The spatial and temporal consistency of seasonal air temperature and precipitation in eight widely used gridded observation-based climate datasets (CANGRD, CRU-TS3.1, CRUTEM4.1, GISTEMP, GPCC, GPCP, HadCRUT3, and UDEL) and eight reanalyses (20CR, CFSR, ERA-40, ERA-Interim, JRA25, MERRA, NARR, and NCEP2) was evaluated over the Canadian Arctic for the 1950-2010 period. The evaluation used the CANGRD dataset, which is based on homogenized temperature and adjusted precipitation from climate stations, as a reference. Dataset agreement and bias were observed to exhibit important spatial, seasonal, and temporal variability over the Canadian Arctic with the largest spread occurring between datasets over mountain and coastal regions and over the Canadian Arctic Archipelago. Reanalysis datasets were typically warmer and wetter than surface observation-based datasets, with CFSR and 20CR exhibiting biases in total annual precipitation on the order of 300 mm. Warm bias in 20CR exceeded 12°C in winter over the western Arctic. Analysis of the temporal consistency of datasets over the 1950-2010 period showed evidence of discontinuities in several datasets as well as a noticeable increase in dataset spread in the period after approximately 2000. Declining station networks, increased automation, and the inclusion of new satellite data streams in reanalyses are potential contributing factors to this phenomenon. Evaluation of trends over the 1950-2010 period showed a relatively consistent picture of warming and increased precipitation over the Canadian Arctic from all datasets, with CANGRD giving moistening trends two times larger than the multi-dataset average related to the adjustment of the station precipitation data. The study results indicate that considerable care is needed when using gridded climate datasets in local or regional scale applications in the Canadian Arctic.

No abstract

While surface station observations of downwelling radiation offer accuracy at high temporal resolution, they do not easily allow an evaluation of model surface radiation budgets (SRB) over a wide geographical area. We evaluate three gridded SRB data sets against detailed observations from six surface radiation sites from the US surface radiation (SURFRAD) network. We subsequently use the most accurate surrogate observational data set for evaluation of modelsimulated SRB. The data sets assessed are: ERA40 -reanalysis of European Centre for Medium-Range Weather Forecasts (ECMWF), North American Regional Reanalysis (NARR) -regional reanalysis of National Centres for Environmental Prediction (NCEP) and the surface radiative budget (SRB) from the International Satellite Cloud Climatology Project (ISCCP). Due to varying constraints with respect to temporal coverage of each data set, the evaluation period used in this study is 1996-2001, inclusive.The ERA40 downwelling longwave radiation (DLR) appears the most accurate surrogate observation, while both ERA40 and ISCCP show accurate results when the incoming shortwave radiation (ISR) is considered across the annual cycle. Winter DLR is less accurate in ISCCP with a positive bias and lack of very low (<200 Wm −2 ) flux values. The NARR SRB shows a large positive bias in the ISR throughout the annual cycle, linked to a significant underestimate of cloud cover.The ERA40 data are subsequently used to evaluate the simulated SRB in three regional climate models across North America. With respect to solar radiation, cloud cover biases are seen to be crucial, while for longwave fluxes both cloud fraction and in-cloud water content are important to simulate correctly. Inclusion of trace gases beyond H 2 O, CO 2 and O 3 appears necessary for an accurate calculation of clear-sky longwave radiation. Error compensation frequently occurs between the various components contributing to a model total-sky SRB. This is important to consider when trying to identify the underlying causes of errors in the simulated total SRB.