Planejamento florestal espacial
Resumo
O planejamento florestal espacial é uma opção ainda nova no Brasil, porém antiga nos países desenvolvidos. A sua importância está relacionada à melhor compreensão e análise do talhão dentro da floresta, entendendo ainda suas relações espaciais com outros talhões e áreas vizinhas. A restrição de adjacência é um exemplo de controle espacial na colheita florestal e contribui para a redução de danos ao ambiente e à paisagem. Existem dois tipos de restrições de adjacência: l) URM (Unít Restriction Modef) ~ que impede a colheita de talhões adjacentes no mesmo período de tempo; e 2) ARM (Area Restriction Modef) ~ que permite a colheita deste tipo de situação, desde que não ultrapasse um limite de área máxima estabelecida. Assim, o estudo teve como objetivo geral aplicar o planejamento florestal espacial, levando em consideração o problema de agendamento da colheita florestal. Além disso, analisar o grau de impacto econômico no VPL e na produção volumétrica de madeira, entre as opções de planejamento florestal tradicional e espacial e ainda testar a utilização do algoritmo genético (AG) e simulated annealing (SA) na resolução de cenários florestais, comparando com o método exato via programação linear inteira (PU). Foram gerados 4 tipos de casos de planejamento distribuídos em 32 cenários florestais, tendo como referência o modelo tipo I de Jonhson e Scheurmann (1977) e as restrições URM, ARM50 e ARM70. Os dados foram obtidos de um reflorestamento de eucalipto com 1.490 ha, contendo 52 talhões, pertencente à empresa VCP (Votorantim Celulose e Papel) na região de Caçapava-SP. Os métodos de resolução foram o AG, SA e PU, porém houve a necessidade de modelar os parâmetros das metaheurísticas, através de testes para diferentes configurações. Os resultados mostraram que as metaheurísticas foram influenciadas pela configuração testada, onde a melhor configuração para o AG foi: população inicial (300 indivíduos), operador de seleção (amostragem estocástica universal), taxa de trocas gênicas pelo crossover (50%), taxa de alteração gênica na mutação (20%) e replacement (steady state 3 ou 50% de substituição); e para a metaheurística SA: o uso da opção busca3, função3 de resfriamento e porcentagem de 10% na taxa de vizinhança de busca. A utilização de restrições de adjacência aumenta o número de restrições dos modelos, principalmente quando são empregadas as restrições do tipo ARM 50 e 70. O uso do planejamento florestal espacial promoveu uma redução do VPL na faixa de 2,10% (caso 4) a 3,74% (caso 2), proporcionado pelas perdas volumétricas de madeira colhida. As metaheurísticas foram eficientes e eficazes na resolução dos cenários, onde os desvios percentuais de VPL em relação à PLI foram: AG 0,76% e SA 2,45% (caso 1), AG 13,60% e SA 6,02% (caso 2), AG 4,18% e SA 7,66% (caso 3) e AG 2,91% e SA 3,31% (caso 4) para os cenários com HP de 7 anos. No caso de problemas onde a PU apresenta problemas em tempo de processamento, as metaheurísticas podem ser aplicadas como uma opção viável. Os casos de planejamento florestal espacial demonstram ser perfeitamente aptos e viáveis. As restrições de adjacência URM, ARM50 e ARM70 controlam as relações espaciais entre talhões e podem ser empregadas como uma opção no planejamento florestal. Spatial forest planning is still a new option in Brazil, but old in the developed countries. Its importance is related to the better understanding and analysis of the stand within the forest, understanding, still, their space relationships with other stands and neighboring areas. The adjacency constraints are an example of space control in the forest harvest, and it contributes to environmental and landscape damage reduction. Two types of adjacency constraints exist: 1) URM (Unit Restriction Model) -it avoids the harvest of adjacent stands at the same period of time; 2) ARM (Area Restriction Model) -that it allows the harvest of this type of situation, as long as it does not cross an established limit of maximum area. As such, this study had, as general objective to apply the spatial forest planning, taking into account the forest harvest scheduling problem. An objective was also to analyze the degree of economic impact on the NPV and on the volumetric production of wood, between the traditional and spatial forest planning options. Another objective was to test the use of the genetic algorithm (GA) and simulated annealing (SA) on the resolution of forest scenarios, comparing with the exact method via complete linear programming (lLP). For this, 4 types of planning cases were generated, distributed in 32 forest scenarios, having as a reference, the type I model of Jonhson and Scheurmann (1977) and the URM, ARM50 and ARM70 constraints. The data were obtained from a reforestation with 1,490 ha of eucalyptus, containing 52 stands and belonging to the VCP (Votorantim Celulose e Papel) company in the area of Cayapava-SP. The resolution methods were GA, SA and ILP, however there was the need to model the metaheuristic parameters, through tests for different configurations. The results showed that the metaheuristics were influenced by the configuration tested, where the best configuration for GA was: initial popUlation (300 individuals), selection operator (stochastic universal sampling), gene exchange rates by crossover (50%), rate of gene alteration in the mutation (20%) and replacement (steady state 3 or 50% of substitution); and the metaheuristic SA: the use of the option busca3, cooling funya03 and percentage of 10% in the search neighborhood rate. The use of adjacency constaints increased the number of restrictions for the models, mainly when employing the constraints of the ARM 50 and 70 type. The use of spatial forest planning promoted a reduction of NPV in a range from 2.10% (case 4) to 3.74% (case 2), proportionate to the volumetric loss of harvested wood. The metaheuristics were efficient and effective in the resolution of the scenarios, where the percentile deviations of NPV in relation to ILP were: GA 0.76% and SA 2.45% (case 1), GA 13.60% and SA 6.02% (case 2), GA 4,18% and SA 7.66% (case 3) and GA 2.91% and SA 3.31% (case 4) for the 7 year HP scenario. In the case of large problems, where ILP is not applied, the metaheuristics can be applied as a viable option. The cases of spatial forest planning demonstrate to be perfectly capable and viable. The URM, ARM50 and ARM70 adjacency constraints control the space relationships among stands, and can be used as an option in forest planning.
Collections
- Teses & Dissertações [10808]
