Characterizing and predicting biomass production in sugarcane and eucalyptus plantations in Brazil

Resumo

For major emerging countries with significant land resources such as Brazil, the Agriculture, Forestry and Land Use (AFOLU) sector is one of the major sources of greenhouse gas (GHG) emissions. At the same time, this sector offers a large potential for climate change mitigation through best management practices. São Paulo state is the main producer of both eucalyptus and sugarcane in Brazil, and there is potential for expansion in the area managed under both crops. These land uses can have a large impact on the regional carbon balance, both though carbon fixation in the vegetation and soils and though offsetting fossil fuel emissions by the production and consumption of biofuels. Process-based models, calibrated and validated previously, and applied spatially could help quantifying the fluxes and stocks of carbon at the field level, with different time scales (from years to decades) and spatial scales (from stands to regions). The main objective of this project is to take advantage of satellite and field data collected in the past decade; state-of-the-art process-based models; and computational tools that allow processing large amounts of data to assess the carbon dynamics of eucalyptus and sugarcane in São Paulo state. A bottom-up approach will be used, by parameterizing and testing process models based on field measurements, and then upscaling to São Paulo State. Images from Landsat will CBERS, Terra and Aqua satellites will be registered, radiometrically corrected and organized into a data set covering the 2000-2015 period in São Paulo State, with the associated metadata. Soils data will be compiled from published soil surveys, and meteorological variables will be collected from weather stations and global models. Different vegetation indices time-series will be produced, like the normalized difference vegetation index (NDVI) and the enhanced vegetation index (EVI). A time-series classification method will be used, in which the algorithm will use the seasonal and/or pluriannual vegetation indices profiles to classify the vegetation through time series pattern analysis. Estimation of vegetation structural parameters and in particular Leaf Area Index (LAI) and/or the fraction of absorbed photosynthetically active radiation (FAPAR) will also be derived from remote sensing data. Data collected over the last decade by EMBRAPA, CIRAD and CTBE on Eucalyptus plantations and sugarcane fields will be used to calibrate and validate models such as the G'Day process-based model. Both the remote sensing correction and processing, the classification procedure (calibration and application), process-based modelling at the site scale and the upscaling procedures will require a large amount of calculations and data processing. Therefore, novel computer science tools and techniques will be used in this project, including cloud-based computing, machine learning and visualization interfaces for spatial data. The expected outcome of accurate predictions of carbon fluxes and dynamics with satellite-data constrained crop models is in high demand from the scientific community, policy makers, and the forestry and agricultural sectors. Additionally, the science developed in this project will be useful as input to applications in other crops and regions. (AU)