Soil Organic Carbon Distributions in Tidal Wetlands of the Northeastern USA

This dataset provides estimates of soil organic carbon (SOC) in tidal wetlands for the northeastern United States. The data cover the period 1998-2018. Northeastern U.S. tidal wetlands and bordering areas were harmonized from government agencies [U.S. Department of Agriculture - Natural Resources Conservation Service (USDA-NRCS), National Cooperative Soil Survey (NCSS), USDA-NRCS - Rapid Carbon Assessment (RaCA), U.S. Environmental Protection Agency - National Wetland Condition and Assessment (EPA-NWCA)] and published studies. Point data for carbon stocks (in kg m-2) at four soil depths (0-5, 0-30, 0-100, and 0-200 cm) are included. SOC for the four depths was predicted for eight regional zones using regression models driven by environmental covariates. Two methods were used to estimate parameters for these models, a Random Forest (RF) Ranger method and a Quantile Regression Forest (QRF) model. The distribution of SOC was predicted for tidal wetland cover types mapped by Correll et al. (2019). Predictions and uncertainties are available at a 3 m resolution.

There are 100 data files included in this dataset: 96 in GeoTIFF (*.tif) format and four in comma-separated values (CSV; *.csv) format. Also included is one companion file in Portable Document (*.pdf) format.

This dataset provides estimates of soil organic carbon (SOC) in tidal wetlands for the northeastern United States. The data cover the period 1998–2018. Northeastern U.S. tidal wetlands and bordering areas were harmonized from government agencies [U.S. Department of Agriculture - Natural Resources Conservation Service (USDA-NRCS), National Cooperative Soil Survey (NCSS), USDA-NRCS - Rapid Carbon Assessment (RaCA), U.S. Environmental Protection Agency - National Wetland Condition and Assessment (EPA-NWCA)] and published studies. Point data for carbon stocks (in kg m-2) at four soil depths (0–5, 0–30, 0–100, and 0–200 cm) are included. SOC for the four depths was predicted for eight regional zones using regression models driven by environmental covariates. Two methods were used to estimate parameters for these models, a Random Forest (RF) Ranger method and a Quantile Regression Forest (QRF) model. The distribution of SOC was predicted for tidal wetland cover types mapped by Correll et al. (2019). Predictions and uncertainties are available at a 3 m resolution.

Project: North American Carbon Program

The North American Carbon Program (NACP) is a multidisciplinary research program designed to improve understanding of North America's carbon sources, sinks, and stocks. The central objective is to measure and understand the sources and sinks of Carbon Dioxide (CO2), Methane (CH4), and Carbon Monoxide (CO) in North America and adjacent oceans. The NACP is supported by a number of different federal agencies.

Related Publications

Correll, M. D., W. Hantson, T.P. Hodgman, B.B. Cline, C.S. Elphick, W.G. Shriver, E.L. Tymkiw, and B.J. Olsen. 2019. Fine-scale mapping of coastal plant communities in the northeastern USA. Wetlands 39:17–28. https://doi.org/10.1007/s13157-018-1028-3

Holmquist, J. R., L. Windham-Myers, N. Bliss, S. Crooks, J.T. Morris, J.P. Megonigal,  T. Troxler, D. Weller, J. Callaway, J. Drexler, M. C. Ferner, M. E. Gonneea, K. D. Kroeger, L. Schile-Beers, I. Woo, K. Buffington, J. Breithaupt, B.M. Boyd, L.N. Brown, N. Dix, L. Hice, B.P. Horton, G.M. MacDonald, R.P. Moyer, W. Reay, T. Shaw, E. Smith, J.M. Smoak, C. Sommerfield, K. Thorne, D. Velinsky, El. Watson, K.W. Grimes, and M. Woodrey. 2018. Accuracy and precision of tidal wetland soil carbon mapping in the conterminous United States. Scientific Reports 8:9478. https://doi.org/10.1038/s41598-018-26948-7

Related Datasets

Hinson, A.L., R.A. Feagin, and M. Eriksson. 2019. Tidal Wetlands Soil Organic Carbon and Estuarine Characteristics, USA, 1972-2015. ORNL DAAC, Oak Ridge, Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/1742

Holmquist, J.R., L. Windham-Myers, N. Bliss, S. Crooks, J.T. Morris, P.J. Megonigal, T. Troxler, D. Weller, J. Callaway, J. Drexler, M.C. Ferner, M.E. Gonneea, K. Kroeger, L. Schile-beers, I. Woo, K. Buffington, B.M. Boyd, J. Breithaupt, L.N. Brown, N. Dix, L. Hice, B.P. Horton, G.M. Macdonald, R.P. Moyer, W. Reay, T. Shaw, E. Smith, J.M. Smoak, C. Sommerfield, K. Thorne, D. Velinsky, E. Watson, K. Grimes, and M. Woodrey. 2019. Tidal Wetland Soil Carbon Stocks for the Conterminous United States, 2006-2010. ORNL DAAC, Oak Ridge, Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/1612

Acknowledgments

This project received support from NASA’s North American Carbon Project and a National Science Foundation (NSF) CAREER Award (grant 1652594).

