Adaptation Actions for a Changing Arctic: Perspectives from the Barents Area

60

Adaptation Actions for a Changing Arctic: Perspectives from the Barents Area

Box 4.1 Where does our knowledge come from? Atmospheric models and downscaling There are multiple sources for our knowledge of the Arctic climate, including past observations, model results, analysis, and experience. Observations are necessary for empirical analysis as well as for model validation. Although the long- term ground-based observational network is sparse in the Arctic, the World Meteorological Organization (WMO) Polar Prediction Project (PPP) has made some progress in connecting models and observations (Jung et al., 2016). Inoue et al. (2015) proposed a cost-benefit observing frequency of Arctic radiosonde observations. Observations derived from instruments on satellites provide better spatial coverage than the land-based network of in-situ measurements, but do not provide long-term records since many satellite missions are recent. Historic data provide information about climate phenomena, how they are connected, and their sensitivity to changing conditions (Benestad et al., 2016). A rough view of possible future climates is provided by the range of global climate models (GCMs), but these model results must be seen in the context of empirical data such as observations. The GCMs do not provide local and regional detail, and their results must be downscaled to provide a more detailed view of regional and local climate. There are two main downscaling approaches, each with different strengths and weaknesses (Takayabu et al., 2016). The first involves regional climate models (RCMs) with around 10–50 km resolution runs. The higher resolution and adapted physics of RCMs, which includes coupling between the atmosphere, sea ice and ocean, is suitable for studying regional Arctic change and related processes. The Arctic RCMs intercomparison project (Arctic CORDEX) aims to provide high resolution projections and to quantify their uncertainty. However, a large RCM ensemble with forcing from many GCMs must be used to achieve robust conclusions with uncertainty estimates. The other approach, empirical-statistical downscaling (ESD) requires little computer power but needs long historical records for model training. One problem with ESD is that there are few long data records available from the Arctic. While ESD can only provide a similar type of information to that provided by data sampled historically, it does facilitate a means for downscaling a large ensemble of different GCMs, and so can capture the range of natural variation. ESD also makes use of information from different sources to that of RCMs: professional experience, GCMs, empirical data/observations, and statistical theory. Ocean models and downscaling Downscaling two different GCMs for the present climate (20C3M) brings the results closer to observations than the ocean components of global models.The results obtained and compared include sea ice, salinity, temperature, ocean volume and

heat transport. These findings were made in the NorClim and RegScen projects regarding high-resolution regional downscaling experiments for the Barents Sea, based on two CMIP3 global models (GISS-AOM, NCAR-CCSM). Scenarios were downscaled using a ‘delta-change’ type approach (whereby changes simulated by models are added to the observed state and where the models do not necessarily correctly represent the present state, but despite systematic biases it is assumed that the simulated change is nevertheless plausible).The Norwegian Institute for Marine Research has also conducted new simulations for the CMIP5 experiment following RCP4.5 based on the NorESM model. A Russian- Norwegian project ‘Isfjord’ (Gjelten et al., 2016) will focus on coastal aspects.

iStock.com/3DSculptor Olaf Doering / Alamy Stock Photo B&C Alexander / ArcticPhoto