Enhancing Satellite-Based Estimations of Carbon Flux in Caxiuanã Forest Using Artificial Intelligence

The great capacity of the Amazon Forest in absorbing carbon dioxide of the atmosphere is primordial to control climate change. This natural carbon sink performs a significant role mitigating CO2 levels. From this perspective, measuring CO2 flux variables such as Gross Primary Productivity (GPP) and Net Primary Productivity (NPP) is decisive to assess the Amazon's carbon sequestration capacity accurately and contribute to the adoption policies for conservation and sustainability. These measures can be acquired, or calculated using flux towers, greatly equipped stations with a variety of sensors to measure meteorological information. Nevertheless, the limited quantity and short temporal range is a difficult challenge to overcome. To support this measuring, the values of GPP and NPP also can be estimated by the means of satellite images, using sensors such as the MODIS associated with an algorithm that treats the data for approximating them to the ground level measures. However, uncertainties and generalizations associated to the global calibration in the algorithm can cause inaccuracy in the estimations when compared to the in situ  data of the towers. Therefore, the present article has the objective of reducing the discrepancies between the estimations of GPP and NPP obtained by the MOD17 algorithm and the real data measured in the Amazon region using artificial intelligence, executing a case study of the application of this method in the National Forest of Caxiuanã, known as the oldest of these conservation units in the brazilian Amazon.