Synergistic use of very high-frequency radar and discrete-return lidar for estimating biomass in temperate hardwood and mixed forests

Table 4 Regression results between stand-based biomass and lidar (PALS or scanning lidar) and/or BioSAR variables

Variables	R ²	Adjusted R ²	C _p	RMSE	CV-R ²	PRESS	Best model
BioSAR only	0.57	0.54	−2.6	31.0	0.49	28,955	\( Y = 151.2 - 4.5\left( {\hbox{a15}} \right) + 4.4\left( {\hbox{b20}} \right) \)
PALS only	0.55	0.51	−2	31.6	0.47	29,596	\( Y = - 1.7 + 7.5\left( {{k_{\rm{al}}}} \right) + 6.1\left( {{p_{{{60}}}}} \right) \)
BioSAR + PALS	0.80	0.78	4	21.3	0.70	17,358	\( Y = - 163.3 + 9.0\left( {{k_{\rm{al}}}} \right) + 9.7\left( {{p_{{70}}}} \right) - 8.4\left( {{\hbox{b4}}{{0}_{\rm{c}}}} \right) \)
Scanning lidar only	0.64	0.61	2.8	28.5	0.55	25,491	\( Y = 178 - 14.3\left( {{p_{\rm{c30I}}}} \right) - 84.8\left( {{k_{\rm{cI}}}} \right) \)
BioSAR + scanning lidar	0.76	0.72	−1.4	24.2	0.67	18,816	\( Y = 144.9 + 10.3\left( {{p_{\rm{c30I}}}} \right) - 49.8\left( {{k_{\rm{cI}}}} \right) - 2.8\left( {\hbox{a15}} \right) + 2.4\left( {\hbox{b20}} \right) \)

The first column shows the type of independent variables used in the model. The second, third, fourth, and fifth columns show the model coefficient of determination (R ²), adjusted R ², model Mallows C _p, and RMSE (in tonnes/ha), respectively. The sixth and seventh columns show the CV-R ² and PRESS, respectively
Y, biomass (tonnes/ha); a15, mean NRCS response at forward 15° to 20° angle bins (decibels); b20, mean NRCS response at backward 20° to 25° angle bins (decibels); b40, mean NRCS response at backward 40° to 45° angle bins (decibels); p₆₀, 60th percentile of PALS height, all hits (meters); p₇₀, 70th percentile of PALS height, all hits (meters); k_al, kurtosis of PALS height, all hits; k_cI, kurtosis of scanning lidar height, all return canopy hits; p_c30I, 30th percentile of scanning lidar height, all return canopy hits (meters)

ISSN: 1297-966X