Alteração da cobertura florestal e biomassa em área de manejo florestal no Estado do Acre integrando dados de campo e sensores remotos
Logging based on the principles of forest management in the Amazon region is a promising management strategy for biodiversity conservation and carbon sequestration. However, the control methods currently used are insufficient to monitor frequently and on a large scale the occurrence of areas for log...
| Autor principal: | Pantoja, Nara Vidal |
|---|---|
| Grau: | Tese |
| Idioma: | por |
| Publicado em: |
Instituto Nacional de Pesquisas da Amazônia – INPA
2020
|
| Assuntos: | |
| Acesso em linha: |
https://repositorio.inpa.gov.br/handle/1/4983 http://lattes.cnpq.br/7250980918375600 |
| Resumo: |
|---|
|
Logging based on the principles of forest management in the Amazon region is a promising management strategy for biodiversity conservation and carbon sequestration. However, the control methods currently used are insufficient to monitor frequently and on a large scale the occurrence of areas for logging. An area of tropical forest under forest management was mapped in the Antimary State Forest in the state of Acre, Western Brazilian Amazon, to assess the potential for the detection of impacts produced by forest operations (roads and landings) using different levels of data acquisition: field, aerial and orbital. Within the Antimary State Forest, roads and landings were mapped in the field with GPS devices. The remote mapping of these structures was achieved using the nonphotosynthetic vegetation (NPV) fraction images obtained from the mixing model of the Landsat images from years 2009 to 2015 processed in the CLASlite program. RapidEye images from the years 2012 to 2015 were used to identify areas degraded by selectivelogging. Airborne LiDAR data were used to create a high-resolution canopy relative density model (RDM) and to identify the logged areas. The mapping of logging in the study area showed 398 ha and 1,428 ha using automatic classification and visual interpretation, respectively. The overall accuracy was estimated at 0.50 ± 0.060 for the classification of Landsat images and 0.788 ± 0.149 for the classification of the RapidEye images. The underestimated logged area according to the reference data was 4,537 ha using Landsat and 705 ha using RapidEye. The size of the landings affected the Landsat detection since it detected only 40% of the landings, while 98% of landing were detected by LiDAR. The mean area of detected logging ladings was 435 m2 while the mean area of those undetected was 302 m2, with a significant difference in detection being the function of the size of the patios (t = -4.0076, df = 38, p≤0.01). Monitoring of permanent plots showed differences in forest cover stocks before and after selective logging. These results emphasize the need for research related to forest management in order to understand the spatial variability of roads, landings and harvested tree gaps being detected by remote sensing. While GPS is more reliable for the mapping of forest infrastructure, LiDAR and Landsat data are effective in remotely quantifying the extent of exploitation impacts in tropical forests by subsidizing forest management and monitoring |