Collects weather observations and disaster data as CSV files. 한국어
The *_data/ directories contain only a few example snapshots, not a complete archive. Install dependencies with pip install -r requirements.txt, then set the required API keys as environment variables (see src/config.py). Call collectors in src/api/ to fetch more data.
PYTHONPATH=. python -c " from src.api.nasa_api import get_firms_data; get_firms_data(5)" This fetches the previous five days of FIRMS data (NASA_FIRMS_MAP_KEY required). For specific dates, change inqDt in src/api/kfs_api.py, time or date in src/api/kma_api.py, or ymdhm in src/api/moe_api.py. AWS and real-time wildfire calls request the latest data; historical availability depends on each provider.
Wildfire — KFS real-time wildfire API
UV exposure — KMA living weather index API
Flood — MOE real-time flood API
Landslide — KFS landslide forecast API
Heavy snow — data source not specified
Fog — KMA AWS minute visibility API
Data: temperature, pressure, cloud height and cover, visibilityUpdate interval: one minute (approximately 5–7 minutes of publication delay)
Data: regional UV index forecasts up to 75 hours aheadUpdate interval: eight times daily, every three hours (00:00, 03:00, …, 21:00)
Update interval: real time, when a fire is reported
Update interval: twice daily (01:00–03:00 and 13:00–14:00)
Update interval: every two minutes (approximately 10 hours of publication delay)Reference: GK2A metadata
Update interval: every 10 minutes (approximately five minutes of publication delay)Data: river levels at 381 sites nationwide and flood risk (caution, warning, severe)Reference: Flood Information System
Update interval: annually, at year-endData: flood trace PNGs (124–132°E, 33–39°N)Reference: Flood Trace OpenAPI OpenCV extracts flood trace pixel coordinates into CSV (safemap_data/csv/*.csv); error rate: 20%.
KFS landslide forecast API
Update interval: documented as every five minutes (observed updates appear to occur each morning)Data: forecasts for areas with high landslide probabilityReference: KFS landslide forecast OpenAPI
1. Korea Forest Service (KFS) 1-1. Wildfire API (kfs_data/*.csv) Real-time wildfire information
Field
Example
Meaning
frfrInfoId
398504
Unique wildfire incident ID
frfrFrngDtm
2025-08-22 14:41:22
Fire start time
frfrSttmnDt
20250822
Report date (YYYYMMDD)
frfrSttmnHms
144122
Report time (HHMMSS → 14:41:22)
frfrSttmnAddr
충청북도 음성군 감곡면 영산리
Reported administrative address
frfrSttmnAddrDe
충청북도 음성군 감곡면 영산리 산55-3임
Detailed address
frfrLctnXcrd
127.67243000000057
Fire longitude
frfrLctnYcrd
37.08104000024055
Fire latitude
frfrSttmnLctnXcrd
127.67243000000029
Reported longitude
frfrSttmnLctnYcrd
37.08104000012027
Reported latitude
lgdngCd
4377037026
Administrative area code
frfrPrgrsStcd
02
Progress status code
frfrPrgrsStcdNm
진화중
Progress status (firefighting in progress)
frfrPotfrRt
70
Containment rate (%)
frfrStepIssuCd
00
Response stage code
frfrStepIssuNm
초기 대응
Response stage (initial response)
frfrOccrrTpcd
05
Incident type code (lookup needed)
frfrOccrrStcd
31
Detailed cause code (lookup needed)
1-2. Fire warning API (kfs_data/warning_list/*.csv) Real-time wildfire warning areas
Field
Example
Meaning
AREA
영덕군
City/county/district
AREA_ENG
Yeongdeok-gun
English city/county/district name
LOCATION
