Evaluation of Numerical Weather Prediction Models Using Continuous Meteorological Observations at the Iqaluit Supersite, Nunavut, Canada

Ground-based meteorological observations collected at Environment and Climate Change Canada’s (ECCC) Iqaluit supersite (64^oN, 69^oW; Nunavut, Canada) between January and December 2016 are used to evaluate the Global Environmental Multiscale (GEM) models, including the GEM 2.5 km, GEM 10 km, and GEM Global models in a near-Arctic environment.

Forecast skill scores for surface temperature, dew point temperature, wind speed, and precipitation were calculated using automated weather station data and METAR observations at Iqaluit. Numerical weather prediction (NWP) model forecast skill varied between model resolution and seasons through the period of interest. In general, forecast skill decreased with increasing lead time, and increased with increasing model resolution. The GEM 2.5 km model showed the strongest linear correlation for all evaluated variables. Negative daily averaged surface temperature bias was observed for all models for April-December, while positive daily averaged surface temperature bias was observed for January-March; similar results were obtained for the dew point temperature. Daily surface winds were overrepresented by GEM 2.5km, and underrepresented by the GEM 10 km and GEM Global models; the bias was largest in November-December-January. The timing of precipitation events was the main source of forecast error, especially during summer, with GEM 2.5 km under forecasting precipitation and GEM 10 km over forecasting precipitation.

Evaluation of NWP forecast was further addressed using continuous measurements from a ground-based ceilometer, Doppler Ka-band radar, Doppler lidar, and present weather detection instrument. Continuous Doppler lidar and Ka-band radar measurements detected the presence of stratified wind layer events that occur from near the surface up to 7.2 km AGL. This atmospheric feature was underrepresented by ECCC NWP models. Analysis showed that NWP forecast bias was amplified or reversed for all models and all surface variables during wind layer events. This change in forecast skill suggests that stratified wind layer events influence surface conditions at Iqaluit by affecting vertical mixing of the atmosphere and influencing the radiation balance. A better understanding of the mechanisms associated with stratified wind layer events is required to incorporate this atmospheric feature in NWP models and, ultimately, improve forecast skill over Iqaluit, Nunavut.