This issue is specifically for the /conus_hydrology/hydroviz/<stream_id> endpoint - the hidden endpoint that feeds values to the hydroviz web app.
I was looking into the ratio difference method and whether or not we should use this for all of our statistics, and I noticed a column in the original data release description table that provides exactly this information (!) See the "difference_method" column in this table which I have looked at a hundred times but did not really notice until now:
https://github.com/ua-snap/hydroviz/blob/main/data/streamflow_statistics_description_table.csv
This suggests that we should be using the ratio difference for all variables except those with units of event count, number of days, Julian date, and the one odd variable (sum_mag) that divides by drainage area, and so is somewhat of a spatial index. We would use the absolute difference method for these variables.
I think this makes sense intuitively when the unit is "days" or something that we don't normally talk about in percentages. For example: If GCM shifts from Julian day 150 to 135 (−15 days) and Maurer baseline is day 120:
Multiplicative: 120 * (135/150) = 108 (−12 days applied)
Additive: 120 + (135 − 150) = 105 (−15 days applied, which preserves the actual GCM shift)
Should we use the guidance from that table instead of using ratio for every stat?
This issue is specifically for the
/conus_hydrology/hydroviz/<stream_id>endpoint - the hidden endpoint that feeds values to the hydroviz web app.I was looking into the ratio difference method and whether or not we should use this for all of our statistics, and I noticed a column in the original data release description table that provides exactly this information (!) See the "difference_method" column in this table which I have looked at a hundred times but did not really notice until now:
https://github.com/ua-snap/hydroviz/blob/main/data/streamflow_statistics_description_table.csv
This suggests that we should be using the ratio difference for all variables except those with units of event count, number of days, Julian date, and the one odd variable (
sum_mag) that divides by drainage area, and so is somewhat of a spatial index. We would use the absolute difference method for these variables.I think this makes sense intuitively when the unit is "days" or something that we don't normally talk about in percentages. For example: If GCM shifts from Julian day 150 to 135 (−15 days) and Maurer baseline is day 120:
Multiplicative:
120 * (135/150) = 108(−12 days applied)Additive:
120 + (135 − 150) = 105(−15 days applied, which preserves the actual GCM shift)Should we use the guidance from that table instead of using ratio for every stat?