A map of global peatland extent created using machine learning (Peat-ML)

Abstract

Abstract. Peatlands store large amounts of soil carbon and freshwater, constituting an important component of the global carbon and hydrologic cycles. Accurate information on the global extent and distribution of peatlands is presently lacking but is needed by Earth System Models (ESMs) to simulate the effects of climate change on the global carbon and hydrologic balance. Here, we present Peat-ML, a spatially continuous global map of peatland fractional coverage generated using machine learning techniques suitable for use as a prescribed geophysical field in an ESM. Inputs to our statistical model follow drivers of peatland formation and include spatially distributed climate, geomorphological and soil data, along with remotely-sensed vegetation indices. Available maps of peatland fractional coverage for 14 relatively extensive regions were used along with mapped ecoregions of non-peatland areas to train the statistical model. In addition to qualitative comparisons to other maps in the literature, we estimated model error in two ways. The first estimate used the training data in a blocked leave-one-out cross-validation strategy designed to minimize the influence of spatial autocorrelation. That approach yielded an average r2 of 0.73 with a root mean squared error and mean bias error of 9.11 % and −0.36 %, respectively. Our second error estimate was generated by comparing Peat-ML against a high-quality, extensively ground-truthed map generated by Ducks Unlimited Canada for the Canadian Boreal Plains region. This comparison suggests our map to be of comparable quality to mapping products generated through more traditional approaches, at least for boreal peatlands.

Modeling Carbon Accumulation and Permafrost Dynamics of Northern Peatlands Since the Holocene

“…Uncertainties in the peatland C stock partially arise from uncertainties in peatland coverage. Since PTEM 2.1 does not simulate peatland coverage, we use three different maps covering the pan‐Arctic region (Hugelius et al., 2020; Melton et al., 2022; Xu et al., 2018). All maps were aggregated into 0.5° × 0.5° grid cells with spatially explicit peatland abundance, and their average was used as a fourth map.…”

“…Although multiple process‐based models have simulated the dynamic peatland spatial extent with TOPMODEL, the uncertainty remains an issue (Qiu et al., 2019; Stocker et al., 2014). To quantify these uncertainties, three northern peatland coverage maps are selected (Hugelius et al., 2020; Melton et al., 2022; Xu et al., 2018), and the soil C stock is estimated based on different observation‐based data sets and the mean of these data sets, respectively.…”

A map of global peatland extent created using machine learning (Peat-ML)

“…Python code for the statistical modelling is available at https://doi.org/10.5281/zenodo.6345309 (Melton et al, 2022). A netCDF format version of the Peat-ML dataset is available at https://doi.org/10.5281/zenodo.5794336 (Melton et al, 2021).…”

Supplementary material to "Peatlands and their carbon dynamics in northern high latitudes from 1990 to 2300: A process-based biogeochemistry model analysis"

“…Comparison between the reference and calibrated wetland abundance (%) interpolated with TOPMODEL approach. The reference dataset is the average peatland abundance of three peatland maps (Xu et al, 2018;Hugelius et al, 2020;Melton et al, 2022), and calibration is conducted for IPSL-CM5A-LR and bcc-csm1-1 climate inputs, respectively. The grid cells with less than 1% wetlands are left blank.…”