...

The Copernicus Climate Change Service (C3S) has been set up to provide the European community with climate monitoring products and climate services. Important elements of this service portfolio are reanalysis products, and the various generations of the global "ERA" atmospheric reanalysis products of ECMWF, where the most recent version is ERA5, has seen widespread use for a broad range of applications. In C3S it was also decided to develop and produce an Arctic regional reanalysis, which can add value to and complement the global reanalysis products by adding more detail with a Numerical Weather Prediction (NWP) model system well adapted for use in the Arctic. With climate warming going on more rapidly in the Arctic than lower latitudes, this is of broad interest for describing and understanding Arctic weather and climatology as well as the details in the ongoing change processes in the region.

C3S contract 322 Lot 2 It was aimed to set up and run a regional reanalysis for the European Arctic region at a 2.5 km grid resolution , covering a 24-years period between July 1997 and June 2021 (later the dataset had been extended to cover the period from 1991 to presenttime period from September 1990 to present (including also near real time updates). For convenience we use in the rest of this document the acronym CARRA (Copernicus Arctic Regional ReAnalysis) to refer to this NWP code system and also to the dataset produced.

...

For Greenland, corrections have been made to the land-sea mask, in particular for fjords in the northernmost parts. These corrections are based on coastline and lake data from the Danish Map Supply. Major glacier extent corrections in Northeastern Greenland as used in the Iceland-Greenland operational HARMONIE-AROME are implemented here as well. The glacier extent data in the Danish Map Supply data is of good quality but based on decades old data. Therefore, these have been updated with data from the Greenland Ice Mapping Project (GIMP) (Howat et al. 2014) and OpenStreetmap (Haklay & Weber 2008). The glacier extents are fixed for the whole reanalysis period (1997-2021) at extents as they were approximately in the middle of the period. In Figure 2.2.3.2 the overall effect of the coastline corrections is shown relative to the OpenStreetmap data. Differences in the updated coastline reflect floating glacier tongues that have broken off in recent years. These recent changes, we do not include, since we used fixed glacier extends corresponding approximately to the middle of the reanalysis period.

...

For Greenland and areas in Arctic Canada the Leaf Area Index (LAI) climatology has been updated. In Figure 2.2.3.3 examples of these updates LAI data are shown for three days during summer. These LAI data are based on MODIS (Moderate-resolution Imaging Spectroradiometer) MCD15A2H C6 multi-year mean values (Yang et al. 2006; Yuan et al. 2011) and have been used to update the ECOCLIMAP cover types for Greenland as shown in Figure 2.2.3.4.

Figure 2.2.3.3: MODIS MCD15A2H v.6 Leaf Area Index (LAI) climatology maps (multi-year means) for Greenland and parts of Arctic Canada centered around three days during summer. Data used to generate seasonal change of LAI for different land-cover classes in ECOCLIMAP-II.

...

Daily gap-filled 500 m resolution broadband albedo are used for all glaciers in both of the CARRA domains with the exception of the smaller glaciers in Northern Scandinavia. The MOD10A1 C6 product from the Moderate-resolution Imaging Spectroradiometer ( MODIS ) on the Terra satellite is used for the 20 year period from 2000 to 2019. The Geological Survey of Denmark and Greenland (GEUS) provides a gap-filled and denoised albedo product (Box et al. 2017). This has been tested against in situ measurements on the Greenland ice sheet and found to have a root mean square difference of 0.08 for stations in the ablation zone, and 0.05 for stations in the ice sheet accumulation area (Box et al. 2017). For comparison, in ERA5 the forecast albedo for all glaciers is assumed to be 0.85. In Figure 2.2.5.1 and in the close-up in Figure 2.2.5.2, it is illustrated how much the satellite-derived albedo deviates from that constant value. Also in the left-hand plot of Figure 2.2.5.1, it can be seen that the glacier extents in northern and eastern Greenland and in Svalbard are excessively large. In the left-hand plot in Figure 2.2.5.2 an example of the albedo data for Greenland are shown in grayscale.

Figure 2.2.5.1: Comparison of albedos from the 31st of July 2012 as used in ERA5 (left) and as available from the MOD10A1 C6 product (right). Both plots have the same colour scale as shown in the lower right corner. Only the MOD10A1 C6 albedos over glaciated areas are used in the CARRA reanalysis.

...

Table 2.2.6.1: Overview of SST and SIC input data in CARRA

The SIC fields have been extrapolated along the coasts to cover the fjords with reasonable values. In the processing chain SIC<15% is defined as no ice for the Baltic Sea ice, the OSISAF and the SICCI product. The interface occurring, when merging the Baltic Sea SST product with the ESA CCI SST product, has been smoothened by averaging along the interface. The SST and SIC consistency analysis performed in relation to the development of the CARRA reanalysis showed that L4 SST can be a beneficial filter to remove spurious ice in the sea ice products. The reason is that the SST fields in general have fewer coastal issues, due to the use of higher resolution infrared observations. Sea ice (outside the Baltic region) is removed if the SST exceeds a given threshold in either of the two global SST products (ESA CCI SST and CMC). The SST threshold differs slightly between the two products (depending on the SST average and trend within the domain for each product) and varies linearly over time from about 1.5 to 2.5°C for the period 1995-2017.

...