Spatial Coverage: Tidal wetlands of the northeastern U.S., from Virginia to Maine

Spatial Resolution: 3 m

Temporal Coverage: 1998-01-01 to 2018-01-01

Temporal Resolution: One-time estimate

Study Area:  Latitude and longitude are given in decimal degrees.

Site	Westernmost Longitude	Easternmost Longitude	Northernmost Latitude	Southernmost Latitude
Northeastern U.S.	-76.348	-66.945	45.259	37.08

Data File Information

There are 100 data files included in this dataset: 96 in GeoTIFF (*.tif) format and four in comma-separated values (CSV; *.csv) format. The GeoTIFF files contain spatial predictions of soil organic carbon (SOC) in tidal wetlands for eight regional zones between Virginia to Maine. The CSV files hold point-level measurements of SOC and there is one file for each soil depth stratum (0–5, 0–30, 0–100, 0–200 cm). Also included is one companion file in Portable Document (*.pdf) format (Supplementary_Materials.pdf) that provides details of the geographic zones, covariates from Correll et al. (2019), sources for point data, and R code for generating the GeoTIFF files from point data.

The naming convention for the GeoTiff files is SOC_AAAcm_ZoneB_CC_D.tif, where

AAA = maximum soil depth of the measurement ("5cm", "30cm", "100cm", or "200cm"),
B = geographic zone (1–9),
CC = the method ("RF" or "QRF"), and
D = the statistic ("Mean" or "Uncertainty").

The naming convention for the CSV files is SOC_AAAcm_NE-US_2018.csv, where AAA is maximum depth of the measurement ("5cm", "30cm", "100cm", or "200cm"). The number of data points are 373, 371, 169, and 49 for the maximum depths. Variables are listed in Table 2 and data sources are in Table 3. Point locations in the CSVs are accurate to within +/-500 m.

Table 1. Naming conventions and descriptions for three types of GeoTIFF files. There are 4 depths and 8 zones, creating 32 files, for each type of GeoTIFF.

File Type	File Naming Convention	Units	Description
Random Forest Mean	SOC_AAAcm_ZoneB_RF_Mean.tif	kg m-2	Mean estimates of SOC generated with Random Forest (RF) Ranger method using eleven environmental covariables.
Quantile Regression Mean	SOC_AAAcm_ZoneB_QRF_Mean.tif	kg m-2	Mean estimates of SOC generated with Quantile Regression Forest (QRF) method using eleven environmental covariables.
Quantile Regression Uncertainty	SOC_AAAcm_ZoneB_QRF_Uncertainty.tif	kg m-2	Uncertainty in estimates of SOC generated with QRF method using eleven environmental covariables.

Data File Details

The spatial resolution of the GeoTIFFs is 3 meters and the no_data value is -9999.

Point locations in the CSVs are accurate to within +/-500 m.

The coordinate system for all files is UTM zone 18N, datum NAD83 (EPSG:26918).

Table 2. Variables in the CSV files containing point-level data on soil organic carbon.

Variable	Units	Description
ID	string	Unique label for point based on coordinate or data source
X	m	UTM easting, zone 18
Y	m	UTM northing, zone 18
OCSKGM	kg m-2	Soil organic carbon content
Source	string	Abbreviation for data source as defined in Table 3

Tidal wetlands contain relatively high quantities of soil organic carbon. The estuarine-level analysis could provide a better understanding of the conditions that lead to enhanced or degraded carbon sequestration rates in times of rapid global change and could lead to future conservation efforts within specific estuarine boundaries.

This dataset provides information on the spatial distribution of these carbon stocks in tidal regions of the eastern US from the Delmarva Peninsula (Virginia and Maryland) to Maine.

Point-level data were evaluated to remove outliers and duplicate data. Prior to generating the prediction rasters, a mask was developed to limit the analysis to tidal areas while excluding uplands and non-tidal wetlands. Uncertainty estimates were generated with the Quantile Regression Forest method and are provided in GeoTIFF files of this dataset.

Known Issues

There are lines of missing data in the GeoTIFFs as a result of processing data in 2 km x 2 km tiles. These "no data" lines are ≤6 m wide at the boundaries of tiles.

This project compiled data on soil organic carbon (SOC) content of tidal wetlands of the northeastern United States from publicly available sources and published articles (Table 3). These data were harmonized to create a point-level dataset with SOC stocks for four soil depth increments: 0–5 cm, 0–30 cm, 0–100 cm, and 0–200 cm. Data from the past 20 years (1998–2018) were utilized for the three upper depths. The 0–200 cm dataset included all available data without date restrictions.

The region of interest included coastal areas from the Delmarva Peninsula of Virginia and Maryland, northward to Maine. This region was divided into eight zones as used by Correll et al. (2019). These zones differ with respect to ecological structure, climate, geological parent material, and in some cases, salinity gradients.

