limited reduction in the mechanical strength of the ice cover and rapid increase in channel flow. 566 As the rising flow eventually overcomes the resistance of the ice cover, the latter is mobilized in dynamic fashion and breaks 567 down into slabs and blocks, which eventually are arrested by still-intact ice cover to form ice jams, typically at morphologically 568 conducive locations such as constrictions and abrupt slope reductions. According to an anonymous reviewer, ice jams can also 569 form at morphologically conducive locations even without an intact ice cover stopping the ice run. and persistence, ice jams lodged at or below the gauge site affect the local water levels to a varying degree. A jam lodged 574 upstream of a guage can also have a measurable stage (actual discharge) depressions for several hours before reaching an 575 equilibrium. The release of a jam can generate a sharp wave called a 'jave' (Beltaos, 2013) yet another dynamic mechanism 576 that can generate the identified HM on instantaneous water level recordings. Highly dynamic events, initiated with minimal 577 or negligible ice cover decay, are sometimes referred to as "premature" and typically result from mid-winter thaws 578 accompanied by intense rain-on-snow runoff events (Deslauriers, 1968). It is likely that much of the CRID mid-winter data 579 described above in Sect. 3.4.3 are these highly dynamic events. The less common "overmature" break-up sequence was 580 observed at some CRID stations with less obvious "spiking" of water levels. An example water level with this characteristic 581 on the Peace River in 1982 (Fonstad, 1982) is included in Beltaos (1990) where minor water level perturbations are followed 582 by a generally smooth reduction to open channel conditions. In some cases the HB and HM were interpreted to occur at the 583 same time. available, it is notable that the river ice break-up processes described occur prior to this date. While spring break-up peak 593 water level magnitude and timing in the CRID have high degree of accuracy, classification of events as ice jam or not, was not 594 pursued as this would require local observations and/or photos. The Last B Date is sometimes used to represent break-up for 595 time series analysis (e.g. Zhang et al., 2001; Chen and She, 2019) and a recent publication used B dates and discharge to assess 596 trends in ice jam flooding events (Rokaya et al., 2018). Unlike using the Last B Date as a surrogate and/or index, the water-597 level based data in the CRID provides the science community with a direct and thus more accurate data set towards analysis 598 of spring break-up timing, magnitude and processes. For instance, the identification of HM provides the means to assess change 599 in the flow magnitude driving spring breakup flooding, which would not be possible with discharge analysis alone and/or 600 value at each station along with data quality rating. These data are extracted from the hydrometric archives and are easily 616 verified as NHP web pages generally report both daily and instantaneous maximum annual discharge and timing. In the event 617 of damaged or non-functioning instrumentation, NHP or CEHQ may estimate (data flagged with E) daily discharge values. 618 The S-Q relationship (Sect. 3.1) can be used to estimate the associated water level. Gerard and Karpuk (1979)  The accuracy and precision of extracting water level, discharge and timing of the CRID variables is as follows. For the six 630 grey shaded instantaneous variables in Table 2 (HF, HMWB Station Analysis or published online summaries). The wide-spread use of digital water level recording instrumentation after 637 the year ca. 2000 decreased the temporal resolution (i.e., accuracy) of water level records as data collection interval varied 638 from 5 to 15 to 60 minutes. Some data loggers also recorded hourly to sub-hourly maximum and minimums, which increased 639 the accuracy towards instantaneous events, though selection does require judgement. The vast majority of mean daily water 640 level pages and some of the more recent digital water level recordings were deemed "Preliminary" by NHP. Different methods 641 of collecting requisite information for mean daily water level have existed over the archive from at site station observers who 642 viewed a staff gauge once daily to the more modern arithmetic averages determined from continuous water levels. 643 644 Quality Control (QC) for the CRID has included preliminary data analysis and peer review of associated publications (Table  645 1). CRID station data were initially compiled as single station Excel files which include all extracted water level, discharge, 646 date and time and accuracy rating, average ice thickness along with time series plots for visual identification of outliers. A 647 separate station Excel file contains all available ice thickness measurements and averages calculation. All finalized station 648 data were compiled in to a single .csv file (118 columns x 22,736 rows with 464,891 cell entries) for further QC. This single 649 spreadsheet was examined for data entry errors using the filter and count capabilities inherent to Excel. 650 at 16%. This higher percentage of error is a likely remnant to the multiple rounds of revisions to mid-winter time series and 655 confusion that arises when examining non-consecutive events that can occur across calendar years. For ice seasons when both 656 a First and Last B Date were available, dates were incorrectly transcribed on 7.5% of time series. All erroneous daily discharge 657 and First and Last B Date values were replaced. The remaining CRID data entries are not amendable to automated quality 658 control since they were manually extracted. Based on these QC activities the CRID likely has a 5-10% data interpretation/entry 659 error. The CRID initiation of break-up (HB) time series at site Red River near Lockport (05OJ010) was provided to Becket 660 (2020) who reported: of the 34 years, 3 years of timing were revised based on evidence in newspapers (an ancillary evidence 661 source not included in the CRID), while 2 years were found to be incorrectly interpreted and input to the CRID. One year was 662 12:00 hours too early and one year 2 days too early. While it would be impractical to review the entire database for errors, 663 users are encouraged to undertake their own QC activities and review the data disclaimer in Sect. 7. The data quality ratings open water leads at, upstream or downstream of the gauge, percentage of ice cover at gauge, water flowing between the ice 676 layers and anchor ice at a cross section. While these types of observations are not part the CRID, users should be aware of 677 such