celestrak active satellites

Overview

Source: /home/coolhand/html/datavis/data_trove/cache/celestrak_active_satellites.json

Saturn profiled 13,983 rows across 17 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).

OBJECT_NAME text identifier

This is the satellite/space object designation column (OBJECT_NAME), with 13,982 unique values across 13,983 rows and only one duplicate. Values are short (mean 13.3 chars, median 1 word) and almost entirely uppercase (99.99%), consistent with catalog identifiers; recurring tokens like [DTC], FLOCK, COSMOS, IRIDIUM, JILIN-1 and TRANSPORTER-15 reflect known constellation/launch families. Effectively a near-unique label rather than a modelling feature.

Treatment: Treat as an identifier; drop from modelling or use only for joins/lookup.

OBJECT_ID text identifier

OBJECT_ID is a unique catalog identifier, almost certainly the COSPAR international designator for space objects (e.g., '1964-063c', '1965-034c'). Every one of the 13,983 rows is distinct and non-null, single-token, all-caps, and 9-10 characters long, which matches the YYYY-NNN+suffix convention exactly. There is no duplication or missingness to worry about.

Treatment: Use as the primary key for joins; exclude from modelling features.

EPOCH text timestamp

Despite being typed as text and flagged 'allcaps'/'one_word', EPOCH holds ISO-8601 timestamps of fixed length 26, all on 2025-12-21. The value '2025-12-21T14:00:01.999872' appears 658 times and overall duplicate_rate is 5.3% across 13,235 unique values, suggesting a single-day snapshot with one dominant batch timestamp.

Treatment: Parse to datetime and extract hour/minute features; investigate the 658-row spike at 14:00:01.

MEAN_MOTION numeric feature

MEAN_MOTION is a numeric orbital parameter (revolutions per day) for 13,983 objects, with 13,734 unique values and no nulls. The bulk sits tightly between Q1 14.99 and Q3 15.30 (median 15.09), consistent with low-Earth-orbit satellites, but a heavy left tail pulls the mean down to 14.24 and the minimum to 0.28, producing strong negative skew (-3.62) and kurtosis 11.6. That tail flags 2,440 outliers (17.4%), likely geosynchronous or higher-orbit objects mixed into an LEO-dominated population.

Treatment: Segment by orbit regime or apply a robust/quantile transform before modelling to handle the heavy left tail.

ECCENTRICITY numeric feature

This is the orbital ECCENTRICITY of 13,983 objects, with values ranging from 4.6e-07 to 0.8777294 and a median of just 0.00015248. The distribution is extremely tight around near-circular orbits (Q1–Q3 spans only 0.00017982) but has a heavy right tail, producing skew of 18.65, kurtosis of 356.68, and 2,164 outliers (15.5%). The mean (0.00272) sits an order of magnitude above the median, confirming a small population of highly eccentric orbits dominates the upper end.

Treatment: Log-transform or use a robust scaler before modelling to tame the heavy right tail.

INCLINATION numeric feature

Numeric orbital inclination angles in degrees, spanning 0.0008 to 148.0433 across 13,983 rows with no nulls and 4,236 unique values. The distribution is nearly symmetric (skew 0.13, kurtosis 0.05) and centred around a median of 53.16° with IQR 43.00–80.99°, consistent with a mix of low-inclination and polar/retrograde orbits. Only 3 outliers flagged, and the max above 90° confirms retrograde objects are present.

Treatment: Use as-is as a numeric feature; no transform needed given near-symmetric distribution.

RA_OF_ASC_NODE numeric feature

This column holds the right ascension of the ascending node, an orbital element measured in degrees. Values span almost the full 0.053 to 359.9097 range with a near-uniform spread (skew 0.198, kurtosis -1.25) and no outliers, consistent with an angular coordinate distributed around the celestial sphere. With 13918 unique values across 13983 rows and no nulls, it behaves as a continuous orbital feature rather than a categorical key.

Treatment: Treat as a circular variable (e.g. encode as sin/cos of the angle) before modelling.

ARG_OF_PERICENTER numeric feature

This column holds the argument of pericenter, an orbital element measured in degrees, with values bounded between 0.0094 and 359.9416 across 13,983 rows and 13,898 unique values. The distribution is broad and nearly flat (kurtosis -1.53, skew 0.26) with mean 165.24 and median 113.07, consistent with an angular quantity spanning the full 0-360° range. No nulls, no zeros, and no outliers are flagged.

