Local discrepancies in continental scale biomass maps: a case study over forested and non-forested landscapes in Maryland, USA

Table 1 Summary of biomass products used in this study

Product	Sensor and year	Field data and year	Resolution	Forest mask	Approach	Uncertainty map	References
CMS_RF	DRL 2004–2012	2011–2014	30 m	NAIP high-res tree canopy cover	Random forest, regression tree	Percentile error (QRF)	[12, 18]
NBCD_NCE	Landsat + SRTM 2000	2000	30 m	NLCD 2001	Random forest, regression tree	Quality voids	[13]
Blackard	MODIS 2001^a	2005–2009	250 m	NLCD 1992	Cubist, regression tree	Relative error	[14]
Wilson	MODIS 2002–2008	2005–2009	250 m	NLCD 2001 percent tree canopy 25 %	PGNN, kNN		[15]
Saatchi	MODIS + PALSAR + Landsat	2005	~250 m v1^b ~90 m v2^b	NLCD 2006	MaxEntropy, parametric	Percent error	[16]

^aYear is national maps in eastern US.
^bOriginal maps are in lat/lon, where v1 with 0.00222222 ≅ 250 m and v2 with 0.00083333 deg ≅ 90 m.
DRL Discrete Return Lidar, 1–2 m small footprint lidar aggregate, NAIP National Agriculture Imagery Program, QRF Quantile Regression Forests, SRTM Shuttle Radar Topography Mission, PALSAR Phased Array type L-band Synthetic Aperture Radar, PGNN Phenological Gradient Nearest Neighbor, kNN k-nearest neighbor.

ISSN: 1750-0680