saturn·

← all readings

natural hazards earthquakes

saturn notebook · generated 2026-05-01 Report Notebook

Overview

Source: /home/coolhand/html/datavis/data_trove/data/natural_hazards/earthquakes.json

Saturn profiled 3,742 rows across 11 columns. The stats below are deterministic and machine-readable; the prose is a language-model interpretation of those stats (opt-in, added after the fact, never sees raw rows).

[2]:
!pip install saturn-dissect
import subprocess
subprocess.run([
    "saturn", "analyze", "/home/coolhand/html/datavis/data_trove/data/natural_hazards/earthquakes.json",
    "--findings", "natural_hazards-earthquakes.json",
    "--llm", "anthropic:claude-opus-4-7",
])

Summary confidence: high

This dataset contains 3,742 records of significant earthquakes, with numeric measurements (magnitude, depth, latitude, longitude), location text fields, and a date column. Magnitude is tightly clustered between 4.5 and 5.1 (median 4.8) but reaches up to 8.2, producing 184 high-end outliers worth a closer look. Depth is highly skewed (skew 3.07) with a median of 10 km but a max of 248.7 km and 314 outliers, suggesting a mix of shallow and deep events. Geographically, the data is heavily concentrated around Alaska — 'alaska' appears in 1,991 place names, with Canada and Mexico trailing far behind — and longitudes sit firmly in the western hemisphere (median -144.2). Note that 'category' is a single constant value and 'earthquake_type' is 99.9% 'earthquake', so neither adds analytic signal.

citing: row_count · column_count · depth_km · magnitude · latitude · longitude · place · earthquake_type · category

Out[4]:

