Dataset Viewer
Auto-converted to Parquet Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
fire_id: string
country: string
latitude: double
longitude: double
date: date32[day]
daynight: string
first_seen_utc: timestamp[us]
last_seen_utc: timestamp[us]
frp_mw: double
frp_max_mw: double
bright_main_k: double
bright_second_k: double
confidence: string
fire_type: int8
fire_type_label: string
n_detections: int32
n_passes: int16
modis_detections: int32
viirs_detections: int32
imputed_detections: int32
to
{'country': Value('string'), '2018': Value('int64'), '2019': Value('int64'), '2020': Value('int64'), '2021': Value('int64'), '2022': Value('int64'), '2023': Value('int64'), '2024': Value('int64'), 'sen_slope_per_year': Value('float64'), 'kendall_tau': Value('float64'), 'p_value': Value('float64'), 'trend_direction': Value('string'), 'trend_significance': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/parquet/parquet.py", line 220, in _generate_tables
                  yield Key(file_idx, batch_idx), self._cast_table(pa_table)
                                                  ~~~~~~~~~~~~~~~~^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/parquet/parquet.py", line 156, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              fire_id: string
              country: string
              latitude: double
              longitude: double
              date: date32[day]
              daynight: string
              first_seen_utc: timestamp[us]
              last_seen_utc: timestamp[us]
              frp_mw: double
              frp_max_mw: double
              bright_main_k: double
              bright_second_k: double
              confidence: string
              fire_type: int8
              fire_type_label: string
              n_detections: int32
              n_passes: int16
              modis_detections: int32
              viirs_detections: int32
              imputed_detections: int32
              to
              {'country': Value('string'), '2018': Value('int64'), '2019': Value('int64'), '2020': Value('int64'), '2021': Value('int64'), '2022': Value('int64'), '2023': Value('int64'), '2024': Value('int64'), 'sen_slope_per_year': Value('float64'), 'kendall_tau': Value('float64'), 'p_value': Value('float64'), 'trend_direction': Value('string'), 'trend_significance': Value('string')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Thermora harmonized MODIS + VIIRS fires: one file per country

NASA FIRMS MODIS + VIIRS active fires, cleaned, harmonized to one schema and fused: one row per fire observation (common ~0.01 deg pixel x local date x day/night pass). Source: NASA FIRMS yearly country archives (MODIS C6.1, VIIRS 375 m: Suomi NPP, NOAA-20, NOAA-21).

Every country is a separate file in countries/. Read one, a few, or all of them:

import duckdb
duckdb.sql("SELECT * FROM 'hf://datasets/<user>/<repo>/countries/Egypt.parquet' LIMIT 10")
duckdb.sql("SELECT country, count(*) FROM 'hf://datasets/<user>/<repo>/countries/*.parquet' GROUP BY 1")
country rows first date last date file
Croatia 33,497 2000-11-06 2024-12-30 countries/Croatia.parquet
Cuba 460,573 2000-11-01 2024-12-31 countries/Cuba.parquet
France 183,919 2000-11-02 2024-12-31 countries/France.parquet
Germany 235,138 2000-11-02 2024-12-31 countries/Germany.parquet
Hungary 42,944 2000-11-01 2024-12-30 countries/Hungary.parquet
India 9,414,385 2000-11-01 2025-01-01 countries/India.parquet
Indonesia 4,086,760 2000-11-01 2025-01-01 countries/Indonesia.parquet
Iran 2,107,063 2000-11-01 2025-01-01 countries/Iran.parquet
Iraq 1,795,498 2000-11-01 2025-01-01 countries/Iraq.parquet
Italy 422,294 2000-11-02 2025-01-01 countries/Italy.parquet
Japan 382,154 2000-11-03 2025-01-01 countries/Japan.parquet
Kenya 299,528 2000-11-01 2025-01-01 countries/Kenya.parquet
Kuwait 132,591 2000-11-01 2025-01-01 countries/Kuwait.parquet
Libya 494,055 2000-11-02 2025-01-01 countries/Libya.parquet
Madagascar 5,925,648 2000-11-01 2025-01-01 countries/Madagascar.parquet
Malawi 997,692 2000-11-05 2024-12-31 countries/Malawi.parquet
Malaysia 300,848 2000-11-01 2025-01-01 countries/Malaysia.parquet
Maldives 1,110 2000-11-04 2024-12-28 countries/Maldives.parquet
Mali 2,491,889 2000-11-01 2024-12-31 countries/Mali.parquet
Mexico 4,823,994 2000-10-31 2024-12-31 countries/Mexico.parquet
Mozambique 11,778,862 2000-11-01 2025-01-01 countries/Mozambique.parquet
South_Africa 2,827,208 2000-11-01 2025-01-01 countries/South_Africa.parquet
South_Sudan 9,697,256 2000-11-01 2025-01-01 countries/South_Sudan.parquet
Spain 253,001 2000-11-01 2024-12-31 countries/Spain.parquet
Sudan 2,595,889 2000-11-01 2025-01-01 countries/Sudan.parquet
Thailand 2,811,758 2000-11-01 2025-01-01 countries/Thailand.parquet
Togo 829,362 2000-11-01 2024-12-31 countries/Togo.parquet
United_Arab_Emirates 91,099 2000-11-02 2025-01-01 countries/United_Arab_Emirates.parquet
Zambia 13,142,648 2000-11-01 2024-12-31 countries/Zambia.parquet