Treatment: Encode as a circular variable (e.g., sin/cos transform) before modelling rather than using the raw degrees.

MEAN_ANOMALY numeric feature

MEAN_ANOMALY is a numeric astronomical/orbital field with values spanning 0.0645 to 359.9554, consistent with an angle measured in degrees over a full circle. The distribution is broad (std 94.5, IQR 180.9) and mildly left-skewed (-0.31) with platykurtic shape (kurtosis -1.45), suggesting a fairly uniform spread across the 0-360 range rather than a peaked distribution. Near-unique values (13,884 of 13,983) and zero nulls/outliers indicate a clean continuous measurement.

Treatment: Encode as a circular variable (e.g., sin/cos transform) before modelling since the 0-360 range wraps.

EPHEMERIS_TYPE numeric metadata

EPHEMERIS_TYPE is a numeric field that is entirely constant: across all 13983 rows it holds the value 0, with zero_rate 1.0 and n_unique 1. It carries no information for modelling or segmentation.

Treatment: Drop, constant column.

CLASSIFICATION_TYPE categorical metadata

Constant column holding the single value 'U' across all 13983 rows with no nulls. Entropy is 0 and cardinality is 1, so it carries no information for modelling or segmentation.

Treatment: Drop; constant column with zero entropy.

NORAD_CAT_ID numeric identifier

NORAD_CAT_ID is the satellite catalog number assigned by NORAD; with 13,983 unique values across 13,983 rows and zero nulls, it serves as a primary key for tracked space objects. Values span 900 to 67,228 with a left-skewed distribution (skew -1.29) and median 58,353, reflecting that most catalog entries are recent launches while a small tail of legacy objects pulls the minimum down to 900. The 318 low-end outliers (2.27%) are consistent with these older catalog assignments rather than data errors.

Treatment: Use as a join key; exclude from modelling features.

ELEMENT_SET_NO numeric metadata

ELEMENT_SET_NO is a numeric column that is entirely constant: all 13,983 rows hold the value 999.0, with zero nulls and zero variance. There is no information content here — min, max, mean, and median all equal 999.0, and n_unique is 1.

Treatment: Drop; constant column carries no signal.

REV_AT_EPOCH numeric feature

REV_AT_EPOCH is a fully populated numeric column with 13983 rows and 6964 unique values, ranging from 1 to 99870 with a median of 9780 and mean of 12690.5. The distribution is heavily right-skewed (skew 2.54, kurtosis 9.63) with 523 outliers (3.7%) above the upper fence, and the std (13044.7) exceeds the mean. The name suggests a revenue measurement keyed to an epoch, but no temporal evidence is present in the payload to confirm that.

Treatment: log-transform before regression to tame the right skew.

BSTAR numeric feature

BSTAR is the satellite drag term (B*) from TLE orbital element sets — a numeric ballistic coefficient typically tiny in magnitude. Values cluster tightly (median 9.8e-05, IQR ~4.7e-04) but the range spans -0.133 to 0.0798 with extreme negative skew (-8.50) and kurtosis of 214.79, plus 10.04% flagged as outliers and 5.70% exact zeros. The heavy negative tail is unusual since drag terms are usually positive, suggesting a meaningful subset of decaying or anomalous objects.

Treatment: Apply a signed log transform (e.g., sign(x)*log1p(|x|)) and consider winsorising before modelling.

MEAN_MOTION_DOT numeric feature

MEAN_MOTION_DOT is the first time-derivative of mean motion from TLE-style orbital element sets, capturing how quickly an object's orbital rate is changing. Values cluster tightly near zero (median 8.95e-06, IQR 0.00012452) but the tails are extreme: min -0.03542877, max 0.07662247, kurtosis ~123.8 and skew -1.76, with 13.2% flagged as outliers. The heavy negative skew suggests a subset of objects experiencing strong orbital decay relative to the typical near-stable population.

Treatment: Keep as-is for physics-aware models; for linear methods apply a signed log (e.g., sign(x)*log1p(|x|)) and consider winsorising the 1844 outliers.

MEAN_MOTION_DDOT numeric feature

MEAN_MOTION_DDOT is the second derivative of mean motion from a TLE/orbital element record, but here every one of the 13,983 rows is exactly 0.0 with zero nulls and a single unique value. It carries no information in this snapshot — likely because the source feed never populates this term for the included objects.

Treatment: Drop; constant column with no variance.