The accurate estimation of vegetation biomass over large areas is still a significant challenge. The evolution of low-cost geotechnologies is rapidly facilitating this task, but their use to estimate biomass on tropical dry forests (TDFs) is still scarce. TDFs cover large areas and are currently among the most threatened global ecosystems due to intense anthropic pressures. Therefore, there is a need for the development of tools capable of monitoring the complex spatial-temporal dynamics of these highly variable ecosystems. We developed a methodology to estimate aboveground biomass stocks in a tropical dry forest (known as “Caatinga”), native to the semiarid region of Northeast Brazil, using field and orbital data, regression models, and high-performance computation. We combined twenty types of vegetation spectral indices and isolated spectral bands of the Landsat 8 satellite OLI sensor to generate regression model equations that could predict biomass stocks. The selected models were applied in a cloud processing routine developed using Google Earth Engine (GEE) to estimate the biomass (30 m pixel resolution) in the entire semiarid region of Pernambuco state (85,900 km2), using the years 2017 and 2018 as references. The regression models were capable of predicting the biomass in dense Caatinga (μ 32.6 Mg ha−1 - R2 0.52), open Caatinga (μ 11.4 Mg ha−1- R2 0.56), and pasturelands (μ 4.2 Mg ha−1-R2 0.94). Using the equations previously defined and the map of the Brazilian Annual Coverage and Land Use Mapping (Mapbiomas) we developed a workflow script in GEE to plot a map that represents the biomass stock over the entire semiarid region of Pernambuco. The system plotted the distribution and amount of biomass with similar results to those available in the literature. Given the large area of the region, the biomass was estimated with reasonable precision in spatial and temporal distribution, while demanding little time and hardware costs. Future refinements of the method will be pursued, but the proposed version represents an improvement in the capability to provide reliable estimates and contributes to monitor the biomass dynamics in the in semiarid region of NE Brazil.
Read full abstract