utqiagvik-ros-sar/E_RS_Change_Detection_Report.txt at master · amstringer0303/utqiagvik-ros-sar · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160

================================================================================
UTQIAGVIK EXTREME WEATHER EVENT CATALOG + REMOTE SENSING CHANGE DETECTION
1980-2024 (Events)  |  2015-2024 (Sentinel-2 RS Analysis)
NOAA GHCN-Daily USW00027502  x  Sentinel-2 L2A (Planetary Computer STAC)
================================================================================

PART A: EXTREME EVENT CATALOG  (1980-2024)
--------------------------------------------------------------------------------

DETECTION METHODOLOGY
  Each extreme event-day is detected from the GHCN-Daily station record using
  the thresholds below.  A single calendar day may be flagged for more than one
  event type.  Severity is scored 0.5-3.0 based on the quantitative intensity
  of the primary meteorological variable.

  Event Type      | Threshold                                  | Mode Affected
  --------------- | ------------------------------------------ | --------------
  Rain-on-Snow    | PRCP > 0 mm + TMAX > 0°C, Oct-May         | SM / 4WD
  Rapid Thaw      | 3-day mean TMAX rises >10°C in 3 days,    | All modes
                  | Mar-May or Oct-Nov                         |
  Blizzard        | AWND >= 15.6 m/s (35 mph) + WT09 flag     | SM / 4WD
  Extreme Cold    | TMAX < -40°C                               | Snowmachine
  Glaze/Ice       | WT06 flag                                  | All modes
  High Wind       | AWND >= 12.9 m/s, May-Oct                  | Boat

EVENT COUNTS (1980-2024, 44 years)
  Rain-on-Snow  :   215 days total  (4.9/yr)
  Rapid Thaw    :   124 days total  (2.8/yr)
  Blizzard      :    15 days total  (0.3/yr)
  Extreme Cold  :     9 days total  (0.2/yr)
  Glaze/Ice     :   587 days total  (13.3/yr)
  High Wind     :    37 days total  (0.8/yr)
  Total         :   987 days total  (22.4/yr)

TRENDS (1980-2024, Sen's slope via OLS)
  Rain-on-Snow  : +1.80 days/decade  ***
  Rapid Thaw    : +0.05 days/decade  p=0.867
  Blizzard      : +0.02 days/decade  p=0.779
  Extreme Cold  : -0.18 days/decade  * p<0.05
  High Wind     : -0.12 days/decade  p=0.335

SEASONAL PATTERN
  Winter (Dec-Feb): Blizzard and Extreme Cold dominate.  Peak disruption for
    snowmachine routes.  Polar night prevents optical satellite observation.
  Spring (Mar-May): Rapid Thaw and Rain-on-Snow peak — the highest-hazard
    transition period.  Sentinel-2 optical data starts becoming available in
    April.  This is the KEY WINDOW for RS-based change detection.
  Summer (Jun-Sep): High Wind and coastal fog dominate boat route disruption.
    Best optical data availability.  Sentinel-2 cloud-free probability 40-55%.
  Fall (Sep-Nov):  Rain-on-Snow and Rapid Thaw return.  October is the last
    reliable month for optical imagery before polar night.

PART B: RS OBSERVABILITY ANALYSIS
--------------------------------------------------------------------------------

SENSOR CAPABILITIES AT 71°N
  Sensor          Coverage    Resolution  Key Limitation at Utqiagvik
  --------------- ----------- ----------- ----------------------------------
  Sentinel-2 L2A  Optical     10m         Polar night Nov-Feb; cloud cover
                                           avg 60-80%; ~40-55% cloud-free prob
                                           in June-Aug, declining to <25% by Oct
  Sentinel-1 SAR  All-weather 10-20m      No polar night gap; 6-12 day revisit;
                                           detects surface roughness / wetness;
                                           complex processing required
  MODIS MOD10A1   Optical     500m        Daily revisit; good for snow extent;
                                           resolution too coarse for trail-level
  Landsat 8/9     Optical     30m         16-day revisit; poor for rapid events

OBSERVABILITY GAP
  The most disruptive events (Blizzard: 0/yr; Extreme Cold: 0/yr)
  occur almost entirely in November-March, during or near polar night.  These
  events are NOT observable by optical satellite.  Rain-on-Snow events peak in
  November and March — the shoulder season with marginal optical coverage.

  Only ~66% of detected extreme event-days fall in months with any
  meaningful optical satellite observability (April-October).

