Seasonal and interannual variations in global wetland area is a strong driver of fluctuations in global methane (CH4) emissions. Current maps of global wetland extent vary with wetland definition, causing substantial disagreement and large uncertainty in estimates of wetland methane emissions. To reconcile these differences for large-scale wetland CH4 modeling, we developed a global Wetland Area and Dynamics for Methane Modeling (WAD2M) dataset at ~25 km resolution at equator (0.25 arc-degree) at monthly time-step for 2000-2018. WAD2M combines a time series of surface inundation based on active and passive microwave remote sensing at coarse resolution (~25 km) with six static datasets that discriminate inland waters, agriculture, shoreline, and non-inundated wetlands. We exclude all permanent water bodies (e.g. lakes, ponds, rivers, and reservoirs), coastal wetlands (e.g., mangroves and seagrasses), and rice paddies to only represent spatiotemporal patterns of inundated and non-inundated vegetated wetlands. Globally, WAD2M estimates the long-term maximum wetland area at 13.0 million km2 (Mkm2), which can be separated into three categories: mean annual minimum of inundated and non-inundated wetlands at 3.5 Mkm2, seasonally inundated wetlands at 4.0 Mkm2 (mean annual maximum minus mean annual minimum), and intermittently inundated wetlands at 5.5 Mkm2 (long-term maximum minus mean annual maximum). WAD2M has good spatial agreements with independent wetland inventories for major wetland complexes, i.e., the Amazon Lowland Basin and West Siberian Lowlands, with high Cohen’s kappa coefficient of 0.54 and 0.70 respectively among multiple wetlands products. By evaluating the temporal variation of WAD2M against modeled prognostic inundation (i.e., TOPMODEL) and satellite observations of inundation and soil moisture, we show that it adequately represents interannual variation as well as the effect of El Niño-Southern Oscillation on global wetland extent. This wetland extent dataset will improve estimates of wetland CH4 fluxes for global-scale land surface modeling.

This dataset provides monthly globally gridded freshwater wetland methane emissions from 2001-2018 in nmol CH4 m-2 s-1, g C-CH4 m-2 d-1, and TgCH4 grid cell-1 month-1. The data were derived from a six-predictor random forest upscaling model (UpCH4) trained on 119 site-years of eddy covariance CH4 flux data from 43 freshwater wetland sites covering bog (8), fen (8), marsh (10), swamp (6), and wet tundra (11) wetland classes and distributed across Arctic-boreal (20), temperate (16), and (sub)tropical (7) climate zones. Weekly mean CH4 fluxes were computed from half-hourly FLUXNET-CH4 Version 1.0 fluxes. Each grid cell CH4 flux prediction was weighted by fractional grid cell wetland extent to estimate CH4 emissions using the primary global dataset of Wetland Area and Dynamics for Methane Modeling (WAD2M) product and an alternate Global Inundation Estimate from Multiple Satellites GIEMS version 2 global wetland map. Both WAD2M and GIEMS-2 maps were modified with several correction data layers to represent the monthly area covered by vegetated wetlands, excluding open water and coastal wetlands. The data products are: mean daily fluxes with no adjustment for wetland area (i.e., flux densities assuming hypothetical 100% wetland cover); mean daily fluxes adjusting for WAD2M or GIEMS-2 wetland area; and by-pixel monthly sum of freshwater wetland methane emissions adjusting for WAD2M or GIEMS-2 wetland area. The data are provided in NetCDF4 format.

This dataset presents monthly gridded methane emissions from subtropical freshwater wetlands across the Southeastern United States spanning from 1982 to 2010, measured in nmol m-2 s-1. Each grid cell's methane flux prediction was adjusted based on the fractional wetland extent within that cell, utilizing data from the National Wetland Inventory (NWI) and the Wetland Area and Dynamics for Methane Modeling (WAD2M) product. The dataset includes mean monthly fluxes with no adjustment for wetland area (i.e., fluxes assuming hypothetical 100% wetland cover), as well as mean monthly fluxes adjusted for wetland area based on NWI or WAD2M, along with their respective standard deviations of the daily emissions for each month. Data are provided in NetCDF4 format.

Lat: 25-40°N

Lon: 95-75°W

Period: 198201-201012 (for "CH4_Monthly_SEUS_unweighted.nc" and "CH4_Monthly_SEUS_NWI.nc") and 200001-201012 (for "CH4_Monthly_SEUS_WAD2M.nc")

Temporal resolution: Monthly

Spatial resolution: 0.0083° × 0.0083° (~1 km × 1 km)

Unit: nmol m-2 s-1

Fill value: -9999

This dataset (WetCH₄) contains methane (CH₄) emissions using three different wetland maps, their uncertainties, and underlying flux intensities from northern wetlands (>45° N). The dataset is a data-driven upscaling product using observations from northern eddy covariance CH₄ flux sites and random forest machine learning. WetCH₄ provides daily CH₄ fluxes of northern wetlands at 10-km resolution from 2016 to 2022 and can be used to study regional CH₄ budgets and wetland responses to climate change. The data products are provided in netCDF format files (.nc) with more details in the attributes of the files.

File list:

- fch4_nmol_m2_s_10km_intensity.nc.gz and fch4_nmol_m2_s_10km_uncertainty.nc.gz:

The underlying flux intensities and associated uncertainties.

- fch4_10km_emi_wad2m.nc.gz and fch4_10km_emi_uncertainty_wad2m.nc.gz:

Upscaled CH4 emissions and uncertainties using WAD2M monthly dynamic wetland map.

