A snow stake is established at each site with a metal fence post driven into the ground to support a 2" PVC pipe that slides over the post. Survey rod decals are placed on most of the PVC pipes except at lower elevation sites where shorter stakes are adequate and stream level metal guys are used as the calibration system. Slope and aspect are recorded for all sites, as well as the initial stake tare reading relative to the ground (not all stakes have zero at the ground level).
Before 2014, depth measurements at the stake were recorded, and high and low depth values were visually estimated only. These data were routinely recorded at the stakes while driving to other field activities. Typically, the Upper Lookout stakes on Rd. 1506 and Rd. 350 stakes were read once every 3 weeks. The 1507 stakes were sampled only once or twice a year. A check sheet was established to keep the collected field data organized.
In 2014, timelapse cameras (mounted on trees with a small overhanging shelter) were tested to capture daily images of the snow stakes. In 2015, they were deployed at all snow stake locations as the primary means of data collection, but not replacing in-person site visits and data collection with check sheets. The cameras are set up to capture 3 images a day at 09:00, 12:00, and 15:00 PST. One of the three images, usually 12:00 by default, is interpreted later during data entry for snow depth and snow cover.
A special trip is also made at least once per year to collect snow water equivalent (SWE) measurements as a way of ground truthing the snow in the vicinity of the stake for snow moisture. The nearby snow stake is measured at the same time. SWE Measurements at each stake are conducted by taking three snow core measurements with the Mt. Rose or “federal” snow tube sampling equipment to establish a relationship between snowpack water equivalence and depth. The snow tube core sampler is pushed down to the soil surface interface, depth is recorded, and snow is extracted. The snow core is measured, any soil plug removed, and the core weighed on a spring balance, with the tare weight of the corer subtracted, and the actual snow weight is converted to millimeters of equivalent water.
The snow tube sampling procedure generally follows the NRCS snow sampling procedure except only three cores are taken. If there is a complete snowpack around the snow stake, 3 points are taken in the immediate vicinity of the snow stake (3 points approximately 1 meter equidistant from stake) and averaged. If the snow coverage is patchy, core sample placement becomes more subjective with limited spots where a useful core can be sampled. In the NRCS snow course a particular point not having any snow would be recorded as zero, but here all of our ground truthing cores have some amount of snow because points are selected for representation of the density of the snow (and not for representative depth). The % snow cover is estimated from camera images to document snow patchiness.
Mt. Rose or “federal” snow core sampler. Note that a new snow core sampler was put into use on approximately March 1, 1999. The new scale can measure greater weights. Generally, comparable snow cores weigh less on the new scale in comparison to the older one.
Wingscapes timelapse cameras, including Birdcam Pro WCB-00019 and TimelapseCam Pro WCT-00121, WCT-00122, and WCT-00126 with SD card and C or AA batteries, depending on the model.
The time-lapse cameras are removed every summer (usually May) and reinstalled in fall (usually November), and because of this, the exact orientation of the cameras differs slightly from year to year. Efforts are made annually to maintain the snow stake sites, such as clearing vegetation, trimming low branches, etc., but there is always some environmental variability. A few stakes have been moved, which is indicated by the position attribute in the data.
To ensure quality data, those assessing data in the field and processing snow stake images are trained by more experienced personnel and given guidance on visual interpretation and data entry. However, with assistance from a variety of personnel over the years, the human factor in interpretation, the difficulty in assessing percent cover from a vertical camera, the limits of equipment in winter conditions, and uncontrollable environmental variables, there is a degree of imprecision in the data that should be acknowledged.
After data entry, there is a stepwise review that occurs on both the data entered online and through spreadsheets and programs by data managers and field personnel. The data is flagged and corrected for issues such as mistakes and typos, illogical values, and inconsistencies. The entry notes are also used to help flag data appropriately. The snow depth is eventually converted to metric units and made available. The snow images and data sheets are also archived.
The snow stake data from the field check sheets were originally entered into a spreadsheet. An in-house C# program was developed to make image interpretation and data entry easier for a quantity of images that can exceed 20,000 in each season. This program is used for sequencing images and entry from the field check sheets. Depth is entered in increments of 0.1 foot, and non-zero values less than 0.05 feet are interpreted as a ‘trace’ amount of snow. In some cases, high and low depths were estimated from the images as well, but this was discontinued following snow year 2017 due to its difficulty and low confidence in the numbers produced.
Snow cover is considered only for the relevant open or closed canopy areas with consideration for site specific factors such as vegetation, roads, tracks from vehicles, field of view, camera angle and height, and “stake” well or mound. For example, the snow stake can accelerate melting (e.g. albedo effect), which creates a depression around the stake (i.e. “stake” well), or through wind, can cause a small snowdrift or snow to mound up around the stake, issues which are accounted for by assessing the area around the stake to compensate for its effects. Data obtained from blurry or obscured images (with variability in megapixels between camera models) is sometimes inferred when reasonable and flagged as Q (questionable or estimated) and left blank when not able to read.