factors that add further complexity to wintertime water level interpretation. Furthermore, collection of data using a stilling 678 well (von de Wall, 2011) also could affect resultant water level interpretation. Since river ice processes are site specific users 679 should be aware of possible spatial discrepancy in location of gauge site versus where ice thickness and flow measurements 680 are collected. Access to ice cover and worker safety are field based considerations which can result in a wintertime cross 681 section measurements taken meters or kilometres upstream or downstream from the actual gauge. Another consideration is 682 that many gauges are located near a bridge, which provides a safe platform from which water velocity measurements can be 683 performed. Bridge pilings would change the hydraulics and very likely the ice condition on a river channel such as promoting 684 a thicker ice cover in the deck shadow and promoting ice jamming against abutment or piers. Finally, changes to watershed 685 characteristics such as urbanization and agriculture likely have effects on river ice hydrology. 686 687 CRID users should also bear in mind that all variables were transcribed directly as recorded in the NHP archive. There is no 688 tabulation of: at-station movements, benchmark or datum shifts, or changes to the stage-discharge relationship. Since river ice 689 processes are site specific, prior to time series analysis of phenology or water level data an accounting for these factors towards 690 assessments of station homogeneity are a necessary next step. For example, Fig. 13  In total, the CRID holds 72,595 recorded variables with more than 460,000 data entries of water level, discharge, date, time 716 and data quality rating based on the review of over 100,000 hydrometric archive files. Tabulation of the 6,094 ice thickness 717 measurements required examination on the order of 100,000 cross-sectional measurements and removal of slush affected data. 718 In terms of data completeness, extraction of maximum open-water level (Ho) was the most successful covering 9,705 (94%) 719 of the 10,378 active station years. Similarly, the 8,933 (9,240) first (last) day with backwater due to ice (B dates) and 8,178 720 first minimum winter discharge populate the majority of active station-years and attest to the NHP historical mandate to publish 721 discharge information. Freeze-over water level and maximum spring break-up water level were extracted from 72% and 80% 722 of those years reporting First and Last B Date. This first known attempt to centralize data on mid-winter break-up occurrence 723 includes 467 maximum mid-winter break-up water level and 362 associated mid-winter break-up initiation events. The data 724 quality rating presented in Table 4 confirms that the NHP archives is a high quality source of river ice information with 82% 725 of data rated as '0'. Although some of the data have lower quality ratings, their inclusion increases the population size and 726 helps provide a more complete spatial and temporal coverage over Canada. 727 728 While the CRID represents the largest existing effort to extract river ice variables from hydrometric archives, it does not 729 provide a complete time series of ice events at the near 2,800 active and 5,500 discontinued hydrometric stations in Canada. 730 However, it covers a representative sample with six station types (Table 4)  The CRID can be used for the study of river ice processes and the key characteristics of different ice regimes that are 747 encountered within Canada and how these characteristics may have been changing over time. Break-Up Data From Hydrometric Station Records (Beltaos, 1990). National Hydrometric Program gauge records proved to 788 be very valuable sources of field data for parameterization of ice related hydrologic events on Canadian rivers. This work 789 involved reviewing over 10,000 station years of data from a select subset of 196 stations, covering a range of stream types and 790 climatic regions, to identify and extract recorded data corresponding to 15 variables comprising water levels, discharges, 791 timings, ice thickness, and data quality ratings. Close to 73,000 records of river ice variables are now available to the water 792 research community. For sites not included, the CRID can represent a template to extract pertinent information for various 793 purposes including flood mapping and hydraulic structure design. It is recommended that periodic updates be made to this 794 database since a longer time series record is of more value. Based on the 160 locations in operation up to Dec 31, 2015 (Table  795 A1), a 5 year update of CRID time series (2016-2020) would require 800 person-hours of work. Evaluation of future research 796 priorities are needed to formalize whether this task would be completed by the same group or undertaken by others. It is 797 fortunate that much of the data acquisition tasks, discussed above could be automated using the Aquarius platform currently 798 in use by NHP partner organizations (S. Hamilton, pers. comm). It is also recommended that a tabulation of station movements, 799 benchmark or datum shifts, and changes to the stage-discharge relationship be compiled to rectify the site-specific nature of 800 river ice conditions and non-homogeneous time series. Lastly, the CRID follows on several other notable national and 801 international efforts to compile river ice information. The Global Lake and River Ice Phenology Database to river ice science over the past two decades. The CRID expands on the number of variables considered, as well as, the 805 temporal and spatial scope of these earlier databases for stations in Canada. The work highlights the excellence of NHP 806 agencies in the collection and dissemination of hydrometric data, adds value to the NHP archive and delivers on Environment 807 and Climate Change Canada's commitment to making water science knowledge and data openly available to the scientific 808 community and the general public. The CRID supports continued research on river ice processes and the extreme water level 809 fluctuations common to many cold regions river systems. Environment and Climate Change Canada employs every reasonable effort whenever feasible, to ensure the currency, accuracy 821 and precision of the information provided. However, there are some limitations due to the sources of the data and the 822 technology used in its processing and management. Furthermore, the material or any data derived using the data is subject to 823 interpretation. Users are responsible for verifying that the supplied material is appropriate for the use or application for which 824 they wish to employ it. 825 826