P?s-estratifica??o e sele??o de parcelas para fiscaliza??o de invent?rios florestais na Amaz?nia

O trabalho foi realizado com apoio da Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior - Brasil (CAPES) - C?digo de Financiamento 001. Submitted by Jos? Henrique Henrique (jose.neves@ufvjm.edu.br) on 2021-04-10T13:28:00Z No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) diego_santos_vieira.pdf: 6073690 bytes, checksum: 578b27f60f74c8ec668274d36f32b147 (MD5) Approved for entry into archive by Jos? Henrique Henrique (jose.neves@ufvjm.edu.br) on 2021-04-12T14:09:26Z (GMT) No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) diego_santos_vieira.pdf: 6073690 bytes, checksum: 578b27f60f74c8ec668274d36f32b147 (MD5) Made available in DSpace on 2021-04-12T14:09:26Z (GMT). No. of bitstreams: 2 license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) diego_santos_vieira.pdf: 6073690 bytes, checksum: 578b27f60f74c8ec668274d36f32b147 (MD5) Previous issue date: 2020 Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior (CAPES) O objetivo desse estudo foi avaliar m?todos de p?s-estratifica??o e, a partir do melhor m?todo, avaliar os efeitos da redu??o da intensidade amostral sobre a precis?o e custos de amostragem. Avaliou-se ainda formas de sele??o de parcelas para fiscaliza??o de invent?rios florestais (IFs). Todas as an?lises foram realizadas a partir de um invent?rio amostral sistem?tico, com 204 parcelas de 0,75 ha. A esse invent?rio atribui-se a denomina??o de IF0. As p?s-estratifica??es foram realizadas de cinco formas, a saber: dividindo o conjunto de parcelas em tr?s partes iguais (M1/3), por meio de an?lise multivariada (MSOUZA), univariada Scott-Knott (MSK), geoestat?stica (MGEO) e por meio redes de Kohonen (MSOM). A partir do IF0 foram simulados mais tr?s invent?rios com redu??es de 50, 66,2 e 77,4% do n?mero de parcelas. Todos os IFs foram analisados considerando-se a amostragem casual simples (ACS), sistem?tica (AS) e p?s-estratificada (APE). As alternativas para determina??o de quais e quantas parcelas devem ser fiscalizadas foram obtidas da combina??o de tr?s formas de distribui??o de parcelas entre estratos (1 - igualmente entre os estratos; 2 - proporcional ? ?rea dos estratos; 3 - aloca??o ?tima de Neyman) e tr?s formas de sele??o de parcelas dentro dos estratos (1 - aleatoriamente; 2 - as mais pr?ximas ? m?dia do volume e ?rea basal; 3 - as mais pr?ximas ? m?dia do volume e que estejam a mais ou menos um desvio-padr?o da m?dia). A alternativa que produziu menores diferen?as absolutas para volume, ?rea basal, densidade e m?dia aritm?tica do di?metro, quando comparada ao IF0 com todas as parcelas foi selecionada para simula??es dos percentuais de parcelas para fiscaliza??o de 18, 16, 14, 12, 11 e 10%. Todos os m?todos de p?s-estratifica??o foram eficientes e aumentaram a precis?o do IF em mais de 50%. O melhor m?todo foi o MSOUZA, que apresentou um dos maiores percentuais de acerto (99,1%), maior ganho em precis?o (56,2%) e maior facilidade de aplica??o. A p?s-estratifica??o via m?todo MSOUZA permitiu redu??es de at? 77,4% do n?mero de parcelas, mantendo-se inalterado as estimativas em n?vel de comunidade e classe de tamanho, al?m de precis?o abaixo de 10%. A redu??o dos custos do IF com aplica??o do MSOUZA foi de U$$ 24.057,91, corroborando a hip?tese de que a p?s-estratifica??o produz ganhos em precis?o e custos de amostragem. Os processos ACS e AS permitiram redu??es de intensidade e custo de amostragem de 66,2% e U$$ 20.811,47, respectivamente. A p?s-estratifica??o por meio do m?todo MSOUZA foi uma excelente ferramenta de aux?lio na fiscaliza??o de IFs, pois facilita o controle da fiscaliza??o com a subdivis?o da ?rea em estratos. O percentual de parcelas selecionadas para fiscaliza??o do IF deve ser de 12%, os quais devem ser distribu?dos igualmente entre os estratos volum?tricos obtidos por meio da p?s-estratifica??o. Al?m disso, dentro dos estratos, deve-se selecionar aquelas parcelas mais pr?ximas ? m?dia e a mais ou menos um desvio-padr?o da m?dia. Tese (Doutorado) ? Programa de P?s-Gradua??o em Ci?ncia Florestal, Universidade Federal dos Vales do Jequitinhonha e Mucuri, 2020. The objective of this study was to evaluate methods of post-stratification and, using the best method, to evaluate the effects of the reduction of the sampling intensity on the sampling precision and costs. Ways of selecting parcels for inspection of forest inventories (FIs) were also evaluated. All analyzes were performed based on a systematic sampling inventory, with 204 plots of 0.75 ha. This inventory is called FI0. Post-stratifications were carried out in five ways, namely: dividing the set of plots into three equal parts (M1/3), through multivariate analysis (MSOUZA) and univariate Scott-Knott (MSK), geostatistics (MGEO) and through Kohonen networks (MSOM). From FI0, three more inventories were simulated with reductions of 50, 66.2 and 77.4% of the number of plots. All FIs were analyzed considering simple casual sampling (SCS), systematic sampling (SS) and post-stratified sampling (PSS). The alternatives for determining which and how many plots to be inspected were obtained by combining three forms of distribution of plots between strata (1 - equally between strata; 2 - proportional to the strata area; 3 - Neyman's optimal allocation) and three ways of selecting plots within the strata (1 - randomly; 2 - those closest to the mean of the volume and basal area; 3 - those closest to the mean of the volume and which are more or less one standard deviation from the mean). The alternative that produced the smallest absolute differences for volume, basal area, density and arithmetic mean of the diameter, when compared to the IF with all the plots was the one selected for simulations of the percentage of plots for inspection of 18, 16, 14, 12, 11 and 10%. All post-stratification methods were efficient and increased the inventory accuracy by more than 50%. The best method was the MSOUZA, which had one of the highest percentages of correct classification (99.1%), greater gain in accuracy (56.2%) and easier application. Post-stratification using the MSOUZA method allowed for reductions of up to 77.4% in the number of plots, keeping the estimates at the community level and size class unchanged, in addition to accuracy below 10%. The reduction in the costs of the IF with the application of the MSOUZA was U$$ 24,057.91, corroborating the hypothesis that post-stratification produces gains in precision and sampling costs. The SCS and SS processes allowed reductions in intensity and sampling cost of 66.2% and U$$ 20,811.47, respectively. Post-stratification using the MSOUZA method was also an excellent tool to assist in the inspection of FIs and definition of the number of plots to be inspected, as it facilitates the control of inspection with the subdivision of the area into volumetric strata. The percentage of parcels selected for inspection by the FI must be 12%, which must be distributed equally among the volumetric strata obtained through post-stratification. In addition, within the strata, one should select those plots that are closer to the mean and more or less a standard deviation from the mean.