  SENTINEL-1 SAR is the ONLY sensor capable of detecting the trail-surface
  changes caused by the highest-impact events:
    - Blizzard:     wind-packed snow → surface roughness change in VV/VH
    - Rain-on-Snow: wet snow → ice crust → sharp backscatter signal (dB change
                    typically -3 to -8 dB at C-band)
    - Rapid Thaw:   standing water on tundra → low VV backscatter (<-15 dB)

PART C: SENTINEL-2 CHANGE DETECTION RESULTS
--------------------------------------------------------------------------------

TARGET EVENTS ANALYZED (2015-2024, spring/summer window)

  2015-05-25  [Rain-on-Snow]  PRE: no scene found  |  POST: no scene found
  2019-05-29  [Rain-on-Snow]  pre=?  cloud=0%  |  post=?  cloud=0%
  2015-05-08  [Glaze/Ice]  PRE: no scene found  |  POST: no scene found
  2016-04-26  [Glaze/Ice]  pre=?  cloud=0%  |  post=?  cloud=0%
  2022-10-16  [Glaze/Ice]  pre=?  cloud=0%  |  post=?  cloud=0%
  2023-05-30  [Glaze/Ice]  pre=?  cloud=0%  |  post=?  cloud=0%

NDSI CHANGE DETECTION
  NDSI (Normalized Difference Snow Index) = (Green - SWIR1) / (Green + SWIR1)
  Values > 0.4 indicate snow/ice cover.  Negative ΔNDSI = snow/ice loss.
  Positive ΔNDSI = new snow/ice accumulation.

  Key observable change signatures by event type:
    Rapid Thaw (Apr-May): NDSI drops sharply along trail corridors as snow
      melts and standing water / exposed tundra replaces snowpack.  Expected
      ΔNDSI: -0.3 to -0.7 within 2 weeks post-event.
    Rain-on-Snow (Apr-May): NDSI change is SUBTLE in optical imagery because
      an ice crust (from refreezing) has similar spectral reflectance to snow.
      Optical RS is a poor detector for this hazard — SAR is preferred.
    High Wind (Jun-Aug): No snow signal.  Coastal erosion and wave scouring
      may be visible in RGB and in NDWI (water index) along shorelines.

  RS ANALYSIS SUMMARY
    Target events selected:  6
    Events with imagery:     4
    Events without imagery:  2  (cloud cover or polar night)

PART D: RECOMMENDATIONS FOR OPERATIONAL RS MONITORING PROGRAM
--------------------------------------------------------------------------------

  1. DEPLOY SENTINEL-1 SAR AS PRIMARY SENSOR
     C-band SAR is the only tool that can observe year-round, including the
     winter and fall seasons when the most dangerous events occur.  The
     rain-on-snow signal (wet snow → ice crust) is a well-established SAR
     target (Barber et al. 1994; Nghiem et al. 2012).  Recommended: monitor
     VV backscatter change along trail corridors following every detected
     RoS event from the weather station.

  2. USE SENTINEL-2 FOR SPRING TRANSITION MONITORING (APR-MAY)
     The spring thaw season coincides with good optical coverage and the
     highest trail-disruption frequency.  Establish NDSI time series
     (every available scene, cloud-filtered) along major snowmachine
     corridors to map the snow-free date for each trail segment each year.

  3. COMBINE WEATHER STATION TRIGGERS WITH AUTOMATED RS QUERIES
     Build an automated pipeline: when GHCN-Daily flags an extreme event,
     query Planetary Computer for the next available cloud-free Sentinel-2
     or Sentinel-1 scene.  This closes the observation gap automatically.

  4. ESTABLISH A TRAIL-BUFFER PIXEL EXTRACTION PIPELINE
     Buffer all 670 routes by 100m.  Extract NDSI/SAR statistics inside
     vs. outside the buffer to separate trail-surface change from
     background tundra/coastal change.  This distinguishes trail impacts
     from landscape-wide climate signals.

  5. MODIS AS DAILY MONITORING (REGIONAL CONTEXT)
     MODIS MOD10A1 (500m, daily) provides continuous snow cover context.
     Even at coarse resolution, it captures the snow-free date and sea ice
     extent changes that constrain the travel window.

OUTPUT FILES
--------------------------------------------------------------------------------
  E1_Event_Catalog.png            -- 44-yr event timeline, severity, seasonality
  E2_Event_Trends.png             -- Per-type trend analysis with significance
  E3_RS_Observability.png         -- Observability gap: events vs. satellite data
  E4_NDSI_Change_Detection.png    -- Pre/post Sentinel-2 NDSI imagery
  E5_Change_Statistics.png        -- ΔNDSI trail vs. background statistics
  E_event_catalog.csv             -- Full event catalog (date, type, severity)
  E_RS_Change_Detection_Report.txt -- This report
================================================================================