Biomassa e carbono por meio de inventário convencional e LiDAR em floresta seca no Nordeste do Brasil

Abstract

With the growing environmental concern about climate change in recent years, there has been more demand for efficient alternatives using indirect methods for studies of biomass quantification and forest carbon stock. This work aimed to obtain estimates of biomass and carbon by means of data from conventional forest inventory and LiDAR technology in a dry forest. In the first segment, data were collected from 507 trees distributed among 14 species and 12 genera in an area under forest management with predominantly Caatinga vegetation and characterized by shrub vegetation, located in the municipality of Floresta, in the state of Pernambuco. The green weight above the soil (stem and branches), total and commercial heights and base diameter (0.30 cm of the soil level) were measured for each individual, as well as values of wood density, aiming to adjust and validate three allometric simple regression models (the base diameter as an explanatory variable) and five multiple regression models (explanatory variables: Db, total height and wood density). Two selected local equations were compared with generic pantropical equations and applied to the sample trees of this study. Biomass predictions were performed considering the inventory data of live trees in two areas with different usage histories; one with a history of major disturbance (correntão) and one with less disturbance (transposition). Among the biomass models tested, the Schumacher-Hall logarithms and the Chave-Model I presented better performance. There was a significant improvement in the pantropical models when height and density variables were included; the largest prediction errors were generated by equations of simple local and pantropical inputs. In the second segment, we used conventional forest inventory data from two areas together with LiDAR overfly data, generating local biomass estimates from a developed local equation and the carbon contents obtained from local species. With the data from LiDAR technology, the metrics were extracted from its cloud of points and were used as the independent variable. Next, three types of models were analyzed for data analysis for constructing the allometric models of biomass and carbon per hectare: Multiple Linear Regression with Major Components - PCA, Multiple Linear Regression and Multiple Linear Regression with Stepwise. The generated equations were analyzed by comparing statistical criteria. After selection of the best biomass and carbon equation, the carbon estimates were calculated by area evaluated at the plot level. Both the TAGB and TAGC models of best fit was the Multiple Linear Regression with Stepwise, therefore concluding that the LiDAR data can be used for estimating the biomass and total carbon in a dry tropical forest, as proven by an adjustment considered in the employed models with good correlation between the LiDAR metrics.