No abstract

The analysis of the diameter distribution is important for forest management since the knowledge of tree density and growing stock by diameter classes is essential to define management plans and to support operational decisions. The modeling of diameter distributions from airborne laser scanning (ALS) data has been performed through the two-parameter Weibull probability density function (PDF), but the more flexible PDF Johnson’s SB has never been tested for this purpose until now. This study evaluated the performance of the Johnson’s SB to predict the diameter distributions based on ALS data from two of the most common forest plantations in the northwest of the Iberian Peninsula (Eucalyptus globulus Labill. and Pinus radiata D. Don). The Weibull PDF was taken as a benchmark for the diameter distributions prediction and both PDFs were fitted with ALS data. The results show that the SB presented a comparable performance to the Weibull for both forest types. The SB presented a slightly better performance for the E. globulus, while the Weibull PDF had a small advantage when applied to the P. radiata data. The Johnson’s SB PDF is more flexible but also more sensitive to possible errors arising from the higher number of stand variables needed for the estimation of the PDF parameters.

Ground point filtering of the airborne laser scanning (ALS) returns is crucial to derive digital terrain models (DTMs) and to perform ALS-based forest inventories. However, the filtering calibration requires considerable knowledge from users, who normally perform it by trial and error without knowing the impacts of the calibration on the produced DTM and the forest attribute estimation. Therefore, this work aims at calibrating four popular filtering algorithms and assessing their impact on the quality of the DTM and the estimation of forest attributes through the area-based approach. The analyzed filters were the progressive triangulated irregular network (PTIN), weighted linear least-squares interpolation (WLS) multiscale curvature classification (MCC), and the progressive morphological filter (PMF). The calibration was established by the vertical DTM accuracy, the root mean squared error (RMSE) using 3240 high-accuracy ground control points. The calibrated parameter sets were compared to the default ones regarding the quality of the estimation of the plot growing stock volume and the dominant height through multiple linear regression. The calibrated parameters allowed for producing DTM with RMSE varying from 0.25 to 0.26 m, against a variation from 0.26 to 0.30 m for the default parameters. The PTIN was the least affected by the calibration, while the WLS was the most affected. Compared to the default parameter sets, the calibrated sets resulted in dominant height equations with comparable accuracies for the PTIN, while WLS, MCC, and PFM reduced the models’ RMSE by 6.5% to 10.6%. The calibration of PTIN and MCC did not affect the volume estimation accuracy, whereas calibrated WLS and PMF reduced the RMSE by 3.4% to 7.9%. The filter calibration improved the DTM quality for all filters and, excepting PTIN, the filters increased the quality of forest attribute estimation, especially in the case of dominant height.

In this study, for five sites around the world, we look at the effects of different model types and variable selection approaches on forest yield modelling performances in an area-based approach (ABA). We compared ordinary least squares regression (OLS), k-nearest neighbours (kNN) and random forest (RF). Our objective was to test if there are systematic differences in accuracy between OLS, kNN and RF in ABA predictions of growing stock volume. The analyses are based on a 5-fold cross-validation at five study sites: an eucalyptus plantation, a temperate forest and three different boreal forests. Two completely independent validation datasets were also available for two of the boreal sites. For the kNN, we evaluated multiple measures of distance including Euclidean, Mahalanobis, most similar neighbour (MSN) and an RF-based distance metric. The variable selection approaches we examined included a heuristic approach (for OLS, kNN and RF), exhaustive search among all combinations (OLS only) and all variables together (RF only). Performances varied by model type and variable selection approaches among sites. OLS and RF had similar accuracies and were more efficient than any of the kNN variants. Variable selection did not affect RF performance. Heuristic and exhaustive variable selection performed similarly for OLS. kNN fared the poorest amongst model types, and kNN with RF distance was prone to overfitting when compared with a validation dataset. Additional caution is therefore required when building kNN models for volume prediction though ABA, being preferable instead to opt for models based on OLS with some variable selection, or RF with all variables together.

Pesquisadores da área de mensuração florestal têm incluído com frequência em seus estudos o uso das técnicas de inteligência computacional (IC) para realização de trabalhos de modelagem por serem capazes de manipular um grande conjunto de dados e criar modelos robustos. Dentre essas técnicas, se destacam a Rede Neural Artificial (RNA) e a recente Máquina de Vetor de Suporte (MVS). Dessa forma, buscou-se nesse trabalho avaliar a aplicação dessas técnicas (RNA e MVS) no processo de classificação da capacidade produtiva de povoamentos florestais com a inclusão de variáveis edáficas, de manejo e do povoamento, comparando os seus resultados com os obtidos pelo método da curva guia. Foi possível concluir que as técnicas de IC avaliadas são capazes de classificar a capacidade produtiva do local de forma satisfatória, desde que utilizadas as variáveis adequadas; o uso conjunto das variáveis "tipo de solo", "espaçamento do plantio", "idade" e "altura dominante", foi suficiente para classificar os sítios; a RNA foi mais precisa para classificar a capacidade produtiva do que a MVS; a inclusão de muitas variáveis pouco significativas pode prejudicar ou ser indiferente no desempenho das técnicas. Palavras-chave: classificação da capacidade produtiva; redes neurais artificiais; máquina de vetor de suporte; inteligência computacional.