Comparação da Produtividade Primária Bruta entre os Sensores OLI/TIRS e MODIS

Abstract

Carbon sequestration by terrestrial biomes plays a significant role in the global carbon cycle to mitigate the increase in atmospheric CO2 and the consequences of climate change. Gross Primary Production (PPP) is an important biophysical parameter of any ecosystem and plays a key role in the spatio-temporal dynamics of CO2. Uncertainty in gross carbon sequestration and its understanding at local, regional and global scales can be overcome by monitoring the surface processes of the earth at high spatial and temporal resolutions. Surface methodologies for PPB calculation provide accurate estimates, but their implementation and operational costs are high and their representativeness limited, except in large and homogeneous areas. In this sense, methods that use remote sensing are presented as an advantage to expand spatial coverage of PPB estimation, provide synthetic and consistent observations of vegetation in heterogeneous and low cost areas. Several orbital sensors have been used to determine PPB and to extrapolate local measurements. Of particular note are the Landsat satellite sensors that provide longer time records of earth surface observations and the launch of the Landsat 8 OLI / TIRS satellite in 2013 continues this legacy. The objective of this study is to present an algorithm to estimate PPB by remote sensing techniques, from sensor images, in different soil uses. The methodology is based on the estimative model of the absorbed photosynthetically active radiation - RFFA by the vegetation, combined with the light efficiency model. For this, the atmospheric and radiometric corrections of the images were performed previously and as basic input for the model, the balance of radiation and other components of the energy balance were calculated through the SEBAL (Surface Energy Balance Algorithm for Land) algorithm. Subsequently, the estimation of PPB was compared through the OLI / TIRS sensor, with the product MOD17A2. The comparison was based on the Mean Error (EM), Root Mean Square Error (REMQ). Statistical analyzes were performed in software R, version 3.4.2. The PPB estimated by the sensor made it possible to analyze the spatial distinctions between the land uses in the region. The representative image of the dry season presented the highest values of PPB indicating that the algorithm extrapolates to PPB when radiation supply, NDVI values and surface humidity are high. However, the representative image of the rainy season presented a higher Spearman correlation with the product MOD17A2, still underestimating for all soil uses. The differences between the temporal and spatial resolution of each sensor also influence the comparison, since they present specific surface data of the imaging date. The proposed algorithm, due to its applicability to the OLI / TIRS sensor, can be useful for the more accurate analysis of PPB in local study areas. However, it is necessary to better investigate temporal estimates for OLI / TIRS sensors with surface data, which integrated can determine more accurate values of PPB, especially in the more remote areas.