An official website of the United States government
Peatlands cover 3% of the global land surface, yet store 25% of the world’s soil organic carbon. These organic-rich soils are widespread across permafrost regions, representing nearly 18% of land surface and storing between 500 and 600 petagrams of carbon (PgC). Peat (i.e., partially decomposed thick organic layers) accumulates due to the imbalance between plant production and decomposition often within saturated, nutrient deficient, and acidic soils, which limit decomposition. As warmer and drier conditions become more prevalent across northern ecosystems, the vulnerability of peatland soils may increase with the susceptibility of peat-fire ignitions, yet the distribution of peatlands across Alaska remains uncertain. Here we develop a new high-resolution (20 meter (m) resolution) wall-to-wall ~1.5 million square kilometer (km2) peatland map of Alaska, using a combination of Sentinel-1 (Dual-polarized Synthetic Aperture Radar), Sentinel-2 (Multi-Spectral Imager), and derivatives from the Arctic Digital Elevation Model (ArcticDEM). Machine learning classifiers were trained and tested using peat cores, ground observations, and sub-meter resolution image interpretation, which was spatially constrained by a peatland suitability model that described the extent of terrain suitable for peat accumulation. This product identifies peatlands in Polar, Boreal, and Maritime ecoregions in Alaska to cover 26,842 (4.6%), 69,783 (10.4%), and 13,506 (5.3%) km2, respectively.
more »
« less
A map of global peatland extent created using machine learning (Peat-ML)
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 (ML) 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, and 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-square 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.
more »
« less
- Award ID(s):
- 1355066
- PAR ID:
- 10504792
- Publisher / Repository:
- European Geosciences Union
- Date Published:
- Journal Name:
- Geoscientific Model Development
- Volume:
- 15
- Issue:
- 12
- ISSN:
- 1991-9603
- Page Range / eLocation ID:
- 4709 to 4738
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
The extent and distribution of tropical peatlands, and their importance as a vulnerable carbon (C) store, remain poorly quantified. Although large peatland complexes in Peru, the Congo basin, and Southeast Asia have been mapped in detail, information on many other tropical areas is uncertain. In the Eastern Colombian lowlands, peatland area estimates range from 700 km2 to nearly 60,000 km2, leading to highly uncertain C stocks. Using new field data, high‐resolution Earth observation (EO), and a random forest approach, we mapped peatlands across Colombian territory East of the Andes below 400 m elevation. We estimated peatland extent using two approaches: a conservative method focused on medium‐to‐high peat probability areas and a more inclusive one accounting for large low‐probability areas. Multiplying these extents by below‐ground carbon density yields a conservative estimate of 0.95 (0.6–1.39 Pg C, 95% confidence interval) over 9,391 km2(7,369–11,549 km2) and up to 2.86 Pg C (1.76–4.22 Pg C) across 29,069 km2 (22,429–36,238 km2). Among four potentially peat‐forming ecosystems identified, palm swamps and floodplain forests contributed most to the peat extent and C stock. We found that most peatland patches were relatively small, covering less than 100 ha. We compared our map to previously published global and pan‐tropical peat maps and found low spatial overlap among them, suggesting that peat maps uninformed by local field information may not precisely specify which landscape areas within a peatland‐rich region are actually peatlands. We further assessed the suitability of different EO and climate variables, highlighting the need for high‐resolution data to capture local heterogeneities in the landscape.more » « less
-
Abstract Northern peatlands are a large C stock and often act as a C sink, but are susceptible to climate warming. To understand the role of peatlands in the global carbon‐climate feedback, it is necessary to accurately quantify their C stock changes and decomposition. In this study, a process‐based model, the Peatland Terrestrial Ecosystem Model, is used to simulate pan‐Arctic peatland C dynamics from 15 ka BP to 1990. To improve the accuracy of the simulation, spatially explicit water run‐on and run‐off processes were considered, four different pan‐Arctic peatland extent data sets were used, and a spatially explicit peat basal date data set was developed using a neural network approach. The model was calibrated against 2055 peat thickness observations and the parameters were interpolated to the pan‐Arctic region. Using the model, we estimate that, in 1990, the pan‐Arctic peatlands soil C stock was 396–421 Pg C, and the Holocene average C accumulation rate was 22.9 g C·m−2 yr−1. Our estimated peat permafrost development history generally agrees with multi‐proxy‐based paleo‐climate data sets and core‐derived permafrost areal dynamics. Under Anthropocene warming, in the freeze‐thaw and permafrost‐free regions, the peat C accumulation rate decreased, but it increased in permafrost regions. Our study suggests that if current permafrost regions switch to permafrost‐free conditions in a warming future, the peat C accumulation rate of the entire pan‐Arctic region will decrease, but the sink and source activities of these peatlands are still uncertain.more » « less
