Presentation is loading. Please wait.

Presentation is loading. Please wait.

Seasonal albedo and snow cover evolution of CMIP5 models in boreal forest regions Chad Thackeray CanSISE East Meeting July 25, 2014.

Similar presentations


Presentation on theme: "Seasonal albedo and snow cover evolution of CMIP5 models in boreal forest regions Chad Thackeray CanSISE East Meeting July 25, 2014."— Presentation transcript:

1 Seasonal albedo and snow cover evolution of CMIP5 models in boreal forest regions Chad Thackeray CanSISE East Meeting July 25, 2014

2 Motivation  Previous research found that two features of the canopy snow parameterization within the Community Land Model (CLM4), combine to produce large differences between simulated and observed monthly albedo.  They are the source of a negative bias (~40% weaker than observed) in snow albedo feedback (SAF) over the boreal region.  We found the largest SAF bias to occur in April-May, when simulated SAF is one-half the strength of SAF in observations.

3 Motivation  Previous research found that two features of the canopy snow parameterization within the Community Land Model (CLM4), combine to produce large differences between simulated and observed monthly albedo.  They are the source of a negative bias (~40% weaker than observed) in snow albedo feedback (SAF) over the boreal region.  We found the largest SAF bias to occur in April-May, when simulated SAF is one-half the strength of SAF in observations.

4 Motivation  1) No mechanism for the dynamic removal of snow from the canopy when temperatures are below freezing.  2) When temperatures do rise above freezing, all snow on the canopy is melted instantaneously, which results in an unrealistically early transition from a snow-covered to snow-free canopy.  Since some climate models share common components they are likely to have similar biases (Masson and Knutti, 2011; Flato et al., 2013).  Does this issue exist within other models, and where CCSM4 fits within the hierarchy of CMIP5 models?

5 Methods

6 Monthly Albedo Change  A majority of the CMIP5 models have an albedo decrease that begins one month earlier than observations.  Models that fall outside of the zone of model consensus either struggle with:  The timing of changes  Or the magnitude of those changes.  The ensemble mean reproduces observations from MODIS fairly well during the melt period.

7 Where does CCSM4 fit?  CCSM4 is not only decreasing before the observations, but also well before the multi-model ensemble mean and the model consensus.  The model does well up until when the canopy snow melt begins.  By the observed melt period in spring, CCSM4 albedo has already decreased substantially resulting in the smallest amount of albedo change in Apr-May of any model.

8 Snow Cover Fraction  We also look at SCF because of the direct linkage that it has with model calculations of surface albedo.  A small group of models see a dramatic decrease in snow cover one month before observations (Mar-Apr).  The ensemble mean very accurately captures the timing of snow accumulation and melt.  However, the magnitude of observed changes are much larger than model consensus.

9 Snow Cover Fraction  We also look at SCF because of the direct linkage that it has on model calculations of surface albedo.  A small group of models see a dramatic decrease in snow cover one month before observations during Mar-Apr.  The ensemble mean very accurately captures the timing of snow accumulation and melt.  However, the magnitude of observed changes are much larger than model consensus.

10 Boreal Skill Scores  The creation of this metric allows for model improvements to the treatment of the boreal forest to be measured and tracked.  It tests the models for both their ability to properly simulate the timing of snow accumulation and melt, along with an accurate peak albedo value.  The CMIP5 models are better at simulating SCF (mean of 0.895) over the boreal region than albedo (mean of 0.842). ModelSSalbSSscfSStot ACCESS1.00.876 N/A BCC-CSM1.10.878 0.904 0.891 CanESM20.878 0.912 0.895 CCSM40.845 0.888 0.867 CESM1-BGC0.848 0.895 0.872 CESM1-CAM50.849 0.909 0.879 CRCM-CM50.883 0.893 0.888 CSIRO-MK3.6.00.880 0.896 0.888 FGOALS-g20.840 0.890 0.870 GFDL-CM30.814 0.882 0.848 GFDL-ESM2G0.856 N/A GFDL-ESM2M0.854 0.888 0.871 GISS-E2-R0.911 0.898 0.905 HadCM30.867 N/A HadGEM2-CC0.869 N/A HadGEM2-ES0.885 0.881 0.883 INMCM40.742 0.914 0.828 IPSL-CM5A-LR0.847 N/A IPSL-CM5A-MR0.843 N/A IPSL-CM5B-LR0.841 N/A MIROC50.768 0.895 0.831 MIROC-ESM-CHEM0.714 0.913 0.813 MIROC-ESM0.716 0.914 0.815 MPI-ESM-LR0.847 0.864 0.855 MPI-ESM-MR0.844 0.851 0.848 MRI-CGCM30.864 0.891 0.877 NorESM1-ME0.861 0.909 0.885 NorESM1-M0.858 0.899 0.879

11 Discussion  Lower scoring models in terms of Ssalb (MIROC-ESM, INMCM4) capture the timing of albedo changes, but have max albedos that are far too high (0.5-0.65) when compared to obs (~0.3).  The peak albedo value is important here because a model that is biased high will have a SAF that is too strong (positive bias).  Because the possible decrease from snow-covered to snow-free is larger than observed.

12 Conclusions  CCSM4 falls in the middle of the CMIP5 hierarchy despite its timing discrepancies because of an accurate peak boreal albedo and good simulation of snow cover.  Its very weak albedo change in the melt period makes it the ideal test case to track improvements.  Improvements to the boreal canopy snow scheme within CLM4, as suggested by Thackeray et al. (2014), should result in an increased skill score.  This work has quantified where CCSM4 fits within the CMIP5 hierarchy of models, while also showing that the issues with canopy snow are only also present in NorESM1.  We would have less confidence in assigning skill scores over more northern regions (where SAF biases also exist in CCSM4) because of increased observational uncertainty in satellite retrievals of albedo.


Download ppt "Seasonal albedo and snow cover evolution of CMIP5 models in boreal forest regions Chad Thackeray CanSISE East Meeting July 25, 2014."

Similar presentations


Ads by Google