Point data were screened to remove outliers. Outliers were calculated for organic, sandy, and fine-textured soils. Ranges in values were generated from known values in fundamental soil science literature (Brady and Wiel, 2008; Buol et al., 2003) of bulk densities for the varying soils and soil materials because bulk density determines the maximum values possible for SOC in a given soil sample.

The point-level data were combined with environmental covariates to predict the distributions of SOC (in kg m^-2) in tidal wetlands across the region of interest. Covariates, from Correll et al. (2019), included terrain metrics calculated from DEMs and vegetation cover types derived from remotely sensed Difference Vegetation Index (DVI), Normalized Difference Vegetation Index (NDVI), and Normalized Difference Water Index (NDWI) metrics. Spatially explicit predictions of SOC were calculated by regression with parameters estimated using two random forest methods: Random Forest Ranger (RF) (Wright and Ziegler, 2017) and Quantile Regression Forest (QRF) (Meinshausen, 2006). Both methods used the same model structure to ensure the comparability of their results. Uncertainty was calculated using the standard deviation of QRF results. 

The data were processed in 2 km x 2 km tiles, and predictions were not possible at pixels at the edge of many tiles. This limitation resulted in lines of "no data" up to 2 pixels (6 m) wide at the boundaries of tiles.

Table 3. Data sources for soil organic carbon measurements in the CSV files.

Abbreviation	Source	Files
ISCN	International Soil Carbon Network Dataset: Generation 3 Database 12-2015 https://iscn.fluxdata.org
WoSIS	World Soil Information Service https://www.isric.org/explore/wosis/accessing-wosis-derived-datasets	wosis_201607_attributes.txt wosis_201607_profiles.txt wosis_201607_layers.txt
CCRCN	Coastal Carbon Research Coordination Network https://serc.si.edu/coastalcarbon/data, https://ccrcn.shinyapps.io/CoastalCarbonAtlas	CCRCB_core_data.csv CCRCN_depthseries_data.csv
EPA-NWCA	U.S. Environmental Protection Agency-National Wetlands Condition Assessment https://www.epa.gov/national-aquatic-resource-surveys/data-national-aquatic-resource-surveys	nwca2011_siteinfo.csv nwca2011_soilchem.csv
RaCA	Rapid Carbon Assessment https://www.nrcs.usda.gov/wps/portal/nrcs/detail/soils/survey/?cid=nrcs142p2_054164	RaCA_data_summary.xlsx
NCSS	National Cooperative Soil Survey https://ncsslabdatamart.sc.egov.usda.gov	NCSS_Lab_Data_Mart_01512017.mbd (Geochemistry_query, Bulk_Density_and_Moisture)
Holmquist et al.	Holmquist et al. 2018. Accuracy and precision of tidal wetlands soil carbon mapping in the conterminous United States. Scientific Reports. https://doi.org/10.1038/s41598-018-26948-7, https://repository.si.edu/handle/10088/35684

Brady, N.C. and R.R. Wiel. 2008. The Nature and Properties of Soils (14th ed.). Pearson Prentice Hall.

Buol, S.W., R.J. Southard, R.C. Graham, and P. A. McDaniel. 2003. Soil Genesis and Classification (5th ed.). Iowa State Press; Ames, Iowa.

Correll, M.D., W. Hantson, T.P. Hodgman, B.B. Cline, C.S. Elphick, W.G. Shriver, E.L. Tymkiw, and B.J. Olsen. 2019. Fine-scale mapping of coastal plant communities in the northeastern USA. Wetlands 39:17–28. https://doi.org/10.1007/s13157-018-1028-3

Holmquist, J.R., L. Windham-Myers, N. Bliss, S. Crooks, J.T. Morris, J.P. Megonigal,  T. Troxler, D. Weller, J. Callaway, J. Drexler, M. C. Ferner, M. E. Gonneea, K. D. Kroeger, L. Schile-Beers, I. Woo, K. Buffington, J. Breithaupt, B.M. Boyd, L.N. Brown, N. Dix, L. Hice, B.P. Horton, G.M. MacDonald, R.P. Moyer, W. Reay, T. Shaw, E. Smith, J.M. Smoak, C. Sommerfield, K. Thorne, D. Velinsky, El. Watson, K.W. Grimes and M. Woodrey. 2018. Accuracy and precision of tidal wetland soil carbon mapping in the conterminous United States. Scientific Reports 8:9478. https://doi.org/10.1038/s41598-018-26948-7

Meinshausen, N. 2006. Quantile regression forests. J. Machine Learning Research 7:983-999. https://www.jmlr.org/papers/v7/meinshausen06a.html

Wright, M. N. and A. Ziegler, A. 2017. Ranger: A fast implementation of random forests for high dimensional data in C++ and R. J. Statistical Software 77:1-17. https://doi.org/10.18637/jss.v077.i01