Environmental Data Initiative logo
EDI Staging Environment

Data Package Metadata   View Summary

Coastal landcover change and the associated biomass trends in the mid-Atlantic sea-level rise hotspot

General Information
Data Package:
Local Identifier:edi.1196.1
Title:Coastal landcover change and the associated biomass trends in the mid-Atlantic sea-level rise hotspot
Alternate Identifier:DOI PLACE HOLDER
Abstract:
Climate change is driving worldwide landscape reorganization. In the coastal ecosystem, climate-driven sea level rise is forcing landward marsh migration and forest die-off, with potentially large consequences on coastal carbon balance. Here we used 30 m resolution Landsat images to study coastal landcover change from 1984 to 2020, and analyzed the Normalized Difference Vegetation Index (NDVI, a proxy of plant biomass) trend between 1984 and 2020 in the mid-Atlantic sea level rise hotspot. Our study region stretches across the entire Chesapeake Bay and the Delaware Bay to encompass all areas between 0-5 m above sea level (total area ~12,500 km2). Specifically, the data package includes 3 raster datasets derived from the Landsat images. All datasets cover the identical mid-Atlantic region and have identical spatial resolution of 30 m. The two landcover datasets, named as 'Landcover_year1984.tif' and 'Landcover_year2020.tif', respectively refer to landcover map in 1984 and 2020. Each of the maps has 7 landcover classes differentiated by different integers, and they are: water (0), farmland (1), urban area (2), upland forest (3), transition forest (4), marsh (5) and sandbar (6). Both landcover maps were generated using a combination of random forest classification and manual delineation, and the results were validated with high-resolution aerial photos and satellite images with an overall mapping accuracy beyond 90%. The third raster dataset, named as 'NDVItrend_1984to2020.tif', is the NDVI trend map. The value of each 30 by 30 m pixel in the map represents the slope of the NDVI trendline estimated using annual peak-growing season NDVI images acquired between 1984 and 2020. Negative values in the dataset represent decreases of NDVI (i.e. biomass loss, or ecosystem browning) from 1984 and 2020, whereas positive values correspond to an increase of NDVI (i.e. biomass gain, or ecosystem greening) between 1984 and 2020. The data package is completed.
Publication Date:2022-08-16
For more information:
Visit: DOI PLACE HOLDER

Time Period
Begin:
1984
End:
2020

People and Organizations
Contact:Chen, Yaping (Virginia Institute of Marine Science) [  email ]
Contact:Kirwan, Matthew (Virginia Institute of Marine Science) [  email ]
Creator:Chen, Yaping (Virginia Institute of Marine Science, Postdoctoral Research Associate)
Creator:Kirwan, Matthew (Virginia Institute of Marine Science, Associate Professor)
Associate:Messerschmidt, Tyler (Virginia Institute of Marine Science, Field technician)
Associate:Smith, Alex (Virginia Institute of Marine Science, Graduate Student)

Data Entities
Other Name:
Landcover_year1984
Description:
Landcover in 1984. The landcover map has 7 classes differentiated by different integers, and they are: water (0), farmland (1), urban area (2), upland forest (3), transition forest (4), marsh (5) and sandbar (6).
Other Name:
Landcover_year2020
Description:
Landcover in 2020. The landcover map has 7 classes differentiated by different integers, and they are: water (0), farmland (1), urban area (2), upland forest (3), transition forest (4), marsh (5) and sandbar (6).
Other Name:
NDVItrend_1984to2020
Description:
Map of NDVI trend between 1984 and 2020. The value of each pixel represents the slope of the NDVI trend between 1984 and 2020, which can also be described as change of NDVI per year. Negative values represent decreases of NDVI (i.e. ecosystem browning) from 1984 and 2020, and positive values correspond to an increase of NDVI (i.e. ecosystem greening) from1984 to 2020.
Detailed Metadata

Data Entities

Non-Categorized Data Resource

Name:Landcover_year1984
Entity Type:raster dataset
Description:Landcover in 1984. The landcover map has 7 classes differentiated by different integers, and they are: water (0), farmland (1), urban area (2), upland forest (3), transition forest (4), marsh (5) and sandbar (6).
Physical Structure Description:
Object Name:Landcover_year1984.tif
Size:254534103 byte
Authentication:24a8939b44e9419371eda30a38858d7b Calculated By MD5
Externally Defined Format:
Format Name:tif
Data:https://pasta-s.lternet.edu/package/data/eml/edi/1196/1/0bfb499eac0e8e47cef803de498ef9de

