A mathematical programming approach for downscaling projections from a land use change model

Abstract:

Under the prospects of climate change, conciliating economic development with environmental protection is one of major global challenges today. This balance is particularly vital for Brazil and its agricultural sector: agriculture is a significant driver of deforestation, while at the same time being dependent on environmental services, such as rainfall, that may be harmed by deforestation. In this context, land use and land cover change models are important tools for planning policies aiming to attain such balance. One example is GLOBIOM-Brazil, a global partial equilibrium economic model used to analyze the competition for land between the agriculture, forestry, and bioenergy sector in Brazil. In this type of model, the granularity of the projected land use changes is important, and too coarse of a granularity may lead to ill-suited representations of underlying phenomena. In this work, a downscaling method is proposed for the GLOBIOM-Brazil model, in order to improve the model’s granularity from 0.5 degree square cells to 0.01 degree. The downscaling method employs a mathematical programming approach to disaggregate the multiple layers of the model’s output while respecting a series of constraints, such as preserving the total values from the corresponding original cells. Calibration and validation are performed using GLOBIOM-Brazil projections and other datasets, with the method proposed being effective in generating accurate fine data in the validation studies. By allowing finer land cover projections to be obtained, this method is expected to expand the horizon of possible applications for the GLOBIOM-Brazil model.