Search

In order to use the best suited snow models to investigate snow conditions at ungauged sites and for a changed climate, we have tested four snow models for 17 catchments in Norway. The Crocus and seNorge models are gridded whereas the Distance Distribution Dynamics (DDD) model with its two versions, DDD_CX and DDD_EB, is catchment based. Crocus and DDD_EB use energy balance for estimating snowmelt and SeNorge and DDD_CX use temperature-index methods. SeNorge has calibrated the temperature-index against observed snowmelt, whereas DDD_CX has calibrated the temperature-index against runoff. The models use gridded temperature and precipitation at 1 h resolution for the period 2013–2016. Crocus needs additional forcing from a numerical weather prediction model, whereas DDD_EB calculates the energy-balance elements by using proxy models forced by temperature and precipitation. The threshold temperature for solid and liquid precipitation is common for all the models and equal to 0.5 °C. No corrections of precipitation or temperature are allowed. The snow simulations are validated against observed snow water equivalent (SWE) and against satellite derived snow covered area (SCA). SeNorge and DDD_EB perform best with respect to both SWE and SCA suggesting model structures suited for describing snow conditions at ungauged sites and for a changed climate.

Observed trends in annual maximum snow depth (SD) in Norway are analyzed and examined in the context of changes in winter climate from 1961 until today. Trends are evaluated for the 50-year period (1961–2010) and for three 30-year periods (1961–1990, 1971–2000, 1981–2010). The analyzed dataset is the most extensive and geographically representative for the country so far, and the analysis gives an up-to-date picture of the recent development in snow accumulation. In regions characterized by colder winter climate long-term trends are found to be positive in general, while short-term trends shift from strongly positive in the first period to predominantly negative in the last period. Variation in SD is here mainly linked to variation in precipitation. In regions of warmer winter climate variation in SD is dominated by temperature, and long-term trends are mainly negative. Short-term trends start out weak overall in the first period but become strongly negative most places in the last period. It is likely that, although more and more regions in Norway will experience declining maximum annual SD in a projected wetter and warmer future climate, some inland and higher mountain regions may still accumulate more snow in the coming decades.

Trends in time series of snow water equivalent (SWE) are analyzed for different time periods and as a function of elevation. Since 1914, hydropower companies have measured snow at the expected snow maximum. Most of the measurements are located at high elevations, 700–1,700 meters above sea level (masl), providing a unique dataset in areas poorly covered by meteorological stations. For single stations, significant positive and negative trends are found using the Mann–Kendall trend test. The trends depend on region, observation period and elevation. In southern Norway negative trends are found for sites below 1,350 masl for the periods 1931–1960 and 1991–2009, and below 850 masl for the period 1961–1990. Above these elevations, positive trends are found. For the entire period, 1931–2009, positive trends are found for stations located above 850 masl. These findings can be explained by varying trends in precipitation and temperature in winter and spring. For central and northern Norway, the pattern in trends is similar, although less pronounced. High and low values of the North Atlantic Oscillation (NAO) index correlate differently with SWE. High NAO index is positively correlated with SWE at high elevations and negatively at low elevations. Low NAO index is positively correlated with SWE at all elevations.

A lake model (MyLake) has been used to simulate the impacts of the projected future (2071–2100) climate on the thermodynamic properties of a shallow and a deep lake in Finland. The model has been calibrated using a Bayesian Markov chain Monte Carlo (MCMC) simulation method. The model results show, among others, that the largest change in water temperature in a projected future climate occurs in April–May when a 4.6–7.6°C increase in epilimnion and 2.7–4.4°C increase in whole-lake monthly mean temperature occurs compared to the climate of 1961–1990. This corresponds to a large decrease in the probability of ice cover in March–April. In winter and early spring a negative correlation was obtained between lake water and air temperatures in a projected future climate due to the thermal buffering feature of the lake ice and snow cover. The uncertainties connected to the choice of the general circulation model and the boundary condition forcing for the regional atmospheric climate model were often significantly larger than the uncertainties connected to MyLake model parameter values.