USE OF RGB DRONE SENSORS TO ESTIMATE VEGETATION BIOMASS IN A SEMIARID REGION

Uso de sensores RGB em drones para estimar biomassa vegetal em uma região semiárida

Abstract

Remote sensing techniques are currently widely used in environmental analysis due to the ability to collect accurate data in a cheaper and easier way than conventional techniques. However, estimates of vegetation biomass stocks in rangelands using remote sensing techniques are still uncertain due to the tridimensional and uneven growth pattern of the vegetation. A methodology was defined to estimate biomass stocks in different land cover types in the semi-arid region of Pernambuco state, Brazil, using a high definition RGB camera coupled to a drone. Flyovers 30 m above ground level were performed in three field experiments, in areas of dense and open tropical dry forest and pastures, during the dry and rainy seasons. Biomass measured in field was related to nine visible spectrum vegetation indices as independent variables, using multiple linear regression. The determination coefficients ranged between 0.73 and 0.82. The models proved to be a feasible way to estimate the biomass, considering the spatial and temporal amplitude of the study, the vegetation characteristics and the types of soil cover evaluated, which could be improved with the addition of more sampling points. We conclude that RGB sensors are promising to estimate biomass in semiarid regions, especially integrated with vegetation indices.