PERFORMANCE DA MODELAGEM PARA CLASSIFICAÇÃO DE SÍTIOS FLORESTAIS EM BASES DE DADOS COM OUTLIERS

As informações utilizadas para estimativa da capacidade produtiva de sítios florestais provêm de bases de dados de inventário florestal que podem conter observações discrepantes (outliers). Assim, torna-se necessário a análise de consistência para exclusão destes. Porém, os outliers podem representar determinado padrão de crescimento existente na floresta, logo a exclusão destes pode ser uma ação equivocada. Objetivou-se comparar a performance de diferentes técnicas de modelagem para classificação de sítios florestais, considerando uma base de dados com a presença de outliers. Utilizou-se pares de dados de idade e altura dominante (HD) de parcelas permanentes de Eucalyptus urophyla x Eucalyptus grandis localizadas no norte de Minas Gerais. Foi simulado um outlier de HD. A base de dados foi modelada, com e sem presença de outliers, por regressão linear (RL) e redes neurais artificiais Multilayer Perceptron (MLP) e Radial Basis Function (RBF). Os métodos foram analisados por meio dos critérios estatísticos de precisão: bias, raiz quadrada do erro médio, correlação de Pearson, erro médio percentual e gráfico de dispersão residual. A MLP foi superior para estimativa do índice de sítio. Portanto, a MLP é indicada para classificação de sítios florestais quando há presença de outliers na base de dados. Palavras-chave: índice de sítio; inventário florestal; dados discrepantes. Performance of modeling for classification of forest sites in databases with outliers ABSTRACT: The information used to estimate the productive capacity of forest sites comes from forest inventory databases that may contain discrepant observations (outliers). Thus, consistency analysis is required to exclude these. However, the outliers may represent a certain growth pattern existing in the forest, so their exclusion may be a mistaken action. The objective was to compare the performance of different modeling techniques for forest site classification, considering a database with the presence of outliers. We used pairs of data of age and dominant height (HD) of permanent parcels of Eucalyptus urophila x Eucalyptus grandis located in the north of Minas Gerais. A HD outlier was simulated. The database was modeled, with and without the presence of outliers, by linear regression (RL) and artificial neural networks Multilayer Perceptron (MLP) and Radial Basis Function (RBF). The methods were analyzed by means of precision statistical criteria: bias, square root of mean error, Pearson correlation, mean percentage error and residual scatter plot. The MLP was superior for site index estimation. Therefore, the MLP is indicated for forest site classification when there are outliers in the database. Keywords: site index; forest inventory; discrepant data.

Forest site classification of forest ecosystems in Bohemian Karst (CzechRepublic)

The study focuses on selected forest site units on water uninfluenced sites on limestone in Bohemian Karst. The species composition of phytocoenoses, their potential production and soil properties were assessed. Studied forest site complexes are correctly determined and they provide important information from the aspect of planning and practical nature conservation management. It is necessary to specify the criteria of categoria calcaria in detail, since it is often difficult to distinguish it from categoria mesotrophica. Water retention capacity of soils, a criterion often neglected so far, presents a useful view of forest site water budget and it is a suitable criterion in forest site classification.

Structure and diversity of psocid taxocenoses (Insecta: Psocoptera) in the forest ecosystems of the Fageta abietis s. lat. zone in the Western Carpathian Mts

Structure of psocid taxocenoses (Psocoptera) were studied in forest ecosystems of the Western Carpathian Mts. during 1997–2001. Vegetation tiers were used as a study frame. Lower units of forest site classification system (forest type complexes) were used for a classification of ecological conditions as well. Only a part of material, i.e. individuals that was found in the forest ecosystems of Fageta abietis s. lat. communities (= the 4th beech vegetation tier) was evaluated for purpose of this work. This vegetation tier is widespread in large part of Vsetínské vrchy hills, in the highest parts of Podbeskydská pahorkatina hills, and in the foothills of the Moravskoslezské Beskydy Mts. 1774 adults comprising 32 species were found in total in the 4th vegetation tier. As eudominant species, the following ones were found: Caecilius flavidus and C. burmeisteri, as dominant species, the following ones were found: Philotarsus picicornis, Elipsocus moebiusi and Peripsocus subfasciatus. In natural geobiocenoses with the level of naturalness of 1 or 2, the following species were found: as eudominant species: Caecilius flavidus and Elipsocus moebiusi, as dominant species, the following one was found: Mesopsocus unipunctatum. Taxocenoses of psocids were evaluated by Detrended Correspondence analysis (DCA) and Divisive Cluster analysis (DvClA). This material was compared to another material gained from various vegetation tiers in the Western Carpathians Mts. The characteristic species composition of psocids in the 4th vegetation tier was as follows: Caecilius flavidus – Elipsocus moebiusi – Peripsocus parvulus.

An ecological land classification approach to modeling the production of forest biomass

Forest site classification is a prerequisite to successful integrated forest resources planning and management. Traditionally,site classification has emphasized a phytocentric approach, with tools such as the site index having a rich and longhistory in forest site evaluation. The concept of site index was primarily devised to assess site productivity of an even-aged,single-species stand. Site index has been the primary method of forest site evaluation in support of management for traditionalforest products. However, this method of site classification has been criticized as the needs, perspectives andsocial values of the public regarding forest management have changed the emphasis from timber production to multiplevalueforestry practices. There are alternative approaches to forest site classification that have the potential to meet thegrowing demands placed on forest information for inventory and modeling purposes. Ecological Land Classification(ELC), is a phytogeocentric approach that stratifies the landscape into ecologically meaningful units (ecosites) based onsubstrate characteristics, moisture regime and canopy composition. This approach offers a more holistic view of site productivityevaluation; however, until recently it has been difficult to acquire data to support widespread mapping ofecosites. Remote sensing technology along with predictive modeling and interpretive mapping techniques make the applicationof an ecosite-based approach at the forest landscape level possible. As forest management moves towards the considerationof a broader set of resources (e.g., woody biomass), there is an opportunity to develop new tools for linking forestproductivity to the sustainable production of forest bioproducts with forest ecosites as a solid foundation forsegmenting the landscape. Key words: forest site classification, site index, site productivity, Ecological Land Classification (ELC), ecosites, forest biomass,bioproducts