saturn

This dataset catalogues 80,678 waterfalls worldwide, sourced entirely from OpenStreetMap with latitude/longitude coordinates and minimal descriptive metadata. The most striking feature is how sparse the descriptive fields are: 'category' and 'source' are constant, 'date' and 'country' are essentially empty (country is blank for 80,650 of 80,678 rows), and 89.9% of 'description' entries are simply 'Waterfall'. The 'name' field is similarly thin — 'Unnamed Waterfall' accounts for 48,168 rows and the duplicate rate is 65.7%. The real analytical signal lives in the geographic coordinates, where latitude skews toward the northern hemisphere (median 40.3) and longitude spans the full globe, making this primarily a spatial dataset rather than an attribute-rich one.

Geographic latitude coordinates spanning -77.72 to 78.66, covering nearly the full habitable range of Earth. The distribution is left-skewed (-0.94) with a median of 40.31 sitting well above the mean of 27.15, suggesting a concentration of records in the Northern Hemisphere with a long tail reaching toward Antarctica. Near-uniqueness (80,650 distinct of 80,678) and zero nulls indicate clean, granular point data.

This is a longitude coordinate column spanning the full global range from -179.99 to 179.41, with 80,650 unique values across 80,678 rows and no nulls. The distribution is broad (std 76.86, IQR 100.27) and only mildly skewed (0.29), with the median at 7.80 sitting east of the prime meridian, hinting at a Europe/Africa-leaning sample. No outliers were flagged, consistent with values bounded by valid geographic limits.

This is the human-readable name of a waterfall, averaging 2 words and 16 characters. The column is dominated by the placeholder 'Unnamed Waterfall' (48,168 of 80,678 rows), driving a 65.7% duplicate rate; multiple languages appear in the vocabulary (cachoeira, cascada, cascata, salto, fossen) alongside English 'falls'.

This is a categorical descriptor column, overwhelmingly dominated by the value "Waterfall" which accounts for 72,565 of 80,678 rows (top_rate 0.899). The remaining 774 categories appear to be variants annotated with heights (e.g. "Waterfall, 3m", "Waterfall, 5m"), suggesting a structured suffix pattern rather than free text. Entropy is very low (1.14, ratio 0.119) and the long_tail alert fires, so most signal collapses into one label.

This column is a constant categorical tag, holding the literal value "usgs_waterfalls" for all 80678 rows. With cardinality 1, entropy 0, and a top_rate of 1.0, it carries no information and likely just records the source dataset or ingestion batch.

This column is named 'date' but contains a single value—an empty string—across all 80678 rows. Cardinality is 1, top_rate is 1.0, and entropy is 0.0, so the field carries no information. It looks like a date field that was never populated.

This is an ISO country code field that is effectively empty: 80650 of 80678 rows (top_rate 0.9996) hold the blank string, leaving only 28 actual codes spread across VE (24), and one each for DE, LB, HN, and BR. Entropy is 0.0048 (entropy_ratio 0.0019), so the column carries almost no information despite having no nulls. The non-blank values look plausible but are far too sparse to support segmentation or modelling.

A nominally numeric height field stored as strings, but 89.9% of the 80,678 rows are empty and the remaining values spread across 775 distinct tokens with very low entropy ratio (0.119). The populated values look like small integers (3, 2, 5, 10...) with no obvious unit, suggesting inconsistent or truncated entry rather than a clean measurement.

This column records the data provenance, with every one of the 80678 rows tagged as "OpenStreetMap". Cardinality is 1 and entropy is 0, so the field carries no information for modelling or filtering.