- fch4_10km_emi_giems2.nc.gz and fch4_10km_emi_uncertainty_giems2.nc.gz:

Upscaled CH4 emissions and uncertainties using GIEMS2 monthly dynamic wetland map.

- fch4_10km_emi_glwd.nc.gz and fch4_10km_emi_uncertainty_glwd.nc.gz:

Upscaled CH4 emissions and uncertainties using static GLWD v1 wetland map.

Time range: 2016-01-01 - 2022-12-31

Time steps: daily, 2557

Geographic extent: longitude 180W - 180E, latitude 45 - 90 N

This repository contains three datasets resulting from the reconstruction of global wetland loss over 1700-2020. The three datasets are listed here and described in more detail below:

A. National and subnational statistics of drained or converted areas

B. Regional wetland percentage loss estimates and geospatial polygons

C. Gridded reconstruction

The scripts used to process input data, model and calibrate the wetland loss reconstruction, and produce the figures are publicly available at https://github.com/etiennefluetchouinard/wetland-loss-reconstruction.

A. National and subnational statistics of drained or converted areas

This tabular database containing national and subnational statistics of wetland area drained and peat mass extracted. The database includes four land use types: cropland, forestry, peat extraction and wetland cultivation. These data are used as input to the mapped wetland loss reconstruction. Column descriptions of drainage_db_v10.csv:

unit: Scale of administrative unit ("national" or "subnational").

type: Land use type ("Cropland", "Forestry", "Peat Extraction" or "Wetland Cultivation")

iso_a3: 3-letter code of each country.

region: Name of subnational unit. Blank if data is national scale.

HASC_1: Hierarchical Administrative Subdivisions Codes for the subnational units. Blank if data is national scale.

year: Year of data.

drained_area_1000ha: Cumulative area drained by the year specified, in thousands of hectares.

drained_weight_1000tonsyr: Annual peat extraction rate for the year, in thousand tons per year.

peatland_only: Label indicating whether the drained area applies to all wetlands or peatlands specifically ("Peatland only" or blank).

Comment: Additional description from original data source, or unit conversion, or data corrections.

Source: Reference of data source and/or compilers.

B. Regional wetland percentage loss estimates and geospatial polygons

A shapefile of 151 polygons projected in WGS84. Columns description for polygon shapefile of the regional wetland loss percentage: regional_loss_poly.shp:

id: Numerical identifier.

name: Name of administrative unit, region or water feature the polygon area covers.

country: Name of country.

continent: Name of continent.

wet_categ: Broad category of wetlands included in the estimate (“Peatlands”, “Inland natural wetlands”, “Coastal natural wetlands”, “Unspecified natural type(s)” or “All wetlands”).

yr_start: Start year of the period over which wetland loss is estimated.

yr_end: End year of the period over which wetland loss is estimated.

area_mkm2: Surface area of the polygon, in million square kilometers (Mkm2).

perc_loss: Numerical value of percentage wetland loss (positive value represent loss of wetland area between start and end year.

comment: Additional description of estimate used or estimation method.

source: Citation of original data source.

compiler: Citation of intermediary data compiler.

C. Gridded reconstruction

Gridded outputs are stored in a separate NetCDF file for each of the 12 reconstructions of simulated wetland and present-day wetland maps. An ensemble average was also computed from the 12 reconstructions (only individual reconstructions were discussed in the manuscript). These data consist of global maps generated from the drainage reconstruction methodology for 33 decadal intervals (1700-2020 inclusive) for 9 variables:

The filenames of ensemble members are labelled to with the name of the input present-day and simulated wetland maps:

    “wetland_loss_1700-2020_” + simulated input + “_” + present-day input + “_v10.nc”

The 4 simulated wetland map inputs are: LPJwsl, SDGVM, ORCHIDEE, DLEM. The 3 present-day wetland map inputs are: GIEMSv2, GLWD3, WAD2M.

Description of the 9 variables in each NCDF file:

wetland_loss: Cumulative wetland area lost (km2 per grid cell). This variable is equivalent to the sum of area drained for the seven land uses drained

nat_wetland: Remaining natural wetland area (km2 per grid cell)

cropland: Cropland area drained (km2 per grid cell) leading to wetland loss

forestry: Forestry area drained (km2 per grid cell) leading to wetland loss

peatextr: Peat harvest area drained (km2 per grid cell) leading to wetland loss

wetcultiv: Wetland cultivation area (km2 per grid cell) leading to wetland loss

ir_rice: Irrigated rice area leading to wetland loss (km2 per grid cell)

pasture: Pasture area drained leading to wetland loss (km2 per grid cell)

urban: Urban area drained leading to wetland loss (km2 per grid cell)

All layers were capped below the land pixel area grid (from HYDE 3.2, excl. open water).

Time: 33 slices; numerical years spread at decadal intervals, ranging between 1700-2020 (inclusive)

Extent: Longitude: -180° to 180°. Latitude: -56° to 84°.

See the README file for a more detailed description of this dataset. Anyone wishing to use this dataset should cite Fluet-Chouinard et al. 2023. Please contact Etienne Fluet-Chouinard at etienne.fluet@gmail.com with any questions or comments with regards to the best usage of our dataset.

Fluet-Chouinard E., Stocker B., Zhang Z., Malhotra A., Melton J.R., Poulter B., Kaplan J., Goldewijk K.K., Siebert S., Minayeva T., Hugelius G., Prigent C., Aires F., Hoyt A., Davidson N., Finlayson C.M., Lehner B., Jackson R.B., McIntyre P.B. Nature. Extensive global wetland loss over the last three centuries