Skip to main content

Table 7 Model (echo-based and CHM) evaluation for each study site and power model fitted

From: Impact of data model and point density on aboveground forest biomass estimation from airborne LiDAR

Study site Data model Model Parameters Point density (points m−2) R2 RMSE (Mg ha−1) relRMSE(%)
Sierra Nevada Mountains Echo \(\hat{y} = \alpha X^{\beta }\) α = 2.75; β = 1.52 20 0.70 85.15 43.50
10 0.70 85.00 43.42
5 0.71 84.13 42.98
1 0.70 85.10 43.47
CHM α = 11.72; β = 1.07 20 0.70 84.87 43.36
10 0.70 87.72 44.81
5 0.71 93.90 47.97
1 0.74 120.77 61.69
Echo   \(\hat{y} = \alpha x_{1}^{\beta } x_{2}^{\gamma }\) α = 11.50; β = 1.20; γ = 0.88 20 0.79 70.67 36.10
10 0.79 70.71 36.12
5 0.80 69.66 35.59
1 0.79 70.15 35.83
CHM α = 4.22; β = 1.39; γ = −0.62 20 0.76 73.84 37.72
10 0.77 76.98 39.33
5 0.78 85.43 43.64
1 0.78 128.20 65.49
Barro Colorado Island Echo \(\hat{y} = \alpha X^{\beta }\) α = 1.80; β = 1.61 10 0.71 30.08 12.36
5 0.71 30.12 12.38
1 0.70 30.32 12.46
CHM α = 2.07; β = 1.49 10 0.70 30.42 12.50
5 0.70 31.78 13.06
1 0.69 61.30 25.19
Echo \(\hat{y} = \alpha x_{1}^{\beta } x_{2}^{\gamma }\) α = 9.24; β = 1.12; γ = 0.11 10 0.71 30.22 12.42
5 0.71 30.25 12.43
1 0.70 30.37 12.48
CHM α = 12.32; β = 0.97; γ = 0.13 10 0.69 31.00 12.74
5 0.70 31.75 13.04
1 0.70 55.93 22.98
Serra do Mar Echo \(\hat{y} = \alpha X^{\beta }\) α = 2.69 (3.08); β = 1.88 (0.26) 20 0.44 45.71 10.50
10 0.40 48.02 11.03
5 0.44 46.02 10.57
1 0.58 44.42 10.21
CHM α = 1.68 (2.47); β = 2.00 (0.33) 20 0.45 45.47 10.45
10 0.40 50.63 11.63
5 0.40 63.36 14.56
1 0.46 136.39 31.34
Echo \(\hat{y} = \alpha x_{1}^{\beta } x_{2}^{\gamma }\) α = 0.68 (1.43); β = 2.55 (0.42); γ = −3.3 (1.08) 20 0.62 37.73 8.67
10 0.58 39.77 9.14
5 0.59 39.26 9.02
1 0.69 38.93 8.94
CHM α = 2.70 (6.60); β = 2.14 (0.48); γ = −16.79 (16.34) 20 0.45 45.43 10.44
10 0.38 48.90 11.24
5 0.40 49.53 11.38
1 0.37 67.23 15.45
  1. Model parameters are presented for each model. For Sierra Nevada Mountains and Barro Colorado Island 70% of the plots were used for calibration and 30% for independent validation. For Serra do Mar due to the small size of the sample a jackknife approach was used instead
  2. Standard deviation of the parameters is presented in brackets