Some countries have in addition to the network of synoptic stations also a network of stations observing e.g. precipitation and often also snow depths. On an earlier initiative from ECMWF, snow depth observations from these networks have been made available on GTS; from Swedish "climate" stations since December 2010 and from Norwegian "precipitation" stations since March 2013. The increased number of snow depths observations from North Scandinavia on GTS is visualized in Figure 2.2.78.1 showing observations used in ERA5 as blue circles and observations available for use in CARRA as red circles for January, 2000 (left) and January, 2017 (right).

...

The EUMETSAT H-SAF snow extent product H-32 (Metop/AVHRR) was also considered for use in CARRA. H-SAF achieved operational status in the autumn 2017. Reprocessed data is available since January 2015. The algorithm uses all daily multichannel images from AVHRR on board Metop satellites to classify land pixels as snow covered, partially covered by snow or snow free. More information about the product is available on the LSA SAF web site (https://landsaf.ipma.pt/en/^{Image Modified}); Product User Manual (PUM), Algorithm Theoretical Basis Document (ATBD) and Validation Report (VR). Comparisons of CryoClim and H-SAF for the spring 2016 show that the products agree quite well on the snow/no snow decision, but might differ in the decision of classified/not classified. For CARRA production, the CryoClim data is ultimately selected in view of its extended coverage for the whole period until 2015. For the period after 2015, arrangement has been made with MET Norway so that data for the later period is also produced.

The Surface scheme SURFEX 7.3 as used in Harmonie 40h1.1.1 does not include a separate glacier model, thus the 1-layer snow scheme is used also over glaciers. As mentioned in 2.2.4, the extent of the glaciers coverage is corrected and updated, and the initial amount of snow (SWE) on the glaciers is set to approximately 10 (9.99) tons/m2 to avoid snow free glaciers and unrealistically hot surface during melting season. The snow amount on the glaciers is also reinitialized yearly at 1. September with 10 tons/m2. For the SWE amounts on the glaciers, snow analysis is not applied.

...

CARRA assimilates bending angle from reprocessed GPS radio occultation data sets as provided by the Radio Occultation Meteorology Satellite Application Facility (ROM SAF), which is a decentralized operational RO processing center under EUMETSAT. The RO data as used in CARRA is available at: http://www.romsaf.org^{Image Modified}

The processing of the scatterometer data in HARMONIE-AROME differs from that in ERA5. The data sets are collected in advance from the EUMETSAT OSISAF. Reprocessed climate data records (CDR) for ERS Scatterometer (OSI SAF, 2017), OceanSAT/OSCAT (OSI SAF, 2018), QuikSCAT/SeaWinds (OSI SAF, 2015) and Metop/ASCAT-A (OSI SAF, 2016) are selected in order to achieve consistent data quality. For the time after April 2014, operational near-real time Global OSI SAF "coastal" wind product based on Metop/ASCAT (OSI SAF, 2019) is utilized. The data is first stored at ECFS and fetched by a separate script during pre-processing step in CARRA cycling. The processing and assimilation of scatterometer data in HARMONIE is described in Valkonen et al. 2016.

...

Table 3.1: Summer verification CARRA (CA) and ERA5 (ER). Verification of MSLP, T2m and W10m for June, July and August 2015 for a selection of regions in CARRA domains. Green (red) numbers are used when CARRA is verifying better (worse) than ERA5. Colored cells are used to highlight larger differences (STD differences larger than 20% and bias difference are larger than 0.25). Asterisks denote significant differences (95th percentile confidence intervals calculated by bootstrapping is not over-laying each other).

Table 3.2: Similar to Table 3.1, but verification is valid for winter 2014/2015 (December, January and February). 

Figure 3.2.5: Time series of Equitable Threat Score, the monthly observed 90%-tile is used as threshold, for a) CARRA-East domain (thresholds varies from ~ 8 m/s in summer to ~ 12 m/s in winter) and b) CARRA-West domain (thresholds varies from ~ 8 m/s in summer to ~ 15 m/s in winter). CARRA in red and ERA5 in blue (including 95th percentile confidence interval by bootstrapping).

...

The CARRA system has been developed to perform a regional reanalysis for the European Arctic regions with a 2.5 km grid resolution. The CARRA data sets will cover the 24-years period between July 1997 and June 2021period from September 1990 to present with near real time updates. The reanalysis is performed on two domains covering Greenland and Iceland on the west side, and Svalbard and Northern Scandinavia on the east side.

...

Appendix: Input data for CARRA

1 Orography and Physiographic Database (PGD)

Orography: Baseline dataset with GMTED2010 in ca 300m resolution + ArcticDem v7.0 over Greenland (2 m) with corrections on gaps using dataset of ArcticDem v6.0 , AsterDem and TanDem X + consistency correction about coast lines in Greenland with land-sea mask data from Danish Map Supply.

...

Soil database (sand/clay): SoilGrids except for Iceland.
Daily gap-filled 500 m resolution albedo data, MOD10A1 C6, for period 2000 to 2019, provided by the Geological Survey of Denmark and Greenland (GEUS) (Box et al. 2017).

2 Lateral boundary

ERA5 4DVAR hourly analyses for key atmospheric parameters (wind, temperature, geopotential height, specific humidity), 31 km resolution.

...

Two different SIC products have been used: the ESA CCI SIC product (SICCI) and the EUMETSAT OSISAF SIC product, OSI-450 , with the former used whenever available (/- 5 days), else the latter is used to fill the gap (/- 5 days).

The baseline SST product (used outside the Baltic region) is the ESA SST CCI Analysis Long Term Product (Merchant et al., 2016), which covers the period 1991 to 2010. From 2011 and onwards the Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA; Donlon et al., 2012) is used for SST input outside the Baltic region.

...