(1) The Inversion Lab, Germany;
(2) Ocean Atmosphere Systems, Germany;
(3) Lund University, Sweden;
(4) eolab.dk, Copenhagen, Denmark;
(5) Norwegian Meteorological Institute, Oslo;
(6) Nansen Environmental and Remote Sensing Center, Norway;
(7) Mercator Ocean, France;
(8) ECMWF, UK
One of the tasks of the EU-funded Key Environmental monitoring for Polar Latitudes and European Readiness (KEPLER) project is the identification of research gaps and the development of a roadmap towards an improved European capacity for monitoring and forecasting the Polar Regions. Such a capacity clearly relies on the combination of numerical models with data streams provided by space-borne sensors and in-situ measurements. Within KEPLER a number of observational scenarios are evaluated in terms of their performance in a data assimilation system. In the construction of these observational scenarios we put emphasis on the Copernicus Sentinel satellites with particular focus on the High Priority Candidate Missions (HPCMs) for expansion of the Sentinel fleet.
Our main tool for this evaluation is the Arctic Mission Benefit Analysis (ArcMBA) system, which evaluates in a mathematically rigorous fashion the observational constraints imposed by individual and groups of EO data products by using the quantitative network design (QND) approach. The system was developed within the ESA-funded A+5 study (see http://arctic-plus.inversion-lab.com and the article of Kaminski et al. (2018) available at https://www.the-cryosphere.net/12/2569/2018/tc-12-2569-2018.html). The ArcMBA tool quantifies observation impact (added value) of a (potentially large and heterogeneous) set of observations through the reduction of uncertainties in a set of relevant target quantities simulated by a coupled model of the sea ice-ocean system. The target quantities for the present study are 1-week to 4-week forecasts of sea ice volume (SIV) and snow volume (SNV) for selected regions along the Northern Sea Route and the Northwest Passage as well as for the entire Arctic. Our assessments assume observations are assimilated in April 2015, with the respective 1-week and 4-week forecasting periods starting on May 1. We assume consistently for all observational scenarios that the model provides correct sensitivities. This has two consequences: First, the differences between the observation impacts of the respective scenarios are most pronounced. Second, we provide an optimistic (but consistent) view of the impacts of the respective observational scenarios. Another point worth noting is that our reference for the observation impact is a simulation without assimilation of any observations. This clearly yields higher observation impact than adding a new data stream to a (possibly operational) reference setup that already assimilates many other data streams. Finally the observational uncertainties that we assume for the planned future missions are based on currently available information and plausible, but they can always be refined when further information becomes available.
Our first two observational scenarios address altimeter measurements from the CryoSat-2 (CS2) and the Sentinel 3 (S3) missions. We focus on radar freeboard (RFB) products, because the A+5 study had identified RFB as the data product with the highest impact when compared to Sea Ice Thickness (SIT) or Sea Ice Freeboard (SIFB) products. S3 RFB outperforms CS2 RFB in the selected target regions relevant for marine transportation in the Arctic because of the higher temporal coverage. The larger pole hole of S3 is irrelevant. While this is trivial for the selected target regions relevant for shipping (too far away from the pole hole), S3 outperforms CS2 as well for the Arctic-wide assessment, i.e., the total ice and snow volume in the Arctic (two key variables for monitoring of the state of the Arctic climate system) are better constrained by S3 then by CS2.
The next set of scenarios also builds upon a finding of the A+5 study, which had indicated good complementarity of RFB with snow depth (SND) products. We hence constructed three hypothetical SND products and evaluated them in
combination with the CS-2 RFB product. The first SND product is intended to look like a SND product to be expected from the HPCM CRISTAL (Copernicus Polar Ice and Snow Topography Altimeter). The second SND product is intended to look like a SND product to be expected from the HPCM CIMR (Copernicus Imaging Microwave Radiometer). The third SND product is intended to look like a SND product to be expected
from modelling approaches that combine a dedicated snow model forced by a numerical whether prediction model (reanalysis) with an ice drift product to calculate the temporal and spatial evolution of SND (reanalysis-based product). When combined with CS2 RFB, the CIMR-like and CRISTAL-like SND products yield a strong gain in forecast performance. The same holds for the reanalysis-based product. Although the differences for these assessments are small, CIMR shows the best performance among the three products.
The CRISTAL SND product would be derived from the difference of two freeboard measurements on board the same platform. An alternative is the direct assimilation of two freeboard products into the model. For this purpose we constructed a LFB product that mimics an ICESat-2 product. The combination of CS2 RFB and this ICESat2-like LFB shows the overall best performance for both, SIV and SNV. This is because the assumed accuracy of the LFB (2 cm) was higher than the accuracy of the SND products. Furthermore, a general finding in previous work may come into play here: Assimilation of the raw freeboard products is more beneficial than the assimilation of derived products.
A further observable we evaluate is SST. Our standard scenario is based on the product retrieved by OSI-SAF from infrared (IR) measurements, and is contrasted with a scenario that uses a SST product to be expected from the HPCM CIMR. The performance of the CIMR-like SST product is better than that of the standard IR-based SST product. Although the IR product is more accurate, the better spatial coverage (owing to its capability to penetrate clouds) renders CIMR attractive for predicting SIV and SNV along the shipping routes.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner