Repository logo

Understanding past and future changes in northern Fennoscandian snow cover.



Change log


Vignols, Rebecca Marianne 


In this project, a combination of field measurements, remote sensing data and regional climate model outputs were used to study recent and projected future changes in Northern Fennoscandian snow cover. The research questions considered in this thesis are: What are the uncertainties in remote sensing and climate modelling datasets used in snow studies? How has snow cover been changing since the 1960s and how will it change over the next century, at a regional level over Northern Fennoscandia?

Field measurements were made over two field seasons in the Khibiny Mountains in Arctic Russia. This ground data was used to gain an understanding of snow cover behaviour in the Western Mountain Regions (WMR) of the Kola Peninsula and to ground-truth 500 m resolution satellite data (MODIS: Moderate Resolution Imaging Spectroradiometer) snow products. The overall root mean square error (RMSE) for both MODIS instruments was found to be less than 10 %. The ground-truthed MODIS snow product was then used with station data to analyse past changes in snow cover in the WMR over the past 16 years. Though there is high inter-annual and spatial variability in the long-term snow cover trends in the WMR, overall, the duration of the snow cover season has increased at lower elevations and decreased at higher elevations.

Field measurements and MODIS data were used in the sensitivity analysis of the Weather Research and Forecasting (WRF) regional climate model. Twelve experiments with different physics parameterisations were run over the first field season, and a statistical scores evaluation was undertaken to determine the optimised parameter setup for modelling snow in the region. Three CMIP5 (Coupled Model Intercomparison Project 5) models were used to force WRF in historical (1990 - 1999) and two future climate (2090 - 2099) emission scenarios over Northern Fennoscandia. Outputs from the historical runs were compared to data from 10 stations across Northern Fennoscandia in order to further validate WRF. WRF makes excellent temperature estimates, with a mean bias in the yearly mean temperature outputs of the runs of -1.89 °C. The precipitation outputs are less accurate with values often higher than observations, especially for extreme precipitation events (CMIP5 ‘ensemble’ mean RMSE of 24.0 mm for 20 + mm precipitation events).

Finally, the future runs were compared to historical runs to study projected future changes in temperature, precipitation, snowfall and snow cover. The three models give a range of different future predictions for regional climate change over Northern Fennoscandia. However, all CMIP5 models agree that in both emission scenarios mean snow cover duration will be lower over 2090 to 2099 than it was between 1990 and 1999. Importantly, changes in temperature, precipitation and snowfall are all higher, and snow cover is most impacted, in the higher emission scenario. RCP 8.5 consistently sees a higher decrease in solid precipitation than RCP 4.5 at all stations, and for all models and seasons, for example. Thus, aiming to reduce greenhouse gas emissions is still crucial to reducing anthropogenic impact on Northern Fennoscandian snow.





Marshall, Gareth
Rees, Gareth


snow, climate change, arctic, snow cover, remote sensing, climate modelling, regional climate modelling, polar, northern fennoscandia, MODIS, Khibiny, Khibiny Mountains, Murmansk Oblast, WRF, Weather Research and Forecasting model, ground truthing, sensitivity study, model validation, ERA-Interim, field work, British Antarctic Survey, Scott Polar Research Institute, Kola Peninsula


Doctor of Philosophy (PhD)

Awarding Institution

University of Cambridge
Funded by NERC PhD studentship NE/L002507/1