-
Abstract. Peatlands have often been neglected in Earth system models (ESMs).Where they are included, they are usually represented via a separate, prescribed grid cell fraction that is given the physical characteristics of a peat (highly organic) soil. However, in reality soils vary on a spectrum between purely mineral soil (no organic material) and purely organicsoil, typically with an organic layer of variable thickness overlying mineral soil below. They are also dynamic, with organic layer thickness and its properties changing over time. Neither the spectrumof soil types nor their dynamic nature can be captured by current ESMs. Here we present a new version of an ESM land surface scheme (Joint UK Land Environment Simulator, JULES) where soil organic matter accumulation – and thus peatland formation, degradation and stability – is integratedin the vertically resolved soil carbon scheme. We also introduce the capacity to track soil carbon age as a function of depth in JULES and compare this to measured peat age–depth profiles. The new scheme is tested and evaluated at northern and temperate sites. This scheme simulates dynamic feedbacks between the soil organic material and its thermal and hydraulic characteristics. We show that draining the peatlands can lead to significant carbon loss, soil compaction and changes in peat properties. However, negative feedbacks can lead to the potential for peatlands to rewet themselves following drainage.These ecohydrological feedbacks can also lead to peatlands maintaining themselves in climates where peat formation would not otherwise initiate in the model, i.e. displaying some degree of resilience. The new model produces similar results to the original model for mineral soils and realistic profiles of soil organic carbon for peatlands.We evaluate the model against typical peat profiles based on 216 northern and temperate sites from a global dataset of peat cores.The root-mean-squared error (RMSE) in the soil carbon profile is reduced by 35 %–80 % in the best-performing JULES-Peat simulationscompared with the standard JULES configuration. The RMSE in these JULES-Peat simulations is 7.7–16.7 kg C m−3 depending on climate zone, which is considerably smaller than the soil carbon itself (around 30–60 kg C m−3). The RMSE at mineral soil sites is also reducedin JULES-Peat compared with the original JULES configuration (reduced by ∼ 30 %–50 %). Thus, JULES-Peat can be used as a complete scheme that simulates both organic and mineral soils. It does not requireany additional input data and introduces minimal additional variables to the model. This provides a new approach for improving the simulation of organic and peatland soils andassociated carbon-cycle feedbacks in ESMs.more » « less
-
Northern peatlands are unique ecosystems typically located in boreal latitudes that sustain unique habitats and species. They are a critical component of the global carbon cycle, acting both as a carbon reservoir (via peat accumulation after sequestration of carbon dioxide from the atmosphere) and as a carbon source to the atmosphere (by releasing methane and carbon dioxide, two greenhouse gases produced by microbes that thrive in the conditions found in peatlands). Although the ecology of peatlands has been well studied for many decades, the geological controls on peatland development and groundwater flow patterns are not completely understood. In Maine (USA), peatlands began forming about 10,000 years ago following the retreat of the ice sheets at the end of the last ice age. They formed in depressions, often starting as lakes or wetlands, within the landscape carved by glaciers and draped with sediments. Almost two decades of research in several peatlands in Maine suggest that glacial landforms buried beneath peatlands may play a key role in regulating both the hydrology (including the presence of open water pools at the surface) and release of methane gasses from peatlands. Subsurface images using an array of hydrogeophysical methods (including ground-penetrating radar, GPR) constrained with direct coring has revealed the presence of buried esker complexes beneath (or close to) surface pools. Hydrological measurements further suggest that the presence of these permeable esker deposits (mainly gravel and sand) may enhance the connection of peatland water to underlying groundwater and help sustain these pools. Whereas geological maps show that the presence of esker systems in Maine are widespread, satellite-derived digital elevation models (DEMs) reveal that they are often proximal to peatland boundaries, further suggesting that they may commonly extend below the peat formation and control hydrology and carbon dynamics in more peatlands than previously thought. Better understanding of how the critical zone influences coupled water and carbon cycling in northern peatlands may improve the understanding of their contribution to radiative forcing of climate as the climate warms.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.5194/gmd-15-4709-2022
