Multi-ensemble Biomass-burning Emissions Inventory (MBEI)

MBEI is a monthly, 0.1° global biomass-burning emissions ensemble for 2003–2023 that unifies two complementary pathways: a bottom-up line based on burned area and fuel load and a top-down line based on satellite fire radiative power (FRP). All inputs are harmonized to 0.1°/monthly grids, and emissions are provided for 11 species using land-cover- and region-specific emission factors. In addition to the ensemble mean, MBEI reports grid-cell minima and maxima across members to make uncertainty explicit. Eight sub-inventories and their naming.Each member name encodes four choices: method (BU/TD), active-fire confidence (AC/HC), and the key dataset varied within that method. BU-AC-GB — Bottom-Up, All-Confidence, GlobBiomass: uses all MODIS active fires and the GlobBiomass AGB map to drive fuel load.BU-AC-BC — Bottom-Up, All-Confidence, Biomass_cci: identical to BU-AC-GB except AGB comes from Biomass_cci.BU-HC-GB — Bottom-Up, High-Confidence, GlobBiomass: same as BU-AC-GB but restricted to medium-to-high confidence fire pixels (>30%).BU-HC-BC — Bottom-Up, High-Confidence, Biomass_cci: high-confidence fires with Biomass_cci AGB.TD-AC-LC30 — Top-Down, All-Confidence, 30-class: FRP-to-DM conversion uses a 30-class biome CR map.TD-AC-LC8 — Top-Down, All-Confidence, 8-class: conversion uses the GFAS 8-class biome CR map.TD-HC-LC30 — Top-Down, High-Confidence, 30-class: as TD-AC-LC30 but limited to medium-to-high confidence fires.TD-HC-LC8 — Top-Down, High-Confidence, 8-class: as TD-AC-LC8 with the high-confidence filter.Methods and core inputs. Active fire detections & FRP: MODIS NRT MCD14DL (Terra & Aqua) provides 1-km fire pixels, confidence, and FRP, aggregated to monthly 0.1°. Two confidence streams are run in parallel (AC vs HC >30%).Bottom-up pathway (BU):Build a dynamic annual AGB series by scaling baseline AGB (either GlobBiomass or Biomass_cci) with interannual changes in MODIS NPP (MYD17A3HGF).Compute annual burned dry-matter density (BD) allowing multiple fires per year via sequential consumption with land-cover-specific burning efficiency (BE).Estimate annual emissions as BA × BD × EF, where burned area (BA) is derived from the active-fire mask times pixel area.Downscale to months using the monthly share of cumulative FRE from observed FRP.Inputs: MCD12Q1 land cover (IGBP) to assign BE (Mieville/Shi) and EF by GFED region × land-cover class.Top-down pathway (TD):Reconstruct daily FRE from instantaneous FRP by fitting a Gaussian diurnal cycle (parameters inferred from long-term Terra:Aqua FRP ratios), correct Terra FRP magnitude relative to Aqua, and average Aqua- and Terra-based FRE. Convert FRE to combusted dry matter using biome-specific conversion factors (CR) from either the 30-class or 8-class map, then apply EF and aggregate to monthly totals.Harmonization & outputs: All members share the same spatial (0.1°) and temporal (monthly) grid; emissions are provided alongside the ensemble mean, min, and max to bracket uncertainty.This structure allows users to select a member consistent with their assumptions (e.g., conservative HC filtering, preferred AGB or CR scheme) or to propagate the full ensemble spread in downstream atmospheric or biogeochemical applications.