Non-Categorized Data Resource

Name:Landcover_year2020
Entity Type:raster dataset
Description:Landcover in 2020. The landcover map has 7 classes differentiated by different integers, and they are: water (0), farmland (1), urban area (2), upland forest (3), transition forest (4), marsh (5) and sandbar (6).
Physical Structure Description:
Object Name:Landcover_year2020.tif
Size:254534103 byte
Authentication:69f1c8cc7aacb42b05de27c8100cbeb6 Calculated By MD5
Externally Defined Format:
Format Name:tif
Data:https://pasta-s.lternet.edu/package/data/eml/edi/1196/1/e205538465bde9725bdf289da8dc7dfb

Non-Categorized Data Resource

Name:NDVItrend_1984to2020
Entity Type:raster dataset
Description:Map of NDVI trend between 1984 and 2020. The value of each pixel represents the slope of the NDVI trend between 1984 and 2020, which can also be described as change of NDVI per year. Negative values represent decreases of NDVI (i.e. ecosystem browning) from 1984 and 2020, and positive values correspond to an increase of NDVI (i.e. ecosystem greening) from1984 to 2020.
Physical Structure Description:
Object Name:NDVItrend_1984to2020.tif
Size:508813983 byte
Authentication:6029af8b490f6e35404af5cb1225b690 Calculated By MD5
Externally Defined Format:
Format Name:tif
Data:https://pasta-s.lternet.edu/package/data/eml/edi/1196/1/9a62251ffc3f9484039c1aeac689ecf5

Data Package Usage Rights

This information is released under the Creative Commons license - Attribution - CC BY (https://creativecommons.org/licenses/by/4.0/). The consumer of these data ("Data User" herein) is required to cite it appropriately in any publication that results from its use. The Data User should realize that these data may be actively used by others for ongoing research and that coordination may be necessary to prevent duplicate publication. The Data User is urged to contact the authors of these data if any questions about methodology or results occur. Where appropriate, the Data User is encouraged to consider collaboration or co-authorship with the authors. The Data User should realize that misinterpretation of data may occur if used out of context of the original study. While substantial efforts are made to ensure the accuracy of data and associated documentation, complete accuracy of data sets cannot be guaranteed. All data are made available "as is." The Data User should be aware, however, that data are updated periodically and it is the responsibility of the Data User to check for new versions of the data. The data authors and the repository where these data were obtained shall not be liable for damages resulting from any use or misinterpretation of the data. Thank you.

Keywords

By Thesaurus:
LTER Controlled Vocabularycarbon cycling, seawater, landscape change, climate change, salt marshes, forest dynamics

Methods and Protocols

These methods, instrumentation and/or protocols apply to all data in this dataset:

Methods and protocols used in the collection of this data package
Description:
The data package includes 3 raster datasets: two landcover maps ('Landcover_year1984.tif' and 'Landcover_year2020.tif') and one NDVI trend map (''NDVItrend_1984to2020.tif''). All datasets were derived from the 30 m resolution Landsat images, and cover the mid-Atlantic region between 0-5 m above sea level at the spatial resolution of 30 m. Both landcover maps (1984 and 2020) have seven classes (water-0; farmland-1; urban area-2; upland forest-3; transition forest-4; marsh-5; and sandbar-6), created using the random forest algorithm implemented in R (v 4.1.1). For each mapping, we acquired Landsat images during multiple seasons corresponding to distinct plant phenological phases: the leaf-out season, the peak growing season, and the leaf-off season. For each map, we randomly sampled ~5,000 sites for every landcover type across the study region with a minimum between-site distance of 1 km according to high-resolution images and published delineation. The sites were then randomly divided into training and validation groups in the ratio of 60% to 40%. We ran the random forest classifier on all training sites and selected the best performing models to generate the final landcover maps. For areas that were masked from auto-classification (e.g. pixels contaminated by cloud and cloud shadow, ~5% of all area), they were manually classified by digitizing high-resolution images in ArcGIS (v10.7). The final landcover maps in 1984 and 2020 were validated across the region, which attained an overall classification accuracy of 91.0% and 93.1%, respectively. To generate the NDVI trend map (''NDVItrend_1984to2020.tif''), we first stacked all peak growing season NDVI images acquired annually from Landsat 5, 7 and 8 between 1984 and 2020 in time series. The NDVI is a remote sensing index closely correlated with plant biomass. NDVI value is unitless, ranging between -1 and 1. To determine the NDVI trend for each pixel from 1984 to 2020, we performed trend analysis for every 30 by 30 m pixel using the non-parametric Theil-Sen slope estimator in R (v 4.1.1). The Theil-Sen slope estimator was selected as it is the most widely used approach for quantifying monotonic trends with the advantage of being insensitive to outliers. The value of each pixel thus represents the slope of the NDVI trend between 1984 and 2020, which can also be described as change of NDVI per year. Therefore, negative values represent decreases of NDVI (i.e. biomass loss, or ecosystem browning) from 1984 and 2020, and positive values correspond to an increase of NDVI (i.e. biomass gain, or ecosystem greening) from1984 to 2020.