saturn.schema() · 11 columns

column kind n null% unique alerts
latitude numeric 3,742 0.0% 3,627
longitude numeric 3,742 0.0% 3,668
name text 3,742 0.0% 3,002 multilingual
description text 3,742 0.0% 3,591 near_unique
category categorical 3,742 0.0% 1 imbalance
date text 3,742 0.0% 3,741 near_unique one_word allcaps short_text
country unknown 3,742 0.0% — skipped
magnitude numeric 3,742 0.0% 123
depth_km numeric 3,742 0.0% 1,505 high_skew outliers
place text 3,742 0.0% 3,002 multilingual
earthquake_type categorical 3,742 0.0% 3 imbalance
Fig 1.
magnitude · Most quakes sit between 4.5–5.1; watch the long tail extending to 8.2.
Show data table
bincount
4.5 – 4.593752
4.593 – 4.685601
4.685 – 4.777445
4.777 – 4.87340
4.87 – 4.963286
4.963 – 5.055254
5.055 – 5.147218
5.147 – 5.24177
5.24 – 5.332137
5.332 – 5.425104
5.425 – 5.51885
5.518 – 5.6166
5.61 – 5.70253
5.702 – 5.7953
5.795 – 5.88738
5.887 – 5.9846
5.98 – 6.07225
6.072 – 6.16517
6.165 – 6.25716
6.257 – 6.3511
6.35 – 6.44215
6.442 – 6.53511
6.535 – 6.62710
6.627 – 6.724
6.72 – 6.8126
6.812 – 6.9055
6.905 – 6.9970
6.997 – 7.093
7.09 – 7.1823
7.182 – 7.2753
7.275 – 7.3671
7.367 – 7.460
7.46 – 7.5521
7.552 – 7.6451
7.645 – 7.7370
7.737 – 7.832
7.83 – 7.9222
7.922 – 8.0150
8.015 – 8.1070
8.107 – 8.21
Fig 2.
depth_km · Strongly right-skewed with a median of 10 km but values reaching 248 km — inspect the deep-event tail.
Show data table
bincount
-2.261 – 4.013219
4.013 – 10.291730
10.29 – 16.56370
16.56 – 22.84258
22.84 – 29.11230
29.11 – 35.38250
35.38 – 41.66167
41.66 – 47.93129
47.93 – 54.2156
54.21 – 60.4831
60.48 – 66.7543
66.75 – 73.0327
73.03 – 79.319
79.3 – 85.5829
85.58 – 91.8521
91.85 – 98.1219
98.12 – 104.424
104.4 – 110.712
110.7 – 116.99
116.9 – 123.214
123.2 – 129.513
129.5 – 135.819
135.8 – 14214
142 – 148.35
148.3 – 154.66
154.6 – 160.90
160.9 – 167.17
167.1 – 173.44
173.4 – 179.70
179.7 – 1861
186 – 192.25
192.2 – 198.51
198.5 – 204.83
204.8 – 211.12
211.1 – 217.33
217.3 – 223.61
223.6 – 229.90
229.9 – 236.20
236.2 – 242.40
242.4 – 248.71
Fig 3.
latitude · Latitude clusters in the high northern hemisphere (median ~52°), consistent with an Alaska-heavy sample.
Show data table
bincount
20.02 – 21.2635
21.26 – 22.5128
22.51 – 23.7542
23.75 – 2588
25 – 26.2491
26.24 – 27.4935
27.49 – 28.7343
28.73 – 29.9839
29.98 – 31.2252
31.22 – 32.4683
32.46 – 33.7152
33.71 – 34.9526
34.95 – 36.267
36.2 – 37.4443
37.44 – 38.6954
38.69 – 39.9322
39.93 – 41.18131
41.18 – 42.4252
42.42 – 43.6696
43.66 – 44.91177
44.91 – 46.154
46.15 – 47.47
47.4 – 48.6434
48.64 – 49.89117
49.89 – 51.13151
51.13 – 52.38288
52.38 – 53.62423
53.62 – 54.86349
54.86 – 56.11223
56.11 – 57.35201
57.35 – 58.675
58.6 – 59.84104
59.84 – 61.09116
61.09 – 62.33117
62.33 – 63.58120
63.58 – 64.8233
64.82 – 66.0635
66.06 – 67.3125
67.31 – 68.5529
68.55 – 69.835
Fig 4.
place · Top recurring locations are dominated by Alaska, Canada, and Mexico — useful for spotting regional concentration.
Show data table
charscount
4 – 51
5 – 70
7 – 81
8 – 100
10 – 110
11 – 120
12 – 142
14 – 1511
15 – 1640
16 – 181
18 – 1920
19 – 2019
20 – 228
22 – 23219
23 – 2560
25 – 26122
26 – 27543
27 – 29499
29 – 30823
30 – 32325
32 – 33362
33 – 34378
34 – 36105
36 – 3740
37 – 3837
38 – 4025
40 – 4134
41 – 423
42 – 440
44 – 4515
45 – 4722
47 – 4814
48 – 493
49 – 511
51 – 522
52 – 545
54 – 550
55 – 560
56 – 580
58 – 592
Fig 5.
earthquake_type · Shows that 'earthquake' accounts for 99.9% of records, with only a handful of explosions and one landslide.
Show data table
valuecountshare
earthquake373999.9%
explosion20.1%
landslide10.0%
Fig 6.
Per-column null rate across the corpus. Columns are ordered by input position.
Show data table
columnkindnull %
latitudenumeric0.0%
longitudenumeric0.0%
nametext0.0%
descriptiontext0.0%
categorycategorical0.0%
datetext0.0%
countryunknown0.0%
magnitudenumeric0.0%
depth_kmnumeric0.0%
placetext0.0%
earthquake_typecategorical0.0%
Fig 7.
Language mix across all text columns (per-string detection, sampled).
Show data table
langcountshare
en743899.4%
es320.4%
de60.1%
ja20.0%
ceb20.0%
Fig 8.
Pearson correlation across numeric columns (sampled, bounded).
Show data table
latitudelongitudemagnitudedepth_km
latitude+1.00-0.72-0.12+0.31
longitude-0.72+1.00+0.07-0.37
magnitude-0.12+0.07+1.00-0.03
depth_km+0.31-0.37-0.03+1.00

latitude numeric feature

Geographic latitude coordinates ranging from 20.02 to 69.80, with a mean of 48.53 and median of 52.40. The distribution is moderately left-skewed (-0.76), concentrated in northern mid-to-high latitudes (Q1=41.34, Q3=55.90), suggesting a Europe/North America-heavy sample with no southern hemisphere points. Near-unique values (3627/3742) and zero nulls indicate clean, granular location data.

