In this study, we set out to develop a new algorithm to model the unloading process, based on the conditions that are likely to promote unloading of snow from the canopy. We examined in random order, a series of photographs taken with cameras mounted on towers at the Boreal Ecosystem Research and Monitoring Sites (BERMS) Old Black Spruce (OBS) and Old Jack Pine (OJP) forests, and estimated the relative intercepted snow load and fsnow on a 10 point scale. We then used the changes in relative snow load over time to calculate the e-folding time of the unloading process, and we plotted the results against binned intervals of environmental variables (incoming solar radiation, total incoming radiation, air temperature and wind speed). The resulting plots were used to derive models for the e-folding time of the unloading process, which were then incorporated into CLASS. We discovered that in order to improve the modelled above-canopy albedo, it was necessary to change the way in which fsnow is represented. Rather than using the relative intercepted snow load to represent fsnow, we allowed new snow to refresh fsnow to unity, and modelled the decrease in fsnow over time as the ratio of intercepted snow load to the most recent peak following a snowfall event.
The above changes improved the seasonal trend in the modelled albedo at these forests, but we noticed that the short-term albedo was often not well represented. We investigated this by identifying snowfall events detected in the photographs and by the weighing precipitation gauges. Experience with weighing precipitation gauges at other sites has shown that snow can stick to the side of the gauge orifice, only to fall in later, which delays the measurement. In extreme cases the gauge orifice can cap-over with snow, resulting in a large delay and underestimation of snowfall. It is also possible for a snowfall event occurring between subsequent photographs to go unnoticed, if conditions are such that the snow melts or unloads before the next photograph is taken. We compared precipitation events identified by gauge measurements with those based on photographs, and discovered that indeed more than a third of the events identified by the gauges were not identified in the photographs, and that for the reverse comparison, the agreement was less than 50 %. Indeed, many of the periods of disagreement were during times of warm, high-energy conditions, when false (delayed) gauge readings are likely, and when fast unloading between photographs is likely. A relaxation of the time windows for event corroboration, to allow for clock mismatches and delayed recording of events significantly increased the agreement between gauges and photographs. Comparisons of the winter precipitation with snow survey data suggest that the seasonal precipitation as measured by the weighing gauges is reasonable and not subject to severe underestimation. We believe that the weighing precipitation gauges are adequate for representing the seasonal trends in precipitation, and for forcing a surface scheme, but that care should be taken when validating a model over short time periods of hours to a few days.
![](images/flag.png)
![](images/blue_ribbon.png)