Seasonal Snowline Fluctuations on Mt Hood

Using Sentinel-2 imagery to locate and track the location of the snowline on Mt Hood over 7 years. This is my final project for PCC's GEO246 (Remote Sensing).

TLDR

You basically want to look at visualization.ipynb which uses prepared_data.csv and prepared_weather.csv. In most code files there is a variable near the top for the root data folder that you'd have to change. I also do not provide the elevation slices (contour_rings) geodatabase used in some of the arcpy tools.

Poster

Sources

Data	Source
DEM	National Map 3DEP
imagery	SentinelHub
snotel	USDA
weather	NWAC

Misc stats

Stat	Info
data dates:	2017-01-03 to 2024-02-21
number dates:	443
timespan:	2605 days (7.13 years)
data size:	5 GiB
download time:	108 min (about 15s/request)
processing time:	80 min (about 11s/raster)

Public github items

• sentinel hub data request (uses venv from requirements.txt)
  • senthub_request.py
  • senthub_search.py
  • senthub_util.py
• arcpy processing (uses arcgis conda env)
  • tool_calculate_pct_snow.py
  • tool_pct_clouds_by_date.py
  • tool_rename_bands.py
• analysis notebooks (any environment with pandas and seaborn)
  • explore.ipynb
  • visualization.ipynb
• data
  • prepared_data.csv
  • prepared_weather.csv
• other
  • poster_60x35_100dpi.png
  • hood_aoi_32610.geojson
  • requirements.txt
  • README.md (github)

Data fields

prepared_data.csv

Field	Type	Description
date	date/string	Date of observation in %Y-%m-%d format.
contour	integer	The lower end of a 100m elevation 'slice' of the mountain, in meters above sea level.
pct_snow	float	Percent snow cover of the horizontal area of a elevation 'slice'.
pct_clouds	float	Percent cloud cover of the AOI in each observation.

prepared_weather.csv

Field	Type	Description
place	string	Name of observation site.
date	date/string	Date of observation in %Y-%m-%d format.
precip	float	Total precipitation (inches).
temp_avg	float	Mean temperature during the day (°F).
temp_min	float	Minimum temperature during the day (°F).
temp_max	float	Maximum temperature during the day (°F).
snow_water	float	Snow water equivalent (inches). Only at Snotel.
snow_depth	float	Snow depth (inches). Only at (Timberline) Lodge.
elev	integer	Elevation of the observation (meters).

Validation / Verification

Information used:

snowtel 651     https://wcc.sc.egov.usda.gov/nwcc/site?sitenum=651
                45.316667, -121.716667
                1635m (5370ft)
                1980-present
                snow water equiv, min, max, mean temp, total precip, accum precip
NWAC ski telem  https://nwac.us/data-portal/location/mt-hood/
                ski lifts at TL, MHM, SB
                various elev
                2016-present
                temp, wind, humidity, precip, snow depth, pressure

Issues

Mask classification values

• 'usable' area:
  • NDVI <= 0.0
  • Cloud Prob < 50%

NOTE: NDVI <= 0 seems to work well for winter/snow covered conditions, but undercounts 'usable' area (by ignoring bare pumice) in the summer. < 0.15-0.18 would have worked better for summer. The consequence of this overestimating the pct_usable_snow.

(Note: 'usuable' renamed 'countable' in poster.)

Sudden drops in snowline chart

In 2020, 2021, and 2023 there are odd spots in the snowline chart where the elevation suddenly becomes 1300m even though the heatmap shows nothing of the sort. The issue is that in those times, there is no elevation with > 0.4 snow, so .gt().idxmax() returns the first index of the year (which corresponds to 1300m generally) instead of a meaningful result. A lower snow threshold should solve this issue.