Treatment: Pair with longitude for geospatial features; consider binning or projecting rather than using raw value in linear models.

anthropic:claude-opus-4-7 · confidence high
Out[14]:

saturn.columns["latitude"].stats

statvalue
n3,742
nulls0 (0.0%)
unique3,627
min 20.02
max 69.8
mean 48.53
median 52.4
std 11.58
q1 41.34
q3 55.9
iqr 14.56
skew -0.7591
kurtosis -0.2887
n_outliers 0
outlier_rate 0
zero_rate 0
Fig 9.
Distribution of latitude. Vertical dash marks the median.
Show data table
bincount
20.02 – 21.2635
21.26 – 22.5128
22.51 – 23.7542
23.75 – 2588
25 – 26.2491
26.24 – 27.4935
27.49 – 28.7343
28.73 – 29.9839
29.98 – 31.2252
31.22 – 32.4683
32.46 – 33.7152
33.71 – 34.9526
34.95 – 36.267
36.2 – 37.4443
37.44 – 38.6954
38.69 – 39.9322
39.93 – 41.18131
41.18 – 42.4252
42.42 – 43.6696
43.66 – 44.91177
44.91 – 46.154
46.15 – 47.47
47.4 – 48.6434
48.64 – 49.89117
49.89 – 51.13151
51.13 – 52.38288
52.38 – 53.62423
53.62 – 54.86349
54.86 – 56.11223
56.11 – 57.35201
57.35 – 58.675
58.6 – 59.84104
59.84 – 61.09116
61.09 – 62.33117
62.33 – 63.58120
63.58 – 64.8233
64.82 – 66.0635
66.06 – 67.3125
67.31 – 68.5529
68.55 – 69.835

longitude numeric feature

Geographic longitude coordinates, all negative and ranging from -169.997 to -65.039, placing every record in the western hemisphere. The distribution is wide (IQR ~34.8°) with a slight right skew (0.449) and only 26 mild outliers (0.69%); near-uniqueness (3668 unique of 3742) suggests point-level locations rather than a small set of sites.

Treatment: Pair with latitude as a 2-D geospatial feature; avoid treating as a standalone scalar in models.

anthropic:claude-opus-4-7 · confidence high
Out[17]:

saturn.columns["longitude"].stats

statvalue
n3,742
nulls0 (0.0%)
unique3,668
min -170
max -65.04
mean -140.1
median -144.2
std 21.81
q1 -159.9
q3 -125.1
iqr 34.84
skew 0.4489
kurtosis -0.4302
n_outliers 26
outlier_rate 0.006948
zero_rate 0
Fig 10.
Distribution of longitude. Vertical dash marks the median.
Show data table
bincount
-170 – -167.4368
-167.4 – -164.7161
-164.7 – -162.1243
-162.1 – -159.5241
-159.5 – -156.9113
-156.9 – -154.3129
-154.3 – -151.6241
-151.6 – -149220
-149 – -146.477
-146.4 – -143.883
-143.8 – -141.129
-141.1 – -138.548
-138.5 – -135.941
-135.9 – -133.334
-133.3 – -130.6129
-130.6 – -128428
-128 – -125.4200
-125.4 – -122.889
-122.8 – -120.144
-120.1 – -117.5130
-117.5 – -114.9149
-114.9 – -112.3113
-112.3 – -109.6163
-109.6 – -107134
-107 – -104.433
-104.4 – -101.812
-101.8 – -99.1510
-99.15 – -96.5320
-96.53 – -93.94
-93.9 – -91.281
-91.28 – -88.652
-88.65 – -86.036
-86.03 – -83.411
-83.41 – -80.782
-80.78 – -78.163
-78.16 – -75.537
-75.53 – -72.918
-72.91 – -70.297
-70.29 – -67.665
-67.66 – -65.0414

name text metadata

This column holds short geographic descriptions of event locations, almost certainly earthquake place names (e.g. '104 km SSW of Nikolski, Alaska', 'off the coast of Oregon'), with mean length 29 chars and 6 words. Of 3742 rows, only 3002 are unique and 740 are duplicates (19.8%), with 'off the coast of Oregon' alone repeating 151 times. The language detector flags a multilingual mix but it is overwhelmingly English (3719) with trace de/es/ja/ceb counts likely false positives on place tokens.

Treatment: Parse into structured fields (distance, bearing, place) or geocode rather than using the raw string as a feature.