경상북도 영덕군
Full address (province + city/county/district)
WARNING_LEVEL
관심
Fire warning level (attention)
DATE
2025-09-02 11:00
Warning issue time
LAST_UPDATED
2025-09-02 01:15:59
Last update time
PROVINCE_CODE
47
Province code
CITY_CODE
770
City code
1-3. Landslide API (kfs_data/landslide/*.csv) Real-time landslide forecasts
Production: every five minutes
Parameter
API name
Type
Size
Required
Description
Service key
serviceKey
STRING
50
Yes
API service key
Rows per page
numOfRows
NUMBER
30
No
Rows per page
Page number
pageNo
NUMBER
30
No
Page number
Response format
returnType
VARCHAR
30
No
JSON or XML
Query start date
inqDt
STRING
—
No
YYYYMMDD
Field
API name
Type
Size
Required
Description
Landslide forecast name
LNLD_FRCST_NM
—
20
Yes
Forecast name
City/county/district
SGG_NM
—
100
Yes
Area name
Prediction analysis time
PREDC_ANLS_DT
—
50
Yes
Analysis time
2. Korea Meteorological Administration (KMA) 2-1. AWS (automated weather stations)
Production: every minute (approximately 3–5 minutes of delay)Stations: 510 nationwide (see map_data/grid.csv)
2-1-1. AWS minute observations (kma_data/<timestamp>/AWS_*.csv)
Wind direction and speed (one-minute mean, ten-minute mean, maximum gust)
Air temperature (one-minute mean)
Precipitation (detection; 15-minute, 60-minute, 12-hour, and daily totals)
Relative humidity
Local and sea-level pressure
Dew point
Index
Meaning
Unit / range
WD1
One-minute mean wind direction
degrees (0=N, 90=E, 180=S, 270=W, 360=calm)
WS1
One-minute mean wind speed
m/s
WDS
Maximum gust direction
degrees
WSS
Maximum gust speed
m/s
WD10
Ten-minute mean wind direction
degrees
WS10
Ten-minute mean wind speed
m/s
TA
One-minute mean air temperature
°C
RE
Precipitation detection
0=no rain; nonzero=rain
RN-15m
15-minute accumulated precipitation
mm
RN-60m
60-minute accumulated precipitation
mm
RN-12H
12-hour accumulated precipitation
mm
RN-DAY
Daily accumulated precipitation
mm
HM
One-minute mean relative humidity
%
PA
One-minute mean local pressure
hPa
PS
One-minute mean sea-level pressure
hPa
TD
Dew point temperature
°C
2-1-2. AWS cloud height and cover (kma_data/<timestamp>/AWS_cloud_*.csv)
Low, middle, and upper cloud height
Total cloud cover
Index
Meaning
Unit
Note
CH_LOW
Low cloud height
m
7620 m = no clouds
CH_MID
Middle cloud height
m
—
CH_TOP
Upper cloud height
m
—
CA_TOT
Total cloud cover
%
—
2-1-3. AWS surface and soil temperature (kma_data/<timestamp>/AWS_temp_*.csv)
Air, dew point, grass minimum, and ground surface temperature
Relative humidity
Soil temperature (5 cm–5 m)
Local and sea-level pressure
Index
Meaning
Unit
TA
One-minute mean air temperature
°C
HM
One-minute mean relative humidity
%
TD
One-minute mean dew point
°C
TG
One-minute mean grass minimum temperature
°C
TS
One-minute mean ground surface temperature
°C
TE0.05
Soil temperature at 5 cm
°C
TE0.1
Soil temperature at 10 cm
°C
TE0.2
Soil temperature at 20 cm
°C
TE0.3
Soil temperature at 30 cm
°C
TE0.5
Soil temperature at 50 cm
°C
TE1.0
Soil temperature at 1.0 m
°C
TE1.5
Soil temperature at 1.5 m
°C
TE3.0
Soil temperature at 3.0 m
°C
TE5.0
Soil temperature at 5.0 m
°C
PA
One-minute mean local pressure
hPa
PS
One-minute mean sea-level pressure
hPa
Values of −50 or below indicate a missing observation or error. Soil temperature is measured at selected stations only.
