data trove global volcanism program

Overview

Source: /home/coolhand/html/datavis/data_trove/data/quirky/volcanoes.json

Saturn profiled 200 rows across 9 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).

name categorical label

This column contains volcano names, functioning as a label for individual volcanic entities in the dataset. With 111 unique values across 200 rows, many volcanoes appear multiple times — 'Semeru' leads with 13 occurrences (6.5% of rows), suggesting repeated eruption or activity events per volcano rather than one row per volcano. The high entropy ratio of 0.946 combined with the long-tail alert indicates the distribution is broad but uneven, with a handful of well-known volcanoes (Semeru, Merapi, Etna, Stromboli) dominating while most names appear only once or twice.

Treatment: Group by this column to aggregate per-volcano statistics, or encode as a categorical feature with frequency-based or target encoding given the long-tail distribution.

country categorical label

This column records the country associated with each record — likely the location of a seismic, volcanic, or natural-disaster event given the top countries (Indonesia, Japan, Philippines, Papua New Guinea, Chile). Indonesia dominates heavily at 28.5% of all 200 rows (57 occurrences), followed by Japan at 14.5%, which is a pronounced geographic skew toward the Pacific Ring of Fire. With only 33 unique values and zero nulls, coverage is clean, but the top-heavy distribution (entropy ratio 0.78) means most records cluster around a handful of high-activity nations.

Treatment: One-hot encode or target-encode for modelling; be aware of class imbalance with Indonesia representing 28.5% of rows.

lat numeric feature

This column contains geographic latitude values, spanning from -41.33° (southern hemisphere, e.g., southern South America or New Zealand) to 63.983° (northern hemisphere, e.g., Scandinavia or Canada), consistent with a globally distributed dataset. With only 111 unique values across 200 rows, many locations are repeated, suggesting the dataset references a limited set of geographic points rather than unique coordinates per record. The distribution is nearly symmetric (skew 0.20, kurtosis -0.48) and spans a wide IQR of 40.73°, indicating broad global coverage rather than clustering in one region. No nulls, outliers, or zeros are present.

Treatment: Pair with a longitude column for geospatial analysis; consider binning into regions or using as-is in spatial models.

lon numeric feature

This column is a geographic longitude coordinate, spanning the full valid range of −175.65 to 177.18 degrees, indicating global coverage. Surprisingly, with only 111 unique values across 200 rows (~55% uniqueness), there is notable coordinate repetition, suggesting many records share the same location or coordinates have been rounded/binned. The mean (59.16) is substantially pulled away from the median (112.31) by a left skew (−0.86), implying a cluster of observations in Eastern hemisphere longitudes with a tail of negative (Western hemisphere) values dragging the mean down.

Treatment: Use as-is or pair with latitude for geospatial modelling; investigate duplicate coordinates (111 unique / 200 rows) to determine if binning or data quality issue.

elevation numeric feature

This column represents geographic elevation in metres (or feet) for 200 location records, spanning from -185.0 (below sea level, consistent with places like the Dead Sea or Death Valley) to 5393.0 (alpine/high-altitude terrain). The distribution is broad and fairly flat — IQR of 1809.75 against a mean of 2074.21 — with a slight positive skew (0.39) and near-platykurtic shape (kurtosis -0.57), suggesting a deliberately diverse geographic sample rather than a natural population draw. With only 109 unique values across 200 rows, roughly 45% of values are repeated, which may indicate rounding to nearest metre or binned elevation bands.

Treatment: Use as-is or apply mild normalisation (e.g. standard scaling); the negative minimum requires care if log-transforming — shift first.

type categorical label

This column classifies volcanic structures into 13 morphological types, making it a geological label for each record. 'Stratovolcano' dominates heavily at 69.5% of 200 records (139 occurrences), while the remaining 12 types share the rest — an extreme concentration that yields an entropy ratio of only 0.47. The long tail of rare categories (e.g., 'Cinder cone' and 'Maar' each appearing ≤2 times) may cause class-imbalance problems in any supervised modelling task.

Treatment: One-hot encode or target-encode with caution due to severe class imbalance; consider grouping rare types (frequency < 3) into an 'Other' category before modelling.

vei numeric feature

This column is almost certainly the Volcanic Explosivity Index (VEI), a logarithmic scale rating volcanic eruption intensity. With only 7 unique integer values ranging from 0 to 6 and a median of 3, it behaves more like an ordinal category than a continuous numeric. Notably, 15 outliers (7.5% of rows) sit at the upper end of the scale — VEI 5–6 events are rare in real-world volcanology, so their presence is worth verifying. The IQR of 1.0 and tight Q1–Q3 band of 2–3 confirm most eruptions cluster at moderate intensity.

Treatment: Treat as ordinal; consider one-hot or ordinal encoding rather than raw numeric use in models.

year numeric feature

This column represents a calendar year, spanning 1900 to 1999 — exactly one century of data with no nulls. With 84 unique values across 200 rows, some years appear multiple times, suggesting records grouped by year rather than unique annual entries. The distribution is notably flat (kurtosis ≈ −1.17) and nearly symmetric (skew ≈ −0.21), with the bulk of records concentrated between 1928 and 1977 (IQR = 49.25 years), which is surprisingly wide and uniform for a year field.

Treatment: Use as an ordinal or numeric feature; consider binning into decades if cardinality reduction is needed.

last_eruption numeric feature

This column almost certainly records the year of a volcano's last known eruption, ranging from 1900 to 1999 with a mean of 1952.3 and median of 1955 — consistent with a dataset scoped to the 20th century. The distribution is notably platykurtic (kurtosis ≈ −1.17), meaning eruption years are spread fairly uniformly across the century rather than clustering tightly around any single period. With only 84 unique values across 200 rows, many volcanoes share the same recorded eruption year, which is unsurprising given that annual granularity naturally produces ties. No nulls, no outliers, and near-zero skew make this a clean numeric feature.

Treatment: Use as-is or engineer recency features (e.g., years since eruption relative to a reference year); no transformation needed given near-uniform distribution.