anthropic:claude-opus-4-7 · confidence high
Out[20]:

saturn.columns["name"].stats

statvalue
n3,742
nulls0 (0.0%)
unique3,002
len_min 4
len_max 59
len_mean 29.47
len_median 29
len_p95 36
word_mean 6.293
word_median 6
n_empty 0
n_duplicates 740
duplicate_rate 0.1978
vocab_size 1,036
readability_flesch_mean 69.91
emoji_rate 0
url_rate 0
one_word_rate 0.0005345
allcaps_rate 0
boilerplate_rate 0
alert: multilingual6 languages detected in sample
Fig 11.
Character-length distribution for name.
Show data table
charscount
4 – 51
5 – 70
7 – 81
8 – 100
10 – 110
11 – 120
12 – 142
14 – 1511
15 – 1640
16 – 181
18 – 1920
19 – 2019
20 – 228
22 – 23219
23 – 2560
25 – 26122
26 – 27543
27 – 29499
29 – 30823
30 – 32325
32 – 33362
33 – 34378
34 – 36105
36 – 3740
37 – 3837
38 – 4025
40 – 4134
41 – 423
42 – 440
44 – 4515
45 – 4722
47 – 4814
48 – 493
49 – 511
51 – 522
52 – 545
54 – 550
55 – 560
56 – 580
58 – 592

description text free_text

This is a templated text description of seismic events, with every row mentioning 'earthquake', 'magnitude', and 'depth:' and lengths tightly clustered between 45 and 100 characters (median 72). Despite the formulaic structure, 3591 of 3742 values are unique because the embedded magnitude and depth numbers vary; still, 151 exact duplicates (4.0%) exist. Alaska appears in 1973 rows and a 10km depth in 1216, suggesting the corpus is dominated by shallow Alaskan quakes.

Treatment: Parse out magnitude, depth, and region with regex into structured features rather than embedding the raw string.

anthropic:claude-opus-4-7 · confidence high
Out[23]:

saturn.columns["description"].stats

statvalue
n3,742
nulls0 (0.0%)
unique3,591
len_min 45
len_max 100
len_mean 71.71
len_median 72
len_p95 79
word_mean 12.29
word_median 12
n_empty 0
n_duplicates 151
duplicate_rate 0.04035
vocab_size 2,674
readability_flesch_mean 63.23
emoji_rate 0
url_rate 0
one_word_rate 0
allcaps_rate 0
boilerplate_rate 0
alert: near_unique96.0% of rows are unique strings
Fig 12.
Character-length distribution for description.
Show data table
charscount
45 – 461
46 – 481
48 – 490
49 – 500
50 – 520
52 – 530
53 – 554
55 – 564
56 – 5719
57 – 5915
59 – 6020
60 – 629
62 – 6317
63 – 64158
64 – 6645
66 – 6773
67 – 68380
68 – 70350
70 – 71726
71 – 72389
72 – 74367
74 – 75605
75 – 77204
77 – 7891
78 – 7990
79 – 8121
81 – 8233
82 – 848
84 – 8525
85 – 8631
86 – 8824
88 – 8910
89 – 907
90 – 921
92 – 933
93 – 945
94 – 961
96 – 972
97 – 991
99 – 1002

category categorical metadata

This column is a constant tag labeling every row as "significant_earthquakes" across all 3742 records, with cardinality of 1 and entropy of 0. It carries no information for modelling or analysis since the top_rate is 1.0 with no nulls.

Treatment: Drop; constant column with a single value.

anthropic:claude-opus-4-7 · confidence high
Out[26]:

saturn.columns["category"].stats

statvalue
n3,742
nulls0 (0.0%)
unique1
top_value significant_earthquakes
top_rate 1
cardinality 1
entropy 0
entropy_ratio 0
alert: imbalancetop value is 100.0% of rows
Fig 13.
Top values for category.
Show data table
valuecountshare
significant_earthquakes3742100.0%

date text identifier

Despite its name, this column holds malformed date-like strings rather than parseable timestamps — every value follows a 'NNNNNNNNNNNNN-01-01' pattern with a 13-digit prefix (likely a millisecond epoch) glued to a literal '-01-01' suffix. Values are 18-19 chars, one token each, and 3741 of 3742 are unique with only one duplicate ('1614452365296-01-01' appears twice). Stored as text, not a date type, so no temporal ordering or range is usable as-is.

Treatment: Parse the leading 13-digit epoch as ms-since-epoch into a real timestamp, or treat as a near-unique id and drop.

anthropic:claude-opus-4-7 · confidence high
Out[29]:

saturn.columns["date"].stats

statvalue
n3,742
nulls0 (0.0%)
unique3,741
len_min 18
len_max 19
len_mean 18.95
len_median 19
len_p95 19
word_mean 1
word_median 1
n_empty 0
n_duplicates 1
duplicate_rate 0.0002672
vocab_size 3,741
readability_flesch_mean 121.2
emoji_rate 0
url_rate 0
one_word_rate 1
allcaps_rate 1
boilerplate_rate 0
alert: near_unique100.0% of rows are unique strings
alert: one_word100.0% rows are a single word
alert: allcaps100.0% rows are all-caps
alert: short_text95th-percentile length under 20 chars
Fig 14.
Character-length distribution for date.
Show data table
charscount
18 – 18180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 180
18 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 190
19 – 193562

country unknown feature

The column is named "country" and has 3742 non-null entries with a null_rate of 0.0, but saturn skipped detailed profiling (kind is "unknown") so cardinality and value distribution are not available. Without n_unique or category stats, we cannot tell whether this is a clean ISO code field, free-text country names, or a near-constant. The lack of any descriptive statistics is itself the main signal.

