On the representation of heavy lake-effect snow events for the Laurentian Great Lakes region in a Regional Climate Model

In my previous post (http://cnrcwp.uqam.ca/content/impact-lakes-projected-changes-surface-climate-and-hydrology), I had discussed the impact of lakes on projected changes to the near surface atmospheric fields and streamflows. There, all the lakes were parameterized using a 1D lake model. The results confirmed that lakes are important components of the climate system and affect regional climate by modulating surface albedo, surface energy and moisture budgets. Therefore, they should be realistically represented in climate models. Many climate models are currently representing lakes interactively using 1D models. However, for large lakes such as the Laurentian Great Lakes, 3D models are required, as it is important to simulate the circulation patterns which can impact lake temperature as well as ice freeze/melt onset dates and fractional coverage, and lake-effect snow extent and amount as suggested by recent studies. The aim of this study is to compare lake-effect snow simulated by a regional climate model (CRCM5: Canadian Regional Climate Model, Version 5) with 1D and 3D models for the Great Lakes.

In this investigation, two CRCM5 simulations at 10 km horizontal resolution are performed and analysed over the Great Lakes region for the 1979-2012 period. The first simulation (CRCM5_HL), where the Great Lakes are handled by a 1D lake model (Hostetler), is used as a base to represent the configuration of recent regional climate modelling studies. The second simulation (CRCM5_NEMO), where the Great Lakes are simulated by a 3D ocean model (NEMO), is used to assess the impact of representing circulation in large lakes. The model outputs and available gridded observation datasets are then used to investigate heavy lake-effect snowfall (HLES) in the region. The algorithm, used to identify the HLES events, is based on the one presented in Notaro et al (2015). Briefly, it considers heavy snowfall events at points near the lakes where wind blows from the parts of the Great Lakes that are not covered by ice (i.e. ice fraction is less than 70 %). Additionally, to make sure that the snowfall is amplified by lakes, considered snowfall should exceed the average non-local snowfall (average over the 500-km radius region around each point, excluding the 100-km lake-effect zone around lakes) by 4 cm/day.

The summer 2-m air temperature and lake ice cover are overestimated in CRCM5_HL, which are improved in CRCM5_NEMO. The lower ice cover in CRCM5_NEMO leads to higher snowfall amounts during heavy lake-effect snow (HLES) events (Figure 1). Although the HLES values simulated in the case of CRCM5_NEMO are lower than the ones diagnosed from the observation data, they are much closer to that observed, compared with CRCM5_HL, which highlights the importance of large lake parameterization different from that for smaller lakes for better representation of the physical processes leading to HLES events.

Figure 1. Accumulated lake-effect snowfall (cm) for the November-January months of 1998 derived from the CRCM5_HL simulation (CRCM5 configuration, where the Great Lakes are simulated using 1D lake model Hostetler), from CRCM5_NEMO simulation (CRCM5 configuration, where the Great Lakes are simulated using 3D ocean model NEMO) and from gridded observation datasets of 2m air temperature, precipitation, lake ice fraction and 10m wind. Observed temperature and precipitation values are taken from the ANUSPLIN and UDEL datasets. Observed lake ice data are from CIS and NIC. The wind fields used to derive observed lake-effect snowfall are from the ERA-Interim reanalysis.

Add new comment