In this study we make use of a series of digital photographs taken using cameras mounted on towers at the BERMS (Boreal Ecosystem Research and Monitoring Sites) Old Black Spruce (OBS) and Old Jack Pine (OJP) sites, along with meteorological data, to derive relationships between the rate of change in the intercepted snow mass (I) and coverage (Fsnow) and local weather conditions. The new interception and unloading algorithms are employed in the Canadian Land Surface Scheme (CLASS) Version 3.3 and the modelled snow water equivalent (SWE), and albedo are compared with observations, and with simulations employing algorithms used in previous versions of CLASS.
The results show that interception algorithms from early versions of CLASS benefited from a fortuitous offsetting of errors, with the effects of a small interception capacity, offset by a lack of unloading. Subsequent algorithms offering incremental improvements to specific aspects of the interception and unloading processes do not always show improved agreement between the model and observations because errors no longer offset one another. Increasing the interception capacity (I*) suppressed the range in modelled albedo because Fsnow is modelled as I/I*. Initial estimates of the timescale of the unloading process were too long; the e-folding time was initially set to 10 days but analysis of above canopy albedo and photographs showed that 1-2 days was more appropriate. At OBS, total incoming radiation was the most important variable for modelling the unloading rate, followed by global solar radiation and air temperature. At OJP, air temperature was the most important variable, followed by total incoming radiation and wind speed. Initial implementations of an unloading rate modelled based on meteorological conditions did not show an obvious improvement over a time-based unloading rate. Canopy photographs also showed that modelling Fsnow as I/I* results in an underestimate because of the widespread coverage that results from new snow. Employing a polynomial of the average relationship derived from photographs improved the modelled albedo. Modelled SWE is sensitive to whether or not unloading takes place, to large changes in the unloading rate and to the interception capacity, but is less sensitive to the parameters used to model the unloading rate.