Treatment: Re-profile or manually inspect to determine cardinality before deciding whether to one-hot encode or normalize to ISO codes.

anthropic:claude-opus-4-7 · confidence low
Out[32]:

saturn.columns["country"].stats

statvalue
n3,742
nulls0 (0.0%)
unique—
alert: skippedno profiler for kind=unknown

magnitude numeric feature

This is an earthquake-style magnitude reading, bounded between 4.5 and 8.2 with a mean of 4.92 and median of 4.8. The distribution is heavily right-skewed (skew 1.97, kurtosis 5.58) with a tight IQR of 0.5 and 184 outliers (4.9%) — consistent with a catalog truncated at magnitude 4.5 where large events are rare but extreme.

Treatment: Treat as a right-skewed numeric feature; consider modelling tail events separately rather than transforming, since magnitude is already on a log scale.

anthropic:claude-opus-4-7 · confidence high
Out[34]:

saturn.columns["magnitude"].stats

statvalue
n3,742
nulls0 (0.0%)
unique123
min 4.5
max 8.2
mean 4.917
median 4.8
std 0.462
q1 4.6
q3 5.1
iqr 0.5
skew 1.97
kurtosis 5.583
n_outliers 184
outlier_rate 0.04917
zero_rate 0
Fig 15.
Distribution of magnitude. Vertical dash marks the median.
Show data table
bincount
4.5 – 4.593752
4.593 – 4.685601
4.685 – 4.777445
4.777 – 4.87340
4.87 – 4.963286
4.963 – 5.055254
5.055 – 5.147218
5.147 – 5.24177
5.24 – 5.332137
5.332 – 5.425104
5.425 – 5.51885
5.518 – 5.6166
5.61 – 5.70253
5.702 – 5.7953
5.795 – 5.88738
5.887 – 5.9846
5.98 – 6.07225
6.072 – 6.16517
6.165 – 6.25716
6.257 – 6.3511
6.35 – 6.44215
6.442 – 6.53511
6.535 – 6.62710
6.627 – 6.724
6.72 – 6.8126
6.812 – 6.9055
6.905 – 6.9970
6.997 – 7.093
7.09 – 7.1823
7.182 – 7.2753
7.275 – 7.3671
7.367 – 7.460
7.46 – 7.5521
7.552 – 7.6451
7.645 – 7.7370
7.737 – 7.832
7.83 – 7.9222
7.922 – 8.0150
8.015 – 8.1070
8.107 – 8.21

depth_km numeric feature

This is almost certainly the focal depth in kilometres for seismic events, with values ranging from -2.261 to 248.7 and a median of 10.0. The distribution is heavily right-skewed (skew 3.07, kurtosis 11.6) with 314 outliers (8.39% of rows) and a small number of negative depths that warrant inspection. Half the records sit between q1=10.0 and q3=29.1015, suggesting a strong concentration of shallow events with a long deep tail.

Treatment: Clip or investigate negative values, then log-transform (e.g., log(depth+c)) before modelling.

anthropic:claude-opus-4-7 · confidence high
Out[37]:

saturn.columns["depth_km"].stats

statvalue
n3,742
nulls0 (0.0%)
unique1,505
min -2.261
max 248.7
mean 23.71
median 10
std 28.79
q1 10
q3 29.1
iqr 19.1
skew 3.072
kurtosis 11.61
n_outliers 314
outlier_rate 0.08391
zero_rate 0.002672
alert: high_skewskew=+3.07
alert: outliers8.4% rows beyond 1.5 IQR
Fig 16.
Distribution of depth_km. Vertical dash marks the median.
Show data table
bincount
-2.261 – 4.013219
4.013 – 10.291730
10.29 – 16.56370
16.56 – 22.84258
22.84 – 29.11230
29.11 – 35.38250
35.38 – 41.66167
41.66 – 47.93129
47.93 – 54.2156
54.21 – 60.4831
60.48 – 66.7543
66.75 – 73.0327
73.03 – 79.319
79.3 – 85.5829
85.58 – 91.8521
91.85 – 98.1219
98.12 – 104.424
104.4 – 110.712
110.7 – 116.99
116.9 – 123.214
123.2 – 129.513
129.5 – 135.819
135.8 – 14214
142 – 148.35
148.3 – 154.66
154.6 – 160.90
160.9 – 167.17
167.1 – 173.44
173.4 – 179.70
179.7 – 1861
186 – 192.25
192.2 – 198.51
198.5 – 204.83
204.8 – 211.12
211.1 – 217.33
217.3 – 223.61
223.6 – 229.90
229.9 – 236.20
236.2 – 242.40
242.4 – 248.71

