1Leo-Juhani Meriö*, 1Pertti Ala-aho, 1Hannu Marttila, 1Bjørn Kløve, 2Pekka Hänninen,
2Jarkko Okkonen, 2Raimo Sutinen
1Water Resources and Environmental Engineering Research Group, University of Oulu, FINLAND
2Geological Survey of Finland, FINLAND
*jmerio@student.oulu.fi
KEYWORDS
Snowmelt, snow measurements, spatial variability, degree-day model, low-cost temperature loggers
1. INTRODUCTION
The snowpack properties typically have large spatial variability depending on the catchment and climate conditions. The measurements are usually discontinuous and the network is often very sparse due to high resource needs. The spatial and temporal accuracy of the results depend of the variability of the snowpack properties and the representativeness of the measurements points. Inexpensive temperature loggers can provide a cost efficient method to get more accurate spatial coverage of the snowpack including its variability in real time. In this study, the low-cost temperature loggers were tested, calibrated and utilized to measure the local and microscale variability of the snowmelt in subarctic fell at Pallastunturi region (Figure 1).
Figure 1. Experiment area and test plot locations. One test plot zoomed at the bottom right.
61
2. METHODS
The loggers were installed on the ground and on fixed height 30 cm above ground in six locations with different terrain type, topography and vegetation (Figure 1). Daily standard deviation of the logger temperatures (Figure 2) were used to determine the melting processes and rates. Snow depth data from adjacent measurement stations equipped with acoustic snow height sensors were used to verify the results.
Air temperature and precipitation data and melt rates determined using the temperature loggers were used in an empirical degree-day model (DeWalle & Rango 2008, 279-281) to study the validity of the results. Snowpack measurements on 16/17th of April 2014 were used to calibrate the model. The goodness of the model was evaluated using root mean squared error (RMSE) between the modelled and measured dates for the ending of the permanent snow cover.
Figure 2. Daily standard deviation of the air and logger temperatures. Snow depth at the adjacent reference measurement site. The vertical lines on the x-axis show the dates when snow depth is 30 cm (red) and 0 cm (green) determined using the loggers
and measured at reference site (blue).
3. RESULTS AND DISCUSSION
The results revealed difference in the timing and variability of the snowmelt (Figure 3). The variability was highest in plots with forest cover (ST2, ST3 and ST7) and lowest in the plot located on an open mire (ST6). The melt timing was earlier in open areas than in forest. Delay in melt timing at the northern slope (ST7) was also observed. The timing agreed reasonable well with the acoustic measurement data and gives additional information about the spatial variability of snowmelt.
62
Figure 3. Determined spatial variability of snow melt timing at snowpack height of 30 cm and 0 cm. GTK reference measurements adjacent to ST2 and ST3 are marked in
blue.
The modelled variability of snow water equivalent (SWE) on experiment winter is shown in Figure 4. The difference between modelled and measured date for the end of permanent snow cover is presented in Table 1. The RMSE at the end of permanent snow cover was 3.74 days.
Smallest error was found on the open mire. In areas with more complex topography and forest the error was larger. Model was also run with median degree-day factors determined from the five test points at each test plot. The RMSE improved to 3.19 days. The error at the open mire was increased whereas in most of the other sites it was decreased.
Table 1. Difference between the modelled and measured date for the end of permanent snow cover in days. Overall mean and median: -0.29 & -1 days. Overall SD and IQR:
3.8 & 6 days.
ST2 ST3 ST4 ST5 ST6 ST7
1 -2 -1 4 -5 1 -5
2 -2 NA -2 4 1 -2
3 -5 4 -5 -1 -1 3
4 NA 5 5 2 NA NA
5 -5 NA NA 5 2 -7
Mean / SD -3.5 / 1.73 2.66 / 3.21 0.5 /4.80 1 / 4.06 0.75 / 1.26 -2.75 / 4.35 Median / IQR -3.5 / 3 4 / 3 1 / 7 2 / 5 1 / 0.75 -3.5 / 4.75
63
Figure 4. Modelled snow water equivalent using melt rates determined from each temperature logger pair. Box plots of measured snow water equivalents (orange) and end of permanent snow cover (horizontally in the x-axis). Boxplots on the right show the measured variability of SWE in each test plot. SWE measurements from snow
courses provided by Finnish Environment Institute are shown in greyscale.
Natural variation of solar radiation absorbed by the temperature loggers and consequent heat emission increasing melt rate in the proximity of the logger is assumed to be the largest individual source of uncertainty for the method. Unknown snow properties during the melt were also expected to cause inaccuracy in melt rate determination. Possible displacement of the sensors due to snow pack metamorphosis during the melt can cause additional uncertainty.
4. CONCLUSIONS
The melt rates determined using the low-cost temperature loggers were found to be reasonable accurate on open, flat and relatively homogenous terrain conditions, such as open mire. In more complex topography and forested areas the microscale accuracy decreased, thus usage of median determined melt rates improved the accuracy in test plot (mesoscale) level. The timing of the snow melt agreed reasonably well with the reference measurement sites. Overall, the usage of inexpensive temperature loggers for snow melt measurements was observed to be reasonably accurate method to gain information about spatial variability of the snow melt timing and rates. The method could especially be suitable for areas where available regional snow measurements are not representative and in remote ungauged basins. Using wireless connections with logger’s measurement method could be utilised also in real time operational use.
5. REFERENCES
DeWalle, D. R., & Rango, A. 2008 Principles of snow hydrology. Cambridge University Press.
64