Dataset of Factors Affecting Forest and Grassland Fires in Southwest China

Dataset of Factors Affecting Forest and Grassland Fires in Southwest ChinaFengyun-3D (FY-3D) global active fire productWildfires have a strong negative effect on the environment, ecology and public health. However, the potential degradation of mainstream global fire products leads to large uncertainty in the effective monitoring of wildfires and their influence. To fill this gap, we produced Feng yun-3D (FY-3D) global fire products with a similar spatial and temporal resolution, aiming to serve as an alternative to and continuity for Moderate Resolution Imaging Spectroradiometer (MODIS) global fire products. Firstly, the sensor parameters and major algorithms for noise detection and fire identification in FY-3D products were introduced. For visual-check-based accuracy assessment, five typical regions with a large number of fire spots across the globe, Africa, South America, the Indochinese Peninsula, Siberia and Australia, were selected, and the overall accuracy exceeded 94 %. Meanwhile, the consistence between FY-3D and MODIS fire products was examined. The result suggested that the overall consistence was 84.4 %, with a fluctuation across seasons, surface types and regions. The high accuracy and consistence with MODIS products proved that the FY-3D fire product is an ideal tool for global fire monitoring. Based on field-collected reference data, we further evaluated the suitability of FY-3D fire products in China. The overall accuracy and accuracy without considering omission errors were 79.43 % and 88.50 % higher, respectively, than those of MODIS fire products. Since detailed local geographical conditions were specifically considered, FY-3D products should be preferably employed for fire monitoring in China. Meteorological factorsMeteorological factors were extracted from the monthly ERA5-Land climate reanalysis dataset on the GEE platform, including monthly average precipitation, wind speed, and temperature for the study area. Relative humidity was calculated based on air temperature and dew point temperature. The 3σ principle was applied to each meteorological factor to filter the extracted data. The mean (μ) and standard deviation (σ) of the monthly averages were calculated, and values ​​exceeding μ ± 3σ were replaced with boundary values. This dataset combines long-term series data with high spatiotemporal resolution, providing reliable spatiotemporal data support for the construction of meteorological factors in subsequent forest and grassland fire susceptibility models.Land surface environmental factorsLand surface environmental factors include vegetation index, vegetation cover, normalized differential moisture index (NDMI), snow cover, land surface temperature (LST), leaf area index (LAI), and soil moisture content (SM). All of these factors were acquired and preprocessed using the GEE platform. Monthly average NDVI was extracted for the study area based on the MOD13A1.061 (Terra) and MYD13A1.061 (Aqua) datasets from the GEE platform. Based on the extracted NDVI values, FVI values ​​were calculated using a pixel-based bisection model. This paper utilizes the MOD09GA.061 data from the GEE platform to extract the red band (Band 4, 620–670 nm) and shortwave infrared band (Band 5, 1230–1250 nm), with units of reflectance (0–1). Monthly average NDMI for the study area was calculated. Snow cover was extracted using the NDSI thresholding method based on the MOD10A1.061 snow cover product from the GEE platform (NDSI > 0.4 indicates snow cover). The characteristics of snow cover are highlighted by calculating the ratio of the difference and sum of the reflectance in the green band and the shortwave infrared band (SWIR). Snow cover has high reflectance in the green band and low reflectance in the SWIR band, so snow-covered areas will exhibit a high NDSI value under this index calculation. The monthly mean land surface temperature of the study area is calculated using the MOD11A1.061 (Terra) and MYD11A1.061 (Aqua) datasets from GEE. As a key land surface environmental variable, land surface temperature can be used to assess regional environmental thermal conditions. The monthly mean LAI of the study area is calculated using the MOD15A2H.061 (Terra) and MYD15A2H.061 (Aqua) datasets from GEE. As a key land surface environmental variable, LAI can effectively characterize the amount of combustible load in the region. The surface soil moisture of the corresponding time period within the study area is screened using the GLDAS-2.2 land surface dataset from GEE, and the monthly average value is calculated to construct the soil moisture factor. This factor effectively reflects the regional soil moisture status and provides key environmental variable support for the assessment of forest and grassland fire susceptibility.Topographic factorsUsing a 1 km precision DEM obtained from the Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, slope, aspect, and elevation of the study area were extracted through GIS analysis to construct topographic feature factors and reveal the impact of topography on forest and grassland fires in Southwest China.Human activity factorsHuman activity factors include nighttime light data and population density data, both acquired through the GEE platform. The NPP-VIIRS nighttime light imagery data used is sourced from satellite sensors of the National Oceanic and Atmospheric Administration (NOAA). Extraction was performed using vector data of the Southwest region as a mask, ultimately obtaining historical nighttime light data. As a key factor reflecting human activity, nighttime light effectively characterizes the intensity of regional human activity, laying a solid foundation for quantifying the socio-economic factors of fire risk. LandScan global population distribution data, provided by Oak Ridge National Laboratory (ORNL), utilizes advanced spatial modeling techniques and advanced geospatial data sources to provide detailed information on population size and density at a resolution of 30 arcseconds, enabling accurate and timely insights into global human settlement patterns. This paper uses LandScan global population distribution data from GEE to obtain population grid data for the study area and calculate the monthly average population density. As a key human activity factor, population density can be used to assess the impact of population distribution on fire exposure and risk.Historical fire factorsUsing MCD64A1 data from GEE, the fire frequency density of the study area was calculated through local mean filtering. As a historical fire factor, the fire frequency density can be used to reveal the spatiotemporal distribution characteristics of high-frequency combustion zones.