Evaluating the influence of spatial resolution of Landsat predictors on the accuracy of biomass models for large-area estimation across the eastern USA

Ram K. Deo, Grant M. Domke, Matthew B. Russell, Christopher W. Woodall, Hans Erik Andersen

Research output: Contribution to journalArticlepeer-review

4 Scopus citations

Abstract

Aboveground biomass (AGB) estimates for regional-scale forest planning have become cost-effective with the free access to satellite data from sensors such as Landsat and MODIS. However, the accuracy of AGB predictions based on passive optical data depends on spatial resolution and spatial extent of target area as fine resolution (small pixels) data are associated with smaller coverage and longer repeat cycles compared to coarse resolution data. This study evaluated various spatial resolutions of Landsat-derived predictors on the accuracy of regional AGB models at three different sites in the eastern USA: Maine, Pennsylvania-New Jersey, and South Carolina. We combined national forest inventory data with Landsat-derived predictors at spatial resolutions ranging from 30-1000 m to understand the optimal spatial resolution of optical data for large-area (regional) AGB estimation. Ten generic models were developed using the data collected in 2014, 2015 and 2016, and the predictions were evaluated (i) at the county-level against the estimates of the USFS Forest Inventory and Analysis Program which relied on EVALIDator tool and national forest inventory data from the 2009-2013 cycle and (ii) within a large number of strips (∼1 km wide) predicted via LiDAR metrics at 30 m spatial resolution. The county-level estimates by the EVALIDator and Landsat models were highly related (R 2 > 0.66), although the R 2 varied significantly across sites and resolution of predictors. The mean and standard deviation of county-level estimates followed increasing and decreasing trends, respectively, with models of coarser resolution. The Landsat-based total AGB estimates were larger than the LiDAR-based total estimates within the strips, however the mean of AGB predictions by LiDAR were mostly within one-standard deviations of the mean predictions obtained from the Landsat-based model at any of the resolutions. We conclude that satellite data at resolutions up to 1000 m provide acceptable accuracy for continental scale analysis of AGB.

Original languageEnglish (US)
Article number055004
JournalEnvironmental Research Letters
Volume13
Issue number5
DOIs
StatePublished - May 2018

Bibliographical note

Funding Information:
This work was supported by the US Department of Agriculture Forest Service Information Resources Direction Board through the Landscape Change Monitoring System and the NASA Carbon Monitoring System (NASA reference number: NNH13AW62I), in addition to the US Forest Service—Northern Research Station (#14-JV-11242305-012), and the Minnesota Agricultural Experiment Station (project MIN-42-063). We thank three external reviewers for their comments that improved this manuscript.

Publisher Copyright:
© 2018 The Author(s). Published by IOP Publishing Ltd.

Keywords

  • Landsat data
  • LiDAR
  • above-ground forest biomass
  • design-based estimates
  • large-area estimation
  • spatial resolution of predictors

Fingerprint

Dive into the research topics of 'Evaluating the influence of spatial resolution of Landsat predictors on the accuracy of biomass models for large-area estimation across the eastern USA'. Together they form a unique fingerprint.

Cite this