Prediction error aggregation behaviour for remote sensing augmented forest inventory approaches

Abstract In this study we investigated the behaviour of aggregate prediction errors in a forest inventory augmented with multispectral Airborne Laser Scanning and airborne imagery. We compared an Area-Based Approach (ABA), Edge-tree corrected ABA (EABA) Individual Tree Detection (ITD). The used 109 large 30 × m sample plots, which were divided into four 15 subplots. Four different levels aggregation examined: all subplots (quartet), two diagonal (diagonal), edge-adjacent (adjacent) without aggregation. noted that at aggregated depend on selected predictor variables, therefore, effect was studied by repeating variable selection 200 times. At subplot level, EABA provided lowest mean root square error ($\overline{\mathrm{RMSE}}$) values repetitions for total stem volume (EABA 21.1 percent, 23.5 ITD 26.2 percent). also fared best adjacent ($\overline{\mathrm{RMSE}}$: 17.6 17.4 percent), followed 19.3 18.2 percent) 21.8, 21.9 Adjacent less correlated than subplots, resulted clearly lower RMSEs This appears to result from edge tree effects, where omission commission cancel trees leaning one other. performance achieved quartet as expected fundamental properties variance. had similar level ($\overline{\mathrm{RMSE}}$ 15.5 15.3 poorer 19.4