place text metadata

Short geographic descriptors of earthquake locations, typically formatted as ' km of , ' (e.g., '104 km SSW of Nikolski, Alaska'), with mean length 29 characters and median 6 words. Alaska dominates (1,991 of 3,742 rows mention it), and a single string 'off the coast of Oregon' repeats 151 times, contributing to a 19.8% duplicate rate across 3,002 unique values. The 'multilingual' alert is driven by a tiny non-English fringe (16 es, 3 de, 1 ja, 1 ceb) against 3,719 English rows, so it is effectively monolingual.

Treatment: Parse into structured fields (distance_km, bearing, place_name, region) rather than using the raw string as a categorical.

anthropic:claude-opus-4-7 · confidence high
Out[40]:

saturn.columns["place"].stats

statvalue
n3,742
nulls0 (0.0%)
unique3,002
len_min 4
len_max 59
len_mean 29.47
len_median 29
len_p95 36
word_mean 6.293
word_median 6
n_empty 0
n_duplicates 740
duplicate_rate 0.1978
vocab_size 1,036
readability_flesch_mean 69.91
emoji_rate 0
url_rate 0
one_word_rate 0.0005345
allcaps_rate 0
boilerplate_rate 0
alert: multilingual6 languages detected in sample
Fig 17.
Character-length distribution for place.
Show data table
charscount
4 – 51
5 – 70
7 – 81
8 – 100
10 – 110
11 – 120
12 – 142
14 – 1511
15 – 1640
16 – 181
18 – 1920
19 – 2019
20 – 228
22 – 23219
23 – 2560
25 – 26122
26 – 27543
27 – 29499
29 – 30823
30 – 32325
32 – 33362
33 – 34378
34 – 36105
36 – 3740
37 – 3837
38 – 4025
40 – 4134
41 – 423
42 – 440
44 – 4515
45 – 4722
47 – 4814
48 – 493
49 – 511
51 – 522
52 – 545
54 – 550
55 – 560
56 – 580
58 – 592

earthquake_type categorical label

This is a categorical event-type label with only 3 distinct values across 3742 rows and no nulls. The distribution is essentially degenerate: 'earthquake' covers 99.92% of rows, leaving just 2 'explosion' and 1 'landslide' records, yielding an entropy ratio of 0.006. The column carries almost no information as-is.

Treatment: Drop or collapse to a binary rare-event flag; near-constant for modelling.

anthropic:claude-opus-4-7 · confidence high
Out[43]:

saturn.columns["earthquake_type"].stats

statvalue
n3,742
nulls0 (0.0%)
unique3
top_value earthquake
top_rate 0.9992
cardinality 3
entropy 0.01014
entropy_ratio 0.006396
alert: imbalancetop value is 99.9% of rows
Fig 18.
Top values for earthquake_type.
Show data table
valuecountshare
earthquake373999.9%
explosion20.1%
landslide10.0%

How to cite

click to copy

BibTeX
@misc{saturn-natural-hazards-earthquakes-2026,
  author       = {Steuber, Luke},
  title        = {Saturn reading: natural hazards earthquakes},
  year         ={2026},
  howpublished = {\url{https://dr.eamer.dev/saturn/view/natural_hazards-earthquakes}},
  note         = {Profiled with saturn-dissect v0.2.0, prompt saturn-insight-v2, model anthropic:claude-opus-4-7},
}
APA
Steuber, L. (2026). Saturn reading: natural hazards earthquakes. Source: /home/coolhand/html/datavis/data_trove/data/natural_hazards/earthquakes.json. Profiled with saturn-dissect v0.2.0 (saturn-insight-v2, anthropic:claude-opus-4-7). Retrieved from https://dr.eamer.dev/saturn/view/natural_hazards-earthquakes