2-1-4. AWS visibility (kma_data/<timestamp>/AWS_vis_*.csv)
Visibility (one-minute and ten-minute mean)
Fog and present weather
Index
Meaning
Unit / note
S
Instrument type
1=fog network; 2=modernized equipment
VIS1
One-minute mean visibility
m; one-second sampling, modernized equipment
VIS10
Ten-minute mean visibility
m; fog network only
WW1
One-minute instantaneous weather code
one-second sampling, modernized equipment
WW15
15-minute mean weather code
fog network only
Code
Meaning
0–2
Clear
4
Haze
10
Mist
30
Fog
40–42
Rain
50–59
Drizzle
60–68
Rain
71–76
Snow
2-2. GK2A satellite API (GK2A_data/csv/*.csv)
Data: suspected fire locations, delivered as NetCDF (.nc)Production: approximately every two minutesData delay: approximately 10 hoursCoverage: Korean Peninsula and extended areaPerformance: true positive 87.17%; false positive 15.20%
Column
Example
Meaning
lat37.48554
Fire detection latitude (degrees)
lon129.0597
Fire detection longitude (degrees)
FF0, 1
Fire flag (0=no fire; 1=fire)
DQF_FF10
Data quality flag (0–13)
Code
Meaning
0
Invalid; outside observation range (SZA > 70°)
1
Invalid; masked area or missing input
2
Land
3
Water
4
Cloud
5
Rejected by cloud test
6
Rejected by bare soil, urban, and water test
7
Potential fire
8
Fire
9
Absolute fire
10
Industrial heat detection
12
Stability test
13
Probably cloud
2-3. Living weather index: UV (kma_data/uv/*.csv)
Data: UV index (0 to 11+)Production: every three hours, eight times daily (00:00, 03:00, …, 21:00)Coverage: Korean Peninsula
Column
Example
Meaning
areaNo
110000000
Administrative area code (see map_data/korea_administrative_zone_code.csv)
date
2025090118
Request time (YYYYMMDDHH)
h0–h75
8
UV index n hours ahead in field h<n>
Level
Index
Extreme
11+
Very high
8–10
High
6–7
Moderate
3–5
Low
0–2
3. Ministry of Environment (MOE) 3-1. Real-time flood API (moe_data/*.csv)
Data: river levels and flood risk (caution, warning, severe)Production: every ten minutes (approximately five minutes of delay)Coverage: 381 sites nationwide
Column
Meaning
Example
lon
Longitude
128.27
lat
Latitude
35.728
obsnm
Observation site / bridge
고령군(회천교)
ymdhm
Date and time
2025-09-07 14:30
wl
Current water level
3.25
wrnwl
Warning level (fixed)
4.5
almwl
Caution level (fixed)
5.0
Output: firms_data/*.csvData: satellite-based fire monitoringProduction: approximately every 12 hours (twice daily)Coverage: Korean Peninsula
Column
Example
Meaning
latitude37.48554
Fire detection latitude (degrees)
longitude129.05978
Fire detection longitude (degrees)
frp10.3
Fire Radiative Power (MW; fire intensity)
daynightD
Day/night (D=day, N=night)
acq_date2025-08-08
Satellite observation date (UTC in API response)
acq_time0950
Satellite observation time (UTC, HHMM, in API response)
satelliteTerra
Observing satellite
instrumentMODIS
Observing sensor
version6.1NRT
Algorithm version (NRT=near real time)
VIIRS_NOAA20_NRT — confidence n
Fire confidence (l=low, n=normal, h=high)
VIIRS_NOAA20_NRT — bright_ti4 331.48
Band 4 brightness temperature (K)
VIIRS_NOAA20_NRT — bright_ti5 302.14
Band 5 brightness temperature (K)
MODIS_NRT — confidence 65
Fire confidence (%)
MODIS_NRT — brightness 313.81
Fire-pixel brightness temperature (K)
MODIS_NRT — bright_t31 301.66
Band 31 brightness temperature (K)
The collector converts acq_date and acq_time from UTC to Korea Standard Time in saved CSV files.
