Перейти к основному содержанию

page search

Library Estimating upper soil horizon carbon stocks in a permafrost watershed of Northeast Siberia by integrating field measurements with Landsat-5 TM and WorldView-2 satellite data

Estimating upper soil horizon carbon stocks in a permafrost watershed of Northeast Siberia by integrating field measurements with Landsat-5 TM and WorldView-2 satellite data

Estimating upper soil horizon carbon stocks in a permafrost watershed of Northeast Siberia by integrating field measurements with Landsat-5 TM and WorldView-2 satellite data

Resource information

Date of publication
декабря 2015
Resource Language
ISBN / Resource ID
AGRIS:US201500211073
Pages
131-157

Arctic soils contain three times as much carbon (C) as all aboveground biomass distributed globally, much of which is stored in permafrost soils. Here, we (1) determine the predictability of estimating soil organic carbon (SOC) using different satellite data, classifications, and methods; (2) estimate the quantity and distribution of SOC for the top 10 cm for the Ambolikha River watershed (~121 km ²) in northeast Siberia, a sub-watershed of the Kolyma River; and (3) produce a hybrid SOC map through data fusion, combining the strengths of each data type. Land cover maps were produced using a pixel-based classification with Landsat-5 Thematic Mapper (TM) data and an object-based classification using WorldView-2 data. Spectral mixture analysis (SMA) was performed on both data types to calculate the fraction of four vegetation types in each pixel, and land cover maps were combined with field measurements of SOC in the top 10 cm. The overall classification accuracy was 69% for Landsat, 81% for WorldView-2, and 82% for the hybrid map. The hybrid map estimated 490.5 Gg of SOC for the top 10 cm within the Ambolikha River watershed, with less severe over- or under-estimated than the Landsat or WorldView-2 maps alone. The results suggest that (a) higher spatial resolution satellite data and object-based classification should be used for land cover classification in the Arctic; (b) integrating multiple sensors maximizes the strengths of different sensors for estimating SOC; and (c) previous studies performed at the regional or pan-Arctic scale using lower spatial resolution data likely underestimate total SOC.

Share on RLBI navigator
NO

Authors and Publishers

Author(s), editor(s), contributor(s)

Broderick, Dylan E.
Frey, Karen E.
Rogan, John
Alexander, Heather D.
Zimov, Nikita S.

Publisher(s)
Data Provider