For all the snotel sites we use to calculate our metrics, we want to combine them with both mean and median aggregation. So, for every individual hourly (or daily) metric, we want the mean/median of all values for all sites.
Desired Data Output:
Date
Max Air Temp 24hr
Max Air Temp 72hr
Total Snowfall 24hr
Total Snowfall 72hr
Max Windspeed 24hr
Weighted Snowfall 96hr
Min Temp Diff 48hr
Was Heavy Snowfall 24hr
Was High Winds 24hr
Sum Max Temp 72hr
Delta SWE 24hr
Yesterday Danger
Danger Rating
1/1/2000
32.0
33.0
13
23
27
40
-10
True/1
False/0
63
0.5
MODERATE/2
LOW/1
(These are absolutely random values)
Data Points and Why They Are Necessary:
Nov 1 2022 - Apr 30 2023
Max Air Temp 24hr: Maximum air temperature in the past 24 hours
Max Air Temp 72hr: Maximum air temperature in the past 72 hours
This is similar to 1, but shows warm temperatures over a longer time frame.
Yesterday Danger: Previous day's danger rating
If the previous day had a high danger rating, the current day will likely have a high(ish) danger rating.
Ideally we wouldn't use this, because it would be cool to generate predictions based only on weather, but we should make our prediction good before optimizing completely.
Total Snowfall 24hr: Total snowfall in the past 24 hours
Lots of new snow is a red flag.
Total Snowfall 72hr: Total snowfall in the past 72 hours
Prolonged high snowfall adds a lot of stress to the snow. Imagine putting a brick onto a pile of packing peanuts that's sitting on an incline: the heavier the brick the more peanuts it will displace when the friction gives out and the pile (of peanuts/snow) begins to slide.
Max Windspeed 24hr: Maximum wind speed in the past 24 hours
High wind creates "wind slabs": blocks of cohesive snow that stick together. These blocks of snow sit on top of older snow, and as in the above example, they are another "brick on top of packing peanuts".
Was Heavy Snowfall 24hr: Threshold snowfall: was there > 7cm (convert to in) in the past 24 hours
This shifts us from linear/numerical to categorical data. We might not be able to use both, so this is a less important metric. Basically, there are "thresholds" that if they are passed it is an immediate red flag. This is one of them.
We should try with both a 12 in threshold, which is the standard for avalanche forecasting, and a 7 cm threshold, which is the value that one of the DS projects identified as being the best measure.
Was High Winds 24hr: Did the wind speed exceed 20 mph in the past day?
This is another boolean value like the one above that we may or may not be able to use.
As like the max windspeed metric, high winds can be a bad sign
It would also be cool to try this with:
Use 25 mph instead
Use maximum gust instead of speed
Delta SWE 24hr: Water content of new snow (use SWE)
How much did the SWE change in the past 24 hours?
Snow Water Equivalent (SWE) measures how much water actually fell from the sky. This is a measure of the density of new snow. 10" of super light snow could weigh the same as 2" of super dense snow, so this might actually be a better metric to use than actual snowfall, which is just the depth of snow.
Sum Max Temp 72hr: Sum of maximum temperature of the past 3 days
indicative of a warm spell
Min Temp Diff 48hr: Difference in minimum temperature from previous day
This can tell us if there are large temperature swings (rapid warming leads to rapid thawing which creates rivers under the snow that destabilize the snowpack, rapid cooling does some bad stuff too but I can't remember what exactly).
Today's temperature - yesterday's temperature
Weighted Snowfall 96hr: Weighted sum of snow for the past 4 days
1.0 yesterday snowfall + 0.75 two days ago snowfall + 0.5 three days ago snowfall + 0.25 four days ago snowfall
This is similar to total snowfall, but gives a better view of the past few days, with more recent snowfall being more "important" than snowfall 4 days ago.
How Do We Get These Data Points:
mean/median: Can we do this with SNOTEL Report Generator? Or do we have to do ourselves?
It seems that Snotel doesn't like to do both Delta from Previous and Mean/Median, so we have to pick one and deal with the other. I think it's easier to generate reports for every site and combine them ourselves than to generate the mean/median and calculate deltas ourselves.
air temperature maximumValueNoneDaily
Same as above, then calculate max of 3 days ourselves
Probably have to calculate ourselves (shouldn't be terribly hard)
snow depthDelta from Previous ValueNoneDaily
Same as above, then calculate total of 3 days ourselves
Have to calculate ourselves
Use 4 to calculate
Ourselves, probably
snow water equivalentDelta from Previous ValueNoneDaily
Use 1 to calculate
air temperature minimumDelta from Previous ValueNoneDaily
In light of the below information, figure out how to get the "Desired Data Output" dataframe
What Our Data Looks Like At Present:
Data Inputs:
etc,etc for all desired snotel sites
Combining SNOTEL Data
For all the snotel sites we use to calculate our metrics, we want to combine them with both
mean
andmedian
aggregation. So, for every individual hourly (or daily) metric, we want the mean/median of all values for all sites.Desired Data Output:
(These are absolutely random values)
Data Points and Why They Are Necessary:
Nov 1 2022 - Apr 30 2023
Max Air Temp 24hr
: Maximum air temperature in the past 24 hoursMax Air Temp 72hr
: Maximum air temperature in the past 72 hoursYesterday Danger
: Previous day's danger ratingTotal Snowfall 24hr
: Total snowfall in the past 24 hoursTotal Snowfall 72hr
: Total snowfall in the past 72 hoursMax Windspeed 24hr
: Maximum wind speed in the past 24 hoursWas Heavy Snowfall 24hr
: Threshold snowfall: was there > 7cm (convert to in) in the past 24 hoursWas High Winds 24hr
: Did the wind speed exceed 20 mph in the past day?Delta SWE 24hr
: Water content of new snow (use SWE)Sum Max Temp 72hr
: Sum of maximum temperature of the past 3 daysMin Temp Diff 48hr
: Difference in minimum temperature from previous dayWeighted Snowfall 96hr
: Weighted sum of snow for the past 4 daysHow Do We Get These Data Points:
mean
/median
: Can we do this with SNOTEL Report Generator? Or do we have to do ourselves?It seems that Snotel doesn't like to do both Delta from Previous and Mean/Median, so we have to pick one and deal with the other. I think it's easier to generate reports for every site and combine them ourselves than to generate the mean/median and calculate deltas ourselves.
air temperature maximum
Value
None
Daily
snow depth
Delta from Previous Value
None
Daily
snow water equivalent
Delta from Previous Value
None
Daily
air temperature minimum
Delta from Previous Value
None
Daily