natural hazards storms

This column holds geographic latitudes, ranging from -14.3236 to 70.1269 with a mean of 37.28 and median of 37.12, consistent with degrees north/south. The distribution is tightly clustered (IQR 7.50, std 5.25) around the mid-30s to low-40s, suggesting most records sit in northern temperate zones. Skew is mild (-0.18) but kurtosis of 3.34 plus 159 outliers (1.08%) hints at a long southern tail extending into the southern hemisphere. Treatment: Pair with longitude as a geospatial feature; consider binning or projecting rather than treating as a plain scalar. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

7,810

min

-14.32

max

70.13

mean

37.28

median

37.12

std

5.247

q1

33.63

q3

41.13

iqr

7.499

skew

-0.1787

kurtosis

3.341

n_outliers

159

outlier_rate

0.01077

zero_rate

0

Geographic longitude coordinates, with 8828 unique values across 14770 rows and no nulls. Values span -170.73 to 171.47 (full global range), but the distribution is tightly concentrated around a median of -90.22 with an IQR of just 12.17, indicating most records cluster in the Americas. Heavy kurtosis (55.6) and 623 outliers (4.2%) reflect a small set of points scattered far from this North/Central American core. Treatment: Pair with latitude for geospatial features; do not standardize as a plain scalar. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

8,828

min

-170.7

max

171.5

mean

-90.94

median

-90.22

std

11.7

q1

-96.4

q3

-84.23

iqr

12.17

skew

1.286

kurtosis

55.61

n_outliers

623

outlier_rate

0.04218

zero_rate

0

This column holds short structured event labels of the form ' in , ' describing US severe weather incidents (tornadoes, floods, hail, thunderstorm winds). It is highly repetitive: 8110 of 14770 rows are duplicates (54.9% duplicate_rate) with only 6660 unique values, and 'Hail in TEXAS, TARRANT' alone appears 59 times. The 'multilingual' alert is largely a false positive driven by place names — 4796 rows classify as English versus tiny counts like 134 Spanish and 25 German. Treatment: Parse into structured fields (event_type, state, county) rather than treating as free text. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

6,660

len_min

17

len_max

134

len_mean

30.22

len_median

29

len_p95

41

word_mean

4.588

word_median

4

n_empty

0

n_duplicates

8,110

duplicate_rate

0.5491

vocab_size

1,980

readability_flesch_mean

31.16

emoji_rate

0

url_rate

0

one_word_rate

0

allcaps_rate

0

boilerplate_rate

0

This is a templated event-summary field describing storm or disaster impacts (magnitude, injuries, fatalities, property damage in dollars), not free-form prose. With 14,770 rows but only 5,796 unique values and a 60.8% duplicate rate, the text is highly formulaic — the single string 'Magnitude 0; $2.5M property damage' alone appears 1,055 times. Language detection flags 5 French, 1 Japanese, and 10 Norwegian rows against 4,984 English, almost certainly false positives from numeric/currency tokens rather than real multilingual content; mean Flesch of 29.9 reflects the terse template, not difficult prose. Treatment: Parse with regex into structured fields (magnitude, injuries, fatalities, damage_usd) rather than treating as free text. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

5,796

len_min

3

len_max

259

len_mean

50.09

len_median

36

len_p95

166

word_mean

7.393

word_median

5

n_empty

0

n_duplicates

8,974

duplicate_rate

0.6076

vocab_size

4,289

readability_flesch_mean

29.86

emoji_rate

0

url_rate

0

one_word_rate

0.0002708

allcaps_rate

0.0002708

boilerplate_rate

0

This column is a constant categorical tag identifying the dataset source: every one of the 14770 rows holds the single value "significant_us_storms". With cardinality 1, entropy 0, and top_rate 1.0, it carries no information for modelling and only serves as a provenance marker. Treatment: Drop before modelling; retain only as a source tag if merging with other datasets. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

1

top_value

significant_us_storms

top_rate

1

cardinality

1

entropy

0

entropy_ratio

0

This is a date column stored as ISO-formatted text (YYYY-MM-DD), with every one of 14770 rows exactly 10 characters and a single token. Values span at least 1965 to 2021, and duplicates are heavy: 9712 rows (65.8%) repeat, with 1974-04-03 appearing 126 times and 2011-04-27 appearing 105 times, suggesting clustering around specific event dates rather than a uniform timeline. The 'allcaps' flag is a false positive since digits and hyphens trigger it. Treatment: parse to date dtype and use for temporal grouping or joins. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

5,058

len_min

10

len_max

10

len_mean

10

len_median

10

len_p95

10

word_mean

1

word_median

1

n_empty

0

n_duplicates

9,712

duplicate_rate

0.6575

vocab_size

5,058

readability_flesch_mean

121.2

emoji_rate

0

url_rate

0

one_word_rate

1

allcaps_rate

1

boilerplate_rate

0

This column records the country of origin for each record, but every one of the 14770 rows holds the value "USA". Cardinality is 1 and entropy is 0, so the field carries no information. Treatment: Drop; constant column with zero variance. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

