
Fig. 1
General weather situation in Europe on 28 January 2012 12 UTC. Mean sea level pressure (contours) and 2-m temperature (shaded) are based on ECMWF analysis. The boxes show selected observed temperatures. The black rectangle box shows the area of interest of this study. (Source: Finnish Meteorological Institute)

Fig. 2
Observed 2-m temperature at two stations, Konnunsuo (02733) and Tohmajärvi (02832) in Eastern Finland from 24 January to 5 February 2012. The distance between stations is about 150 km.

Fig. 3
NOAA AVHRR thermal IR images over Finland and Karelia on 28 January 06 UTC (a) and on 29 January 00 UTC (b) 2012. The low-level cloud cover, shown with dark-grey shades, spreads first northward (a) and later westward (b) from Lake Ladoga. In the single-channel images, the cloud over Lake Ladoga cannot be distinguished from the dark water surfaces. The stations Konnunsuo and Tohmajärvi, referred to in Fig. 2, are marked with red and blue dots, respectively.

Fig. 4
MODIS-Terra colour composite image over Lake Ladoga on 28 January 2012 9:20 UTC. The dark colour in the northern part of the lake shows the ice-free part. South of it, an area of fractional ice can be seen. The white plume of the low-level cloud cover on the northeastern part of the lake is spreading northward.

Fig. 5
Simulated ice concentration (%, scale at bottom) and surface water temperature (°C, scale on the right) on 28 January 00 UTC from the three experiments: OLD (a), TRU (b), and NHA (c). The red dots marked with 1 and 2 in (a) show the two observation stations: 1=Ilomantsi, 2=Joensuu (discussed in Sections 4.2 and 4.3). Point L over Lake Ladoga shows the gridpoint discussed in Section 4.3. The embedded small map in (a) shows the whole integration area of the experiments and the red rectangle box the area of interest of this study.

Fig. 6
Six-hour forecasts of instantaneous low-level cloud cover (octas, upper panels) and screen-level temperature (°C, lower panels) from the three experiments OLD (left), TRU (middle), and NHA (right). The analysis time (starting time of the forecasts) is 28 January 00 UTC. The red dots denote the two observation stations, Ilomantsi and Joensuu (see Fig. 5).

Fig. 7
Same as Fig. 6 but for 24-hour forecasts.

Fig. 8
Observed and predicted 2-m air temperature (°C, blue line, left y-axis) and instantaneous low-level cloud cover (octas, green bar, right y-axis) at Ilomantsi (WMO station number 02939): Observed (a), predicted by OLD (b), predicted by TRU (c), and predicted by NHA (d).

Fig. 9
As Fig. 8 but at Joensuu (WMO station number 02928).

Fig. 10
Twenty-four-hour predicted temperature (solid line) and dew point (dotted line) (a) and wind speed (b) profiles at the gridpoint over Lake Ladoga marked with L in Fig. 5 from three experiments: OLD (red), TRU (green) and NHA (blue). The valid time of the profiles is 29 January 00 UTC.

Fig. 11
Predicted surface fluxes (unit Wm−2, left y-axis), averaged over 3 hours, and temperatures of ice and water (unit °C, right y-axis) during 28 January 2012 at the gridpoint over Lake Ladoga marked with L in Fig. 5 as given by the experiment TRU. No clouds were predicted by TRU at this location. All fluxes are denoted positive towards the surface, both from above and below.

Fig. 12
As in Fig. 11 but for Ilomantsi and temperature of forest surface instead of LSWT. Cloudy time period is shown with shading. Ice temperature represents the local small lakes in the Ilomantsi gridbox. Note that the scale in the y-axis is different to Fig. 11.

Fig. 13
Bias and rms-error of 2-m temperature as a function of forecast length for the period from 25 January to 5 February and averaged over 10 observing stations westwards and closer than 250 km from Lake Ladoga. Forecasts from 00 UTC and 12 UTC analysis are included in the statistics.

Fig. 14
Scatterplot of observed temperature vs. predicted temperature for the same 10 stations as in Fig. 13 for OLD (a), TRU (b) and NHA (c). The time period is from 25 January to 5 February and +06-, +12-, +18- and +24-hour forecasts are included. Colours show the number of cases at each point as shown in the legend.