People and Organizations

Publishers:
Organization:Environmental Data Initiative
Email Address:
info@edirepository.org
Web Address:
https://edirepository.org
Id:https://ror.org/0330j0z60
Creators:
Individual: Yaping Chen
Organization:Virginia Institute of Marine Science
Position:Postdoctoral Research Associate
Address:
2385 Jacqueline Dr,
Apt. 501C,
Hayes, VA 23072 United States
Phone:
2178192135 (voice)
Email Address:
ychen@vims.edu
Individual: Matthew Kirwan
Organization:Virginia Institute of Marine Science
Position:Associate Professor
Address:
PO Box 1346, 1375 Greate Road,
Gloucester Point, Virginia 23062
Phone:
8046847054 (voice)
Email Address:
kirwan@vims.edu
Contacts:
Individual: Yaping Chen
Organization:Virginia Institute of Marine Science
Address:
2385 Jacqueline Dr,
Apt. 501C,
Hayes, VA 23072 United States
Phone:
2178192135 (voice)
Email Address:
ychen@vims.edu
Individual: Matthew Kirwan
Organization:Virginia Institute of Marine Science
Address:
PO Box 1346, 1375 Greate Road,
Gloucester Point, Virginia 23062
Phone:
8046847054 (voice)
Email Address:
kirwan@vims.edu
Associated Parties:
Individual: Tyler Messerschmidt
Organization:Virginia Institute of Marine Science
Role:Field technician
Individual: Alex Smith
Organization:Virginia Institute of Marine Science
Role:Graduate Student

Temporal, Geographic and Taxonomic Coverage

Temporal, Geographic and/or Taxonomic information that applies to all data in this dataset:

Time Period
Begin:
1984
End:
2020
Geographic Region:
Description:The study region stretches across all areas between 0-5 m above sea level along the US mid-Atlantic coast, including the Chesapeake Bay and the Delaware Bay.
Bounding Coordinates:
Northern:  39.74Southern:  36.64
Western:  -77.76Eastern:  -74.22

Project

Parent Project Information:

Title:The Virginia Coast Reserve Long-Term Ecological Research
Personnel:
Organization:Virginia Coast Reserve Long-Term Ecological Research Project
Role:pointOfContact
Abstract:The Virginia Coast Long-Term Ecological Research (VCR/LTER) project's research activities focus on the mosaic of transitions and steady-state systems that comprise the barrier-island/lagoon/mainland landscape of the Eastern Shore of Virginia. Primary study sites are located on Hog Island, Parramore Island, mainland marshes near Nassawadox VA and lagoons behind Hog and Wreck Islands. The VCR/LTER has its field facilities at the Anheuser-Busch Coastal Research Center in Oyster, VA.

Maintenance

Maintenance:
Description:Data are updated as needed.
Frequency:asNeeded
Other Metadata

Additional Metadata

additionalMetadata
        |___text '\n    '
        |___element 'metadata'
        |     |___text '\n      '
        |     |___element 'emlEditor'
        |     |        \___attribute 'app' = 'ezEML'
        |     |        \___attribute 'release' = '2022.08.10'
        |     |___text '\n    '
        |___text '\n  '

Copyright 2024 Environmental Data Initiative. This material is based upon work supported by the National Science Foundation under grants #2223103 and #2223104. Any opinions, findings, conclusions, or recommendations expressed in the material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. Please contact us with questions, comments, or for technical assistance regarding this web site or the Environmental Data Initiative. Please read our privacy policy to know what information we collect about you and to understand your privacy rights.

EDI is a collaboration between the University of New Mexico and the University of Wisconsin – Madison, Center for Limnology:

UNM logo UW-M logo