1

top_value

USA

top_rate

1

cardinality

1

entropy

0

entropy_ratio

0

Categorical label of severe-weather event types across 14,770 rows with no nulls and 17 distinct categories. Tornado dominates at 42.9% (6,334 rows), followed by Flash Flood, Thunderstorm Wind, Flood, and Hail; entropy ratio of 0.57 confirms the distribution is concentrated in a few classes. The long tail (Heavy Rain, Marine Strong Wind, Debris Flow, Marine Thunderstorm Wind) has very thin support, which will hurt per-class modelling. Treatment: Use as categorical target or feature; consider grouping rare classes given the heavy Tornado skew. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

17

top_value

Tornado

top_rate

0.4288

cardinality

17

entropy

2.336

entropy_ratio

0.5715

US state names in uppercase, with 65 distinct values across 14,770 complete rows — more than the 50 states, suggesting territories, military codes, or 'UNKNOWN'-style entries are mixed in. Distribution is broad (entropy ratio 0.86) with Texas leading at 9.8% (1,450 rows), followed by Missouri, Arkansas, Mississippi, and Georgia, indicating a southern/central US tilt. Treatment: Normalize to standard state codes and one-hot or target-encode; investigate the 15 extra categories beyond 50 states. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

65

top_value

TEXAS

top_rate

0.09817

cardinality

65

entropy

5.182

entropy_ratio

0.8605

Numeric magnitudes stored as strings, with 170 distinct values ranging from '0' to two-decimal figures like '70.00'. Over half the rows (51.78%) are null, and of the non-nulls the value '0' dominates at 54.24%, leaving real magnitudes in a minority of records. Entropy ratio of 0.48 confirms most signal is concentrated in a few buckets. Treatment: Cast to numeric, treat '0' and nulls as likely missing/sentinel, and consider a presence flag before modelling. high · anthropic:claude-opus-4-7

n

14,770

nulls

7,648 (51.8%)

unique

170

top_value

0

top_rate

0.5424

cardinality

170

entropy

3.586

entropy_ratio

0.484

This is an injury count per record, stored as strings but numeric in nature (top values are '0' through '12'). The distribution is dominated by zeros: 68.1% of 14,770 rows report '0' injuries, with 893 at '1' and a long tail spanning 178 distinct values. Entropy ratio of 0.33 confirms heavy concentration at the low end. Treatment: Cast to integer and consider log1p or zero-inflated modelling given the heavy zero mass. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

178

top_value

0

top_rate

0.6814

cardinality

178

entropy

2.468

entropy_ratio

0.3301

Counts of fatalities per event, stored as strings with 49 distinct values across 14,770 rows and no nulls. The distribution is heavily zero-inflated: 69.1% of records are "0" and the next bucket "1" covers 3,208 rows, leaving a long thin tail (e.g., 25 rows at "9", 24 at "10"). Low entropy ratio (0.25) confirms most variance lives in the 0/1 split. Treatment: Cast to integer and consider modelling as a count (Poisson/negative binomial) or binarise to any-fatality given the zero inflation. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

49

top_value

0

top_rate

0.6912

cardinality

49

entropy

1.423

entropy_ratio

0.2535

This column encodes property damage as short magnitude strings with a K/M suffix (e.g., '2.5M', '250K', '0.00K'), not free text — every value is one word and max length is 8. Formatting is inconsistent: '1M' and '1.00M' coexist, and '0.00K' appears 1229 times alongside 368 empty entries, conflating true zeros with missingness. With 93.1% duplicate rate across only 1013 unique tokens, this is a coarse categorical-looking encoding of a numeric quantity. Treatment: Parse the K/M suffix into a numeric dollar amount and treat empty strings as missing (distinct from 0.00K). high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

1,014

len_min

0

len_max

8

len_mean

4.381

len_median

5

len_p95

7

word_mean

1

word_median

1

n_empty

368

n_duplicates

13,756

duplicate_rate

0.9313

vocab_size

1,013

readability_flesch_mean

117

emoji_rate

0

url_rate

0

one_word_rate

1

allcaps_rate

0.8724

boilerplate_rate

0

This column records the data provenance, holding the constant string "NOAA Storm Events Database" for all 14770 rows. With cardinality of 1 and entropy of 0.0, it carries no information for modelling and only serves as a dataset-level annotation. Treatment: Drop before modelling; retain in documentation as a provenance tag. high · anthropic:claude-opus-4-7

n

14,770

nulls

0 (0.0%)

unique

1

top_value

NOAA Storm Events Database

top_rate

1

cardinality

1

entropy

0

entropy_ratio

0