Border pixels: each country is fused on its own, so a pixel on a border can appear in two countries' files with the same fire_id (GROUP BY fire_id merges them).

Columns

column meaning
fire_id stable id of the pixel/date/pass
country country of most detections in the pixel
latitude, longitude centre of the common pixel
date local solar date of the observation
daynight D or N pass
first_seen_utc, last_seen_utc first and last detection fused into this row
frp_mw fire radiative power, MW: pixels summed within a pass, passes averaged
frp_max_mw strongest pass, MW
bright_main_k fire-band (~4 um) brightness temperature, radiance-averaged, K
bright_second_k background-band (~11 um) brightness temperature, radiance-averaged, K
confidence highest detection confidence: low / nominal / high
fire_type, fire_type_label most frequent FIRMS type: 0 vegetation, 1 volcano, 2 static land source, 3 offshore
n_detections, n_passes detections and satellite passes fused into the row
modis_detections, viirs_detections how many of them came from each sensor
imputed_detections detections that had a missing value filled before fusion

Fusion rules

  • Common pixel: a GRID_DEG grid (0.01 deg, ~1.1 km, the MODIS resolution). The VIIRS 375 m pixels that fall inside a MODIS pixel land in the same cell. latitude/longitude of a row = centre of its cell.
  • Same observation = same cell + same local solar date + same day/night pass. Day and night stay separate rows: they are different passes of a burning fire.
  • FRP: within ONE satellite pass, the pixels in a cell each see part of the fire, so their FRP is SUMMED (total fire power of the cell at that moment). Across passes (e.g. Terra 10:30, VIIRS 13:30) those totals are AVERAGED -> frp_mw. frp_max_mw = the strongest pass.
  • Brightness temperature is not additive. Each detection's temperature is turned into radiance (Planck's law, 3.9 um for the fire band, 11 um for the background band), averaged weighted by pixel area within a pass, averaged across passes, then turned back into Kelvin. A hot sub-pixel fire therefore counts as much as it radiates.
  • Confidence: the highest of the detections (low < nominal < high).
  • Fire type: most frequent type in the cell (0 vegetation, 1 volcano, 2 other static land source, 3 offshore). Nothing is filtered.
  • Times: first_seen_utc / last_seen_utc of the detections that were fused.
  • Provenance counts: n_detections, n_passes and how many detections came from MODIS and from VIIRS (needed to measure what the legacy MODIS record missed).

Cleaning (per detection, before fusion)

  • Dropped: no valid latitude/longitude/date, or unknown satellite.
  • Impossible values set to NULL, then filled: frp < 0, brightness outside 200-1000 K (background band 150-1000 K), scan/track outside (0, 10] km, confidence outside 0-100, fire type outside 0-3, invalid HHMM time.
  • Duplicates (same satellite + position to 4 decimals + date + time) removed, most complete copy kept, before filling.
  • Missing numbers filled with the average of the same country + sensor + satellite + day/night (then country + sensor, then country). Missing time: the satellite's mean local crossing time. Missing day/night: from local solar time. Missing type/version: most frequent value.
Downloads last month
354